The optimal solution / (feature request) would be to add a third option to the Globals 3 / Parameter 4... so instead of two choices (Slop, Vintage), you have three (Slop, Vintage 1, Vintage 2)... Based on the analysis I did on P6, it appears that there's probably just a small array/table of static values that represent max position of the knob, so I'm assuming it probably wouldn't be a gigantic undertaking to do so... could be as simple as pointing the knob to a separate set of a couple dozen static values. (though I could be wrong... I'm just projecting based on being a programmer for day job... it could be more involved)
Really great analysis! I hope Sequential checks this out. I’d love to hear your recommendations implemented. I own a Prophet 10 and I do feel like turning vintage all the way up makes it pretty unplayable. It would be great if turning it all the way up made it sound super vintage, but still musical. I think your filter cutoff recommendations would help achieve that.
Fascinating investigation. What I'm beginning to realise is that there are patterns in tone and various concepts of sound that are more appealing to oneself once subtle nuances are present but indefinable to the ear. It seems you are exploring the asthetic value of sound in a scientific way. I applaud you and keep it up.
Thanks Marc! Yeah, a few years back I was frustrated with the lack of objectivity in synth discussions on various forums... so much synth bashing and praise based on total subjective adjectives - "organic, warm, fat, lush, gritty, harsh, woody, cold, etc..." I set out to try and find the specific, objective differences between DCO and VCO oscillators initially. (some other videos on my channel about phase jitter / harmonic jitter and other related topics)... but that lead to investigating and measuring per voice variances. Anyways, I've written up a lot more on these topics here: www.presetpatch.com/articles/voice-component-modeling Cheers, Jason
(Sonic Neutronic) - Wow some real nice work here man. Glad to see you picked up the Prophet 6 - I think it is probably my favourite synth if I am honest - sounds so good. i hope Sequential implement some of your recommendations!
Thanks dude! yeah, i'm excited about the P6! Great sound! I actually ordered it while I was on the waitlist for PolyBrute, and frustrated with waiting. Then two days later it was announced the PBs were inbound finally... So I've got these two amazing 6-voice polys vying for my attention... first world problems...
Thanks for this gem of a video. Very informative, on point and yes: geeky. As the topic deserves. I am working on a video game soundtrack with heavy Prophet-driven sounds and I am itching to get a Prophet keyboard instead of using virtual instruments. This video made me sit back comfortably on my chair 30 seconds in, because I knew I had hit exactly what I wanted to analyze. Thanks for taking the time to do this as it's - I know that - super time consuming to be this accurate. Now it's time to go off-budget and spend some money before I even make it from this commission ;)
I got a Rev2 after your recommendation, and I have to say it sounds fantastic. Your VCM techniques make it very customizable as to what sound I want, but I may still be hoping Sequential adds the vintage knob to the Rev2 as well as a shortcut if I don't feel like making a specific VCM set up for a sound (or to add VCM techniques to parameters that the vintage knob doesn't affect) since it seems to do some similar things. Loving the modulation capabilities of the Rev2 though so if I don't get that as well I will be alright haha :)
A couple extra notes here regarding OB-6... Disclaimer: I did these tests on Prophet 6... I have not tested an OB-6, and its possible the value ranges are different on the various instruments. Also worth noting for OB-6 users: Since the OB uses a 2-Pole filter, the extreme dissonance of the cutoff deltas, and large ADR filter variance would be less evident on the OB, because the filter slope is half as steep. With a 4-Pole filter (P6), you audibly hear the cutoff point much clearer, as there is a steep drop-off (24db/octave). With a 2-Pole, the audible location of the cutoff points is less clear since the drop-off is half as intense. Also, the resonance/Q on P6 is more intense, so that is another difference that makes this all much more evident on P6. If the values are the same on OB-6, I would still expect it to be audibly dissonant on patches with a decent amount of resonance, or longer ADR timings, but yeah, probably less evident. My analysis and recommendations are mostly focused on P6.
Confirmed. I rarely crank up the vintage knob to 100% though, mostly 50%. No problem with the OB-6 filter then. BTW: Doesn‘t VCM also affect the VCA envelope? Why should that be more precise than the filter ENV?
@@RayyMusik I usually target both the filter envelope and amp envelope, for most realistic vintage flavor. The filter envelope produces a more tangible / perceptible sound though... so I would always prioritize filter env over amp env for vintage modeling. On an instrument like Polybrute, you have basically unlimited mod slots for doing this type of modeling, so its easy to target everything. With other synths, there may be a tradeoff with sound design real estate.
Ideally, you want to be able to randomize the table, at will (e.g. by pressing a button) and stop the process when you found your dream vintage sound, and then store that table in the patch itself.
Having access to multiple tables of values is something I've definitely been asking for / advocating for with the synth manufacturers. However, using random number generation is not the best approach... You can get better results by curating the value sets with "modeled disorder" --- by curating the offsets, your can push them further for more wild vintage sounds, while avoiding extreme dissonance or wonkiness in certain areas... for instance, the offset between any two oscillators in a given voice, the offsets in the low bass octaves being too high and muddying up 100hz-800hz range with their overtones, etc... There are several areas where its better to curate the variance. (this is also evident if you measure vintage synths... when users or shops manually calibrate / tune filters, oscillators, and other circuitry, they naturally make choices on where to focus the synth with the best tuning.)
@@CreativeSpiral Yes, I understand that. But this is a long (and maybe boring) programming. The random-at-will suggestion, solves the problem by its simplicity : if you are not satisfied with how it sounds across all voices, just perform a new randomisation again.
