Software vs. Analog in 2025 – Has the Balance Shifted?

[ghettosynth]

Also, another thing to consider is that hardware isn't as efficient at warming your room, unless of course it has toobs. So, in the winter, you can load more instances of Softube Modular to help heat up your space and keep the juices flowing.

So yes, again, point software!

[pekbro]

A lot of folks don’t really get how hw modular actually works around here.

[pdxindy]

Analog filters don't just sound different, they sound better.
...

They sound better. That is the perfect excuse that can't be argued against, isn't it? Until someone comes along and asks you to empirically define what makes a sound "better."

It's an easy question to answer. And it has been answered multiple times in this thread.

Analog filters don't alias. They don't fall apart when pushed to extremes. They have a naturally complex character that tends to be pleasing to the human ear. They just sound better.

[pdxindy]

Digital or analog is not a binary choice, hybrids are a thing too and have led to some truly outstanding gear. Digital into analog filters can be a wonderful combo.

I agree, digital osc's or a combo of analog and digital osc's into a then fully analog signal path is a great combo. There are lots of great sounding hybrids these days.

[ghettosynth]

I wanted to respond to this, also another post of yours. I'll get to them in time.

...Urs conceded that compromise was made to facilitate 16 voices of Diva on whatever their min-spec computer of the time was. Repro does a great job, but has fewer voices and was created at a time when processors were faster.

But Repro is not without compromises. It's per-voice distortion effect, not something that is a feature on a Prophet 5, sounds good... until you get up to the high notes. That's something actual analog distortion will never do.
...

And as has been mentioned, the plugins that really get things like the Model D right, including feedback, are voice limited, as were the inspiration for them. I brought up how great Anthem sounds, and that is also monophonic or 4 voice paraphonic. I'm sure those design decisions were made to deal with the limitations of modern computers, which contrary to IvyLeague's claims, are not capable of "unlimited voices."

So yeah, there is no such thing as "no compromise."

Exactly and this has always been the case. This was very common with hardware in the early 80s. People think that some features were actually features invented for them when, in fact, the features were side effects of the need to compromise.

DCOs for example were sold as never needed tuning and much more stable. The fact is that they dramatically reduce calibration costs which was needed to meet price points. One of the last classic synths of the era was the Matrix 12 which uses the CEM 3374. This chip is designed to be tuned via computer. So, let's think about that. The Matrix 6, which was much cheaper used DCOs but the flagship Matrix 12 used VCOs, surely if DCOs were the bestest possible thing the flagship would use those as well? But wait, there's more, the Matrix 1000 used a single crystal oscillator to drive all of the countdown timers where as the Matrix 6 used two high frequency Hartley HF oscillators, which could be tuned by voltage control via a varactor diode, in order to have some phase variation in the oscillator pairs in a given voice. Note, this is not the same as a VCO at audio rate controlled by an exponential converter. This is a linear oscillator at radio frequencies and the divide down process smooths the variations dramatically. Still, the entire set of choices were cost driven.

With the DX7 Yamaha tried to sell the narrative, at some level, that FM was capable of producing any sound and so filters weren't needed. Well, that's certainly not the case in practice and it's, in essence, a trivial result in theory. The reality is FM is just a complex "complex oscillator" and of course it can benefit from filters. Only about a half decade later, in fact, Yamaha included filters in the SY77. Again, this was a cost saving measure. The costs of converting 16 voices to analog and running them through filters and then mixing them to the outputs wasn't practical and digital filters, in practical terms, could not run on CPUs of the day.

The trend of improvement in DSP from Ace -> Diva -> Repro highlights the need to balance capability with market realities. This is far from over. Models will still have limitations and will continue to have them for some time to come. Yes, the limitations are more in the noise than they once were, but they still exist. Whether or not they matter, again, at the risk of sounding like a stuck record, is context dependent.

The only time that I upgraded my Arturia collection was when they included the digital classics. They could have my money again if they did a take on the Fizmo, or the K2000.

