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

[wintoid]

One thing I've brought up that keeps getting ignored is that in high quality analog emulations, the feature set is considerably lower than a Moog One. I consider the UAD Minimoog the "gold standard" for analog emulation, and it's a mono synth that's not much more than a hardware Minimoog. Why isn't there a great analog emulation that has 4 LFOs and 3 EGs? This idea that software always trumps hardware is just incorrect. Not when doing a good approximation of analog, it doesn't.

That doesn't bother me at all, I just use Bitwig's LFOs and EGs. All I need from the synth is the analog emulation. But then Bitwig is sort of an instrument to me.

Just to wade into this discussion, I have 8 hardware synths and one small modular case. I have software emulations of some of those synths. Sometimes the convenience of the software is enough to override any advantage the hardware might have. Sometimes the physicality of the hardware overrides the additional features of the software.

Sometimes it's just about which seat I want to sit in, seriously!

I have zero hardware effects units. All my effects are done in software. I don't see that changing.

I occasionally feel silly for having spent money on hardware when the software is really good enough. I occasionally feel silly for having spent money on software when I already had hardware that did the job very well. Most of the time, though, I'm just having fun.

The real winner is the consumer.

[ghettosynth]

Now, one can argue that in some abstract sense that FPGA designs are "software", but that's a distraction because FPGA implementations require hardware to execute, however, they are digital.

All software requires hardware to execute.

Specifically, FPGA "software" requires an, wait for it, FPGA. You don't have an FPGA that is user programmable for user applications in your PC and you likely never will.

FGPA are great if you want low cost, low power consumption for parallel processing. This is an advantage in hardware synth designs that are software based compared with the bargain basement CPUs that usually go into hardware synths that are just running software for example the many Korg models that use Raspberry pi

While that's not intrinsically wrong, it's sidestepping the point that George was making. There are specific advantages to FPGA use for audio, which is why it's sometimes used for projects like the UDO.

Once you enter the desktop space however a modern CPU will easily meet or exceed anything you want to do relative to the FGPAs that exist in mass produced hardware synths

Nope. You cannot achieve the same latency, in microseconds, or the same sample rate on a PC. This is from a talk from CCRMA at Stanford.

https://ccrma.stanford.edu/~rmichon/tal ... -oct22.pdf

When used with audio codec chips optimized for latency such as the ADAU 1787,
ultra-low latency performances can be obtained with our system.

The lowest “round-trip” latency that we managed to achieve so far is 11µs (at a
sampling rate of 768kHz).

Multiple ADAU 1787 codecs can be used on one FPGA. Hence, implementing a
system with 32 audio inputs and 32 audio outputs whith such performances can be easily done on a basic Zybo Z7 board.[/list]

They (CCRMA) are making progress on making FPGAs easier to program using Faust, but we are a long long long way from them finding sufficient application to be included in general PCs such that they would provide the same benefits to your software workflow that they provide to custom hardware. In fact, if I were to put money on applications that would drive this kind of hardware first, it wouldn't be music, but money, i.e., crypto.

[zerocrossing]

One thing I've brought up that keeps getting ignored is that in high quality analog emulations, the feature set is considerably lower than a Moog One. I consider the UAD Minimoog the "gold standard" for analog emulation, and it's a mono synth that's not much more than a hardware Minimoog. Why isn't there a great analog emulation that has 4 LFOs and 3 EGs? This idea that software always trumps hardware is just incorrect. Not when doing a good approximation of analog, it doesn't.

That doesn't bother me at all, I just use Bitwig's LFOs and EGs. All I need from the synth is the analog emulation. But then Bitwig is sort of an instrument to me.

Sure, but if it's not CLAP, you're dealing with global modulation, which is sometimes fine, sometimes not as good.

Just to wade into this discussion, I have 8 hardware synths and one small modular case. I have software emulations of some of those synths. Sometimes the convenience of the software is enough to override any advantage the hardware might have. Sometimes the physicality of the hardware overrides the additional features of the software.

Sometimes it's just about which seat I want to sit in, seriously!

8-10 hardware synths is the correct number of synths you should have. I've calculated this using vigorous science. I only have one chair, though. I'm not sure if that's too few. I do like it, though. It's the identical chair to the "CEO chair," immortalized in Silicon Valley.

I have zero hardware effects units. All my effects are done in software. I don't see that changing.

I sort of do all effects in software... would a UAD Apollo count as hardware? I do have one. I also have a couple of distortion pedals, a wah pedal and a Line6 HX Stomp XL I use for exotic effects pedal sounds and other synth stuff. I use it because I have a dual axis expression pedal that's analog and allows for cool performance stuff to happen. So... I guess I don't do all effects in software.

I occasionally feel silly for having spent money on hardware when the software is really good enough. I occasionally feel silly for having spent money on software when I already had hardware that did the job very well. Most of the time, though, I'm just having fun.