Using a Unison patch with 6 voices on max Pan Spread and max 'vintage' is really difficult currently as one side always ends up FAR brighter than the other. I believe this is due to the same voices always appearing in the same place in the stereo field (1 through 6 going from left to right?). Would be far more useable if the most extreme variations were balanced accross the field so that it all blends together nicely.
This video is fantastic...thank you Creative Spiral for the analysis. As an owner of several analog synths with individually calibrated voice boards, I think your recommendations are great. Of note, I noticed with my OB-6 this weekend the same points you make; some voices and chords are just way out of whack when the others sound nicely variable so there needs to be an option for a more balanced distribution of offsets when jumping through voices. My only counterargument to your points is that some people might still want to push the vintage character in to the "broken synth" range for sound design, so I wouldn't get rid of that option. I certainly hope that Sequential makes these values editable in the future, or at least give us a few different sets of tables to choose from, on a per-parameter basis. Thanks again for starting this conversation and thanks to Sequential for the implementation of what we've all been wanting!
One option could be to increase the max detune of the oscillators to match the broken-ness of the envelopes etc. Then you can get broken synth if you want, but still have a "sane" vintage setting as well.
Yeah, the issue is that the Osc Tuning Max Values are perfect in my eyes, but Filter ADR timing are 2x what they should be and Filter Cutoff is like 5x what it should be... since they are all on a "different scale", in my opinion, we're missing out on a LOT of great vintage character. If they want to keep the current value ranges as is, perhaps they could just create a third variant of this available via the Globals, so you'd have SLOP, Vintage 1, and Vintage 2 options. My guess, based on this analysis I did, is that its literally just stored in a small table of values, and it should be a trivial/easy task to add another variant of max values... if my interpretation of how the max values is stored is correct (I'm a programmer/db engineer by trade, so just guessing how its implemented)
This was very helpful. Fantastic work. Only issue is that you are making assumptions about the usefulness of the extreme / dissonant side of the vintage knob. Having extreme offsets can be inspiring for people working on the more experimental side of the spectrum. I bet DSI left those extreme settings in there for weirdos to make unconventional patches. If you consider the history of synths, strange extremes may end up producing distinctive sounds. So it may not be useful for you, but for me it does the following: 1. Allows for me to create "wonky" versions of patches for dramatic effect. 2. Like when using a compressor, the exaggerated/ extreme ranges help me dial in the sound, before toning it down to what is musical.
Glad you enjoyed. I haven't checked out Obsession yet, but the demos sound great. There's been several new VSTs released in the past year that are using voice modeling type of setups... I'm very excited for the direction things are headed... In my eyes this is the most underrated, and important factor in capturing Vintage Character. I think we'll see all hardware and software synths offering up per-voice variance controls over the next few years... or at least I hope so.
Excellent analysis. I've recently been making measurements on a Prophet 12, which only has a slop parameter (which I'd argue does have a limited use at its minimum setting) but it has a large enough modulation matrix that a lot of things close to this improved vintage modelling can be implemented while still leaving room for sound design (though the mods cannot be voice-linked, so the max. variance would have to be lower to avoid dissonances). A couple of questions. On the polarity of the filter envelope offsets to prevent clipping, is there no precedent for minimum attack and decay times being delayed on some voices, in which case you'd need positive offsets? Also, is it possible that the excessive values for the filter cutoff frequency come from DSI testing the sound with low resonance settings, where perhaps you won't hear as much voice dispersion with lower values? If that's the case, there's perhaps a difficult trade-off to be made with how much variance you can give to the different parameters while retaining a useable sound across many styles of patches.
Thanks. I did a bunch of analysis on Seq's Slop algorithm a few years back and cataloged it on the Rev2 forum. (forum.sequential.com/index.php/topic,3203.0.html) Slop is a sort of a multi-lfo setup. The issues with slop are: 1. it only targets oscillators and not the rest of the voice characteristics. (unless used as a mod source for other routings) 2. It is random in nature, so it results in more dissonance at random times, when phase is at extremes. 3. It introduces artificial drifting motion. What I found from doing research on all the vintage synths was that the offsets should really be Stable and Curated. Stability is more realistic than artificial motion / constant changing of offsets. And by curating the values per-voice and per-component, you can get a lot more character, pushing the offset ranges, while strategically controlling the values / differences between oscillators in a given voice, or deltas of offsets between consecutively triggered voices. Slop, applied directly and as a mod source, can do in a pinch to get some of that vintage voice character though... and with patches that have shorter timings and lower res values, it may be sufficient. I'm not sure why cutoff variances are so high. Perhaps it was just tested by ear with low resonance, where you don't notice it as much. It could possibly be attenuated with an inverse relationship to resonance amount. Though, the best case would probably be to just introduce separate vintage controls for different aspects of the synth engine, so users could control them on a per-patch basis.
Yeah, that was my general thought as well. Prophet 6 does have additional features, and an extremely similar core sound. The P10r4, of course has ten voices which is a nice option. But vs P5r4, you get an extra voice on P6, plus extra sound design options. The P6 does not currently have the old P5 voice allocation scheme, or the two filter choices... which are the main advantages on P5. if you really want the same voice cycling and filter choices of the Rev2 and Rev3 from the 80s. Here's a comparison chart: www.presetpatch.com/articles/Sequential-Prophet-6-vs-New-Prophet-5-and-10-Rev4
Great analysis, but it’s weird to make specific prescriptive suggestions without actually listening to the results of the suggested changes. Final tuning of these values should be done by ear vs. theory. I also don’t agree that clipping of envelope timing values is an issue. Very few sounds will encounter the upper limit (it’s an edge case) and a mix of positive and negative offsets from nominal will likely yield a more natural outcome for the vast majority of sounds. My guess is the best result would come from a subset of your recommendations: cut the envelope timing variances in half, do the same for cutoff frequency (reducing them to 1/5 only makes sense for sounds that rely on filter self-oscillation, which is an edge case) and shuffle the voice order. Overall, I think your recommendations are directionally correct but focus too much on edge cases. Picking the “best” values is ultimately a subjective exercise and requires listening to a wide range of sounds with various proposed offsets.