I think their Synclavier II, Prophet VS, SQ-80 are fantastic, and actually a lot of their other synthesizers are pretty damn good as well.

Indeed, that release (synclavier) was when I upgraded. I can't remember now, I also paid for the Buchla, which is fun. Weirdly, I even like playing with the 2600, even though I don't think that it's very good. The reason is that the ARP sequencer is fun. As long as you aren't modulating the filters too hard, it's usable.

I think that the M12 emulation is very usable. However, I still prefer my own Matrix 12 filter that I built in Reaktor core years before the Arturia synth came out. Now, it's not accurate, in fact, it's clearly wrong because I naively applied the pole mixing without analysis to a different filter. But, that's of no moment, whatever the actual response of the filter is, it's a delicious sizzly variant including an allpass with a lowpass and I love it on pads. In other words, as a model, it's completely wrong, but it's useful, especially on my stereo supersaw oscillator, which is also just an approximation.

So, in fact, even models that are more correct are not necessarily more useful to everyone in context than models that are less correct. I'm fairly certain that the Virus has some aspect, that is more wrong than current models, that people like. That doesn't make them a good model however, and it certainly doesn't make them sound closer to their analog counterparts.

Finally, one of the reasons that analog can make such a resurgence is that the manufacturing costs of surface mount is dramatically lower than that of through hole. The Behringer Moog and Arp circuits are all very discrete and, presumably, accurate to the original short of the exact nature of the parts owing to how technology has changed. This means that new analog designs that may have, at one time, been impractical are now more possible. This gives us Sequential Circuits synths with discrete 2040 implementations and multiple filters. Thus, the goalposts will continue to move as long as there is demand for analog. Because you can't just wave your hands and poof, new accurate model, we will still see software playing catchup in model accuracy for years to come. Feature parity is a different story, that will always cost more to produce in analog hardware and so those products aren't going to really try and compete with software directly in that sense. This is less true in digital hardware, where the big problem is justifying putting your software into hardware based on market acceptance and higher up front costs, e.g. Maschine+.

[pdxindy]

IME analog eurorack sounds better than digital counterparts, when they exist, which is not as often as you would think.

Also, there is a huge advantage, which we’ve gone over before, cpu boundaries…

Eurorack sound quality is astounding. Truly in a league of its own compared to software.

[D-Fusion]

https://musictech.com/news/gear/tom-obe ... mulations/

[seafire]

https://musictech.com/news/gear/tom-obe ... mulations/

Tom Oberheim on synth emulations: “Anything that brings new sounds is exciting

Oh the irony...

[pdxindy]

IME analog eurorack sounds better than digital counterparts, when they exist, which is not as often as you would think.

Also, there is a huge advantage, which we’ve gone over before, cpu boundaries…

Cpu limitations doesn't matter when it comes to a hardware vs software debate.

You can only use 1 instance of a hardware unit and you have to bounce to audio if you want to use another sound while you can use more instances on a Mac or pc before you need to freeze a track to audio.

CPU limitation matters more than anything else.

Take the Elektron Rytm. There is no plugin remotely close to it, but if some company made such a plugin and didn't compromise the sound quality to save CPU, it would not be playable on todays computers (in realtime).

I can only use one instance of a hardware synth in realtime, but I can add more hardware synths. I have more simultaneous analog voices in hardware than my computer could play in realtime with say Repro level quality. And Repro itself made various compromises in order to not crush your cpu.

Take one of my eurorack cases, and do a no compromise software emulation of the modules I have in it, and that one monophonic voice would kill your cpu and be unplayable.

[IvyBirds]

With the DX7 Yamaha tried to sell the narrative, at some level, that FM was capable of producing any sound and so filters weren't needed. Well, that's certainly not the case in practice and it's, in essence, a trivial result in theory. The reality is FM is just a complex "complex oscillator" and of course it can benefit from filters. Only about a half decade later, in fact, Yamaha included filters in the SY77. Again, this was a cost saving measure. The costs of converting 16 voices to analog and running them through filters and then mixing them to the outputs wasn't practical and digital filters, in practical terms, could not run on CPUs of the day.