The real winner is the consumer.

Oh, it's all silly. This I know. Again, though serious and rigorous scientific study.

[IvyBirds]

Now, one can argue that in some abstract sense that FPGA designs are "software", but that's a distraction because FPGA implementations require hardware to execute, however, they are digital.

All software requires hardware to execute.

Specifically, FPGA "software" requires an, wait for it, FPGA. You don't have an FPGA that is user programmable for user applications in your PC and you likely never will.

FGPA are great if you want low cost, low power consumption for parallel processing. This is an advantage in hardware synth designs that are software based compared with the bargain basement CPUs that usually go into hardware synths that are just running software for example the many Korg models that use Raspberry pi

While that's not intrinsically wrong, it's sidestepping the point that George was making. There are specific advantages to FPGA use for audio, which is why it's sometimes used for projects like the UDO.

Once you enter the desktop space however a modern CPU will easily meet or exceed anything you want to do relative to the FGPAs that exist in mass produced hardware synths

Nope. You cannot achieve the same latency, in microseconds, or the same sample rate on a PC. This is from a talk from CCRMA at Stanford.

https://ccrma.stanford.edu/~rmichon/tal ... -oct22.pdf

When used with audio codec chips optimized for latency such as the ADAU 1787,
ultra-low latency performances can be obtained with our system.

The lowest “round-trip” latency that we managed to achieve so far is 11µs (at a
sampling rate of 768kHz).

Multiple ADAU 1787 codecs can be used on one FPGA. Hence, implementing a
system with 32 audio inputs and 32 audio outputs whith such performances can be easily done on a basic Zybo Z7 board.[/list]

They (CCRMA) are making progress on making FPGAs easier to program using Faust, but we are a long long long way from them finding sufficient application to be included in general PCs such that they would provide the same benefits to your software workflow that they provide to custom hardware. In fact, if I were to put money on applications that would drive this kind of hardware first, it wouldn't be music, but money, i.e., crypto.

Awesome but you are speaking in generalities, I am taking about the specific FPGA that is used in the Super Six compared with modern CPUs used in desktop computers

If you want to talk about cutting edge scientific research at the university level awesome, but that is not what is in the Super 6 and that is not what is running the software native to that hardware

So to recap what is occuring in the software in the Super Six is not running on cutting edge research level prototype FPGA's and could easily be done on modern CPUs in desktop computers

Please stop talking about FGPAs in generalities. Again if budget were no concern you could easily make instruments that executed software that a single CPU could not, but you would find it cheaper to just run multiple CPUs to execute DSP for musical instruments

I am a giant fan of FPGA and any other digital computer technology that moves synthesis forward, currently in June almost July of 2025, the most cost effective way to do that is with CPUs, using FPGA is doable if you want to keep on buying FPGA hardware over and over for no sonic advantages

[wintoid]

Sure, but if it's not CLAP, you're dealing with global modulation, which is sometimes fine, sometimes not as good.

Good point. OK for a monosynth though I guess. I have that monosynth/modular mentality I guess. Polys are "special".

8-10 hardware synths is the correct number of synths you should have. I've calculated this using vigorous science. I only have one chair, though. I'm not sure if that's too few. I do like it, though. It's the identical chair to the "CEO chair," immortalized in Silicon Valley.

I've got a piano stool for playing, but often sit on an armchair with the laptop on my lap, right in the middle of the speakers

I sort of do all effects in software... would a UAD Apollo count as hardware? I do have one. I also have a couple of distortion pedals, a wah pedal and a Line6 HX Stomp XL I use for exotic effects pedal sounds and other synth stuff. I use it because I have a dual axis expression pedal that's analog and allows for cool performance stuff to happen. So... I guess I don't do all effects in software.

Well if you count those, you probably have to count the DI for my bass guitar. Nah, don't think it counts.

Oh, it's all silly. This I know. Again, though serious and rigorous scientific study.

It's the only way to prove these things, I agree.

[_leras]

Oh, but I didn't say The Legend HZ sounds as good as a Moog One. I just want the fight to happen

However, all the Moog One sounds in videos didn't show me anything that I would need beyond all the software synths I have now because I can get close enough with those. By the way, for me, I think Blamsoft's Viking VK-2 (not a VST) sounds way more analog than The Legend HZ (which I also have and like a lot as well).

Well don't worry I wasn't replying just for you. I was just curious and it's easier than ever to find great examples of everything.

Software has no doubt improved, and it is the sound of electronic music today in many ways. And does a bunch of stuff not that practical to recreate in hardware.

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet.

Does it matter? Maybe not for everything.

Hearing all these demos has made me want some more hardware synths though


edit: sorry that blamsoft viking sounds pretty digital 'analog like' to me...

