Do I need a new soundcard?
Hi Guys,
I need some advice! I use a few VSTi and as I add more instruments, the output of the system begins to crackle and pop, this becomes steadily worse the more load I put the system under. I can cure this by inceasing the latency in the ASIO control panel, however once the latency is greater than 5ms the lag between playing a note on my keyboard and hearing it makes recording impossible.
My question is this: Will upgrading my soundcard e.g. M-Audio 2496 improve this situation. Will I be able to keep my latency less than or equal to 5ms, or would I be better off just upgrading the PC itself?
Secondly if I do upgrade my soundcard, can I just add it to my system (and have two soundcards) or should I remove the Soundblaster first (and have only one)?
My PC Spec:-
AMD XP3000+
3Gb RAM
Win XP Pro
Soundblaster Audigy 2 ZS (I know it’s pants, hence this post ) 
P.S. A friend of mine with more money than sense has given me a full version of Cubase SX 2 (dongle and all), because he has upgraded to Cubase SX 3 on the mac. I have found that Cubase SX 2 suffers the same lag problems as nTrack.
Well, how many vsti are you running at once? The cpu is fast enough, seems to me. I don’t think the soundcard change will help a whole lot, if you are trying to run a whole bunch at once.
Are you recording midi and then plyaing that back while you record another midi track with a new vsti - that sort of thing? If that is it, then you can simply render the tracks you’ve already played, that will free up some resources. I probably misunderstand what you are doing, however… 
| Quote (Paulie @ Dec. 15 2004,02:49) |
|
I need some advice! I use a few VSTi and as I add more instruments, the output of the system begins to crackle and pop, this becomes steadily worse the more load I put the system under. I can cure this by inceasing the latency in the ASIO control panel, however once the latency is greater than 5ms the lag between playing a note on my keyboard and hearing it makes recording impossible. My question is this: Will upgrading my soundcard e.g. M-Audio 2496 improve this situation. Will I be able to keep my latency less than or equal to 5ms, or would I be better off just upgrading the PC itself? |
Whew! 5 ms is pretty good (but I'm not by any stretch a good keyboard player). to get substantially lower latency, you'd have to go to a stand-alone instrument, wouldn't you?
Some people use an outboard MIDI GM box or a soundcard with MIDI synth to give them no-latency sound when they play during recording, then switch to a VSTi for playback and mixdown.
Secondly if I do upgrade my soundcard, can I just add it to my system (and have two soundcards) or should I remove the Soundblaster first (and have only one)?
I highly recommend having both. My SoundBlaster is the primary Windoze soundcard. It's normally connected to the typical beige plastic PC speakers. My good card (Audiophile 2496) is connected to my mixer & the big monitors.
With this arrangement, any windoze warning sounds never blast out through the big speakers, and I have a bonus Soundfont player for when I need one (eg as a low-latency MIDI voice during recording)
usually it’s how well the asio drivers are written and not necessarily the card itself-of course each card has it’s own drivers, so they go hand in hand.
if you’re interested in vsti’s the Steinberg VSL2020 was designed almost specifically for that purpose. it’s an adat card so you’ll need a/d and d/a converters-like the behringer ada8000. the driver buffering can be set as low as 32 samples in some cases. it’s also got a program called v-stack which works as a vsti program that can also link mulitple computers together-so your main cpu is doing n-track, and the other is running the vsti’s and they’re all run into n-track.
kinda’ crazy. I have one, but I don’t use vsti’s very often, so I’m considering getting a different card. it’ll run you $200 for the one I have.
The oft quoted message: “Drivers, drivers, drivers!” may be in order here. Make sure that all of your drivers are current.
Also, defrag your HD before every session and use “EndItAll” to shut down all background programs. A RAM utility like “FreeRAM” will help get your RAM freed up too. I have 512 Meg of RAM and FreeRAM frees up about 280 Meg. It is freeware as is EndItAll. I use the limited version of “Diskeeper Lite” which is also free. It will defrag my 120GB HD in about 10 minutes. You would be surprised at how much fragmentation happens in a normal day.
