SSL DUENDE MINI - FireWire DAW Expander - Now with conclusions!

[Anderton]

SSL Duende Mini Pro Review: Prologue

Native processing that's done inside the computer is certainly convenient and cost-effective, but eventually, you'll run into limitations: The computer might not be powerful enough to run as many plug-ins as you want, or you may find that really high-quality plug-ins require so much CPU power you settle for ones that are "good enough."

There are various ways to get around these problems, but one of the most popular ones is to add hardware DSP assistance to your computer, either in the form of a card that plugs into a motherboard slot, or an external box that communicates with the computer via a high-speed port. Digidesign popularized this trend with their original Sound Accelerator board for the Mac, and since then, many other solutions have come along: Creamware's SCOPE system, TC Electronics' PowerCore, Universal Audio's UAD-1 (and now UAD-2), Focusrite's Liquid Mix, and others.

But now Solid State Logic (http://www.solid-state-logic.com) (a/k/a "the company that makes those really great mixers you'll never be able to afford") has thrown its hat into the outboard DSP ring with Duende and Duende Mini, two FireWire-based boxes that contain significant DSP and run SSL's custom plug-ins. This is part of an interesting re-alignment of the company that has seen it extend its reach into lower-end markets, without forsaking the clientele that put it on the map. The single rack space Duende (http://pro-audio.musiciansfriend.com/product/Solid-State-Logic-Duende-DAW-Plug-in-Platform?sku=245510) lists for $1,875, while Duende Mini (http://pro-audio.musiciansfriend.com/product/Solid-State-Logic-Duende-Mini-DSP-Powered-DAW-Plug-In-Platform?sku=482193) - the subject of this Pro Review - is functionally similar to Duende, but fits in a third-rack space and retails for $995. Furthermore, a PCIe card version called Duende PCIe (http://pro-audio.musiciansfriend.com/product/Solid-State-Logic-Duende-32-Channel-PCIe-Card?sku=708756) retails for $1,495. (The links give a good idea of street prices.) In this review, I'll simply refer to "Duende" so I don't have to type "Mini" all the time...

Incidentally, in case you wondered, duende is a Spanish word that relates to the arts, and connotes emotion and authenticity. Well, SSL's Duende is certainly authentic, and it really does seem that many design decisions were made to maximize "musicality" - like Duende itself, a difficult work to define precisely. And to give a little foreshadowing, they've succeeded in creating very "musical" software but perhaps not for the reasons you might expect. In fact, one reason I'm looking forware to doing this review is I get to record a bunch of audio examples :)

Unlike most Pro Reviews, where the review starts when I open up the package from UPS or Fed Ex, I've been using Duende for a few months now in my music and soundtrack work. However, Duende isn't just about what comes with the package (a fine-sounding channel strip; see the attached image), but the optional-at-extra-cost programs you can run on the platform. For example, if you've always craved having the famous SSL bus compressor as a software plug-in, now you can. Best of all, Duende is cross-platform, so you can run VST, AU, and (with a little assist from Fxpansion's wrapper) RTAS.

While we'll cover the basic package, we'll emphasize the optional plug-ins so you can more easily decide which ones are most important to what you need. Although outfitting Duende to the max isn't cheap, one advantage of the Duende approach is that you start off with the channel strip (which can cover a lot of your needs and shouldn't be underestimated), but then purchase additional plug-ins as your needs dictate and your finances permit. It's even possible to expand Duende itself to run more plug-ins.

Next, let's look at the construction, and compare the advantages and disadvantages of using an external box as opposed to a PCI card.

[Anderton]

The external box containing the DSP is quite plain, as you can see from the attached image), although it's compact and its all-metal construction gives a reassuring feel regarding reliability. (The second image shows the PCIe card, which I thought you might find interesting). The front panel has an on-off switch, and the rear panel, two FireWire ports and a jack for the included AC adapter (however, I've been able to bus power Duende without problems, other than when using it with my laptop, which has only a four-pin FireWire connector and therefore doesn't provide power).

Duende has a fixed amount of DSP power available for plug-ins. This is unlike native plug-ins, which face a constantly fluctuating amount of available power from the host CPU. As a result, with native systems, it's important to leave some "CPU power headroom." With Duende, you can load it up fully and know it will always have enough power.

Duende comes from the factory able to run 16 mono plug-ins at 44.1/48kHz, however you can also run stereo plug-ins instead of two mono instances. The maximum number of stereo plug-ins is 8, but you can have any combination of mono and stereo plug-ins as long as the combined number of channels is 16 or less.

However, an additional $399 (250 Euros) upgrade, available online, can expand Duende to 32 slots. Once you start getting into Nuendo, this is well-worth it: You can add, for example, 32 SSL channel strips to the mixer in your DAW without loading down your computer's CPU. That's powerful stuff, particularly because of all the channel strip can do (by the way I should mention the channel strip models SSL's C200 digital console). And if you really get into Duende, you can run two mini simultaneously, and even run two PCIe cards with the two Duende Minis. I doubt, however, that you'll need this much stuff!

With high sample rates (88.2 - 96kHz), you have half the number of slots. So, a maxed-out Duende would give 16 slots (16 mono, 8 stereo). Duende doesn't support 176.4 or 192kHz sample rates...is that a deal-breaker for anyone? I didn't think so.

One point worth mentioning is that Duende is truly "disconnected" from your computer. The advantage is that this takes a huge load off your processor, but the disadvantage is that if you track-bounce in a DAW with Duende effects, you need to bounce in real time. This is because hooking up Duende is essentially like having, say, a hardware reverb or whatever patched into your DAW as an external audio effect - but the signal travels on FireWire instead of standard cables. Of course, this isn't a deal-breaker but if you have the unexpanded Duende and plan to do track-bouncing to add effects to tracks without having to get the upgrade, this will take longer if you're used to programs that do faster-than-real-time bouncing. This is also why I found it better to bypass any processing at the plug-in itself, which happens instantly, as opposed to bypassing at the DAW, which sometimes led to a brief hiccup.

Getting Duende set up and working was simple - it's just a matter of installing the software (check online for the latest version), and plugging in Duende. SSL's documentation never left me scratching my head, and the on-screen prompts made sense. Then again, I always follow the golden rules of FireWire audio devices: Give them their own ports, don't daisy-chain anything with them, and use only approved chip sets.

The only problem I encountered during installation was having Duende say certain plug-ins weren't unlocked when in fact they had been. This was a minor bug in the authorization process which I believe has since been remedied.

Zooming out to the big picture, Duende is part of the trend to move some hardware "out of the box," turning the computer into more of a recorder/editor than a do-all studio. You'd think that with things like 8-core computers this wouldn't be an issue, but Duende adds an element of stability to a system because its operation is so predictable.

Before getting into the plug-ins themselves, I should add that SSL is very helpful - the FAQs on their web site cover a broad range of topics and are easy to understand, the documentation is comprehensive without being tedious, and even the installation routine screens made sense. (Note that I installed it on a PC; although come to think of it, it's FireWire so I can just install the appropriate software on my Mac, plug in Duende, and have it work there as well.)

More tomorrow...

[Anderton]

Let's check out the channel strip. For completeness, though, let's take a quick look at the control panel software. There's not much to it, and no adjustable parameters. The first attached image shows the panel in action: The top, highlighted section shows the 32 available slots, spread over 4 DSPs. The zeroes indicate active slots that can host plug-ins. Seems to me it would have made more sense to label an active slot 1, but no big deal.

