PDA

View Full Version : in layman's terms, what is dsp?


prophetik music
02-16-2010, 01:00 AM
and why should i pay a lot for it in a soundcard or interface?

Blue Magic
02-16-2010, 01:58 AM
Digitial signal processing?

From what I can remember from school, DSP takes a real-world signal, like a sound input, and mathematically alters it to improve the sound. For instance, when you talk on a phone DSP takes in your voice, and removes the background noise allowing the person you are speaking to to hear crystal-clear responses from you.

If you plan on doing any recording, music making, or game playing on your pc with a surround sound system on it, then its is VERY important. You want the best sound quality you can get.

zircon
02-16-2010, 02:56 AM
Hmm, not quite what he's looking for Blue Magic, though it's a reasonably accurate general definition.

When people refer to DSP cards or DSP hardware, in terms of music production, they mean gear that can process, effect and/or generate audio without requiring any resources from your DAW. Stuff that runs on your DAW, be it a PC or a Mac, is called "native". Hence, "native plugins" vs. Pro Tools plugins (RTAS/TDM), which run off the physical Pro Tools hardware. The PT hardware is thus DSP.

The obvious reason why DSP is nice is because it saves RAM and processing cycles, but this is much less of an issue now than it used to be, since we now have abundant quad and octo cores. Still, it's a nice benefit to be able to run say 15 EQs without using ANY CPU whatsoever.

The next reason is actually far MORE important to some people. The quality of the plugins and audio processors in DSP hardware is, in the opinion of many, superior to native plugins. Manufacturers of PCI DSP cards, such as Universal Audio and SonicCore (formerly Creamware), typically have amazing emulations right out of the box, with the ability to get some extremely high end stuff that simply is not available anywhere else. For example, the SonicCore platform is the only way you can run a plugin version of the famed Virus hardware synth.

Some soundcards and audio interfaces claim to have DSP capabilities. This might be misleading. For example I had an EMU card (0404) that could run EQs, verbs and compressors, but they were fairly low quality, and could only be used in the EMU interface - they weren't VSTs or anything like that. The advantage of UAD cards, on the other hand, is that they actually integrate directly with your host and are designed to be used for music production, not just messing around on Skype or whatever.

Should you pay a lot? Only bother with DSP hardware (UAD, SonicCore, Muse Receptor, etc.) if you either (a) REALLY need offloaded processing and/or (b) want the sound of plugins only available for hardware DSP platforms. Keep in mind integrating any hardware into your setup will add latency and force you to manually record (no offline rendering) though there may have been advances that have fixed these issues lately. Some people really do swear by DSP stuff, particularly bLiNd, so you might talk with him to get his take on it.

SnappleMan
02-16-2010, 03:01 AM
+1 for UAD.

Meteo Xavier
02-16-2010, 03:22 AM
Should you pay a lot? Only bother with DSP hardware (UAD, SonicCore, Muse Receptor, etc.) if you either (a) REALLY need offloaded processing and/or (b) want the sound of plugins only available for hardware DSP platforms. Keep in mind integrating any hardware into your setup will add latency and force you to manually record (no offline rendering) though there may have been advances that have fixed these issues lately. Some people really do swear by DSP stuff, particularly bLiNd, so you might talk with him to get his take on it.

Do you think it would be worth it for someone like me to get something like that? You have a fairly intimate idea of what kind of tracks I work on now, and now I got a bunch of EastWest stuff to use.

Harmony
02-16-2010, 04:02 AM
Another potential benefit to dsp: Lately I've been recording a vocalist who must have reverb and light compression in her headphones. That's easy to do within the sequencer when I'm tracking with a single guitar track, but when I bring her in on a project that's already got dozens of tracks, a bajillion FX, and she wants 7 part harmony...say goodbye to 2.7ms latency, or anything that a singer would be able to handle.

When I bought my M-Audio Profire 2626 a little while ago, I knew this would be one of the bigger downsides to it over my other top choice (Lexicon FW810s) which had at least the dsp dynamic fx (not reverb.) While I think I made the right choice in the end for other reasons, the lack of dsp has really bothered me lately as I've had to go out and buy a an external multi-fx unit to get the fx in the monitoring chain with zero latency.

