Renewed appreciation for rev2 after trying the Peak

[Merlijn Nash]

So I had the Peak for a month (to see if I'd like the Summit), and it can do a lot of stuff, really enjoyed all the stuff you can do with those FPGA's.
Somehow though, the filter just didn't gel with me. It seemed a bit whiney  So whenever I initialized a patch, I sort had to "get over" the filter to coax some nice sounds out of it. I ended up returning it today for this reason.

Anyway, all the "bad" stuff about the rev2 doesn't seem so bad now, because at least when I initialize and do a filter sweep, it makes me want to do some design or play some James Blakey chords.

How are your experiences with trying some synths? Could you sometimes get over a not-so-good basic sound because you loved the functionality? Have you ever sent a synth back even though it was perfect-except-for-the-filter?

Cheers!

[Tugdual]

I returned a Peak for a REV2 a year ago. The Peak is looking good and robust design; on the paper, wave table and analog filters were absolute features I wanted. I had it for 2 weeks and then had to decide if I wanted to keep it or not. Was a tough decision to be fair but I found the sound... dull. I wanted something organic, plain, imperfect so I decided to return the Peak and picked a REV2 for its analog design and the many modulations. I never regretted my choice.
In the mean time I also purchased a Subsequent 37  and, lord, this one totally fulfills my expectations. I used to say that if synths were motorbikes, the REV2 would be a BMW and the Subsequent a Harley. I wish Moog did an affordable poly synths, looks like the One is an overkill with its fans, there is room for smaller synths.

[creativespiral]

I've had the same experience after picking up other synths.   I have fallen back in love with the Rev2 on more than one occasion.   Haven't tried the Peak, though I've been considering a Summit when the hit streets.

The presumption is that the base sound of the Rev2 isn't good.   This seems to be based a few objective points: 

1.  Some people don't like the Curtis filters.   This all just comes down to taste.   I personally think the DSI curtis filters sound amazing!..   Especially for poly-playing - pads, strings, brass.    I do like other filters types too, and prefer Ladder for some bass and lead sounds, but the Curtis filters are very musical and versatile.   There have been countless hit songs made that use these Curtis filters.

2a.  The Rev2 uses DCOs instead of VCOs.   The raw sound of a single DCO oscillator compared with a single VCO has an almost imperceptible difference -- but there is a very small amount of harmonic jitter in all VCOs... I wrote an article about it here:  https://www.presetpatch.com/article/vco-harmonic-jitter  You can use the modulation matrix in the Rev2 to match the same sort of harmonic jitter as found in VCOs.    I've found that for leads/bass, this can be useful... In my patch bank I have built in this modulation to some patches.

2b.  The other effect of using DCOs is that they are "too-perfect" from a tuning perspective for some sound design.   The frequency clock is perfectly in tune for every oscillator and across every voice.   If you analyze classic VCO poly synthesizers, or acoustic instrument ensembles, you'll find that every oscillator has slightly different tuning performance, and furthermore that every voice has different tuning offsets (as well as other offsets).   This is what I'd say is the one downside of DCO poly architecture (if you want to call it a downside).    This can also be addressed by using the mod matrix... and in fact, the end-result is that you can model just about any classic synth voice character, even better than with some modern VCO poly synths without the mod matrix and gated seq capabilities.   

This type of voice modeling completely closes the gap for me, and I feel like it can accomplish any type of synth character, from a Memory Moog on a hot and humid day, to an OBX or Jupiter, to more modern stable tuning synths.    Plus, you still have the option to keep the DCOs in their inherit perfect tuning, and achieve pristine, glassy sounds that a VCO poly synth could never accomplish.   It's the best of both worlds to me, and I hope Seq and other manufacturers continue to develop the DCO-poly architecture, and just expand upon voice modeling capabilities.  http://www.voicecomponentmodeling.com/

If you want to try some patches out using the voice modulation methods described above, I've uploaded some free ones here:   (all the ones with VCM in the patch name)
https://www.presetpatch.com/user/CreativeSpiral

[Razmo]

Well... I have them both, and I've also played a bit with the PEAK sound design... I agree that the PEAK may sound a tad bit more sterile, but if you know how to program it right it certainly can sound "just  as good" as a REV2, and be a hell of a lot more flexible tonewise.