The suggestions are based on several years of doing research on this topic of voice variances and taking measurements from a variety of vintage synths. I have consulted with other synth manufacturers as well, to curate voice variance options on other hardware synths. In addition, I've built a software poly synth where I've tested all the value ranges both by ear, and technically. So, yeah, I'm not just randomly making suggestions. Anyways, seems you agree with most of them anyways. The "best" sound will vary, depending on the patch type you are creating (pad, strings, brass, pluck, lead, resonant sweep, etc)... the difficulty in designing a single knob (the Vintage Knob) that controls all of the offsets together, is that you have to try to balance the behavior across all the different circuit components that are offset. It would be best to just have separate controls of osc variances, filters variances and time variances... then the current value ranges could even be left as-is... and the user would have the ability to dial in the variances to different sections of the synth architecture, based on the patch type. Or alternatively, as I mentioned, having a couple variants of the Vintage Knob offset tables (Vintage 1 Mode, Vintage 2 Mode, ...) would be a potential solution. The vintage knob is a good implementation overall... but it could be better.
@@CreativeSpiral I didn't say your suggestions were random. I said they're overly specific and prescriptive. There are many other factors to consider beyond what you've tested and decided are the most accurate values to simulate the synths you've measured. For one, there are quite a few vintage synths in the world. I doubt you've measured a significant percentage of them. Second, the objective of a vintage knob is ultimately to obtain a pleasing sound, which is of course a subjective goal. The designers at Sequential -- or any manufacturer -- likely have different taste than yours based on their own experience and expertise and will likely arrive at a different conclusion in tuning the final behavior of this type of control. It's not necessarily "bad" or "good" as you've framed it. Third, it's often useful to provide a wider range of adjustment than what might be considered "musically useful" just so users can easily hear the effect of a given control. An overly wide range reduces utility by limiting the resolution of the control, but an overly narrow range reduces utility by making it difficult to quickly interact with a control, if the act of adjusting it is difficult to hear. Sequential likely applied this and other UX principles to some degree in choosing the values they did. Not to miss the forest for the trees, yes, I agree the vintage control could probably be further refined to good effect. But I also think you're not considering the broader range of sonic and usability goals that inform implementation details, like the exact offset values to use in this methodology. For example, I find the vintage knob implementation on the OB-6 to be nearly perfect from a practical and musical standpoint. Everything up to 50% or so is fully usable. That makes the halfway point on the knob an easy to find sweet spot and a useful reference when designing sounds. Above 50%, both the detuning and other values quickly become dissonant and exaggerated, but they sound relatively proportional to each together even if the numbers say otherwise. The vintage knob on the P5 sounds like it has similar behavior to my ear. Maybe the P6 implementation is markedly different but I'd be surprised. The point is that you assume an ideal implementation should be musically useful all the way up to 100%. While I wouldn't mind a wider range of useful adjustment, I definitely want some range beyond the maximum I would ever use, just so I can easily dial past that point and back to where it sounds "right" for a given sound. We can assume that Sequential made an informed design decision in choosing the values they did and, therefore, probably had a similar usability goal in mind. Simply labelling it as a "flaw" that should be fixed misses a learning opportunity to try to understand why they made different choices than you would recommend. As a musician and synth buyer, my own preference is probably about halfway between what Sequential implemented and your prescribed values. One last detail: did you measure whether the vintage knob response is linear, exponential or somewhere in between? I think that's another interesting option to consider: providing higher resolution within a musically useful range while ensuring there's a wide enough range for extreme settings or special effects.
Heh... yeah, probably could have turned off fx on that... at the last minute I quickly decided that I might preview some patch examples, while I went through the excel doc of analysis... the point of that specific test wasn't really to feature the resonance sound, but just to approximately count out the seconds between attack peaks. The point I was making remains though... my feeling is that if you ever have 8-9 seconds variance between voice attack times that's a seriously broken synth in need of envelope circuit tuning/repair. There's actually another point I should have made: since the longer ADR timings are the same longer values per voice, if you combine the Attack and Decay timings together, you could get upwards of 16-18 seconds of difference between consecutive voices to complete the attack and decay stages.
Jason, to your knowledge, have any if these recommendations been acted upon by Sequential? Have you ever heard from anyone at Sequential about your VCM research? They must have seen it discussed/referenced on the Sequential forum, if nowhere else! Cheers! PS - I assume these recommendations would apply equally to the ‘vintage algorithms’ in the OB6, Tri6, P5/10 & the OBX8, as well - do you know?! Perhaps your recommendations here have in fact been implemented in one or more of these (either upon release, for the later-released units, or via later updates?!).
Yeah, they're aware of it at Seq. I think the vintage mode implementation is similar on all their instruments, but there may be some slight variations.
Great analysis of the vintage knob ! Thank you for this awesome work... but it seems nobody has pointed out the fact that the polymod section is actually doing its work in the vintage process too. Have you tried changing the values of the Filt Env and Osc 2 knobs in the polymod section while you were testing ? I mean with absolutely NO destinations engaged of course... Am I correct thinking it should have no effect then ? Because on my P6 it's clearly doing something on the filter cutoff offsets from the vintage knob. Or am I missing something evident ? In any case this "vintage" functionality is fascinating !
thank you for this analysis! anyone have a prophet 5 (or 10) rev4 too? is the vintage knob 'the same' ? or are there still reasons to get the prophet 5 ;-)
Hello and apologies if you cover this in the vid. In your Rev 2 component modeling video on the BladeRunner patch, you show that the bass can get "wobbly" and solve this by patching the amount to the key note (hope I'm explaining that correctly). Do you know if the P6 (or P5) vintage knob compensates for this wobbliness? If not, then it would seem your method on the Rev 2 may be superior? Thanks - this is another great video!
