Problem with WDM drivers for Delta

[bdemenil]

I've encountered a problem with the newest WDM drivers for the MAudio delta series soundcard. It seems that when using these drivers, newly recorded audio tracks manifest a delay relative to previously recorded tracks. I suppose the problem to be driver related because it occurs under both Ntrack and Sonar. Below is a link to my test results.

http://www.iasorecords.com/public_html/images/album/maudia_dma.html

Also, this discussion was initialy introduces by Slackmaster2K under the ntrack forum :
http://www.homerecording.com/bbs/showthread.php?threadid=58387

 

[c9-2001]

I've encountered a problem with the newest WDM drivers for the MAudio delta series soundcard. It seems that when using these drivers, newly recorded audio tracks manifest a delay relative to previously recorded tracks. I suppose the problem to be driver related because it occurs under both Ntrack and Sonar. Below is a link to my test results.

http://www.iasorecords.com/public_html/images/album/maudia_dma.html

Also, this discussion was initialy introduces by Slackmaster2K under the ntrack forum :
http://www.homerecording.com/bbs/showthread.php?threadid=58387

i don't have a problem with the drivers, but then agian i don't run sonar or n0track, i use SX and nuendo...
i use a 1900+ w/delta 1010, under xppro...

it could be your motherboard.. does your MB have a via chipset...

 

[bdemenil]

Cubase and Nuendo use ASIO drivers. As stated, the problem is only with WDM drivers.

 

[Slackmaster2K]

Also, it should be stressed that this problem is not an *obvious* problem. The WDM drivers do technically work. The problem is specifically related to recorded track offset (or lag) being anywhere from about 3 to 50+ms depending on buffer settings, whereas ASIO yeilds a consistant 1-2ms regardless of buffer settings. The ASIO results are correct and demonstrate how the card should technically be working in WDM.

A basic test procedure can be found in my original thread in the n-Track forum, and read bdemenil's writeup because it really demonstrates the problem well.

Oh, and he's running XP while I'm running 2K, so the problem could be NT related. bdemenil saw the problem both on Sonar and n-Track, whereas I'm only using n-Track.

I'd advise anyone with a Delta using WDM to try these tests, and please post your results.

Slackmaster 2000

 

[bdemenil]

Has anyone sent all this stuff to M-audio and asked about it yet? Is it correctable by dragging it over and lining it up? Is it caused only by the Delta drivers, or even just the current version? Has anyone tried this test with other WDM soundcards besides the Delta?
Post this stuff in a common area, like computers, that way everyone will see it. I dont use N-Track, therefore never peruse this forum.
I will send all these links and some questions to M-audio if no one has yet. I just now got mine, too, and this is quite disturbing since it never came up until I buy one. And of course, I use Sonar 2.0 with WDM.
A question though... this "lag" is happening when you feed it back into itself. If we are recording everything into the computer from a sound source, every track should have the same lag on it, and therefore be just fine. Right?
I'll try to look here again, but try to make a post in the computers forum where more people can see it and comment.

- Posted by Tubedude to the original thread in the NTrack forum

The first tracks you record will not be affected by this problem since, as you noted, they will all be equally delayed. But as soon as you try to add to existing tracks, the problem will manifest - the new track(s) will be delayed relative to the originals.

While I have tried sliding new tracks over manually in order to compensate, I don't like this solution - besides being a pain in the ass, I'm probably not able to place them 'exactly' where they should be.

Tubedude, please test your new system when you have a chance. I'd like to see if there are certain configurations under which the problem does not occur. I've only tested it on computers using AMD chips, I'd like to see if it happens with an Intel chip.

Also, I think the problem only occurs on machines running Win2K or WinXP. I don't think there are WDM drivers for win98/Me.

 

[Slackmaster2K]

Tube, the lag will have an impact on you no matter what. The tests that we did only quantify the problem. The extent to which the problem will impact you depends on how you set your system up.

Because the Delta only has one buffer setting for all driver modes, I typically leave it fairly high for my ASIO applications (~50-60ms or so). What prompted me to originally do this test was that I was recording some new tracks (ntrack, WDM) and no matter how hard I tried I couldn't seem to be on time....not having a great sense of time to begin with I figured it was just a personal problem until I basically just got fed up. I had done this exact same experiment with ASIO a long time ago and I had really assumed that WDM would be the same. Much to my suprise, there was a 50-60ms offset between my playback and recorded tracks! This is definately a noticable delay!

If you use WDM while recording, I suggest setting the Delta buffers to 64, and then increase the buffers in your software to taste. This, if what bdemenil and I are seeing is true for everyone, should only yeild a 3ms offset which is probably acceptable.

ASIO, on the other hand, is consistant at ~1-2ms, regardless of any buffer settings.

This should have absolutely no impact on mixing, which is where latency comes into play. Since WDM performs better for me than ASIO in n-Track, I'm going to record tracks using ASIO and mix using WDM....at least for the time being.

Anyhow, give the tests a try on your system and let us know how it turns out. Put some screen shots up if possible. I'm going to throw something up as well (after company leaves tomorrow). The more people we have with this problem who can DEMONSTRATE it, the more likely it'll be that somebody will pay serious attention. I'm betting they'll just write it off as being a WDM issue though...a feature.

If it's possible, I'd at the very least like to see seperate buffers for both ASIO and WDM. I'm not sure where the delta buffers are physically located...are they just in the driver or is there physical memory on the card itself?