The sentence "just as good" is probably not the right way to put it... a digital front end will probably always have a more clinic sound to it, but to say that it does not sound "just as good" is troublesome in my opinion... it sounds DIFFERENT... then the choice is what the user likes best... i actually like BOTH's sound equally well... I'm not inclined to compare them, I chose them both because they compliment each other, and I sure as hell would not be without my PEAK, that's for sure.

I can do sounds and sound design with PEAK that I will not be able to get with the REV2... sounds that I actually WANT! ... I've made bass patches with the PEAKs filter that sound just as deep a the REV2, but just even more "controlled" (hard to describe this, but it just sounds like the PEAK's sound has a better "grip" on the speaker, a stronger one, where the REV2 seems a bit more "loose" and "fuzzy" ... some users love that, other's do not... I just love to have both options.

Instead of comparing the PEAK to a REV2, it would probably be more fair to compare it to a Prophet 12? ... I've had that one, and for sure, audio wise, the PEAK kicks it off the table... the P12 is filled with a digital metallic sound as soon as you start using FM or other audio rate modulations, the PEAK do not have this because of the FPGA engine.

Summa Summarum; I don't find it fair to compare a REV2 and PEAK... at least not unless you're forced to only have one of them, and even in that situation, it's just down to personal taste...

[OceanMachine]

Could you get over a not-so-good basic sound because you loved the functionality?

Nope, which is why I sold my REV2 in favor of other DSI/Sequential offerings with VCOs and different filters, albeit with less overall functionality. I'm interested in the Peak & Summit strictly for digital tones, but I'll probably not make a purchase unless Sequential does something new along those lines.

[Tugdual]

QQ: I keep reading that the Peak has a FPGA and operates @ 24MHz. What does it mean exactly? As far as I understood the Peak output is Delta Sigma so the Niquist principle doesn't apply and you might not expect 12MHz bandwidth 
Also I suspect that this is the FPGA clock and this has nothing to see with the sample rate.

One other reason why I decided to get rid of the Peak is that it was as good as any VST while a decent analog synth can only be hardware. Of course if you want a digital machine with knobs, this is one option otherwise it is much cheaper to go for VSTs.

[Razmo]

QQ: I keep reading that the Peak has a FPGA and operates @ 24MHz. What does it mean exactly? As far as I understood the Peak output is Delta Sigma so the Niquist principle doesn't apply and you might not expect 12MHz bandwidth 
Also I suspect that this is the FPGA clock and this has nothing to see with the sample rate.

One other reason why I decided to get rid of the Peak is that it was as good as any VST while a decent analog synth can only be hardware. Of course if you want a digital machine with knobs, this is one option otherwise it is much cheaper to go for VSTs.

The FPGA thing is, that when it is running at that speed, audio rate modulations like Ringmod, Frequency Modulation etc. will sound much clearer without aliasing... on normal 44.1/48KHz sample rate synths you can usually hear the degradation in the audio especially when using very high frequencies.

Now I've had the Prophet 12, and I have the PEAK, and I can assure you, that the PEAK sound much much cleaner with audio rate modulations, even when playing at very high pitches.. this is crucial for FM sounds, especially if you want these sounds to be very clean sounding.

VSTs!? ... PEAK is not a VA, it's a hybrid... all but the digital oscillators and the control engine are analog all out... so in no way can you just take on a VST instead! ... The filters sound clearly analog to me, and I've not heard any VSTs capable of presenting the same analog tone yet.

[creativespiral]

As Razmo noted, doing audio rate frequency modulation (FM, RM, AM) seems to be one of the biggest advantages to FPGA oscillators over traditional digital DSP oscillators. 

Some VSTs do over-sample the oscillators and filters, but the general consensus is that there is still aliasing in most, the frequency modulation results in artifacts, and there is still a sense of "digital approximation", rather than realistic analog tone.   My prediction is that the "next generation" of FPGA synths (in 2-5 years) will model the entire signal chain: VCO, VCF, and VCA.   It will all be done at high rates (no-aliasing or artifacts), and the circuits will be modeled down to individual component level (resistors, diodes, transistors, caps, etc), resulting in ultra-realistic emulation of any (and all) classic analog circuits... including getting into the nitty-gritty details of electrical circuit behavior (like non linear electrical quirks, power sag, component threshold variance, etc..)   