Thanks Jon! The key scaling of PWM or Vibrato is a trick that can only be done with a synth that has a mod matrix capable of routing key/note number to the LFO amount. I do often use it on Rev2, and now on Polybrute, as it allows you to really push PWM and Vibrato values without getting flubby bass. On P6, the LFO is directly applied, so at higher values, you do lose a bit of the stability in the bottom end. The P6 is a great sounding synth though.
Idk... I'm not really hearing what the big fuss is over the vintage knob. It just sounds basically like some kind of off pitch/tape effect/detune and you can get that with various effects and LFOs
It's a big improvement over previous generation slop, drift, or LFO based methods. Less artificial motion, less dissonance at higher values, and repeatable from session to session. I've done tons of research on the topic... If you check out www.VoiceComponentModeling.com you can read more about how it compares to other approaches. Also, its not just targeting osc pitch on a per-voice, per-osc basis, but also filter and amp parameters, and Envelope ADR timings. The net effect is that greatly helps to capture the character of vintage 70s/80s poly synths, and the per-voice offsets to various parameters, based on electrical tolerance differences between circuits/chips. Also, check out this video, starting at about 6:30 I do some examples of what per-voice variance does to character: ua-cam.com/video/m1kzNV_JK60/v-deo.htmlua-cam.com/video/m1kzNV_JK60/v-deo.html
@@CreativeSpiral thanks for the explanation and links. I am by no means even remotely close to being a piano/keyboard player. Not even slightly. Lol. But I love the sounds of synths and have a few good ones. Definitely chasing the late 70s and all thr 80s to mid 90s sounds of synths. All just to do my 3 finger not playing lol
@@CreativeSpiral I guess this applies to the “diverge” & “drift” in the Summit/Peak, then? No step sequencer either, I don’t think … so, it’s kind of stuck with the old, ‘slop’ effect, unfortunately? 🤔
@@kierenmoore3236 Yeah, don't think the Peak/Summit can achieve curated, stable per-voice offsets / vintage voice modeling like this... though I haven't totally kept up with their firmware updates over the past year or two.
Thanks Peter. I'm not sure about amp sustain variance per voice. Did you test it on P5/10, or have a source for that info? Would be interested in reading more if someone has done these type of tests with P5/10... I imagine amp sustain would be a hard parameter to test with accuracy.
@@CreativeSpiral I didn’t do any measurements, I only know that with the vintage knob turned up on my P10, some voices don’t fade to silence even when amp sustain is set to zero, which would indicate offsets on that parameter; if I dial the vintage knob back to 4, then the voices fade to silence as expected (all other settings being equal).
@@CreativeSpiral I kind of wish there weren’t offsets applied to the amp sustain to be honest (or could be disabled) as it can be really irritating when I’ve dialled in a nice sound with a touch of the vintage knob that then won’t fade out as desired, especially when doing piano/struck-string-type sounds.
@@peterkenney9158 Good point... I never target sustain in my emulations. A better parameter to target that is related would just be the Env Amount... which would just give small offsets to the peak and sustain when its at a higher level, but any zero'd out sustain would resolve to fully quiet.
yeah... the distortion/grunge on P3 is definitely aggressive, and a little bit goes a long way... P3 fortunately has tons of ways to control gain staging. For some controls having really extreme max values is okay, because you can just dial back the knob and just use a shorter area. In this case, since its controlling multiple parameters (Osc Tuning 1/2, Filter Attack, Decay, Release, Cutoff, etc) and the parameters are all "off by a different scale" it makes it that you can't just adjust the dial and get the specific vintage tone you may want. As I mention in the video, the Osc Tuning offsets are spot on in my opinion, as far as total range goes... but Filter ADR needs to be scaled down by a factor of 2x or so, and Filter Cutoff really needs to be scaled down by 4x to 5x what it currently is. So in this case, you can't just dial back the knob and get the perfect sweet spot. If you dial back to where filter cutoff is not insanely dissonant, you loose out on almost all of the interesting tuning offsets.
Creative Spiral So do you find the original slop more practical in use then the vintage feature? I agree 100% with your findings, I hope sequential is listening.
@@fuenstock No, I'd say the Vintage Knob is definitely a big improvement over Slop/Drift, even with the current value scaling... I've just found that for the majority of patches, I can't use more than 1/4 of the knob at max. With the updates I've suggested, it could be significantly better throughout the value range... at lower levels it would be more balanced, with less cutoff variance (would work better for high resonance patches), and at medium to higher levels it would be way more usable, without so much dissonance.
The optimal solution / (feature request) would be to add a third option to the Globals 3 / Parameter 4... so instead of two choices (Slop, Vintage), you have three (Slop, Vintage 1, Vintage 2)... Based on the analysis I did on P6, it appears that there's probably just a small array/table of static values that represent max position of the knob, so I'm assuming it probably wouldn't be a gigantic undertaking to do so... could be as simple as pointing the knob to a separate set of a couple dozen static values. (though I could be wrong... I'm just projecting based on being a programmer for day job... it could be more involved)
This video is like a job application for Sequential. Nice work!
Really great analysis! I hope Sequential checks this out. I’d love to hear your recommendations implemented. I own a Prophet 10 and I do feel like turning vintage all the way up makes it pretty unplayable. It would be great if turning it all the way up made it sound super vintage, but still musical. I think your filter cutoff recommendations would help achieve that.
