SynthTone PWM - aliasing?

I played a number of the issues arising to my (completely non musical non synthesist) sister to see if she could hear the issues I pointed to. She heard them all, instantly.

The PWM Aliasing (on all variations) is the most serious issue. You can hear it far below C7 - definitely from round G5 on almost all sounds. And - though I'm guessing - probably 60 - 80 % of live sets (and potentially 90% + registrations) use some synth tone with PWM - so it's everywhere.

I am not griping here - but this is genuinely serious. I hope Roland will take this seriously and implement a fix- but it will require a concerted effort from many users to email them to ask for it.


I also expect it to the prevalent on the FA06 and FA08 - so I'm intending emailing the likes of Soundonsound and Sonicstate.com to check it out when they fully review those instruments (perhaps they have already done so?).

In any case, it renders the use of PWM as utterly useless in any professional vein - and - other oscillators are not far behind. Aliasing-wise, my Roland GAIA is better - and this is one of the worst VA synth engines I've ever come across in this regard (my AN1x, OASYS, NL2 and Z1 suffer from none of these issues, nor does virtually every VA plugin I own (all of Arturia, Spectrasonics, NI, Korg Legacy, Logic Pro X, G-Gorce, ....). Even the DX7 does not suffer as bad as this - I have never witnessed Aliasing like this in the digital age. It's damming.

To demonstrate just how bad this issue is, try this:

- Enter "Single" Mode

- Edit the live set

- Deactivate all but one synth tone, and choose Synth Tone from Synth Legends 0016: SLT-JP8 16

- Deactivate all effects

- Play, for example, the chord of C starting on C5 on the Keyboard

- The aliasing is so bad and so dominant audibly, that you can barely hear the intended sound

Deadlines guys! Its about managers..

Synth techs and music guru's do not run Roland; sociopaths run Roland. Why do think they are still in business.

Its a horrible market to be in; I think! the music electronic instrument business!

They charge mass money for this product because they must stay afloat...

They cannot spend their money on engineers overtime to fix a few aliasing parcels...

Your answer is to be happy and not right all the time. In order to do this, you will get creative and use work-a-rounds...

No more arm chair quarter backing from the side lines; time to get in the game and throw the balls yourself!

You can do this!


Many companies like Kurzweil have voluntary engineers who work after hours! They love it! I do not believe they are getting payed. This should give you an understanding of this business and how easy it would be not to have these instruments... people are working on a shoe string budget at times; but they love it! The results; a new synth for you n me!

knolan wrote:

To demonstrate just how bad this issue is, try this:

- Enter "Single" Mode

- Edit the live set

- Deactivate all but one synth tone, and choose Synth Tone from Synth Legends 0016: SLT-JP8 16

- Deactivate all effects

- Play, for example, the chord of C starting on C5 on the Keyboard

- The aliasing is so bad and so dominant audibly, that you can barely hear the intended sound

The Synth Legend Tone 0016, which you've mentioned, uses three Partials. All of them use the second variation of the PWM wave. As I've already explained in my previous posts, those two selectable PWM wave variations reveal a crazy amount of loud aliasing artifacts, while the basic PWM wave is suffering much less from this issue.

Take Kevin's above example and change the PWM wave in all three Partials from the second variation (green virtual wave select button) to the basic PWM wave (grey virtual button), which reduces the aliasing noise very drastically, but it still remains audible in higher range, especially in that badly programmed Synth Legends Tone 0016!

The two selectable variations of the PWM wave are only usable for bass sounds, unless you are looking for digitally crushed sound effects. But the Roland programmers obviously weren't even aware of this fact!

Kevin is absolutely right! The Synth Legends sounds are a collection of pure ALIASING HELL!!!
It is even quite hard (as I've tried by myself) to recreate that amount of aliasing from some Synth Legends sounds, by starting from scratch with a single initialized SN Synth Tone Partial, because the amount of audible aliasing is also dependent on the combination of the PWM wave's modulation amount, LFO, Pitch, Filter and Envelope settings. You can avoid some excessive aliasing noises by sensible programming, using your ears.

But the programmers of the Synth Legends even managed to emphasize the aliasing to the highest possible extent (which is quite hard to achieve from scratch - try it yourself!), by their extremely stupid programming... It's just beyond belief!!!

As always, I can only highly recommend to program your OWN sounds from scratch, which is basically as easy and fast on the JP-80, as on a Juno-6! This will lead to much better results (also considering the aliasing issue), than those incredibly bad programmed Roland sounds!