These next-gen FPGA synths will fool even the "analog purists" in terms of synth tone, and you'll be able to buy a single super synth that can model a variety of different VCO and VCF topologies. You'll be able to switch between VCO and VCF circuits to create any classic synth character, as well as new hybrids.  This is the path that Fractal Audio has taken in the guitar amp and effects modeling world, and it is truly a game changer.   It's disrupting a giant part of the guitar amp/effects market, and being adopted by a who's who of professional guitarists.   

In terms of current generation of VSTs, my two VA favorites are uHe Diva and SonicProjects OP-X.   Part of the reason may be due to oversampling to prevent aliasing, but also, these are two synths that use a sort of voice-modeling like I've discussed - that can be achieved with the Rev2 and Deepmind.   To me, this is low hanging fruit - in the short term, every new synth could provide control over... It's just a matter of designing a lookup table / mod table for voice-by-voice offsets to parameters. It's a key aspect of getting classic analog synth character, and modeling real world acoustic instruments and ensembles.

[Razmo]

One thing to note also is, that even some custom sound chips of the 80's worked at much faster rates actually... and you can hear the difference! ... how many have actually listened to the raw sound of a DX7's FM custom YM chips? ... these run at MHz speed as well... even the Commodore 64 SID chip was clocked at 1Mhz and could even be clocked up to 3MHz on some chips.

I've been talking about this even before the FPGA became reality, not understanding why they kept using these 44.1/48KHz bandlimited approaches, when you could clearly hear the better audio quality of the older custom chips... anyone who has compared a VST FM synth's sound to even a simple YM2164 in the cheap old FB01 from Yamaha can hear the difference... even when it's bit depth was sometimes 12 or even 10bit (that gives a different kind of "aliasing" but still sounds good because the sample rate are in the megahertz range).

FPGA WILL be the future of hardware synths, and untill the VST world gets some kind of FPGA standard, then even hardware FPGA synths (like for example the KYRA) will just sound better than any VST trying the same thing.

[OneLittleFonzie]

Very interesting topic. I didn't know about FPGA.

What I do see is that there isn't really much progress in the VST area. Some are better than others. I do really like Valhalla effects; they are strong competitors for expensive plug-ins and even hardware.

What I think (and that is also what creativespiral tries to say as well). The people who make those hardware/software  synths/effects, don't really make the stuff the way the typical users would like it; whereas some extra things would already make a huge improvement. Example for Rev2:

A "few" missed opportunities.
1. Separate outputs
2. Detune individual oscillators
3. Pan each voice in the stereo field
4. More flexible split (0/8, 2/6/, 4/4, and more for 16 voice)
5. Midi channel per voice (seems like each voice of the rev2 is a stand-alone synth anyway)
6. The mentioned lookup table
7. Extravagance!! 64 pots for the (gate) sequencer
8. Thrill/ratched for gate sequencer
9. Midi out for gate sequencer
10. Panel mode like on my Sub Phatty (probably on other moog synths too); What you see is what you get.

Personally, I've been using the SonicProjects OP-X PRO-II (from older versions). And that piece of software is amazing. You can model any analog synths you want.

[creativespiral]

The oversampling can be done with traditional DSP... as mentioned in this thread, there are examples of both VSTs and digital hardware synths that user oversampling to reduce aliasing / artifacts...  it just takes the right design and processing power.     

The real strength of FPGA is that it's designed to model low level electrical circuitry.    That's what Verilog/HDL is all about (Hardware Description Language).    By building "virtual 1:1 electrical models" in HDL, that's what is going to separate next gen FPGA synths from modern digital/vst synths.   Rather than modeling a "mid level approximation" of a VCO circuit or VCF circuit, FPGA architecture and HDL excels at modeling the circuit down to actual individual components (transistors, resistors, caps, diodes, etc)... and capturing the little quirks of electrical circuits.   