The same for me. I want a rather reasonable amounr of dissonance at max value. I don't want a out of tune 10 voice synth. No way )
Fascinating investigation. What I'm beginning to realise is that there are patterns in tone and various concepts of sound that are more appealing to oneself once subtle nuances are present but indefinable to the ear. It seems you are exploring the asthetic value of sound in a scientific way. I applaud you and keep it up.
Thanks Marc! Yeah, a few years back I was frustrated with the lack of objectivity in synth discussions on various forums... so much synth bashing and praise based on total subjective adjectives - "organic, warm, fat, lush, gritty, harsh, woody, cold, etc..." I set out to try and find the specific, objective differences between DCO and VCO oscillators initially. (some other videos on my channel about phase jitter / harmonic jitter and other related topics)... but that lead to investigating and measuring per voice variances. Anyways, I've written up a lot more on these topics here: www.presetpatch.com/articles/voice-component-modeling Cheers, Jason
(Sonic Neutronic) - Wow some real nice work here man. Glad to see you picked up the Prophet 6 - I think it is probably my favourite synth if I am honest - sounds so good.
i hope Sequential implement some of your recommendations!
Thanks dude! yeah, i'm excited about the P6! Great sound! I actually ordered it while I was on the waitlist for PolyBrute, and frustrated with waiting. Then two days later it was announced the PBs were inbound finally... So I've got these two amazing 6-voice polys vying for my attention... first world problems...
@@CreativeSpiral Well you know with summer coming up and all you will need something to keep you busy in the house for long hours ;) hehe
Where do you live, Jason - Arizona? ☀️
Awesome analysis, thanks for the great video.
Thanks for this gem of a video. Very informative, on point and yes: geeky. As the topic deserves. I am working on a video game soundtrack with heavy Prophet-driven sounds and I am itching to get a Prophet keyboard instead of using virtual instruments. This video made me sit back comfortably on my chair 30 seconds in, because I knew I had hit exactly what I wanted to analyze. Thanks for taking the time to do this as it's - I know that - super time consuming to be this accurate.
Now it's time to go off-budget and spend some money before I even make it from this commission ;)
I got a Rev2 after your recommendation, and I have to say it sounds fantastic. Your VCM techniques make it very customizable as to what sound I want, but I may still be hoping Sequential adds the vintage knob to the Rev2 as well as a shortcut if I don't feel like making a specific VCM set up for a sound (or to add VCM techniques to parameters that the vintage knob doesn't affect) since it seems to do some similar things. Loving the modulation capabilities of the Rev2 though so if I don't get that as well I will be alright haha :)
Excellent work, as I have noticed this on my Ob6 compared to my OB8. I hope your work and suggestions get through to Tony! 👍
Thank you, very interesting and excellent research ...
A couple extra notes here regarding OB-6... Disclaimer: I did these tests on Prophet 6... I have not tested an OB-6, and its possible the value ranges are different on the various instruments. Also worth noting for OB-6 users: Since the OB uses a 2-Pole filter, the extreme dissonance of the cutoff deltas, and large ADR filter variance would be less evident on the OB, because the filter slope is half as steep. With a 4-Pole filter (P6), you audibly hear the cutoff point much clearer, as there is a steep drop-off (24db/octave). With a 2-Pole, the audible location of the cutoff points is less clear since the drop-off is half as intense. Also, the resonance/Q on P6 is more intense, so that is another difference that makes this all much more evident on P6. If the values are the same on OB-6, I would still expect it to be audibly dissonant on patches with a decent amount of resonance, or longer ADR timings, but yeah, probably less evident. My analysis and recommendations are mostly focused on P6.
Confirmed. I rarely crank up the vintage knob to 100% though, mostly 50%. No problem with the OB-6 filter then.
BTW: Doesn‘t VCM also affect the VCA envelope? Why should that be more precise than the filter ENV?
@@RayyMusik I usually target both the filter envelope and amp envelope, for most realistic vintage flavor. The filter envelope produces a more tangible / perceptible sound though... so I would always prioritize filter env over amp env for vintage modeling. On an instrument like Polybrute, you have basically unlimited mod slots for doing this type of modeling, so its easy to target everything. With other synths, there may be a tradeoff with sound design real estate.
Ideally, you want to be able to randomize the table, at will (e.g. by pressing a button) and stop the process when you found your dream vintage sound, and then store that table in the patch itself.
Having access to multiple tables of values is something I've definitely been asking for / advocating for with the synth manufacturers. However, using random number generation is not the best approach... You can get better results by curating the value sets with "modeled disorder" --- by curating the offsets, your can push them further for more wild vintage sounds, while avoiding extreme dissonance or wonkiness in certain areas... for instance, the offset between any two oscillators in a given voice, the offsets in the low bass octaves being too high and muddying up 100hz-800hz range with their overtones, etc... There are several areas where its better to curate the variance. (this is also evident if you measure vintage synths... when users or shops manually calibrate / tune filters, oscillators, and other circuitry, they naturally make choices on where to focus the synth with the best tuning.)
@@CreativeSpiral Yes, I understand that. But this is a long (and maybe boring) programming. The random-at-will suggestion, solves the problem by its simplicity : if you are not satisfied with how it sounds across all voices, just perform a new randomisation again.
Using a Unison patch with 6 voices on max Pan Spread and max 'vintage' is really difficult currently as one side always ends up FAR brighter than the other.
I believe this is due to the same voices always appearing in the same place in the stereo field (1 through 6 going from left to right?).
Would be far more useable if the most extreme variations were balanced accross the field so that it all blends together nicely.