Okay, now for the EQ and Dynamics Channel strip. It has several building blocks:

1. Low cut and high cut filters.
2. 4-band equalizer with low shelf, high shelf, and two parametric stages; the shelving stages have "Bell" switches that select more of a wide bandpass response.
3. Compressor
4. Gate/expander.

There's also an input stage and output stage, each with metering.

The Filter, highlighted in the second attached image, has low cut (up to 500Hz) and high cut (down to 3kHz) controls. The section also includes a button that chooses whether the EQ comes before or after the filter, and a Side Chain button.

Seems simple, right? Well, it is, but already you can start to hear there's a particular SSL quality to the sound. In fact, let's post a couple audio examples.

The source file is a loop from the first Discrete Drums sample library. The Duende Low Cut audio example automates the low cut to go from the lowest frequency to the highest. The Duende High Cut audio example automates the high cut knob so that it goes from the highest to the lowest frequency (i.e., where it cuts the most). Note that the examples are 320kbps MP3 files.

[Anderton]

First of all, I apologize for the delay between posts. I had a freak accident last week - there was a snowstorm and as I was climbing down from the roof to do a repair, the ladder's crossbrace broke and I landed pretty hard on my head and hips. That made it difficult to work or sit for extended periods of time. I'm doing better now, so back to the review - and we'll continue where we left off, with the channel strip. Let's look at the channel strip equalizer in more detail.

Referring to the first attached image, The EQ has four bands: Low shelf, high shelf, and two fully parametric midrange bands. However while all bands have the expected Frequency and boost/gain controls (with the parametric stages having Q controls), the shelving filters also have a "bell" option that adds a resonant peak.

The low shelf covers 40Hz to 600Hz, with up to 16.5dB of boost or cut. The lower mid parametric stage covers 200Hz to 2kHz, and the high mid parametric stage, 600Hz to 7kHz. Both offer a boost/cut of +/-20dB, with variable Q from 0.5 to 2.5. The high shelf covers 1.5kHz to 22kHz, also with +/-20dB of boost or cut.

However, in typical Duende fashion, SSL added some "under the hood" options. The innocent-looking "E" button actually selects between two entirely different EQ characteristics. In this case, let's paraphrase from the manual as the explanation of the differences between the two is quite succinct.

The EQ defaults to the G Series EQ curve, which has a degree of overshoot/undershoot (depending on whether you are boosting or cutting) below the selected HF frequency or above the selected LF frequency. At any Q setting, the bandwidth of the High Mid and Low Mid sections varies with gain, whereby an increase in boost or cut increases the EQ's selectivit.

I find this EQ most effective for "sweetening" the sound, and when using moderate amounts of EQ. For example, if you want the high end of a drum kit to "shine" without sounding shrill, the G-Series EQ is ideal.

The E setting emulates the SSL E-Series EQ. With this curve, the bandwidth of the High Mid and Low Mid sections remains constant at all gains, so at lower gains the EQ curves are comparatively narrower for a given Q setting.

SSL characterizes this sound as more "aggressive," and I can't argue with that. They mention it being useful for drums, as you can obtain relatively high Q values at low gain settings. In fact...check out the two audio examples. The first one is an unprocessed drum loop. In a dance mix it was too "thick" in the midrange, and competed too much with other instruments, muddying the sound.

The second example uses the channel strip settings shown in the attached image. In this case, the bass is boosted a lot so that the kick gives a solid, low-end "thud," with the High Mid Filter adding a considerable amount of boost (at just under 3kHz) to bring out the "crack" of the snare, yet not get in the way of other "midrangey" instruments. The High Frequency shelf is adding a bit of a resonant peak in the high end to really bring out the "air." Overall, the processed version "cut" through a track with tons of midrange a lot better.

[aliensexist112]

thanks for the info btw thank god that you wasn't harmed that bad :)

[Phil O'Keefe]

Yeah, glad to hear you're OK Craig.

I'm using an "original recipe" Duende, which comes with the channel strip and the bus compressor - both are excellent in my opinion - but I'm looking forward to reading your impressions about some of the other optional plugins too.

[Anderton]

Given that the EQ can add real "character" to tracks if you push it a bit, I was a little taken aback that the Compressor section leans toward a more natural, smooth sound - at least that's what I'm hearing. But on another level, that makes sense because the Bus Compressor, alluded to by Phil in the previous post, has some of that "glue the tracks together" attitude associated with SSL compression. So, you have choices. However, as with everything else in Duende, there are also a few surprises.

The attached image shows a screen shot, with the Compressor highlighted. Actually SSL considers the Gate/Expander section part of the overall dynamics control that includes the Compressor, but for purposes of this review it makes sense to consider them separately.

Control-wise, there are the expected controls for Threshold, Ratio, and Release Time - we'll get into the more esoteric ones shortly, but for now, let's start off with a fairly plain application.

The audio example has six measures of an unprocessed synth part, followed by a very brief gap, followed by six measures of the same part but compressed gently with the channel strip compressor. The difference may not be obvious at first, but if you loop the file, you'll hear that while the "strong" parts of both sections are the same, in the first part there are several notes that are weaker and "drop out" a bit compared to the file as a whole. In the compressed section, these notes are given equal strength; overall, the compressed sound is stronger and more consistent.

Given that this is intended to be more of a background part, the consistency helps tremendously as the dynamics in the lead parts can take more of the spotlight. But the other interesting thing with these two sections is that the compressed version does so very subtly and gently. This isn't at all an "in your face" compressor, but more of a "consistency-izer" (to coin a very awkward word!).

The more I worked with the compressor, the more I found it excelled at giving gentle, subtle lifts to parts. It doesn't make the parts "jump out" as much as it causes them to, well, "stand up straight" in the mix. Very cool. And for narration, it helped give consistency to my voice without adding artifacts that said "HEY! THIS SIGNAL IS BEING COMPRESSED!"

However, you're not limited to simply making nice sounds at all, as we'll see in the next post.

[Anderton]

While the channel strip compressor can play nice, it can also get nasty. The key to this is the PK button, which chooses between peak and average response. Another button, Fast Attack, selects between a fixed fast attack (3ms to reduce the gain by 20dB) or program-dependent attack that ranges from 3 to 30ms.

Note that the PK button also changes the compressor "knee." If the ratio is relatively low, then the compressor knee is fairly soft ("over easy"). If the ratio is all the way clockwise, then the compressor becomes more of a limiter and when PK is enabled, this adds a "hard knee" characteristic.

I have two main tests for a compressor: One is how artfully it can compress a signal, and as we saw in the previous post, Duende passes that test. My other one is to see if I can get that "squashed" drum sound that was so popular in the 60s with groups like the Beatles and Traffic, and has now made a resurgence in dance mixes by musicians like Eric Prydz but has also become a bit of a hallmark for the "French" dance music sound. While most compressors can do this, what I look for is a smooth, "pillowy" quality that "breathes" instead of coughs.

Once again, Duende comes through. The attached image shows the settings I used for the audio example, which has a drum part being squashed big-time through the compressor. The key here is using Peak detection, but also, using a fast attack and setting a short release. (Program dependent attack reacts more rapidly to changes, so snare hits become more prominent instead of squashed, and the overall sound is more "percussive" instead of "squishy.")

In any event, listen to the audio example and hear for yourself. The compressor may look basic, but those buttons can add quite a change to the sound.

[Anderton]

The Gate/Expander is the final main section of the EQ and Dynamics Channel Strip; all that remains to be covered for the channel strip are some of the routing and sidechain options, and then we'll move on to the optional plug-ins.