Technically, this can be done with traditional DSP as well, but it's just not practical, because the processing power needed to model a poly synth down to the individual electrical component level would be ridiculous.    FPGA is meant for this type of low level electrical modeling though.   So next gen synths will use FPGA for the modeling of various VCO, VCF, VCA circuits, while probably still using traditional DSP for the "control layer"... ie: the higher level stuff like managing the interface, inputs, etc..   We'll be able to select models of any classic synth architecture, or hybrid combos of different style oscillators and filters.

The question is which manufacturer will be first for market with an end-to-end FPGA modeled synth engine, and which will do it best.  I thought Fractal Audio would be a good candidate, as they have shown a proof of concept in guitar amp/effect market, but it seems they don't have the resources or interest to branch into synth market currently.   Novation and Waldorf are beginning to use FPGA, but their current generation is just touching the surface.   

[Razmo]

I believe that Roland is also using FPGA in their AIRA range of products (?)

[OceanMachine]

All I know is that Roland’s ACB drum machines all have a thin hard plastic character to me. They don't sound nearly as good as the analog sides of current drum machines such as the Tempest, Rytm, and Alpha Base. Ofc all of those are much pricier and I'd take a TR-8S over what the B word are doing.

[Tugdual]

The oversampling can be done with traditional DSP... as mentioned in this thread, there are examples of both VSTs and digital hardware synths that user oversampling to reduce aliasing / artifacts...  it just takes the right design and processing power.     

The real strength of FPGA is that it's designed to model low level electrical circuitry.    That's what Verilog/HDL is all about (Hardware Description Language).    By building "virtual 1:1 electrical models" in HDL, that's what is going to separate next gen FPGA synths from modern digital/vst synths.   Rather than modeling a "mid level approximation" of a VCO circuit or VCF circuit, FPGA architecture and HDL excels at modeling the circuit down to actual individual components (transistors, resistors, caps, diodes, etc)... and capturing the little quirks of electrical circuits.   

Technically, this can be done with traditional DSP as well, but it's just not practical, because the processing power needed to model a poly synth down to the individual electrical component level would be ridiculous.    FPGA is meant for this type of low level electrical modeling though.   So next gen synths will use FPGA for the modeling of various VCO, VCF, VCA circuits, while probably still using traditional DSP for the "control layer"... ie: the higher level stuff like managing the interface, inputs, etc..   We'll be able to select models of any classic synth architecture, or hybrid combos of different style oscillators and filters.

The question is which manufacturer will be first for market with an end-to-end FPGA modeled synth engine, and which will do it best.  I thought Fractal Audio would be a good candidate, as they have shown a proof of concept in guitar amp/effect market, but it seems they don't have the resources or interest to branch into synth market currently.   Novation and Waldorf are beginning to use FPGA, but their current generation is just touching the surface.

As you said, FPGAs are digital circuits just like VSTs so in theory they totally cover the same field of application. I wonder how CUDA and FPGA compare in terms of FLOPS but ultimately that is equivalent in terms of technology and you'll need a mathematical model for any of the things that you want to mimic.
Now I don't think that anybody has an intention to completely mimic analog circuits, I even wonder if that is a good idea. Digital is another domain that has things to offer, such as granular synthesis and I don't think that its sole purpose is to represent legacy analog imperfection (while that could always be an objective of course).

Back to the original topic, I doubt that Novation had any intention to mimic an analog circuit with the Peak.
I also very much wonder what means that 24Mhz selling argument. I have been reading a lot of documentation on Delta Sigma and a lot of people report that Delta Sigma sounds dull compared to a proper DAC which is exactly what I don't like with the Peak, it misses a soul. Also Delta Sigma is clearly a huge cost saver since you barely need a Resistance-Capacitor pair to integrate the output signal so I keep seeing a lot of marketing argumentation instead if features. Probably I'm a bit biased because I did purchase a Peak with very high expectations that were not met.

Analog is also not a holy grail. I have a legacy Akai VX90 which really sounds terribly meh compared to the REV2  . I also restored a Roland SH101 which sounds a bit meh compared to the subsequent and (totally looks like a plastic toy  ).