The SN Synth engine has its obvious strength (very versatile, expressive, musical and responsive, with a lot of "mojo" and fat warmth, with a very "analogish" and organic behavior), but it also has its obvious limits. I think those limits are not necessarily bad engineering, but the result of an engineering compromise, due to its lack of processing power, a typical Roland issue...



;)



PS:

knolan wrote:

Also,

The Supersaw wave is not a VA wave - it's a sample. If you isolate it from all effects in just one tone and play it in the upper octaves, you can hear the looping (and - Aliasing!)...

I've just found this video demo of the new AIRA SYSTEM-1 Synth:

http://www.youtube.com/watch?v=R97kP-52y4k#t=57

Skip to 0:57, where he demonstrates the SYSTEM-1's Super Saw, Super Square and Super Triangle without increasing their "Detune" values! You can hear exactly the same kind of alleged "sample loop", that you hear in the JP-80's Super Saw. This "loop" effect disappears by increasing the JP-80's (and SYSTEM-1's) Super Saw "Detune" amount...

Following your above quoted speculation, does this mean, that the SYSTEM-1's Super Saw is also based on a badly looped sample? Does this mean that the new ACB (Analog Circuit Behavior) technology is using looped samples?

Or does it just mean, that you are wrong with your assumption, by misinterpreting the loop effect of the Super Waves (which only occurs with Detune value = "0") as a sample loop, when it is possibly the inevitable result of Super Saw VA modeling? I seriously doubt, that the SYSTEM-1's ACB technology is using bad sample loops. But who really knows?


Addendum: After further testing, there appears to be at least two detectable sample key split points of the JP-80's Super Saw. So it must be sample based! It isn't comparable with the original VA Super Saw of the JP-8000 and the aliasing is indeed very annoying!

We have to ask Roland some questions...

I am told that the SH01 Gaia synth used sampled waveforms, of which the JP80 osc are basically 28 osc of the SH01. I would say these are indeed sampled, wasn't the SH201 exactly the same? For the price point the System-1 is, I'm sure it isn't a new TRUE virtual analogue but i'm sure further investigations will give us the truth once people get their hands on one.

Listen to this Jupiter-80 sound demo of the SuperNATURAL analog modeling VA Synth engine:

http://www.gearslutz.com/board/attachments/electronic-music-instruments-electronic-music-production/388636d1394573346-goodbye-roland-jupiter-80-sn-synth-engine-aliasing-issue-revealed-jupiter-80-aliasing-issue-demo.wav

As you can hear, the annoying aliasing noise artifacts are completely dominating the initial sound!!!

I've really tried to avoid to "throw the baby out with the bathwater", but the renewal of my JP-80 love just failed after several tries and it died!

I finally got rid of both of my JP-80 and just sold them for a symbolic price, as money wasn't the issue in the first place anyway.

Unfortunately, I was blinded and wrong all the time! But the serious aliasing issue of the Roland SuperNATURAL analog modeling VA Synth engine will finally prevent any Roland product, which includes this badly flawed engine, from ever becoming a digital Roland classic! Those products will inevitably find their place on the waste dump of synth history!

As a former flamboyant proponent of the JP-80, I have to apologize to all of you, as I was unknowingly misleading you all. I was blinded for a long time, but now I've seen the light of truth!

I cannot recommend any Roland product, which includes this flawed SN Synth engine anymore. Only if you are looking for the SN Acoustic preset tones, then buy an Integra-7...

Roland, you've got me with the Juno-6, when I started as a kid! But now, you've lost me forever with your Jupiter-80 fraud!




Long live the ultra low aliasing synth engineering masterpiece, the OASYS!



Goodbye my friends and stay happy!

(Pharrell Williams' Oscars performance of "Happy": http://www.youtube.com/watch?v=GTvbD2MsvYc)

;)

I was just about to ask what you would be replacing your JP-80's with. I myself haven't heard the aliasing, although to be honest I've not tried to hear them. I do suspect and was told anyway that the SuperNatural Synth engine is based on Sampled waveforms derived from the V-Synth, and Gaia. The only difference with the v-synth was how in manipulated the waveforms using the variphrase technology so my guess (having a V-Synth XT) the synthesis was more flexible in terms of sound manipulation.

I wondered about the JP80 but since I only got mine for £1700 a year ago it's not too much of an issue for me, it stays in the studio as my main board (it replaced the Korg M3) and I mainly got it to help deeply edit the synthesis with the large screen, so far it's worked for me but I often think if I didn't have this, what would I replace it with? there isn't much out there that would give me the main board to integrate into my DAW based studio. I have the Integra-7 but the horrible workflow of having an ipad editor wouldn't want me to use this for deep editing. Then again if something major was released that catched my eye from the big three that was a 76 key mainboard I would probably think about changing since Roland's support for the JP80 ended a year ago.