Anyway, the Gate/Expander has a mix of familiar controls with a few extra goodies (hmmm, I think we've heard that before in this review...). Referring to the attached image, there's a Threshold control, Range (determines the amount of attenuation when the gate is closed, from 0 to -40dB), Release (100ms to 4 seconds), Hold (0 to 4 seconds - keeps the gate open for the specified amount of time regardless of what happens at the input), and two buttons. The EXP button, when enabled, turns the Gate into an expander with a 1:2 ratio. Enabling the Fast Att button gives an attack time of 100 microseconds per 40dB, and when disabled, there's a 1.5ms linear attack per 40dB.

As I tested the expander, I was surprised at how well it reacted to signals: No chatter, and a sort of "smooth" quality. Upon checking the documentation, I found out why: The threshold has variable hysteresis. What this means is if the gate opens at a particular setting, it will close at a lower setting; the end result is a more natural decay characteristic. Furthermore, the amount of hysteresis increases with lower thresholds.

Let me go offline for a bit and record an audio example...

[Anderton]

And here are some audio examples.

The first example is of the unprocessed drum loop used for the examples (from the Discrete Drums library).

The second example is gated reverb (reverb done using IK Multimedia's CSR Hall, just to prove I'm ready for the 80s revival :) The gate setting uses a high threshold, no release, but some hold to keep the gate open long enough to let some reverb ring through.

The third example is of gating without the reverb - just the drums and Duende.

The fourth example shows off the expander, set for a fairly extreme setting. Subtler settings give less of a "dynamics rolloff."

[Jim@ssl]

Hi all!

Just thought that I'd chime in here. If there are any questions that Craig's thorough test doesn't answer I'll try to be around to help. Great to hear such good examples of the Duende in use:thu:

[Jim@ssl]

While Craig is on the subject of dynamics (and forgive me if I'm preempting the next post), one of the tools I use most often with the Duende channel strip is putting the filters into the sidechain of the dynamics section. This is done by pressing the DYN SC button in the filter section. What this does is to remove any frequencies you filter out from the gain reduction detection circuit.

Why would you want to do that....?

Well, imagine that you have a drum loop you want to compress. using a "normal" compressor, when the kick drum plays the hi-hat gets very quiet, and when the kick is not playing the hi-hat jumps in volume. What you want is a nice consistent level on the hats.

By filtering out the low frequencies with the filters in the dyn sidechain, the high energy of the low frequencies of the kick drum no longer trigger the compressor. The low frequencies are still compressed, but they do not trigger the compressor (only frequencies above the filter will trigger gain reduction).

This means that you get a smooth consistent level of the HF content, while keeping a big bottom end.

great trick - well worth trying on all sorts of signals!

[Anderton]

While Craig is on the subject of dynamics (and forgive me if I'm preempting the next post)...

I think of it more as a segue :)

While you're here, I have a question about real-time bouncing. It seems that is necessary with any device that sends audio through FireWire because (correct me if I'm wrong) it's sending the audio out in real time to be processed, and returns to the computer in real time.

Pro Tools LE of course can only do real time bounces, so no worries there. With Sonar, you can uncheck "fast bounce."

But the other day, I wanted to apply the SSL Duende Bus Compressor in Wavelab, and render the file...but I couldn't find any option for a real time bounce, which precluded doing it (there were too many clicks and dropouts in the rendered file).

It seems to me that with more and more devices letting you use external hardware as a "plug-in," one would face the same kind of issues as you can't bounce through a hardware processor in faster than real time either. Do I have this right? And while we're at it...do you know any workaround for programs like Wavelab?

And...welcome to the thread! Feel free to keep coming up with useful tips. Either I don't know them, in which case your input is invaluable, or I do, in which case you've covered it so I'll have time to cover something else!

[Jim@ssl]

You are completely right with your assumptions. Duende is optimized to process the Samples in realtime (this is what DSP's are best at)...

This is also called synchronous sample transfer, which you will find in any DSP environment.

Bouncing to Disk is asynchronous or as fast as possible, which speeds up native/CPU based effects when the CPU can switch off the Audio Playback and only "concentrate" on the plug-ins you are running in the selected track.

There is no real workaround for applications that do not have a realtime bouncing setting, it will try to push the "DSP" Hardware to process faster, but this won't happen and apparently there is a flaw in the VST spec, that does not provide safety to still process any sample and/or process accurately in blocks and always start where the last sample processing ended.

Therefore you lose or duplicate samples and the result are clicks.

The only way to capture Wabelab's Out in realtime is to use another system or a VST Plug-In, that streams the Output to Disk, like: http://www.silverspike.com/?Download
->down to Tape it 1 (free)...

You have to cut the ends afterwards...but it works...

[Anderton]

Thanks for the tip on Tape It!!!! That solves my problem with Windows programs that only do fast bounce.

Well, AFAIC, you've more than justified your existence here :thu:

[Anderton]

Jim mentioned putting the filters in the sidechain. The first image shows the "default" routing, with Input off (i.e., the Filter isn't the first stage after the output) and Dynamics Side Chain off. Note that in all these examples, the Dynamics section is post-EQ.

The second image shows what happens if you enable the Input switch: The filter goes before the EQ.

The third image shows Jim's scenario, where the filters go into the sidechain for the Dynamics processing.

But there's more...

[Anderton]

If you put the Dynamics pre-EQ instead of post-EQ as in the previous examples, the routing options change once more.

The first image shows the "default" routing, with Input off and Dynamics Side Chain off.

The second image shows what happens if you enable the Input switch: The filter goes before the Dynamics section, which goes before the EQ.

The third image shows Dynamic Sidechain enabled for the filter. It now provides the sidechain signal for the Dynamics; the Dynamics section is followed by the EQ.

And there are still more routing options...

[Anderton]

So far, we've looked at routing options involving the Filter section. But there are also routing options involving the EQ if you enable Dynamic Sidechain within the EQ module.

The first image shows the "default" routing for using EQ in the sidechain, with Input off, Filter Dynamics Side Chain off, and EQ Dyanmics Sidechain on. The routing is the same whether the Dynamics section is preo or post EQ, as the EQ stage only relates to sidechaining.

The second image shows what happens if you enable the Input switch: The EQ section is post-Filter instead of pre-Filter.

The third image shows what happens if you enable the Filter Dynamic Sidechain option: Now both the Filter and the EQ are part of the sidechain.

Finally, the fourth image shows the most important part of the sidechain: The option to monitor the sidechain signal by itself. For example, if you want to have only the kick appear in the sidechain, as per Jim's example, you can monitor the sidechain signal and tweak the filters to isolate the kick as much as possible.

And that takes care of the routing options in the Channel Strip, unless I've missed something!

[Anderton]

All right--time to have some fun! We'll isolate just the kick using the sidechain, and set an outrageous amount of compression, so that every time the kick hits the compression goes insane and gives an ultra-squashed sound. Then we'll mix that with the original sound to get something not quite as anarchic as the compressed sound by itself, but something that's still pretty rude.

The first audio example has a drum loop from the Discrete Drums library without any processing at all.

The second audio example has the output of the sidechain monitor. The Filter and EQ sections are in the sidechain, and tweaked to really emphasize the kick.

The third audio example shows what happens when you super-compress the drums with only the kick. Squash-o-rama! Granted this isn't all that musically useful (hmmmm...then again, maybe it is) but we can tame it.

The fourth audio example mixes the original drum loop with the super-squashed loop. The end result is a pretty aggressive drum loop that retains some dynamics, courtesy of the original loop, but also folds in the squashed background to add some fatness and nastiness.

[allanh]

Say u have 32 channels in use with the ssl channel strip, and u find that the 4 band EQ is somewhat not enough, can u add 32 instances of the X-eq?
What are the limitations of the X-EQ in combi with the channel strip amount of tracks/channels?