HTH
Don
Before you upgrade your sound card, try the kX drivers in place of the Creative Labs drivers.
http://come.to/sblive
It would appear that my problems are due to my high expectations of the current technology. My investiagtions across the Net have given me pretty much the same information as you guys.
Whatever system you have it seems you will have to render your VSTi output to WAV before you finish your track. Playing the track with a different sound e.g one on my keyboard then assigning the midi track to a VSTi is an idea I’d thought of, and seems to be a good middle ground.
The track I was working with had piano, electric piano, bass, drums, organ and percussion. I have since rendered all of these tracks to WAV and my CPU usage is at 5%.
Thanks for replying
I’d do the KX drivers and then get a utility to check the PCI latency of all the devices plugged into/built on your motherboard. I use a payware called PowerStrip, but there’s freebies out there from what I hear. Some video cards are total hogs (especially the top dollar ones) and can hose up the audio. I lowered the PCI latency to 8 (which is exceptionally low) on all of my devices including video and can now run 64 sample latency on my soundcard (C-Port) no problem with VSTi’s and a good number of tracks and FX’s. Machine: ASUS p4p800, 1 gig ram, 3.2 P4, XP. Of course YMMV since your system & card are different, but it’s probably worth trying.
I bet the total latency is well over 5 ms. Many MIDI synth sounds have that much latency when played standalone. You can test this by setting a mike by the keyboard, whacking a key, recording the result, and looking at it in n-Track or a wave editor. You can see the two sounds – whacking the key, and hearing the note, pretty clearly. Many ROMpler keyboards have sounds where there’s a bit of waveform before the note really begins, and 5 msec isn’t an unusual amount. For example, on many piano sounds, there’s a bit of the noise of striking the key before the hammer strikes the piano strings.
[Note: to measure milliseconds using n-Track, right click on the timescale in the timeline view, pick “custom”, and enter “1000” for “fps”. Then the number to the right of the decimal point will be in milliseconds. Just note that 1.20 is not 1 second and 20 tenths, it’s 1 second and 20 milliseconds (1.020 sec).]
I bet your latency problem is more than just wave device latency. You can measure the total latency as I mentioned above, using different sound sources to see what variations you get.
The things that introduce latency, in order of occurrence in the processing chain:
1) MIDI keyboard’s response time – varies, but 1 ms isn’t unusual. This isn’t easy to measure without an oscilloscope.
2) MIDI message transmission – about 1 ms for a single “note on” message. 5 note chord? 5 msec for the last note, assuming you strike them all at the same instant.
3) MIDI message receipt & conveyance to synthesizer - varies. Subject to PC interrupts. Usually not a major problem, though. I’d say this is generally under 5 msec, but I’m guessing. Not easy to measure.
4) synthesizer response (emitting the waveform data) – this varies from near zero to a half-second (for MS Wavetable synth, which is obviously useless for real-time play).
5) audio buffering – in your case, 3 to 5 msec, which is excellent.
6) soundcard response (I suspect this is nearly instantaneous, unless it’s a USB or Firewire device. Even in the latter case, I bet it’s rather low – perhaps a msec?)
7) distance between speaker and your ears. Sound travels at about 1 foot per msec. If you’re that sensitive, you’d better use headphones. Typical values – 3-8 msec.
I’m ignoring the speed of light here, which is close enough to instantaneous to be ignored. Light goes about 200 miles per millisecond. I’m also ignoring the latencies introduced by analog circuitry. I haven’t measured it, but I expect it to be less significant than the distance to the speaker unless you’re wearing headphones.
Bottom line here: the 5 msec wave buffer latency by itself isn’t the problem, or you wouldn’t be able to tolerate playing any electric keyboard with the speaker more than 5 feet from your ears. It’s the total latency, and I suspect that if you’re having trouble playing it’s over 10 msec, and more likely over 15 or 20 unless you’re particularly finicky. See how far away you can put the speaker and still manage!
Until you find the biggest latency and address that, you’ll be fighting a battle of diminishing returns.
Sorry for the long post, but it’s not a trivial subject.