This video is fantastic...thank you Creative Spiral for the analysis. As an owner of several analog synths with individually calibrated voice boards, I think your recommendations are great. Of note, I noticed with my OB-6 this weekend the same points you make; some voices and chords are just way out of whack when the others sound nicely variable so there needs to be an option for a more balanced distribution of offsets when jumping through voices. My only counterargument to your points is that some people might still want to push the vintage character in to the "broken synth" range for sound design, so I wouldn't get rid of that option. I certainly hope that Sequential makes these values editable in the future, or at least give us a few different sets of tables to choose from, on a per-parameter basis. Thanks again for starting this conversation and thanks to Sequential for the implementation of what we've all been wanting!
One option could be to increase the max detune of the oscillators to match the broken-ness of the envelopes etc. Then you can get broken synth if you want, but still have a "sane" vintage setting as well.
Yeah, the issue is that the Osc Tuning Max Values are perfect in my eyes, but Filter ADR timing are 2x what they should be and Filter Cutoff is like 5x what it should be... since they are all on a "different scale", in my opinion, we're missing out on a LOT of great vintage character. If they want to keep the current value ranges as is, perhaps they could just create a third variant of this available via the Globals, so you'd have SLOP, Vintage 1, and Vintage 2 options. My guess, based on this analysis I did, is that its literally just stored in a small table of values, and it should be a trivial/easy task to add another variant of max values... if my interpretation of how the max values is stored is correct (I'm a programmer/db engineer by trade, so just guessing how its implemented)
This was very helpful. Fantastic work. Only issue is that you are making assumptions about the usefulness of the extreme / dissonant side of the vintage knob. Having extreme offsets can be inspiring for people working on the more experimental side of the spectrum. I bet DSI left those extreme settings in there for weirdos to make unconventional patches. If you consider the history of synths, strange extremes may end up producing distinctive sounds. So it may not be useful for you, but for me it does the following: 1. Allows for me to create "wonky" versions of patches for dramatic effect. 2. Like when using a compressor, the exaggerated/ extreme ranges help me dial in the sound, before toning it down to what is musical.
Thanks for doing this research and sharing. This very helpful for Synapse Audio's Obsession which has per voice offsets.
Glad you enjoyed. I haven't checked out Obsession yet, but the demos sound great. There's been several new VSTs released in the past year that are using voice modeling type of setups... I'm very excited for the direction things are headed... In my eyes this is the most underrated, and important factor in capturing Vintage Character. I think we'll see all hardware and software synths offering up per-voice variance controls over the next few years... or at least I hope so.
Wonderful! I would love to see how this compares to the Prophet 5. I believe that the algorithm is the same but the values are different
Yeh, I would love to see a comparison... I do hear that the vintage knob elevated the Prophet 6 above the 5 though!
Excellent analysis. I've recently been making measurements on a Prophet 12, which only has a slop parameter (which I'd argue does have a limited use at its minimum setting) but it has a large enough modulation matrix that a lot of things close to this improved vintage modelling can be implemented while still leaving room for sound design (though the mods cannot be voice-linked, so the max. variance would have to be lower to avoid dissonances).
A couple of questions. On the polarity of the filter envelope offsets to prevent clipping, is there no precedent for minimum attack and decay times being delayed on some voices, in which case you'd need positive offsets? Also, is it possible that the excessive values for the filter cutoff frequency come from DSI testing the sound with low resonance settings, where perhaps you won't hear as much voice dispersion with lower values? If that's the case, there's perhaps a difficult trade-off to be made with how much variance you can give to the different parameters while retaining a useable sound across many styles of patches.
Thanks. I did a bunch of analysis on Seq's Slop algorithm a few years back and cataloged it on the Rev2 forum. (forum.sequential.com/index.php/topic,3203.0.html) Slop is a sort of a multi-lfo setup. The issues with slop are: 1. it only targets oscillators and not the rest of the voice characteristics. (unless used as a mod source for other routings) 2. It is random in nature, so it results in more dissonance at random times, when phase is at extremes. 3. It introduces artificial drifting motion. What I found from doing research on all the vintage synths was that the offsets should really be Stable and Curated. Stability is more realistic than artificial motion / constant changing of offsets. And by curating the values per-voice and per-component, you can get a lot more character, pushing the offset ranges, while strategically controlling the values / differences between oscillators in a given voice, or deltas of offsets between consecutively triggered voices. Slop, applied directly and as a mod source, can do in a pinch to get some of that vintage voice character though... and with patches that have shorter timings and lower res values, it may be sufficient.
I'm not sure why cutoff variances are so high. Perhaps it was just tested by ear with low resonance, where you don't notice it as much. It could possibly be attenuated with an inverse relationship to resonance amount. Though, the best case would probably be to just introduce separate vintage controls for different aspects of the synth engine, so users could control them on a per-patch basis.
Even though I own an OB-6 I’ve been thinking about picking up a Prophet 6. I mean, it seems it can do more than the Prophet 5 reissue now.
Yeah, that was my general thought as well. Prophet 6 does have additional features, and an extremely similar core sound. The P10r4, of course has ten voices which is a nice option. But vs P5r4, you get an extra voice on P6, plus extra sound design options. The P6 does not currently have the old P5 voice allocation scheme, or the two filter choices... which are the main advantages on P5. if you really want the same voice cycling and filter choices of the Rev2 and Rev3 from the 80s. Here's a comparison chart: www.presetpatch.com/articles/Sequential-Prophet-6-vs-New-Prophet-5-and-10-Rev4
Great analysis, but it’s weird to make specific prescriptive suggestions without actually listening to the results of the suggested changes. Final tuning of these values should be done by ear vs. theory. I also don’t agree that clipping of envelope timing values is an issue. Very few sounds will encounter the upper limit (it’s an edge case) and a mix of positive and negative offsets from nominal will likely yield a more natural outcome for the vast majority of sounds.