[NewSc2]

Thanks for the review so far. I hope you'll end up going over the X-EQ, X-Comp, and the extra plugins =)

With the NAMM announcement, I'm seriously looking at the new packages. I know you touched on this, but is the latency or processing power any different between an expanded Duende Mini and Duende PCIe?

[Jim@ssl]

Duende has 32 channel processing slots.

You can run 32 channel strips in mono
or
16 channel strips in stereo
or
16 channels in mono and 8 in stereo

etc etc

X-EQ also takes 1 DSP slot in mono or 2 in stereo, so you could run 16 mono channel strips plus 16 mono X-EQs

Hope this answers the previous question

I had a good chat with Craig at the NAMM show, and will be getting some more details up in the coming days once my jetlag subsides!

[shandor]

Hey, Anderton, great job on reviewing Mini-Duende! Thank you!

I actually have PCIe and it works fine now but only after a big fght with plugins authorizations (I have them all). It was a really nasty one and this problem somehow has gone by itself... for now at least, I guess. It has already happened once to me.

Anyway, just to say it's a pleasure to use Duende in every way, especially sonically speaking. X-Eq is a killer, Channel Strip is a good all around tool, Drumstrip is really really good integrated drumstrip, X-Comp: well, I haven't mastered this one yet, pretty dark horse for me. BusComp is perhaps the one that wouldn't meet my expectations, but can come handy. I guess for this particular purpose I still prefere my API 2500 (hardware).

Actually, I wanted to post about a possible workaround with Wavelab and realtime bouncing. I won't lie: I didn't do it because I have never needed to do this trick on purpose (to fool Wavelab) mainly because I'm mixing pug-ins and hardware in wavelab master section. Still not clear? Well, in Wavelab 6 one of the greatest function is to be able to integrate your external gear (I'm doing it with my SPL Tube Vitalizer for premasters). And what do you know, it is actually the only way to put the render function in realtime mode. Now, when you're rendering, you're actually running all the master section plugins in real-time and Wavelab is printing it in a file. All this, just by inserting the External Gear plug-ins under /ASIO and setting up the routing in connections.

So, my solution is, with some advanced routing of your asio (to hook up your digital out on a digital in) you actually don't transform your sound staying in digital domain. It's like using a "null" gear, i.e. just hooking up an AES OUT on an AES IN. If you can't route it internally, you can always do it phsically with a cable by creating a digital loop (just be careful :))). And it's done. No fancy plugins (even if it can be handy, of course!). The only thing to remember is that AES works in 24 bit, so if you're in 32 bit you may want to dither it before or better, just put a hard brickwall limiter at 0 to prevent the overshoots. Wavelab will compensate for the lookahead delay. If you don't have AES IN/OUT, you solution can be a virtual cable app' though I've never used it personally.


I hope it can come handy to some who's looking for realtime printing under Wavelab.

Jim, big fun of you and your company. As SAE teacher, I can only advise to my guys to go for your brand!

sH

[IkeTurner]

Holy smackdown, you can get the PCIe card for $700 right now??? That is insane deal.

[IkeTurner]

I'd also like to hear how stable Duende PCIe is with Cubase 4 and Fireface 800

[Anderton]

I'd also like to hear how stable Duende PCIe is with Cubase 4 and Fireface 800

I've used it with Cubase while on the same FireWire bus as the MR816csx, without problems so far. I'll see if I can "break" it as I continue with the MR816csx/CC121 Pro Review.

[Anderton]

I wanted to post about a possible workaround with Wavelab and realtime bouncing. I won't lie: I didn't do it because I have never needed to do this trick on purpose (to fool Wavelab) mainly because I'm mixing pug-ins and hardware in wavelab master section. Still not clear? Well, in Wavelab 6 one of the greatest function is to be able to integrate your external gear (I'm doing it with my SPL Tube Vitalizer for premasters). And what do you know, it is actually the only way to put the render function in realtime mode. Now, when you're rendering, you're actually running all the master section plugins in real-time and Wavelab is printing it in a file. All this, just by inserting the External Gear plug-ins under /ASIO and setting up the routing in connections.

So, my solution is, with some advanced routing of your asio (to hook up your digital out on a digital in) you actually don't transform your sound staying in digital domain. It's like using a "null" gear, i.e. just hooking up an AES OUT on an AES IN.

sH

That's GREAT!! I liked Jim's suggestion of the VST plug-in, but your idea of doing a real-time bounce by going "outside the box" is a very universal solution. I had considered that but I didn't want to do the analog in/analog out conversions. I'm embarrassed to say that I didn't even think of using SPDIF or AES/EBU because two of the three interfaces I use the most don't have it. But of course, that's a way to do the transfer without loss.

I could probably figure out a way to do this internally with the Creamware SCOPE system and avoid going outside the box at all or using a plug-in.

Thanks for the tip!! Interaction is why I love Pro Reviews.

[Anderton]

All right - now that my NAMM videos are over, it's time to dig further into Duende. Let's move on to the Vocalstrip, shown in the first image.

I'll start off with a negative: If money was a big issue, this plug-in would be last on my list. Not because it's not good, but because it has some stiff competition: the bundled Channel Strip we just discussed. I used the Channel Strip a lot for vocals prior to the appearance of the Vocalstrip, and had become very comfortable with which settings to tweak to get good results with my voice.

However, there's a reason why SSL made the Vocalstrip plug-in, and my guess is it's about speed - it takes a lot less time & effort to get a good vocal sound with Vocalstrip than with the more general-purpose Channel Strip. We'll zoom out and look at the elements that make up Vocalstrip, then examine the elements in more detail.

The De-Esser does what de-essers do, but the De-Ploser is a less common processor unique to Vocalstrip. It's sort of like a "smart lowpass filter" for getting rid of plosive sounds - very handy if your singer didn't use a pop filter!

There's a three-band EQ, with the curves and ranges optimized for voice. Although you can get similar sounds with the Channel Strip (or other EQs), it's very easy to dial in a good EQ setting relatively fast.

The Compander is fairly typical, but its drive control has some interesting uses. The metering, on the other hand, is again unique to the Vocalstrip. This is definitely something we'll need to cover in detail.

Also, note that you can change the order of the processors -- see the second image. The highlighted area in the lower right is where you change the order; click on a module, then use the arrow keys to move it left or right in the chain. This takes care of the "compression before EQ or EQ before compression?" question once and for all -- try them both, and see which one you like.

More to come...

[Anderton]

I was looking for some vocals to process through the Vocalstrip, but gave up when I realized that with virtually all the music I've done, the vocals aren't "raw" but recorded with processing, or rendered to include processing. The kind of source material I'd want to use for applying Vocalstrip would be raw vocals compared to processing through Vocalstrip.

So, I think I'll work on a short song that can be used for demoing the Vocalstrip and some of the other plug-in effects - for example, the bus compressor on the overall mix, Drumstrip on the drums, X-Comp on guitar/bass, etc. I'll whip something up in the next day or two we can use in the review...stay tuned.

[Anderton]

Well, it took a while to get back here...sorry, my travel schedule has been pretty hectic lately. Anyway, at long last, now you hear Vocalstrip at work, turning a really unappealing vocal into something, well, more appealing :)

But first...in the latest version of Sonar, I've been able to do the "fast bounce" thing with Duende Mini, which I had not been able to do before. This contradicts my previous findings that I could only get Duende to work with real-time bounces. I do know that Cakewalk has been doing a lot of optimization of their audio engine, and maybe this is one result.

