Hardware Reverb vs Plug-in Reverb

Don't sleep on the SoundToys Little Plate.

I love my pcm70 on guitar and have used it on a snare in a heavy rock song. I would drive the input into the red and there is beauty and brilliance in that unit when you drive that input just right!!! Bricasti M7...Exponential Audio verbs and H-Verb. I prefer hardware usually, but I do use some plugs and sometimes blend to taste.

Not really wrong. But for me the hardware digital units still have their own A/D and D/A chips and they usually...well suck. Which imparts a sound in its own way which often is a important to the sound.

Even so, when i use my M7s in digital mode so their is no conversion happening...they still sound better than any plugin I have. But thats more of a not all code is the same comparison. 

It's a case of the hardware reverb having dedicated DSP hardware and algorithms that are efficiently coded for that hardware, but to run those algorithms on a general purpose CPU would be very time and resource intensive and wouldn't be practical as a plug-in that needs to run in near real-time.

I'm afraid that's simply not accurate.  I worked on a lot of those dedicated DSP chips and while they were the best that could be made at the time, they were crap. You had 16-18 bit accumulators, you had--even in the mighty 480L--an 8-bit multiply coefficient.  All of that vaunted depth is really quantization error.  Just pop a 30 hz sine wave into one and listen to the grinding. Even the best of the bunch could only conjure up 256 steps per sample and that was halved every time the sample rate was doubled.

The only advantage for chips of 30 years ago was address generators that could sort of get around the problems of DRAM.  As soon as you got to SDRAM and DDRAM, that problem went away. A properly-coded plugin can beat an HSP or Lexichip by a couple of orders of magnitude.

The only difference in then and now is the algorithms.  Algorithms for the old DSPs were written around those limitations and the compromises are audible.  They were audible (and undesirable) to David Griesenger and they had the same problem for Barry Blesser.  I could have done a spot-on emulation of any of the old Lex verbs, but I never wanted to.  I had sounds in mind that those older boxes could never get to.  And somebody's going to say the same about my stuff in 20 years.

It's perfectly acceptable if you prefer the sound of an older device. Whatever floats your boat.  You don't even have to conjure a rationale for it.  I give a huge amount of credit to Barry and Dave for their serious brainpower.  But if they'd had better processors, there's no way those verbs would have sounded the way they did.

I had a real EMT 140’s left channel crap out mid mixing, and we didn’t catch it until after we got home. Was glad to have Altiverb to put verb back into the left side.

That one also buzzed and needed recapped.

But repeatability is where plugins are exceptional. I love hardware gear while recording, but it can be a pain when you have mix revisions.

True its always fun to read about someone elses work.

I also love videos about it.

Special if its explained in easy words.

Exp-Audio verbs are too long on my shortlist.

I like them on the other side I am overhlemed with too many options.

The crux of the problem here.  I don't use my real plates as much as I used to because of client demand for 300 tweaky mix revisions.  I've tried printing the reverb to separate tracks, but the asks are somehow always something that requires a reprint of the reverb return.  

Yes. Unfortunately/fortunately the real plate sounds obviously better. I can turn a lead vocal off and only use the reverb return if I want, and it still sounds real, non-artificial, non-cheesy. I haven’t encountered a digital version that doesn’t fall apart to my ear past a certain volume level.

Then I'll be more specific.  The latency of all of the Exponential Audio plugins is 32 samples at 44/48.  64 samples at 88/96.  And so on.  In all cases, round-trip latency is under a millisecond.  When I consider conversion time and buffering in and out of a hardware box, it's pretty close to equivalent (hardware might be a little slower).  And as far as efficiency goes, there are dub-stages all over the world that set up templates with 2-3 dozen surround/3D versions of the plugins. These are typically running on cheese grater Macs that are also running Pro Tools with hundreds of tracks.  I had one tester with 256 copies of PhoenixVerb on a trashcan Mac.  And latency was still under a millisecond for all of them. It's absolutely true that some plugins (those based on convolution) would be much slower.  But a hardware convolver has the same issues.

Could I have this efficiency running the same old algorithms with the DSP chopped down to emulate a 25-year old device?  Sure, but why bother.

I worked with a lot of high-speed DSP processors myself back in the day, going back to bitslice machines and array processors.  There may be a few very specific algorithms that still benefit from standalone DSP, but that number is shrinking.

I appreciate your experience with high-speed DSP, but I know those old algorithms inside and out.  They're elegant in their own way, but they're not particularly demanding of resources. If you know how to code efficiently and if you know how modern desktop processors work, there's no problem at all running dozens of them in a modern machine.



 

I have a Scamp Parametric module in my rack that is seriously underutilized - sounds like a perfect application for it if I ever get my hands on an EMT.