@Dany2 I am not entirely sure whether your are joking or not. It does seem a bit over-the-top. :)

@skinmechanic I was looking for a synth for a long time without finding a good match. When I discovered the Jupiter-80 (some years after its release) it had a combination of features and a build quality that I finally felt justified the price. I haven't found anything out there that could replace it. The closest I have come is a Prophet 12 + a Tyros 5. That would cover both the analog-ish domain and the acoustic/articulation domain. But that combination is more than twice the price, includes a lot of stuff that I don't need, and still wouldn't give me a good keybed to play on. The Jupiter-80 is unique and will probably not be surpassed in the near future.

@knolan Thank you for the detective work you have put into this, and the clear explanation of how to reproduce the aliasing. I will examine this on my own synth and see if I can hear these issues.

Too much doom and gloom for me lads! I have to think that the aliasing only affects the SN VA engine with those sampled PWM waveforms(which I don't use anyway) so I won't be getting rid of my Jupiter-80 anytime soon. The SN strings and horns alone are enough for me to be satisfied.

I'm not giving a pass to the Roland engineers though and I do feel for those who do need those VA's and are sorely disaapointed! Kudos to Kevin for finding this obvious engineering weakness and it would seem that Roland released that part of the JP-80 maybe a little too soon. It is hard to believe that an ace like Scott Tibbs would have missed something like that but oh well, it is not a perfect world and mistakes do happen.

Meanwhile, if I need some killer VA sounds without aliasing, I have an old synth that is an oasis in a desert of sound![attachment: Ancient Aliasing_1.jpg] Sorry...I couldn't resist. Had to lighten the mood a wee bit! ;-D

Just got thru reading Dany2's thread over at gearslutz. I don't quite understand why you would use the alternate version of the PWM wave to highlight flaws in the Jupiter synth engine when you said yourself most of the artifacts are not present when using the standard wave setting. Why leave that out?

This new revelation makes no difference to me. Why give up all the greatness that is the Jupiter 80 over not being able to play in the higher registers using the alternate PWM settings? What I do is simply use a different piece of gear like a Moog Voyager. Isn't that why we own different synthesizers? Since I don't own a Oasys or Kronos I will have to accept these units do not alias with the same programming. If like Kevin described they use higher sampling rates & filtering to remove unwanted artifacts what is really happening to the sound? Does it get thin & buzzy up there? With the JU80, if you move your hands down a few octaves there is plenty to love. What does the Oasys sound like in those lower ranges? Does it have the same richness & fatness as Jupiter? Maybe this is one of the keys to the "roland sound".

In the end, play what you like. Enjoy the Korg koolaid. If you keep trying hard enough you'll eventually ingratiate yourself with the rest of the GS kiddos.

I haven't heard it, but I haven't looked for it either. The JP80 does have an excellent keybed and fits in well as my main board in the studio, along side my Integra-7, V-Synth XT and now the MoXF 6 9 (A Cheap way to get some of the Yamaha string sounds I like).

Coupled with getting it cheaper a few years in, and been able to edit right down to the partial and split individual osc so it does what I need it to do and like others have said it is a unique synth, if I got rid of it, there is nothing out there to replace it with. I did look at the Prophet 12 but I need the extra keys so 61 keys isn't enough.

It looks like I'll be able to get my hands on a JP-50 at long last, and get to spend some quality time with one. In anticipation, I've grabbed the additional patches from the Axial site, save for the dancey stuff, as well as the OS update and manuals. Now I just have to locate all my pedals from the recent move, oy. Still living out of boxes. ;P

But I will have to say that I still lust after the Jupiters and Integra more than the Kronos or new Kurzweils. As much as I love them, and synths in general, the SN stuff Roland put into their gear is simply awesome sauce. Now, if there was just a simple explanation of how to navigate around the thing and have easy access to the Live Sets. Hmm... the Manual button seems to be the magic key...

I can confirm what knolan and Dany2 says. The PWM oscillators sound buggy in the upper octaves. I had heard this "swirling" effect on my Jupiter-80 before, but I didn't know this was aliasing, i.e. unintended. On the presets I assumed the programming was not supposed to work outside a certain range. In synth edit I assumed I did something wrong, since it is quite possible to create weird sounds in many ways.

The aliasing is definitely triggered by increasing the pitch. The swirling is caused by the "Analog Feel" parameter, which simulates analog instability. Without the "Analog Feel" the aliasing is much more stable/periodic, but no less audible.

