Have Modern VST Instruments Replaced Your Hardware Synths ?

[ScrLk]

Some things seem to come down purely to throwing CPU cycles at it. Since RePro is getting bandied about here a lot, for good reason, it sounds fantastic, you can hear for yourself that for the distortion effects, they sound great low on the scale, but as you go up, they get more and more shrill as they start aliasing. Obviously, Urs knows how to write a saturation algorithm that doesn’t alias (just listen to the saturation in Presswerk and Satin) but they had to make cuts somewhere, and there you have it. Other plugins, like Massive X have beautiful sounding distortion, but it’s not per voice and the rest of the synthesis engine isn’t trying to model the behavior of analog synths.

I'd like to see a spectrogram of Repro's aliasing

[chk071]

I wonder how much better the sound of plugins could already be if developers wouldn't have to think of the CPU load at all, as if every customer had a Cray at home.

My guess: Not much. There are already developers which don't comprise much (NI, u-he). I also would say that sound quality is also always up to the skills of the developers. I don't think any developer develops deliberately compromised sounding plugins.

Some things seem to come down purely to throwing CPU cycles at it. Since RePro is getting bandied about here a lot, for good reason, it sounds fantastic, you can hear for yourself that for the distortion effects, they sound great low on the scale, but as you go up, they get more and more shrill as they start aliasing. Obviously, Urs knows how to write a saturation algorithm that doesn’t alias (just listen to the saturation in Presswerk and Satin) but they had to make cuts somewhere, and there you have it. Other plugins, like Massive X have beautiful sounding distortion, but it’s not per voice and the rest of the synthesis engine isn’t trying to model the behavior of analog synths.

It's pretty close though, isn't it? I mean, if you take the Blue Monark filter in MX, for example, the filter sounds a bit better in Monark, might have more or different non-linearities, but, with MX, you get gazillions of voices (which you won't get in analog anyway), so, yeah, if you want, that's compromised due to the size and feature set of the synth.

Might be better to compare with actual emulations, like Monark, RePro, Oberhausen, whatever. At least I would talk about such plugins, when we ask the question how much developers compromise, and where plugins would be if developers wouldn't compromise. (Again, I don't think that's really the primary thinking in such soft synths. At some point, it also gets so complex that you have to ask yourself whether you want to implement every weird fart an analog devices does, to get a "believable" sound. Capturing every single thing going on in the physical world isn't possible anyway.)

[gentleclockdivider]

Presswerk and satin’s dist. algo isn’t per voice either
These are stereo processors , processing left/right.

[Shiek927]

Yeah: would def hold onto your GT! I have an XT that I won’t get rid of either. Hell no.

That’s one synth where there have been many requests to create a vsti. Probably won’t happen anytime soon, tho.

Haha, you're telling me!!

Roland Cloud will 'eventually' get to it..sometime this century haha. I used to have one, but let it go because honestly hardware just isn't for me -- too much money, hassle, the spaghetti of cabling drove me insane, and the workflow of plugins is just too fast with too many advantages. I fall into the "music producer" category more than a "synth enthusiast" which was talked about earlier in this thread. Still, I really appreciated it and am looking to seeing it in my computer..someday

[e-crooner]

I think we could close the gap, if not totally eliminate it. There was this one back-and-forth between Urs Heckman and I, where I posited that Diva’s filter FM sounded rather muted and dull, in comparison to RePro’s. He got quite angry at me and yelled and stamped around the forum a bit, but then after he calmed down, he did concede that he cut some corners so that Diva could hit a 16 voice count and still be kind to the average CPU of the time it was released. I never meant it as an insult, just an observation. There’s generally always a price for audio quality, though one could surmise that audio character is free. I’m usually willing to sacrifice some ideal sonic quality for things that are interesting. I bought a physical modeled cello this Christmas and it’s f’n ridiculous sounding... but also maxes out my i7. I work in the video game world and it’s always a struggle to make something look like movie quality effects, but the truth is, those movies are made with render farms and often a single frame of a movie can take hours or even days to render.

The scene in which Elsa walks out onto the balcony of her newly constructed ice palace is 218 frames long, and includes the film's longest frame to render. The single frame took more than 132 hours to render (that's more than five days).

So, it seems like CPU power isn’t quite making the leaps it once was... so it’s hard to say when a new consumer grade computer can really make a huge difference. I am getting the feeling that my machine needs an upgrade. Maybe in April for my birthday, I’ll assess. Like I said earlier, I was able to match the quality of my Modal 002 using DS’s Diversion, but I could only get a voice or two out of it because I had the oversampling cranked up. Now... could I replace the character of the 002 in software? Hard to say. It does this thing they call “pole morphing” where you can sweep the filter between 24dB per octave, through band pass to one pole 6dB per octave. There’s also a very good sounding per voice transistor overdrive and a “DeRez” down sampling effect that has a very unique sound compared to other bit crushers.

My “plan” is to keep things as they are, because it’s not hurting anyone in any way, but if things change drastically, like we move to a home with less space, I have no doubt that I’d considerably lighten my hardware synth load... maybe altogether. Having hardware is a luxury, no doubt. I’ll indulge myself until I no longer can.

And can't U-he increase the oversampling now that computers are much more powerful than 8 years ago? At least the parts where oversampling makes a big difference. In Sylenth1 different parts of the synth use different oversampling rates if I remember correctly.

You mean this?