Anyway, this vocal was recorded live through a Shure SM58 dynamic mic, and I wanted to use Vocalstrip to give it a more polished, "studio" quality. The first attached image shows the EQ settings I used. Basically, I added a bit of a boost in the upper mids to give more intelligibility, and cut away some of the low end to reduce the somewhat muffled, boomy quality you'll hear in the audio examples. I didn't use the midrange control; it's a bit too "peaky" for my tastes with most vocal material.

The vocal level varied quite a bit, and the second attached image shows the compressor settings. I went for minimum attack (0.1ms) to even things out. The ratio is 3.6:1, with a threshold of -9dB. I went for a soft knee response, 222ms of release, and also set the expander threshold at -48dB - I wasn't singing down there anyway, but I figured it could help get rid of any background noise.

Next, the audio examples.

[Anderton]

The following examples not only show what Vocalstrip can do, but also give an idea of how to produce a vocal to improve its sound.

Example 1 is the original vocal and backing track. In all of these examples except the last one, I normalized the vocal in each case so you could hear how the processing was affecting perceived level.

Example 2 adds the Vocalstrip EQ and no other processing. You can hear the vocal become more intelligible and intimate, and lose some of its "tubbiness."

Example 3 adds the Vocalstrip Compander and no other processing. This definitely helps even out the level variations.

Example 4 is the Vocalstrip EQ and Compander. Now we're getting someplace! The voices cuts better through the track and the levels are smoother.

Example 5 doubles the Example 4 vocal (the one with EQ and compression), pans the two channels a little left and a little right, and adds a bit of reverb. I think you'll hear that overall, the vocal's tonal quality and dynamics are far better than the original sound in Example 1.

[Anderton]

I didn't use the midrange peak/notch function in processing the vocal with the previous examples, because it didn't seem like it was needed (remember, the most important control on any signal processor is the bypass switch!). However, the midrange section does indeed create some interesting effects - it harks back to what I was saying about Duende including extra unusual things that take what might appear to be a normal function into a different realm.

So, to give you an idea of how this feature affects the vocal. I took a portion of the previous equalized/compressed vocal, and automated a sweep of the midrange control over 200Hz to about 7kHz while applying various amounts of boost and cut. (The range actually goes up to about 10kHz, but anything over 7kHz made no significant difference to the vocal sound.)

The attached image shows the pertinent part of Vocalstrip - the middle control (with the notch frequency set to 3257Hz) and toward the upper right, the frequency response curve.

Regarding the audio examples:

Example 1 sweeps the midrange with a 6dB boost. You'll note this is a pretty subtle effect.

Example 2 sweeps the midrange with a 12dB boost.

Example 3 sweeps the midrange with a -18dB notch.

Example 4 sweeps the midrange with a -36dB notch.

All of these add a distinctive "character" to the vocal that you may or may not want to use, but the point is, it's there if you want it, and the -36dB notch is certainly more than you'll find on most general-purpose parametric EQs. It's unfortunate there isn't a bandwidth control, although you do have the option for a gentler boost in the third equalization stage.

[Jim@ssl]

For all those waiting for an update, the new Duende V3 driver is now in beta. Duende V3 is an entirely new architecture, and shares the processing load between the Duende and the host machine. This allows users to run up to 128 processing slots on their Duende.

More info can be found on our blog here http://solidstatelogic.wordpress.com/

[Anderton]

Jim - Thanks for checking in. I will be re-starting the Duende review next week, between the Frankfurt show, editing 40+ videos for same, and getting really sick for a month, I've kind of gotten behind...but there's much more to come.

[Anderton]

Well, that was quite a hiatus...but no traveling for a while, so it's back to Pro Reviews.

Let's take a look at the Bus Compressor, which is available in mono or stereo versions. When you ask Duende owners exactly what it does, they sort of get this dreamy look in their eyes and start waxing mystical..."It sort of, I don't know, glues the sound together." But really, we're just dealing with a compressor, right? What's the big deal?

We'll get to the audio examples in a bit to demonstrate what the big deal is, but first, let's look at the controls in [b]the attached image[/blue]. It's pretty much what you'd expect: Threshold, Attack, Release, auto-release option, Ratio, and Make-Up Gain. There's also a bypass switch.

You'll note your options are restricted, presumably to be as faithful as possible to the hardware. For example, your compression ratio options are 2:1, 4:1, and 20:1 (the latter is, of course, wonderful for a drum room mic bus). Similarly, the Attack and Release times are stepped. At first, the limitations put me off - dude, this is software, you can do whatever you want! - but after playing with the bus compressor for a while, this actually turned out to be a strength. The settings are pretty much what I would use anyway, and I can't argue with either the sonic results, or the speed with which you can get the sound you want.

If you look at the waveforms behind the picture of the Bus Compressor front panel, you can get a visual idea of the compression. Note that both waveforms have been level-normalized to 0. The upper waveform doesn't have the bus compressor, while the lower one is a bounce of the upper waveform through the Bus Compressor. The settings on the Bus Compressor were the ones used to generate the lower waveform.

It's clear just by looking at the waveform that the lower one is "beefier." There's a higher average level for sure, but you'll also note that the dynamics haven't been destroyed or "flat-lined." Granted, part of that is because I didn't push the compression settings all that hard. But as you'll hear with the audio examples, it seems the main talent of the SSL Bus Compressor is to give a nice "lift" to the sound while retaining a solid sense of dynamics.

Now let's listen to it in action.

[Anderton]

I've created three audio examples.

The first audio example is a little demo piece I put together from Big Fish Audio's Funky House Grooves 2 loop library (I'm reviewing it for the August 09 issue of EQmagazine (http://www.eqmag.com)). There's no processing on this at all, and the peak levels have been normalized to 0.

The second audio example is the same file, but bounced through the Bus Compressor using the settings shown in the image in the previous post. No other processing was used.

The third audioexample alternates between segments. First there's the unprocessed sound, then a second of silence, then the same segment through the Bus Compressor. This makes it a little easier to do a head-to-head comparison.

As I think you'll hear, the Bus Compressor adds a gorgeous lift - not subtle, but it doesn't hit you over the head, either. The dynamics remain intact, and the sound is wonderful.

I would not recommend using the Bus Compressor as a substitute for a "maximizer" plug-in. The harder you push it, the easier it is to get the compressor out of its "comfort zone." But it's very transparent in its operation. When you hear the processed example, I doubt your first thought will be "it sounds compressed." Instead, when compared to the unprocessed example, you'll likely think "it sounds better."

There's not really much more to say about the Bus Compressor. It's sweet, effective, and improves your sound without degrading it. I'm not a fan of strapping a compressor across the stereo bus - I leave that for the mastering - but I'm definitely a fan of this plug-in. I've found that using it on a mixed track, and then following it with a dB or two of level maximizing (e.g., WAVES Lx series), gives a "loud," punchy sound that's can compete in the loudness wars, but without the sonic destruction you get using heavy maximization alone.

This plug-in gets a definite :thu: :thu: (that's two thumbs up).

[Anderton]

I like drums. Hey, who doesn't? And I'm always for ways for them to sound bigger, brighter, better, and ever-more bodacious. Having had good luck with Duende so far, it seemed like a good time to check out Drumstrip.

The attached image give a pretty good idea of what's going on. The first thing that struck me: This is a company that understands the virtues of parallel processing on drums - check out the Dry/Wet controls at the output, and for the Listen Mic Compressor.

Let's start with the HF (High Frequency) Enhancer and the LF (Low Frequency) enhancer, because it's easy to wrap your head around it: Pick a frequency, throw in some drive (which adds more of an edge to the sound), then dial in the amount of enhancement you want.