The SY77 was a completely different synth from the DX7. It's not a pure FM synth like the DX7. Yamaha called it Realtime Convolution and Modulation Synthesis (RCM) where it combined Sample Playback with FM designed to be used together

The SY77 wasn't just sine waves doing FM, it had Yamaha's AWM2 sample playback engine that they still use on their flagship Synths and the ability to use samples as operators

The DX7 started with simple sine waves that when FM'd created rich harmonic content, the SY77 could do that, but it also had rich harmonic content from the get go that was intended to be used in a Subtractive Synthesis manner hence the reason they added filters

You are trying to make it sound that the DX7 needed filters but Yamaha cheaped out and didn't include them using the SY77 to support your hypothesis but that simply is not true

Filters are used in Subtractive Synthesis, John Chowning Yamaha Style DX FM is not Subtractive at all and doesn't need a filter

indeed, that release (synclavier) was when I upgraded.

Interestingly enough when New England Digital released the Synclavier II in 1980 it did FM only with additive tone generators for the Oscillators as sampling wouldn't be added for several years, they also licensed the John Chowning's FM patent from Stanford University just like Yamaha. A base system costs $13,000 ($54k in today's money) a fully loaded system with increased polyphony, hard drives and other options costs well into the hundreds of thousands. (Half a million or more in today's money)

Those used discreet voice cards that New England Digital spared no expense to build that had individual analog outs for each voice, adding an analog filter to each voice would have been trivial and not very expensive in 1980 but they did not do so. Did they cheap out also on instruments that sold for huge sums of money?

[IvyBirds]

CPU limitation matters more than anything else.

Take the Elektron Rytm. There is no plugin remotely close to it, but if some company made such a plugin and didn't compromise the sound quality to save CPU, it would not be playable on todays computers (in realtime).

There is nothing remotely difficult in an Elektron Rytm that couldn't be done in a modern plugin with a modern CPU, why do you think 8 mediocre analog drum voices would be an impossibly difficult task to pull off? Everything else in the Rytm is software running on a Coldfire CPU )based on the ancient Motorola 68060) thinking along at 300mhz. That's a joke compared with the CPU in a modern PC or Mac

There would be a very limited market for it however in plugin form however and since it's a current product no one has made one

[IvyBirds]

https://musictech.com/news/gear/tom-obe ... mulations/

"the fact that you can spend $1,000 and get a great system in your bedroom is exciting, no matter what’s being simulated. Imagine if you were able to do this in 1970.”

Tom Oberheim on Software Synths

[pekbro]

It would be true that the cpu in a modern pc would blow away the cpu in an Rytm if it only had to do as much as the chip does in the Rytm,
unfortunately, that’s not the case. The cpu in
a pc may be a lot faster, but it has exponentially
more to do when installed in a pc. Not to mention its bottlenecked at nearly every turn.

[ghettosynth]

Filters are used in Subtractive Synthesis, John Chowning Yamaha Style DX FM is not Subtractive at all and doesn't need a filter

You've been sucked in by the marketing. John Chowning didn't actually label a style of FM, per se. FM is not a "new type of synthesis", that is the song and dance that you bought into hook line and sinker. Chowning is very humble about his contributions and gives credit where it is due. He had virtually nothing to do with Yamaha's vision, which was to use the patented idea to create a line of products possible with the technology of the day.

indeed, that release (synclavier) was when I upgraded.

Interestingly enough when New England Digital released the Synclavier II in 1980 it did FM only with additive tone generators for the Oscillators as sampling wouldn't be added for several years, they also licensed the John Chowning's FM patent from Stanford University just like Yamaha.

Not according to Chowning. He claims that Yamaha had discussed legal action with him and that he requested that they do not take action as it came out of a university project and those were his friends.