Looks like any old animation to me

When you morph the filter from 4 poles to 1 pole, isn't that basically a cutoff-modulation? As you morph continuously, you can hardly say whether a certain state of brightness at any moment is due to a certain number of poles or a certain cutoff setting.
What seems more interesting than morphing between poles is morphing between LP and HP for instance.

Well, I am not a professional, for my occasional playing, a single soft synth is enough

[chk071]

Did you try RePro? It's CPU wrecking enough for me.

[audiouser720]

Why can't software synth developers do what UAD does. Make a dedicated hardware processor box. I know most of you think this would be just like a hardware VA but no. Imagine a box like UA Apollo Twin that size... Say if U-he had an external processor like this and you would run RePro on that box rather than on your CPU (maybe you could extend it). Would that not close a gap or at least help with? It would be more expensive of course but still judging a real Pro 5 rev 4 for £3,300 or a 'dedicated U-he box' for £600-700 is better (if the sound delivers with the help of the external processing e.g. more oversampling not cutting corners in component modelling etc.). Sorry if I'm chatting sht I'm no software developer and nor have substantial knowledge in computer science, this is just an idea.

[e-crooner]

Did you try RePro? It's CPU wrecking enough for me.

I have the demo version for comparison purposes. With the multi-core support on, I can play one instance without limitations. Without MC support, it's horrible

[FapFilter]

Why can't software synth developers do what UAD does. Make a dedicated hardware processor box. I know most of you think this would be just like a hardware VA but no. Imagine a box like UA Apollo Twin that size... Say if U-he had an external processor like this and you would run RePro on that box rather than on your CPU (maybe you could extend it). Would that not close a gap or at least help with? It would be more expensive of course but still judging a real Pro 5 rev 4 for £3,300 or a 'dedicated U-he box' for £600-700 is better (if the sound delivers with the help of the external processing e.g. more oversampling not cutting corners in component modelling etc.). Sorry if I'm chatting sht I'm no software developer and nor have substantial knowledge in computer science, this is just an idea.

This is no different than having more than enough separate cores inside your CPU.
It's actually worse than the latter as you have to fight with huge bottlenecks as external connections can never be nearly as fast as CPU internal solutions

[e-crooner]

How about using graphics cards for the unintelligent number crunching? I heard they are perfect for that and used in illegal activities such as hacking

[audiouser720]

Why can't software synth developers do what UAD does. Make a dedicated hardware processor box. I know most of you think this would be just like a hardware VA but no. Imagine a box like UA Apollo Twin that size... Say if U-he had an external processor like this and you would run RePro on that box rather than on your CPU (maybe you could extend it). Would that not close a gap or at least help with? It would be more expensive of course but still judging a real Pro 5 rev 4 for £3,300 or a 'dedicated U-he box' for £600-700 is better (if the sound delivers with the help of the external processing e.g. more oversampling not cutting corners in component modelling etc.). Sorry if I'm chatting sht I'm no software developer and nor have substantial knowledge in computer science, this is just an idea.

This is no different than having more than enough separate cores inside your CPU.
It's actually worse than the latter as you have to fight with huge bottlenecks as external connections can never be nearly as fast as CPU internal solutions

Fair enough.

[BONES]

What's the problem with RePro? Here are two instances of RePro 5, each playing a 3 note chord on my laptop. The interesting thing to note is that Studio One reports 29% CPU usage, whereas Task Manager says it's really only 6.6%. That's just 1.1% of my available CPU power per voice. It's nothing so I don't quite see why everyone is making such a fuss.

For some reason the image wasn't displaying in the post, so click on the link to see it.
http://novakill.com/stuff/RePro_CPU.jpg

[e-crooner]

CPU meters are tricky. Somehow they often don't reflect the real load. So, it may show like 15%, yet the sound is already suffering.

CPU meters tend to show the global CPU load, not the load in the core hosting the plugin, which can be much higher then the global CPU load and hence lead to sound outages.

[Cinebient]

Running Repro-5 at buffer 32 without multicore, under HQ mode, with a preset like HS CS-Something is no problem here on a little laptop. So I think its not that bad.
There are other synths which are much more cpu wrecking synths.

[zerocrossing]

I wonder how much better the sound of plugins could already be if developers wouldn't have to think of the CPU load at all, as if every customer had a Cray at home.

My guess: Not much. There are already developers which don't comprise much (NI, u-he). I also would say that sound quality is also always up to the skills of the developers. I don't think any developer develops deliberately compromised sounding plugins.

Some things seem to come down purely to throwing CPU cycles at it. Since RePro is getting bandied about here a lot, for good reason, it sounds fantastic, you can hear for yourself that for the distortion effects, they sound great low on the scale, but as you go up, they get more and more shrill as they start aliasing. Obviously, Urs knows how to write a saturation algorithm that doesn’t alias (just listen to the saturation in Presswerk and Satin) but they had to make cuts somewhere, and there you have it. Other plugins, like Massive X have beautiful sounding distortion, but it’s not per voice and the rest of the synthesis engine isn’t trying to model the behavior of analog synths.

Are you saying that U=he Satin is a software saturation that doesn't alias at all? Can I ask you in what sample rate have you tried it and how far have you pushed it?

To be honest, I’ve not pushed it that far as of yet, but I didn’t notice anything when I put it into feedback self oscillation that stuck out at me. I don’t really run things though spectrum analyzers unless something obvious seems to be grating or harsh. When I started playing with RePro, a quick run up the keyboard and it was pretty obvious that the distortion was creating clear aliasing. It sound great on low notes, though.