CMV: There's little reason for digital hardware synths should have less than 8 voices

[slicetwo]

I’m not computer programmer, so I could be totally naïve in this, but given how cheap processing power is these days, what’s the mindset behind digital hardware synths with less than 8 voices? I was looking at the Modal SkulptSE, but that’s only 4 voices. The DN is 8 voices, but over 4 tracks, so averaging 2 voices per track. The Microfreak is also only 4 voices, too. Alternatively, a base model iPad for $329 can have a crapload of voices, and Apple charges a premium for everything, so is it really even worth $329?. In the analog realm, it makes sense to have limited voices based on cost, but with digital, it seems a bit silly a this point.

Discuss among yourselves

[PeterHanes]

@slicetwo

Discuss among yourselves

Previously on Elektronauts:

DSP austerity in 2018

As a Brit, I know all about austerity but seriously, I would love a genuine explanation to why Elektron’s recent products have such low polyphony and track counts in 2018, is there a shortage of DSP in this world i’m not aware about?

[slicetwo]

To be fair, that thread is now over 3 years old

[Dymaxion]

Someone (with more time than me ) should write up a really clear version of the “microprocessors aren’t microcontrollers” explanation for audio folks that covers all the performance differences and explains manufacturing scale and cost. Would be really nice to just link to it every time this comes up.

[Octagonist]

The fact that we can now have small-run, niche, focused, quirky tools & instruments with novel interfaces, and that don’t cost more than a few month’s savings is the really mind blowing thing.

[craig]

I’m no computer programmer either but there must be reasons why things are limited in digital devices – otherwise, I can’t see why Roland (for example) wouldn’t give the JU-06A six voices like the original Juno 106 it is based on. And totally not scientific but only from my experience using different software synths, it largely depends on how much processing is being used for each voice and its signal path/processing. For example, u-he’s Repro is a pretty accurate component-level Prophet-5 emulation but its 8 voices sure do eat up CPU cycles in a jiffy – to the point that I can barely use any other VSTis without something glitching. And when building my own synths in Bitwig’s The Grid, every single module, LFO, or processor I add begins to eat through CPU.

TL;DR: I’m sure it’s a case of balancing CPU power with stability. No one wants a digital hardware synth that is prone to glitching / dropouts in certain situations. So, voices are capped for stability.

[slicetwo]

Damn you and your logical and well thought out answer.

[bwo]

Well there is the roland system 8 which is essentially 4 ACB chips running 8 voices at twice the resolution of the JU-06A. The real question is probably a price sweet spot, would enough people pay 800 for a 6 voice JU-06A?

I think when compared to something like an iPad, even if apple is overpriced they also are in proper mass production, economy of scale and all that. Even then how many instances of moog model 15 in 4 voice mode can you run at once without drop outs.

I want to say higher quality sound is more of an exponential problem rather than linear, I think people making hardware synths want to make sure there digital synth sounds damn good and are willing to drop voices for it.

[craig]

@bwo

price sweet spot

This! And your excellent points about trying to get the best possible sound out of digital.

[PeterHanes]

@slicetwo

that thread is now over 3 years old

More recently on Elektronauts:

Why not Multi-timbral?

If there is one thing that still confuses me about new synths these days, it’s this-I guess I expect that computer chip technology has expanded possibilities, now that we are here in the future, where you can face time someone on your watch. I mean 25 years ago Nord made 4 part synths with at least 4 voices each (maybe I’m wrong about this, I’ve only owned the NM). Now years later digital synths are bi-timbral at best. I guess in the end, maybe it’s just marketing and the way things are…This is…

[slicetwo]

As always, Peter puts me in my place. Haha.

[guga]

won’t be adding much here as there’s been enough mentioned between comments and links but one should really consider a DSP as a dedicated microprocessor which, nonetheless, has to process things in a rather fast and smooth way to allow for things to happen at the right time and sound the way they should and one should also consider that it might not be the only component taking care of generating/processing audio. as @craig has already mentioned:

@craig

it largely depends on how much processing is being used for each voice and its signal path/processing. For example, u-he’s Repro is a pretty accurate component-level Prophet-5 emulation but its 8 voices sure do eat up CPU cycles in a jiffy – to the point that I can barely use any other VSTis without something glitching. And when building my own synths in Bitwig’s The Grid, every single module, LFO, or processor I add begins to eat through CPU.

TL;DR: I’m sure it’s a case of balancing CPU power with stability