My guess is the best result would come from a subset of your recommendations: cut the envelope timing variances in half, do the same for cutoff frequency (reducing them to 1/5 only makes sense for sounds that rely on filter self-oscillation, which is an edge case) and shuffle the voice order. Overall, I think your recommendations are directionally correct but focus too much on edge cases. Picking the “best” values is ultimately a subjective exercise and requires listening to a wide range of sounds with various proposed offsets.
The suggestions are based on several years of doing research on this topic of voice variances and taking measurements from a variety of vintage synths. I have consulted with other synth manufacturers as well, to curate voice variance options on other hardware synths. In addition, I've built a software poly synth where I've tested all the value ranges both by ear, and technically. So, yeah, I'm not just randomly making suggestions. Anyways, seems you agree with most of them anyways.
The "best" sound will vary, depending on the patch type you are creating (pad, strings, brass, pluck, lead, resonant sweep, etc)... the difficulty in designing a single knob (the Vintage Knob) that controls all of the offsets together, is that you have to try to balance the behavior across all the different circuit components that are offset. It would be best to just have separate controls of osc variances, filters variances and time variances... then the current value ranges could even be left as-is... and the user would have the ability to dial in the variances to different sections of the synth architecture, based on the patch type. Or alternatively, as I mentioned, having a couple variants of the Vintage Knob offset tables (Vintage 1 Mode, Vintage 2 Mode, ...) would be a potential solution. The vintage knob is a good implementation overall... but it could be better.
@@CreativeSpiral I didn't say your suggestions were random. I said they're overly specific and prescriptive. There are many other factors to consider beyond what you've tested and decided are the most accurate values to simulate the synths you've measured.
For one, there are quite a few vintage synths in the world. I doubt you've measured a significant percentage of them. Second, the objective of a vintage knob is ultimately to obtain a pleasing sound, which is of course a subjective goal. The designers at Sequential -- or any manufacturer -- likely have different taste than yours based on their own experience and expertise and will likely arrive at a different conclusion in tuning the final behavior of this type of control. It's not necessarily "bad" or "good" as you've framed it. Third, it's often useful to provide a wider range of adjustment than what might be considered "musically useful" just so users can easily hear the effect of a given control. An overly wide range reduces utility by limiting the resolution of the control, but an overly narrow range reduces utility by making it difficult to quickly interact with a control, if the act of adjusting it is difficult to hear. Sequential likely applied this and other UX principles to some degree in choosing the values they did.
Not to miss the forest for the trees, yes, I agree the vintage control could probably be further refined to good effect. But I also think you're not considering the broader range of sonic and usability goals that inform implementation details, like the exact offset values to use in this methodology. For example, I find the vintage knob implementation on the OB-6 to be nearly perfect from a practical and musical standpoint. Everything up to 50% or so is fully usable. That makes the halfway point on the knob an easy to find sweet spot and a useful reference when designing sounds. Above 50%, both the detuning and other values quickly become dissonant and exaggerated, but they sound relatively proportional to each together even if the numbers say otherwise. The vintage knob on the P5 sounds like it has similar behavior to my ear. Maybe the P6 implementation is markedly different but I'd be surprised.
The point is that you assume an ideal implementation should be musically useful all the way up to 100%. While I wouldn't mind a wider range of useful adjustment, I definitely want some range beyond the maximum I would ever use, just so I can easily dial past that point and back to where it sounds "right" for a given sound. We can assume that Sequential made an informed design decision in choosing the values they did and, therefore, probably had a similar usability goal in mind. Simply labelling it as a "flaw" that should be fixed misses a learning opportunity to try to understand why they made different choices than you would recommend. As a musician and synth buyer, my own preference is probably about halfway between what Sequential implemented and your prescribed values.
One last detail: did you measure whether the vintage knob response is linear, exponential or somewhere in between? I think that's another interesting option to consider: providing higher resolution within a musically useful range while ensuring there's a wide enough range for extreme settings or special effects.
Fun comparison.
Maybe consider turning off the phaser when looking at the resonance peak : )
Heh... yeah, probably could have turned off fx on that... at the last minute I quickly decided that I might preview some patch examples, while I went through the excel doc of analysis... the point of that specific test wasn't really to feature the resonance sound, but just to approximately count out the seconds between attack peaks. The point I was making remains though... my feeling is that if you ever have 8-9 seconds variance between voice attack times that's a seriously broken synth in need of envelope circuit tuning/repair. There's actually another point I should have made: since the longer ADR timings are the same longer values per voice, if you combine the Attack and Decay timings together, you could get upwards of 16-18 seconds of difference between consecutive voices to complete the attack and decay stages.
@@CreativeSpiral It's a very good comparison and I think you're on to something here!
Jason, to your knowledge, have any if these recommendations been acted upon by Sequential? Have you ever heard from anyone at Sequential about your VCM research? They must have seen it discussed/referenced on the Sequential forum, if nowhere else! Cheers! PS - I assume these recommendations would apply equally to the ‘vintage algorithms’ in the OB6, Tri6, P5/10 & the OBX8, as well - do you know?! Perhaps your recommendations here have in fact been implemented in one or more of these (either upon release, for the later-released units, or via later updates?!).
Yeah, they're aware of it at Seq. I think the vintage mode implementation is similar on all their instruments, but there may be some slight variations.