Slackmaster 2000

 

[bdemenil]

I just talked to Midiman tech support. The guy over there thought the problem might have something to do with Microsoft's implementation of ACPI in Win2K and WinXP. He says he's looking into it, but his initial suggestion was to disable ACPI. Unfortunately, this makes it impossible to run dual processors. It may also cause other problems with the OS. Here's a link to some info on ACPI :

http://service.steinberg.net/knowledge.nsf/show/acpi_kills_audio_performance

I think ACPI may have something to do with the less than desirable performance of ASIO under Win2K/XP (Stuttering, occasional dropouts and crackles, ect). I'm not sure though if it explains this problem.

Also, surprisingly, the Midiman techy thought it was perfectly normal that as you increase buffer size, your new tracks become more and more delayed compared to the originals. Makes me question if he really knows what he's talking about.

 

[Slackmaster2K]

That figures. Next time you'll have to give the tech some information about the product he is supporting, like the fact that it supports zero latency monitoring. Sheesh.

I have ACPI disabled, btw, so that's not it.

We have to get past the tier 1 tech support nerds. If anybody out there has a Delta card, ANY delta card, try this out and let us know how it works on your system!

Slackmaster 2000

 

[Slackmaster2K]

fd

 

[Teacher]

i have my buffer setting set to 512 for my delta 66 and i don't 'notice' a difference within sonar with delay of old recorded tracks

BD did u try running wave profile again so sonar can detect the new drivers and settings?

 

[bdemenil]

Yes I ran the wave profiler. Teacher, what O.S. and chipset are you using? Also, try the test described - a 10ms delay is not always immediately noticable. It takes about 15ms for the ear to recognise two distinct sounds.

The tech got back to me - said my motherboard has been tested to work with the delta - both in ntrack and sonar - so he ruled out the ACPI issue. He's trying to blaim the audio software - says I should contact Sonar and Ntrack support. I want to gather a little more testimony before I go back to them.

 

[bdemenil]

Yes I ran the wave profiler. Teacher, what O.S. and chipset are you using? Also, try the test described - a 10ms delay is not always immediately noticable. It takes about 15ms for the ear to recognise two distinct sounds.

 

[Slackmaster2K]

Yeah, this isn't an immediately noticeable problem...it's pretty subtle, but it definately will have an impact on your mixes if every take is 10ms late.

I knew midiman tech support would react this way. They didn't even try to verify it. In fact, the guy probably didn't even talk to anybody.

I'd be very suprised to find that both sonar and n-Track suffer from the same problem.

If you've got a Delta series card, try the test before responding that everything works properly. (e.g. route an output to an input with a patch cable and record a slow click track as it plays back. Then use a wave editor to zoom way in on both the original and the newly recorded track, and see how far back the new track is from the original. In a normal situation it should be < 2ms, regardless of buffer settings. If you have this WDM problem, it will vary from 3ms to >50ms depending on your Delta control panel buffer settings)

Slackmaster 2000

 

[Teacher]

i'm not quite sure what u mean...but i recorded the metronome off my drum machine twice look at both in wavelab and the look the same i'm not too sure what i'm suppose to be looking for though


BD:
I run an MSI amd 762 chipset running dual MP 1900+

 

[Lame Thrower]

I just wanted to confirm that this problem is not limited to NT-based OSes. I've duplicated the results found by S2K and bdemenil in n-Track under Windows 98SE.

Under "Very High Buffering", the lag was about 2ms.

Under "Very Low Buffering", it was about 15ms.

Hope this helps in some way.

 

[bdemenil]

Teacher, this is the procedure :

Take the wav file from the metronome you recorded. Import that into Sonar (If you don't have it there in the first place). Physically patch the output #1 of your delta into input #1. Now when you play back the recorded metronome, it should feed right back into your soundcard. So, insert a new audio track, and set it to record off of delta input #1. Hit record - the original metronome track should play out of your soundcard's out and feed right back into its in - thus being recorded to the new track. The waveform of the new track should look pretty close to the same as that of the original. By comparing the offset of the new one from the original, you can calculate how much delay is introduced by the system. Ideally, this would be 0. Typically it will be 1 or 2 ms - maybe contributed to in part by the time it takes the signal to pass through the hardware. The amount of this delay should also remain constant regardless of buffer size.

What Slack and I have noticed are much larger delays - up to maybe 60ms in the extreme - which vary according to the buffer size chosen in the delta control panel. This implies that somewhere along the line, software is not compensating for buffer size and syncing new tracks as it should. The fact that 2 different audio applications exhibit the same problem with this particular driver leads us to believe the driver might be at fault.

 

[bdemenil]

I just wanted to confirm that this problem is not limited to NT-based OSes. I've duplicated the results found by S2K and bdemenil in n-Track under Windows 98SE.

LT, this is actualy the inverse of what we have been experiencing. Under low buffering we have measured less lag. Under high - more. What drivers are you using - MME?

Also, let me just throw out there again that the problem does not occur when using ASIO drivers.

 

[Teacher]

ok i did it....i see the delay but it isn't audible...it does produce a lil pop when looping though

 

[bdemenil]

Teach, try it with different buffer setting in the delta control panel - see if it changes the amount of lag. You may need to quit and restart the program for control panel changes to go into effect. Anyway, can you quantify how much delay you are seeing?

Under default buffer settings (384 samples)- you may get something like a 10ms delay. You won't hear two distinct sounds when the new and original tracks are played together, but you may notice some phasing - and believe me, for any kind of rhythmic music (which means most music) 10ms does make a difference – for instance, it can affect weather a particular instrument sounds 'on top of' the beat or 'behind' the beat.

 

[Lame Thrower]

LT, this is actualy the inverse of what we have been experiencing. Under low buffering we have measured less lag. Under high - more. What drivers are you using - MME?

D'oh! Yep, I'm using MME alright. And here I thought I was being helpful.

There is no WDM driver for the 1010 under Win95/98 is there?