In use, these two processors can definitely add some low end and high end mojo and the easiest way to show this is with audio. So, the first audio example is the bypassed drum sound, and the second audio example is the same track but processed with the enhancers. By the way, I used a reasonable amount of enhancement - you can take it to a much more extreme level, if you want bass that may trigger small tremors in seismically-active areas.

[Anderton]

On to the Transient Shaper. This is arguably the most "drum specific" part of Drumstrip. It's an interesting effect, because I feel it really needs to be used subtly for best results; it's almost something you "feel" rather than hear.

The first attached image highlights the Transient Shaper module. Gain essentially determines how much of the attack gets "caught" for the enhancement process; more gain means more of the attack gets shaped. Think of it similarly to how you would think of a noise gate threshold.

Speed controls how long it takes for the generated transient to decay, from a "glitch"-type spike to something more substantial. The Amount controls determines how much of the transient gets mixed in with the original signal (and can also be subtracted if you click on the Inv button). Check out the second attached image; the upper waveform (track 2) is the original drum loop, while the lower waveform (track 3) is the processed loop with no gain or change other than adding the Transient Shaper.

The Audition button is one of Drumstrip's coolest features - it's like being able to listen to a sidechain signal. You can hear the changes that happen to the attack in isolation, which makes it easier to adjust the controls.

And of course, you want to hear what it sounds like, right?

The first audio example has the Transient Shaper bypassed, while the second audio example has it active, using the settings shown in the first image.

Note that because the Transient Shaper can generate sharp spikes, I find it's a good idea to lower the drum level to create the headroom for these to be recorded properly. Then, adding a level maximizer can reclaim a few dB by taking off the very peaks of the attacks.

So...is it great? It's useful, but I find it works best if you don't hit people over the head with it, and if you follow it with some kind of maximizer (or the SSL Bus Compressor, if you're so inclined!) so you can still maintain a reasonable average level. It's definitely more of a team player that adds some enhancement rather than taking a sound into a whole different dimension.

[Anderton]

Now let's take a look at the Gate. Probably anyone who's in the market for a Duende is far enough along to know what a gate is, and SSL's implementation isn't all that different from the norm. The attached image shows the Gate controls.

The Attack control sets the time for the gate to ramp open after being triggered, while Release determines how long it takes to reach the full gate off setting after the trigger goes away. Hold controls a gate-on duration that starts when triggered and lasts for the Hold value. This would be useful if you want to make sure you always catch transients of a particular length.

However, it sure seems to me that the calibrations have the decimal point in the wrong place...Attack is said to go from "0.00ms" to "0.10ms," but I think they mean 0-10ms. Ditto Release: It reads 0.01ms to 1.00ms, but it sure sounds to me like 1-100ms. Hold seem to be calibrated correctly - 0ms to 4000ms.

Open sets the level at which the gate opens, while Close sets the level at which the gate closes. Range sets a maximum amount of attenuation when the gate is closed.

While this is all fairly straightforward, the metering is very helpful. The left meter shows the input; the small green line indicates the Open threshold, and the red line shows the Close threshold. As you can correlate this to what's happening with the signal level, the process of setting levels is pretty obvious. The right meter shows the amount of reduction being applied, with the white line indicating the maximum amount of attenuation.

Being able to attenuate by something less than full off is very useful for tightening up drum ambience, without making the drums sound artificially gated. Listen to the first audio example; the Gate is bypassed, and you can hear a ton of reverb on the drums. Now, check out the second audio example - this is with the Gate active, but attenuating by only 15dB or so. You'll hear how the reverb sounds way less prominent, but nonetheless, doesn't go away; in a way it resembles downward expansion. I also tried this gating setup with ringing toms, and the Gate was very effective for that, too.

[Anderton]

This is the last element of the Drumstrip, and it is a truly unusual "special effect." But we need to start with a little background...

First, you can download a free, somewhat less capable version of the Listen Mic Compressor from SSL's web site (http://www.solid-state-logic.com/Music/LMC-1/index.asp) - all you need to do is register. It's available for Mac or Windows, so you can check out this effect for yourself. I'll be interested to see if anyone has any comments on it to add to this thread.

Second, the LMC effect has an interesting history. Here's the story, as quoted from SSL (you can find out more at the link given above for downloading the plug-in):

The SSL Listen Mic Compressor was the secret weapon in many producers' sonic arsenal of recording techniques. Originally designed to prevent overloading the return feed from a studio communications mic, its fixed attack and release curves were eminently suitable for use on ambient drums mics. Of course, we’d like to take all the credit for this great sound, but as usual, it was the creativity of SSL users that led to the idea.

Long-time SSL user Hugh Padgham was one of the first to capture this new drum sound on tape,while working with Steve Lilywhite on Peter Gabriel’s ‘Intruder’, he told Mix magazine: "On a normal console, you have a button to press to talk to the musicians in the headphones, but you did not have a button to press for us to listen to the musicians. To do that, you'd plug a microphone into a spare channel on the desk and listen to your musicians through that. But the SSL had a reverse talkback button and there was a microphone hanging up in the studio already, a dedicated input into the reverse mic input on the console. And on this microphone, they had the most unbelievably heavy compressor, so you could hear somebody who was over in the corner.

"One day, Phil (Collins) was playing the drums,” Hugh recalls, “and I had the reverse talkback on because he was speaking, and then he started playing the drums. The most unbelievable sound came out because of the heavy compressor. I said, 'My God, this is the most amazing sound! Steve, listen to this.' But the way the reverse talkback was setup, you couldn't record it. So I had the desk modified that night. I got one of the maintenance guys to take the desk apart and get a split output of this compressor and feed it into a patch point on the jack field so I could then patch it into a channel on the board. From there, we were able to route that to the tape recorder."

All right...now let's get into the LMC itself.

[Anderton]

Let's look at the difference between the free version and the one included in the Drumstrip plug-in.

The first attached image shows the freebie plug-in. Controls are Compression (goes from "less" to "more"), Input level, and Output level - that's it.

The second attached image shows the version in the Drumstrip. There's no Input level control because of course, the Drumstrip itself has one at the beginning of the plug-in. The Compressor control provides the same function as on the LMC-1, and Makeup is equivalent to the LMC-1 Output control.

The Mix control, which the LMC-1 doesn't have, mixes the LMC in parallel with the dry drum sound. The control ranges from dry only to effect only, or anywhere in between. (If you want to be able to do what the Mix control does with the LMC-1, you'd need to use the LMC-1 as a send effect, and mix in the desired amount of processed sound by changing the aux send or aux return level.)

The LMC-1 includes EQ that provides a more "room-like" sound, but cannot be defeated for full-range operation. The version in the Drumstrip includes an EQ in/out button so you can remove the EQ if desired, and go for a more full-range compression sound.

However, there's a caution here. The compressor has extremely tight time constants because it was designed to pick up musicians talking from far away in a room; no thought was given at the time to using this "reverse talkback" compressor with regular instruments. As a result, you can't lengthen the release time to minimize distortion with low-frequency sounds. If you apply the LMC to drums with the EQ out and there's a massive kick going on, it will have a distorted character that you may or may not like. I realize that the whole point of the LMC is to re-create a specific effect, but I would have much preferred if when the EQ is bypassed, you would have the option to adjust the compressor release time. As a result, I tended to leave the EQ in more often than not.

[Hey SSL - any chance of adding a release control in an update? You could make room for this control in the plug-in by abbreviating "Listen Mic Compressor" to LMC, moving the EQ in/out button up, and putting a Release control underneath the button.]

Okay...time for some audio examples! As a post can have only five attachments and we've already used up two for the images, we'll put the audio examples in the next post.

[Anderton]

Here we go...