Great analysis of the vintage knob ! Thank you for this awesome work... but it seems nobody has pointed out the fact that the polymod section is actually doing its work in the vintage process too. Have you tried changing the values of the Filt Env and Osc 2 knobs in the polymod section while you were testing ? I mean with absolutely NO destinations engaged of course... Am I correct thinking it should have no effect then ? Because on my P6 it's clearly doing something on the filter cutoff offsets from the vintage knob. Or am I missing something evident ? In any case this "vintage" functionality is fascinating !
thank you for this analysis! anyone have a prophet 5 (or 10) rev4 too? is the vintage knob 'the same' ? or are there still reasons to get the prophet 5 ;-)
Hello and apologies if you cover this in the vid. In your Rev 2 component modeling video on the BladeRunner patch, you show that the bass can get "wobbly" and solve this by patching the amount to the key note (hope I'm explaining that correctly). Do you know if the P6 (or P5) vintage knob compensates for this wobbliness? If not, then it would seem your method on the Rev 2 may be superior? Thanks - this is another great video!
Thanks Jon! The key scaling of PWM or Vibrato is a trick that can only be done with a synth that has a mod matrix capable of routing key/note number to the LFO amount. I do often use it on Rev2, and now on Polybrute, as it allows you to really push PWM and Vibrato values without getting flubby bass. On P6, the LFO is directly applied, so at higher values, you do lose a bit of the stability in the bottom end. The P6 is a great sounding synth though.
Idk... I'm not really hearing what the big fuss is over the vintage knob. It just sounds basically like some kind of off pitch/tape effect/detune and you can get that with various effects and LFOs
It's a big improvement over previous generation slop, drift, or LFO based methods. Less artificial motion, less dissonance at higher values, and repeatable from session to session. I've done tons of research on the topic... If you check out www.VoiceComponentModeling.com you can read more about how it compares to other approaches. Also, its not just targeting osc pitch on a per-voice, per-osc basis, but also filter and amp parameters, and Envelope ADR timings. The net effect is that greatly helps to capture the character of vintage 70s/80s poly synths, and the per-voice offsets to various parameters, based on electrical tolerance differences between circuits/chips. Also, check out this video, starting at about 6:30 I do some examples of what per-voice variance does to character: ua-cam.com/video/m1kzNV_JK60/v-deo.htmlua-cam.com/video/m1kzNV_JK60/v-deo.html
@@CreativeSpiral thanks for the explanation and links. I am by no means even remotely close to being a piano/keyboard player. Not even slightly. Lol. But I love the sounds of synths and have a few good ones. Definitely chasing the late 70s and all thr 80s to mid 90s sounds of synths. All just to do my 3 finger not playing lol
@@CreativeSpiral I guess this applies to the “diverge” & “drift” in the Summit/Peak, then? No step sequencer either, I don’t think … so, it’s kind of stuck with the old, ‘slop’ effect, unfortunately? 🤔
@@kierenmoore3236 Yeah, don't think the Peak/Summit can achieve curated, stable per-voice offsets / vintage voice modeling like this... though I haven't totally kept up with their firmware updates over the past year or two.
Good work, Jason, I hope Sequential looks at this.
Does the P6 vintage knob alter the amp envelope sustain parameter? It does on the P5/10.
Thanks Peter. I'm not sure about amp sustain variance per voice. Did you test it on P5/10, or have a source for that info? Would be interested in reading more if someone has done these type of tests with P5/10... I imagine amp sustain would be a hard parameter to test with accuracy.
@@CreativeSpiral I didn’t do any measurements, I only know that with the vintage knob turned up on my P10, some voices don’t fade to silence even when amp sustain is set to zero, which would indicate offsets on that parameter; if I dial the vintage knob back to 4, then the voices fade to silence as expected (all other settings being equal).
@@peterkenney9158 Nice... good measurement logic. I may do another round of analysis on P6 at some point... if I do, will test this. Cheers, Jason
@@CreativeSpiral I kind of wish there weren’t offsets applied to the amp sustain to be honest (or could be disabled) as it can be really irritating when I’ve dialled in a nice sound with a touch of the vintage knob that then won’t fade out as desired, especially when doing piano/struck-string-type sounds.
@@peterkenney9158 Good point... I never target sustain in my emulations. A better parameter to target that is related would just be the Env Amount... which would just give small offsets to the peak and sustain when its at a higher level, but any zero'd out sustain would resolve to fully quiet.
Hola excelente trabajo y análisis puedes hacer lo mismo para el moog one, o si tienes algunos patched similares al prophet rev 2
I also feel like sequential tends to over do some parameters.
The Pro3 grunge setting to me is so aggressive it’s hard to find a good use for it.
yeah... the distortion/grunge on P3 is definitely aggressive, and a little bit goes a long way... P3 fortunately has tons of ways to control gain staging. For some controls having really extreme max values is okay, because you can just dial back the knob and just use a shorter area. In this case, since its controlling multiple parameters (Osc Tuning 1/2, Filter Attack, Decay, Release, Cutoff, etc) and the parameters are all "off by a different scale" it makes it that you can't just adjust the dial and get the specific vintage tone you may want.
As I mention in the video, the Osc Tuning offsets are spot on in my opinion, as far as total range goes... but Filter ADR needs to be scaled down by a factor of 2x or so, and Filter Cutoff really needs to be scaled down by 4x to 5x what it currently is. So in this case, you can't just dial back the knob and get the perfect sweet spot. If you dial back to where filter cutoff is not insanely dissonant, you loose out on almost all of the interesting tuning offsets.
Creative Spiral
So do you find the original slop more practical in use then the vintage feature?
I agree 100% with your findings, I hope sequential is listening.
@@fuenstock No, I'd say the Vintage Knob is definitely a big improvement over Slop/Drift, even with the current value scaling... I've just found that for the majority of patches, I can't use more than 1/4 of the knob at max. With the updates I've suggested, it could be significantly better throughout the value range... at lower levels it would be more balanced, with less cutoff variance (would work better for high resonance patches), and at medium to higher levels it would be way more usable, without so much dissonance.