Those used discreet voice cards that New England Digital spared no expense to build that had individual analog outs for each voice, adding an analog filter to each voice would have been trivial and not very expensive in 1980 but they did not do so. Did they cheap out also on instruments that sold for huge sums of money?

The Synclavier was primarily an additive synthesizer with a limited subset of Chowning's preferred FM structure which is quite simple as compared to Yamaha's FM algorithms, which, he had nothing to do with, his words. This architecture likely followed Chowning's voice development work. He spent significant time, about three years, developing sounds/patches (in Music V) that his simple structure was good at producing.

Synclaiver sold their synths largely to universities and large studios for whom it was for study, or generally did not need to be the only synth.

From Wikipedia:

The early Synclavier I used FM synthesis, re-licensed from Yamaha,[4][additional citation(s) needed][a] and was sold mostly to universities. The initial models had only a computer and synthesis modules;

At this time FM for sound was being studied. After Chowning's paper it was typical low lying research fruit so of course it was going to be studied. Having a computer that could synthsize those waveforms in real time was an advantage over Music V. From a research POV, you don't need filters there because you are researching the timbral possibilities of FM. Calling it a "new kind of synthesis" is the kind of crap you say to sell a paper, or a synthesizer. I find it interesting that the assumption is that Synclavier licensed it. Again, Chowning says that didn't happen.

When you add modulators to modulators you obtain increasingly harsh sounds that benefit from taming. Yamaha knew this, which is why in the SY77 the filters apply to the FM engine and this is called the AFM engine. The wave aspects are a red herring with respect to my point. Yamaha tried to sell this as a replacement for your analog keyboard. Again, synths of the day were as much "keyboards" as they were "synths." It was often quipped by retailers that the piano patch sells a synthesizer.

It's not about cheaping out, Yamaha didn't have a choice at the time regardless of whether there was a uniform belief among their engineering team whether filters were needed or not. There were no 16 voice analog polys either. A big selling point of the DX was the polyphony. However, as soon as the technology supported it at a price point that it could be put in a synth, it was. Shortly after, of course, Synclavier closed up shop.

So, regardless of whether or not Yamaha execs believed the pie in the sky vision, within half a decade the came around and realized that it was false. I give the Yamaha engineers more credit than that. I'm pretty sure more than a few said "this simulated filter sweep sounds like absolute shit, this would be much better with real filters."

[D-Fusion]

IME analog eurorack sounds better than digital counterparts, when they exist, which is not as often as you would think.

Also, there is a huge advantage, which we’ve gone over before, cpu boundaries…

Cpu limitations doesn't matter when it comes to a hardware vs software debate.

You can only use 1 instance of a hardware unit and you have to bounce to audio if you want to use another sound while you can use more instances on a Mac or pc before you need to freeze a track to audio.

CPU limitation matters more than anything else.

Take the Elektron Rytm. There is no plugin remotely close to it, but if some company made such a plugin and didn't compromise the sound quality to save CPU, it would not be playable on todays computers (in realtime).

I can only use one instance of a hardware synth in realtime, but I can add more hardware synths. I have more simultaneous analog voices in hardware than my computer could play in realtime with say Repro level quality. And Repro itself made various compromises in order to not crush your cpu.

Take one of my eurorack cases, and do a no compromise software emulation of the modules I have in it, and that one monophonic voice would kill your cpu and be unplayable.

Ok. That is it.
Now i have to sell all mys vst's so I can fill my wall with modules and spaghetti

On a serious note. Yes Modular is built for raw power and sound so I agree that Software is not there yet and if you like modular setups you can get lots of cool Filters and oscillators and other goodies to create a new synth when you want a change.

I have no cpu issues with Repro 5 and i can play 16 instances with HQ setting enabled with a 4 note chord with a pad called Deep Sky which uses up 89% when i play fast chords but i would never use that many instances of one synth so maybe 4 or 5 would the max before i add something else to the mix so i will still have plenty of cpu left for other things.

The Korg Arp 2600 vst is a cpu eater but i could still manage to use close to 8 in mono mode before i need to freeze a track.