The aliasing sound itself seems to lie within a constant range regardless of the pitch of the played tone. This range appears to always be a lower pitch than the tone itself, since it is only triggered by high pitched tones. It therefore seems possible to filter out these lower frequencies on a higher frequency tone. But I haven't found a way to achieve this on the Jupiter-80 without simultaneously filtering out the intended sounds.

Saxifraga wrote:

So the digital PWM wave is much nearer to the exact mathematical thing. It's a feature not a bug of th JP-80 to give you those waveforms for experiments. Just apply a filter to those waves that cut the aliasing away and you have an analog PWM wave. Voila!

Your explanation of why/how this happens makes sense. Do you have any suggestions how to configure such a filter?

I am thinking something along the line of a high-pass filter which only kicks in above a certain pitch, but I can't find a way to program this on the Jupiter-80. There is an independent high-pass filter per partial, but it appears to be constant, therefore cutting off lower frequencies at all pitches.

Since the Jupiter-80 has so many oscillators, it might work to duplicate the partials into another tone and split the key ranges. One tone could then play the partials below a certain pitch (without the HPF) and the other tone could play the partials above the same pitch (with the HPF). Hmm, I will give this a try.

Reducing the 'Key Follow' parameter on all partials greatly reduces the aliasing effect.

Just to let you know that I contacted Roland UK (as did Cello) and they have forwarded the matter to Roland Japan. Absolutely no guarantee they'll do anything about it - but at least they replied to me, logged the call officially and forwarded it.

I sincerely hope they implement a fix, though I have my doubts. I feel it is important they do so, and i'll articulate these reasons to Roland when next talking with them:

- it will improve the quality of the JP80, JP50, Integra-7 and FA06/08

- it will install much needed confidence in an existing user base and allow us to promote Roland in a positive light to all prospective future Roland users

- not doing so will cast doubt over the quality of the System-1 synthesiser, and Roland could do without that.


I'll give it another week, but if I do not hear back from Roland or they do not implement a fix, I feel it is important to flag that with review sites such as Sonicstate.com for the benefit of prospective Roland purchasers.

All that said, and you all know how much of a struggle I've had with the JP80, I feel some perspective needs to be reinstated. Firstly in this regard, Dany - though you have sold your JP80's and utterly understand your perspective - perhaps you might consider coming back the JP80 at some future juncture - do not rule that out. Underneath it all, while I suspect you are now disappointed with the JP80, I believe you actually like it - and that is a real, potent and important reason to own and play an instrument, whatever its shortcomings.

I have to say - the whole registration-liveset thing annoys me every day - but - while lying against the wall of go my project studio for near on a year, I could not bring myself to sell it. Why? I like it. I like the look of it (even if its interface is a bit pointless in places). But - I do like the sound. PWM waves aside, all other waveforms are pristine, it has the bulk of the JD800 sample set on board, its filters are to die for in my opinion - it is VERY easy to get a rich, thick, warm 'analogue' sound from this instrument - make no bones about it. Its sawtooth waves are gorgeous - those alone are reason enough for me to keep it, but combined with the filters - it's a great sounding VA synth.

But then there are the SN Acoustic Tones. Cello and I have been doing a little work behind the scenes on this and hope to have more to say on that before too long - and we both agree that it is simply in a class of its own on that front.

But for me - the main reason I've kept it, and am now working very hard to tailor it to my needs, is its voice layering capability, and then registration live selection capability. The prospect of layering up to 26+ tones, plus managing live variations of those though the registration select buttons below the keyboard, make it utterly unique in my opinion. Imagine having a rich swirling pad sound made up of countless VA sounds and, say, Timpani in the lower octaves in one registration, and then the same sound without the Timpani in another registration and perhaps with a solo woodwind sound in the solo section. For these sorts of live purposes, this is a magnificent feature. To use it meaningfully will take a lot of configuration, but the possibility is there.

But Dany - the last thing I'll say is - as weird as this sounds - the instruments I love the most in my large arsenal of synths are the ones with flaws. And, for me, I only start to get to love an instrument after it has been discontinued - because something weird psychologically happens - only then do I accept that the remaining flaws will never be fixed, I know the 'warts' and flaws of the instrument, and accept them; and with that I feel liberated in that the challenge is then to make the most of the instrument, to push it and myself to attempt something new and worthwhile musically.

So Dany - if you feel in the near or long term future inclined, pick up a 2nd hand JP80 for a keen price and rekindle your passion for the instrument - it has not been misplaced - and as others have pointed out - it is magnificent physically - in terms of keyed, built like a tank, looks stunning, excellent controllers, iPad editing, ...

It's a flawed 'King' - but it has endless possibilities and is a class act.