If live tracking with FX, even low quality ones, sounds useful, might want to consider a card with dsp capabilities.

prophetik music
02-16-2010, 11:38 AM
Should you pay a lot? Only bother with DSP hardware (UAD, SonicCore, Muse Receptor, etc.) if you either (a) REALLY need offloaded processing and/or (b) want the sound of plugins only available for hardware DSP platforms. Keep in mind integrating any hardware into your setup will add latency and force you to manually record (no offline rendering) though there may have been advances that have fixed these issues lately. Some people really do swear by DSP stuff, particularly bLiNd, so you might talk with him to get his take on it.

in other words, no, i don't need it right now.

awesome, thanks for the good explanation, man.

Arcana
02-17-2010, 01:16 AM
DSP hardware is to real-time audio as GPU hardware is to real-time graphics, pretty much.

Just as a graphics processing unit does super-fast real time processing of graphics (especially when it comes to textures and lighting), your DSP will allow you to do super-fast real time processing of audio (especially effects).


Aside to Zircon:

I have no experience with DSP personally, but based on what I know about computer architecture, I would intuitively believe that adding DSP hardware would reduce your latency, not add to it. This is because instead of processing real-time audio on your general processing unit (which is busy doing stuff like running FLStudio, plugins, your OS, etc.) you can do the real-time audio processing on the DSP instead. Thus, you bypass the processor.

Most latency delay is because the bits associated have to go through a processor, not because they travel throughout your computer.

However if your drivers aren't good or your operating system has no idea how to route the audio correctly to the correct device then it may very well add latency.

zircon
02-17-2010, 01:35 AM
Not exactly. Your latency has nothing to do with how many plugins you're running or what audio processing you're using. It's a value you set your ASIO/WDM/whatever drivers to (alternatively, done within your host.) While it's true that having a huge CPU load at low latencies will produce more audible artifacts (pops, clicks, distortion, slowdown), having a huge CPU load won't affect your latency whatsoever.

Outboard hardware always adds latency because the audio from your computer has to go out and be processed through the outboard and then come back in, all synced to a clock. I believe UAD has a "no latency" mode for processing but it is much more intensive on UAD's own processors and possibly on your CPU's processors too.

dannthr
02-17-2010, 03:14 AM
Not exactly. Your latency has nothing to do with how many plugins you're running or what audio processing you're using. It's a value you set your ASIO/WDM/whatever drivers to (alternatively, done within your host.) While it's true that having a huge CPU load at low latencies will produce more audible artifacts (pops, clicks, distortion, slowdown), having a huge CPU load won't affect your latency whatsoever.

Outboard hardware always adds latency because the audio from your computer has to go out and be processed through the outboard and then come back in, all synced to a clock. I believe UAD has a "no latency" mode for processing but it is much more intensive on UAD's own processors and possibly on your CPU's processors too.

There doesn't have to be latency, as much, say, as there is between typing text and seeing it appear on your monitor.

A lot depends on the chips managing your PCI load (or PCI-e) or the load on your USB or Firewire Hubs if that's the equipment you're using.

In layman's terms DSP is good.

It's like adding a GPU for your graphics processing as opposed to relying upon your CPU to process graphics. GPUs are built for matrix math whereas CPUs are not, they're geared toward more general processing. Matrix math is highly relevant when we're talking about 3D computer generated graphics.

Similarly, DSP is like adding a processor to process highly specific audio instructions.

But it's only going to do you so good as long as you're actually using its functionality.

SnappleMan
02-17-2010, 09:59 PM
I believe UAD has a "no latency" mode for processing but it is much more intensive on UAD's own processors and possibly on your CPU's processors too.

UAD doesn't really have a 0 latency mode, It has a "low latency" mode (probably between 1-3ms) but all their plugins have at least a little bit of latency and the low latency mode is only supported in certain hosts (RTAS versions only and most likely limited to PTHD). I've never been able to use them for monitoring because of that.

For true, guaranteed 0 latency monitoring, I think you still have to use rackmount effects.