[IvyBirds]

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet

Based on what? Define what "general presence_ means and how do we measure that. Define what "top ends" means on obviously you don't mean frequencies since software can easily produce frequencies that exceed human hearing so there would be nothing "clear" about those

As far as Modulations again what are you referring to? Because software easily exceeds the modulation options of hardware

[pekbro]

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet

Based on what? Define what "general presence_ means and how do we measure that. Define what "top ends" means on obviously you don't mean frequencies since software can easily produce frequencies that exceed human hearing so there would be nothing "clear" about those

As far as Modulations again what are you referring to? Because software easily exceeds the modulation options of hardware

You are hitting the cpu bottleneck again here, the things you say betray the fact that you really
don't know what your talking about.

[enCiphered]

Turns out hot air fills pages just fine. Proof that you can generate walls of text without a foundation.

[pekbro]

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet

Based on what? Define what "general presence_ means and how do we measure that. Define what "top ends" means on obviously you don't mean frequencies since software can easily produce frequencies that exceed human hearing so there would be nothing "clear" about those

As far as Modulations again what are you referring to? Because software easily exceeds the modulation options of hardware

Most of the analog bits you're talking about are not calculatory at all, they are a physical process.
They only need the correct voltage to be propagated through a system, e.g. no cpu impact whatsoever, not to mention free lightspeed transmission.

[ghettosynth]

If you want to talk about cutting edge scientific research at the university level awesome, but that is not what is in the Super 6 and that is not what is running the software native to that hardware

So to recap what is occuring in the software in the Super Six is not running on cutting edge research level prototype FPGA's and could easily be done on modern CPUs in desktop computers

Research level hardware? Would you like to define that?

The board in question is a standard dev board available to consumers. You can buy one from Amazon. Granted, it is faster than whatever UDO used six or seven years ago, but, that's beside the point. As I pointed out to you, George discussed the value of FPGAs versus modeling in an interview.

[pdxindy]

Software has no doubt improved, and it is the sound of electronic music today in many ways. And does a bunch of stuff not that practical to recreate in hardware.

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet.

Does it matter? Maybe not for everything.

CPU is the main limitation. U-he for example, could make better emulations but then the cpu use would be so high they could not be played in realtime. The developers purposefully make compromises to sound quality and emulation accuracy in order to keep the CPU use in the high, but still playable range.

When it comes to non-linearities and audio rate modulation it is obvious. There is nothing in software that sounds/reacts like the analog distortion and analog compressor in the Elektron Rytm. No matter how extreme the settings, or how I modulate it, there is never a hint of that unpleasant digital edge. I can devolve the sound to stuttering grungy mess and it always sounds great, with none of that digital unpleasantness. It's glorious!

Eurorack is an interesting case because even when some modules are digital, it is an analog signal that is being wired up. Eurorack has an incredible sound quality. It stands distinct from software.

And yeah, whether it matters is up to each person. There's no one right answer. I'm happy using both.

[IvyBirds]

But although it covers some hardware/analog sounds, it's clear that top ends, the general presence, the filters and modulations are not there yet

Based on what? Define what "general presence_ means and how do we measure that. Define what "top ends" means on obviously you don't mean frequencies since software can easily produce frequencies that exceed human hearing so there would be nothing "clear" about those

As far as Modulations again what are you referring to? Because software easily exceeds the modulation options of hardware

You are hitting the cpu bottleneck again here, the things you say betray the fact that you really
don't know what your talking about.

What CPU bottleneck? The fact you claim I do betrays the fact you have no idea

You claim I am hitting a CPU bottleneck and that's simply not true

[IvyBirds]

If you want to talk about cutting edge scientific research at the university level awesome, but that is not what is in the Super 6 and that is not what is running the software native to that hardware

So to recap what is occuring in the software in the Super Six is not running on cutting edge research level prototype FPGA's and could easily be done on modern CPUs in desktop computers

Research level hardware? Would you like to define that?

The board in question is a standard dev board available to consumers. You can buy one from Amazon. Granted, it is faster than whatever UDO used six or seven years ago, but, that's beside the point. As I pointed out to you, George discussed the value of FPGAs versus modeling in an interview.

Awesome so you say the board in question is not what is used in the Super 6, thanks for proving my point

As for taking in an interview so what, again I think software running in a FGPA is awesome and can do things analog synths can't do

But that is no being used in the UDO Synths, or any hardware synths at this point as such there is nothing in there that can't be done with a modern CPU which is my point

As far as research level why did you post a PDF of a PowerPoint given at one of the most advanced Sound Research Labs in the World affiliated with a major University. Founded by John Chowning the father of FM Synthesis if it's not research level

[zerocrossing]

Most of the analog bits you're talking about are not calculatory at all, they are a physical process.
They only need the correct voltage to be propagated through a system, e.g. no cpu impact whatsoever, not to mention free lightspeed transmission.

*Not quite lightspeed.