The first audio example is the drum loop without any processing, for comparative purposes.

The second audio example has the LMC sound only (no dry mixed in), with about 40% compression.

The third audio example mixes the dry and processed sounds together, with about 40% wet, 60% dry.

The fourth audio example takes out the EQ and monitors the LMC sound only. I've kicked the Compression up to 50% so you can really hear the low-frequency distortion that occurs due to the short time constants.

The fifth audio example has the EQ switched out. The output is a mix of the dry and processed sounds, with about 40% wet, 60% dry; the Compression amount is around 43%.

[Jim@ssl]

I've asked the developer who wrote Drumstrip to check out the issue you mention with the gate, and will post back with some answers.

[Anderton]

I've asked the developer who wrote Drumstrip to check out the issue you mention with the gate, and will post back with some answers.

Thanks Jim!

[Jim@ssl]

Craig is correct about the gate. While the legending on the GUI is in milliseconds the algorithm works in seconds.

we will address this in a future update. Thanks for the eagle-eyed work!:thu:

[Anderton]

First of all, thanks Jim for checking out the Gate calibration issue. I knew my ears were telling me something different from my eyes!

We've pretty covered everything about the Drumstrip except for the routing option, which is similar to other SSL modules. The attached image shows the "Process Order" section of the module, where you can arbitrarily move any module to the left or right in the chain. You just click on the module, it becomes highlighted in yellow, and you move it using the << and >> buttons.

So...let's draw some conclusions. First off, I find this a very useful module. That usefulness is mitigated somewhat by your being able to get a more basic version of the LMC for free, and that the Channel Strip module already includes a Gate (albeit without the same degree of control over threshold) as well as EQ. However, the Transient Shaper has no equivalent in the Duende processing arsenal, and (in typical SSL fashion, I might add!) the HF and LF Enhancers have their own special mojo that the Channel Strip EQ doesn't duplicate.

Being a big fan of parallel processing with drums, I also like the Dry/Wet controls on the LMC and the overall output. Being able to change the Process Order is also very handy.

My "wish list" is limited to two things: Fixing the Gate calibrations, but it seems SSL is already on that, and having time constant controls on the LMC which would do things: 1) allow it to act more like a standard compressor, a feature that Drumstrip lacks; and 2) minimize distortion on low-frequency material. And while I'm in wish list mode, I'd like to see two buttons on the output: Limiter on/off that traps peaks, and Saturation on/off that would add more of a clipping effect with peaks. Of course you could have a plug-in after Drumstrip that did these things, but it would be convenient to have these features on Drumstrip itself.

Overall, though, I have to say that you can indeed make drums sound better with Drumstrip...and that's what it's all about, right?

Next, let's take a look at the X-EQ, and follow that up with the amazing X-Comp.

[Jim@ssl]

Thanks Craig :-) I'll make sure the developers see your comments.

For anyone want ing to stay abreast of the latest V3 progress check our blog www.solidstatelogic.wordpress.com

[Anderton]

Does the world really need another equalizer plug-in? Let's find out.

X-EQ is a 10-band, non linear-phase EQ. The ten bands are highpass, low shelf, six bands of parametric EQ, high shelf, and lowpass. Of course we'll get into details shortly, but let's take an overview first.

The X-EQ's main claim to fame, aside from eschewing linear-phase response, is that there are five different curves for the highpass and lowpass filters and nine different curves for the midrange. Referring to the first attached image, selecting a band causes a drop-down menu to appear toward the upper right (circled in red for clarity); this is where you make your selection.

The overall philosophy reminds me a bit of the Liquid Mix hardware, which allows for multiple equalizer responses. In the case of the Duende Mini, some emulate specific analog responses, while others have responses that could be done only through digital technology. I'm not surprised that SSL has decided X-EQ should have a "character" - that seems to be a common theme here amongst all the plug-ins.

As appropriate for different functions, Gain, Frequency, and Q controls appear. There's also input metering, output metering, on-off switches for each band, and standard load/save/compare functions.

You can also enable a post-EQ spectrum analysis display (second attached image) which makes it easy to see "rogue resonances" and the like right after loading it, rather than listening carefully for them to get "in the ballpark." You'll also see one of the nine possible parametric responses being selected.

[Anderton]

Let's listen to how the choice of the midrange "bell" affects the sound. The sonic character depends on two main factors: Whether the EQ is boosting or cutting, and the amount of boost or cut. I tried to think of the most obvious way to show this, and decided that the constant hiss of white noise would highlight any differences the best. As you can imagine, the same kind of "character" will be impressed upon instruments.

Duende's documentation is very helpful in this respect, giving full descriptions and charts of the various responses, and indicating the goal of the emulation.

The first audio example plays three sections back-to-back, with a slight silence in between: White noise boosted around 2kHz with the Proportional 1 bell, then the same thing with the Proportional 2 bell, and finally, the same thing with the Proportional 3 bell. You won't hear much difference between 1 and 2, but you will definitely hear a difference with 3.

The second audio example again plays three sections back-to-back, but this time white noise is cut around 2kHz with the Proportional 1 bell, then with the Proportional 2 bell, and finally, with the Proportional 3 bell. This time you'll almost certainly hear the difference between 1 and 2, but you will definitely hear a difference with 3.

The third audio example plays four sections back-to-back using the classic curve (standard and asymmetrical). The first is standard cut, the second is asymmetrical cut; the difference is obvious. The third is standard boost, and the fourth is asymmetrical boost. You'll hardly hear any difference at all.

[Anderton]

For the five different highpass and lowpass filter responses, I thought showing the curves themselves would be useful because the main difference is the steepness of the response dropoff; see the five attached screen shots. These show the highpass response; the lowpass response is simply the reverse.

Also, note that there are some sonic differences, so the next post compares the Critical, Butterworth, and Chebychev responses.

[Anderton]

This audio example plays white noise through the Classic, Butterworth, and Chebychev responses (the rolloffs get progressively steeper). The Butterworth has some resonance, while the Chebychev is very close to a brickwall type of response.

[Anderton]

Okay! NAMM is over, the videos (http://namm.harmony-central.com/SNAMM09) are edited, and the NAMM edition of the Harmony Central Confidential (http://www.harmony-central.com/Newsletter/) newsletter is put to bed and will ship in a few days...so I get to return to Pro Reviews!

The remaining Duende module to check out is the X-Comp Compressor. At first glance, you might think "Just what the world needs, another compressor" - particularly because the channel strip includes dynamics control. But, once again, SSL has put a different twist on matters.

Instead of modeling a particular compressor, the X-Comp is designed to cover a wide range of compression scenarios, from transparent to raw/overdriven (like the Listen Mic Compressor). It also has "semi-side-chaining" (i.e., it doesn't accept an external signal, but applies filtering to the input signal), making it easy to add frequency-dependent parallel compression.

Before getting into the details, here's an overview. Referring to the attached image, we can dispense with the input and output metering as it's similar to the other SSL modules. Threshold, Ratio, Attack, Release, and Knee work as expected. The "I/O Diff" is a type of meter (amplitude histogram) that shows how often signals hit particular amplitude levels, while the GR History shows the current amount of gain reduction as well as the gain reduction history for the past second.

Where X-Comp really differs is how it handles gain reduction on peaks, which is what the Max GR control is all about, and the way that you can apply frequency-selective compression through the LF and HF Bleed controls, so let's concentrate on these two features.

[Anderton]

This is one of the most interesting elements of the X-Comp, as it allows emulating some of the compression characteristics associated with optical/vintage compressors. With some of these, dynamic peak control happened only up to a certain level, past which point the circuitry kind of gave up and let more transients through. The end result is that because these don't clamp down on the signal so hard, the sound has more "life" due to retaining some of the percussive transients.