So after all I'm quite happy with the Subsequent and the REV2. I might consider an analog drum box one of these days (the incoming Behringer RD-8?) and some effects but for now I'm happy with NI Battery 4, the ValhallaVintageVerb and NI Replika. It is my way to go to combine the best of the 2 worlds (as Novation advertises  ): true analog devices and some VSTs for the digital domain that is effects and sampled drum boxes.

[jg666]

Last year I went to a music shop and had a quick play around with the Peak. I always take my own headphones there to listen and when I listened to the Rev2 for the first time I immediately loved the sound and bought one on the spot.

With the Peak however, it just didn't do anything for me over and above what I was already able to do with other synths. Having read this thread (and others) I realise that it was myself who was in the wrong here as all I did was go through the first few dozen presets on the Peak and didn't have a 'proper' play around with it.

The presets didn't impress me and it's been the same watching videos of the Summit to be honest. I mentioned this in another thread on these forums and got a great explanation from Razmo about why it's wrong just to listen to the presets on these synths

[Razmo]

Only listening to facrory presets will never show off most synths capabilities, and you may cheat yourself into believing a synth is mediocre if you do... You should study both the synths theoretical capabilities and also its sound character, but i find that hard to do if you do not program a handful of your owns sounds before judging. In many cases people judge on other users sound design insteaad, and sounds not of the tye you would make yourself.

When I bought the REV2 it did not impress me from the factory sounds much bexause they are not my genre that much... But after having programed 128 of my own, I have seriously been surprised at how varied this synth can sound, even for an analog poly... Many of my sounds I would not have thought possible... That taught me that I have to actually try my own sounds before judging.

If you give the same chance on PEAK, I am sure you can get lots of nice sounds... Stop comparing it to something else... If you like else, then get else to do what else does best, and let PEAK be PEAK at what it is best at...

[OceanMachine]

the P12 is filled with a digital metallic sound as soon as you start using FM or other audio rate modulations, the PEAK do not have this because of the FPGA engine.

The P12 has an additional oscillator (per its layers), but the Peak does have the ability for linear FM.

[RobustAmerican]

Nothing to renew here. My appreciation for this synth has only grown since the day I ordered it. (Oct 2017) Part of that has to do with the high quality support offered by Sequential. I think Razmo's point is the most relevant here. Extended hands on time is crucial but with the retailers having less and less in the stores that has become "Buy, try and return" if it doesn't work out. I feel for the retailers a little. Just a little though.

[Dubone]

I have both. I love using them together. I don't think the digital VCOs in the Peak make any more difference than the DCOs in the Rev2. It's mainly unimportant. The rest of the Peak signal path is analog apart from the FX. I can get dirty sounds out of the Peak and I just don't believe anyone who thinks there is something unartistically digitally clinical in either of these two synths. It's really nonsense when it comes to making sounds. My main bass sources are analog and I like that, but I just find this digital versus analog thing stupid. The Peak filter is very vanilla but totally useable and the user interface with the filter knob dominating is excellent. You know some of the classic sounds that people describe as analog warmth were made with digital synths, dub techno stabs are a classic example.

Actually if you think you need a bunch of different synths to get different sounds you may need to focus more on sound design. Most modern synths enable you to make a limitless panoply of sounds that when properly placed in a mix render the digital or analog source oscillator completely unimportant.

[Razmo]

I do not agree, that the oscillator section in a synth is unimportant... Quite the contrary actually... The variation in timbre that you can get before the signal hits the filter is to me crucial in how much timbral variance you can get out of a synth.

But with that said, this is not the only importance of course, the actual modulation options available to modulate the oscillator parameters are just as important... All the different elements of a synth from the oscillators to the filters and amp section all play in tandem with the control engine... So if your oscillator is digital, and has both FM, AM, wavetables etc. Then you have a much broader range of timbral variation if it is flexible and controllable from the control engine... An analog oscillator section is rarely more sofisticated than a few mixable waveforms and some PWM, which hugely limits the raw soundsource...

This does not mean that analog cannot be as flexible it just rarely is on analog synths, unless you go modular... AM, FM, wavefolding etc... Lots of options there, but you rarely find them in the analog polysynths.