X-Comp quantifies this and goes one step further by allowing you to tailor the response. Think of the way traditional compression works: Higher amplitude signals are compressed the most and lower amplitude ones, the least. So by definition, peaks can receive massive amounts of compression compared to the lower-level program material. Max GR sets a limit to the amount of compression, and ranges from 20 to 60dB.

For an illustration of how this affects the signal, look at the first attached image. This shows Max GR set to 20dB, and as the Threshold is low, it wouldn't take much for signals to be compressed by 20dB. Now check out the Compression Law curve, which shows Max GR in action. Up to the threshold (pink dot) the compression acts as expected. However, there's a second threshold (blue dot) and this is where the Max GR starts having an effect. Note that as the input amplitude increases along the bottom of the graph, the output starts increasing more linearly, as if compression wasn't being applied. Increasing Max GR pushes the curve more toward normal compression.

In the second attached image, the threshold is higher. This gives less room for Max GR to have an effect, as reflected in the Compression Law curve. Note that in this case, we've also made the knee sharper.

[Anderton]

Now let's hear how Max GR and Bleed effect the sound with some audio examples.

I've taken a tom beat from the Discrete Drums library and added a lot of compression. The first audio example plays the sound with standard gain reduction - in other words, it works like a traditional compressor. The first attached image is a screenshot that shows the settings used to get this sound.

The second audio example plays the sound with gain reduction set to a maximum of 20dB, which gives the transients a bit more "breathing room" because they aren't compressed by as much as they would be otherwise. The second attached image is a screenshot that again shows the settings used to get this sound, but note that the only difference is the setting of the Max GR control; as you can see, this also changes the shape of the compression law curve.

[Anderton]

Now for one more example - let's show what bleed does. The audio example is the same as the one with minimal Gain Reduction (20dB), but with some HF Bleed added. This means the highs bypass the compression process somewhat.

The attached image is a screen shot that shows the settings used to get this sound. The main difference you'll see is in the HF Bleed Curve.

[Anderton]

Well...we've covered a lot, starting with the Channel Strip plug-in that's included with Duende, and several optional plug-ins - Bus Compressor, Drumstrip, Vocalstrip, X-EQ, and X-Comp. Along the way, we even checked out the freebie LMC-1 from SSL.

The Mini-Duende hardware is currently going for a street price (http://pro-audio.musiciansfriend.com/product/Solid-State-Logic-Duende-Mini-DSP-Powered-DAW-Plug-In-Platform?sku=482193) of $699. Referring to the SSL web store (http://www.solid-state-logic.com/store/), the 16-to-32 channel update is $399., and the optional plug-ins cost as follows:

X-EQ $375
X-Comp $295
Vocalstrip $399
Drumstrip $275
Bus Compressor $370

The whole collection will cost you a shade over $1,800, which isn't exactly cheap - although they do cost less than hardware processors, or dragging an SSL console into your studio :). A mitigating factor is that the channel strip included with Duende does a lot, making the X-EQ, X-Comp, and Bus Compressor more of a luxury than a necessity. And of course, because Duende is hardware, it doesn't load your CPU.

Normally I'd say just pass on the plug-ins that your setup already duplicates, but actually, the whole point of Duende is to supply that "SSL sound." Duende is not a "me-too" product; every plug-in has some kind of unexpected twist, and there's a definite sonic characteristic. So while you might have, say, a compressor in your DAW, it won't have the same characteristics as X-Comp.

To help those who are considering getting into Duende, I've tried to include a bunch of representative audio examples. I hope you find them helpful, but if you have any requests for the sound of specific applications, no promises I'll do it but I'll try my best.

So how does Duende compare to Universal Audio's UAD-2 and TC Electtonic PowerCore, which are the primary competitors? Keep reading....

[Anderton]

I recently compared the three hardware helpers in this review (http://www.eqmag.com/article/how-about-some/may-09/95539) for EQ magazine. Here are my conclusions about their sound and character:

This is of course subjective—generalizations are risky, and as the old saying goes, “comparisons are odious.” However, having worked with and used all three extensively, I feel there are some differentiating characteristics.

TC PowerCore has a precise, clean sound: It’s what I reach for when I want to process a signal with as much transparency as possible. Of course, there are exceptions; for example, you can use tube emulators to “dirtify” the sound. But even these have a certain precision.

UA’s specialty is really nailing the sound of vintage analog gear. I’m not sure how they do it, but if you got laid off from your job and had to put your classic compressors and EQs on eBay, the UAD-2 will take away the pain. I know some very picky engineers who also feel that UA has the analog thing down.

Duende makes no secret of the fact that it’s all about adding SSL’s special “character”—and if you like that character, then Duende is the ticket to getting it. This isn’t to say you can’t get some fairly neutral sounds, but that’s not what separates Duende from the pack.

I still think that's a pretty reasonable summary, and there's no question that Duende delivers a distinctive - and popular - sound.

The hardware aspect has its pluses and minuses: On the minus side, the FireWire bus probably won't last forever, and you need to bounce through Duende in real time - no fast bounce allowed. On the other hand you can run a predictable number of instances without using CPU, thus leaving it free for demanding task such as software synhts; and if you change platforms, just plug into the new platform's FireWire port and keep going.

[Anderton]

It's worth asking whether external hardware boxes represent an idea whose time has come and gone, thanks to increasingly powerful computers and more plug-ins being bundled with host programs. Let's address each point separately.

As to computers, I've been using card- or bus-based hardware expanders for over 10 years, and even though computers keep getting more powerful, programs and algorithms keep getting more complex. It's always great to be able to call up a bunch of sophisticated processors and not see the CPU consumption meter move. Furthermore, increasingly powerful computers can mean compatibility issues, and in that respect, the "brains are in the box" with Duende, which is a big help.

It's also clear that SSL doesn't see hardware as a dead end. They've released Version 3 software (http://www.solid-state-logic.com/support/duende/downloads.asp), which expands the standard number of channels to 32 (with optional-at-extra-cost upgrades to go to 64 or 128), and makes several other improvements. However, system requirements are stiffer: Power PC Macs, those running Mac OS X 4.x, or any computer that's not multi-core need not apply. I haven't checked out the Version 3 software yet as it's very new, but once I've had a chance to experience what it does, I'll add some supplementary notes to this thread.

So, do you really need the plug-ins if you have a host that's already "fully loaded"? That depends. Host plug-ins are generally fairly "safe," both sonically in the sense that they provide "bread and butter" functions, and in terms of CPU consumptions. The Duende plug-ins are more adventurous, and I would think that if they were native instead of hardware, they'd make a lot of demands on your CPU.

In terms of other options, Waves offers an SSL 4000-based native bundle (no hardware required) for about $750 street - not much more than the Duende Mini package. However, the Waves SSL 4000 Collection models the Solid State Logic SL 4000 series consoles, while Duende's plug-ins are taken from the Solid State Logic C200 digital production console. As a result, the characters are quite different, although they're both SSL.

The bottom line is Duende brings the SSL sound, as approved and designed by SSL, into a desktop environment for far less than any other hardware alternative. It's been a real pleasure to use in the studio, because the software is versatile enough to do the expected and the unexpected - that's true even of the channel strip that comes with Duende.

[Anderton]

And that pretty much wraps things up - I think there's really nothing much left to cover! As with other pro reviews, this one will be kept open for a while - particularly because there's a new version of software out, as you might want to add some comments of your own. And of course, if you have particular questions, or if you want to know something we haven't covered here, chime in!