KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/21 15:35:25
(permalink)
ORIGINAL: jlgrimes
What about you guys? Does this problem (timing problems between different recorded/bounced audio tracks) only occur when softsynths are added to the equation, or if you record each track at different latencies?
I do mixed softsynth/audio projects, and haven't noticed any problems, but then again, I don't see the delay on audio-only loopback recording with my RME cards/drivers. I'll try some more exacting experiments when I have the time...
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/21 15:42:36
(permalink)
KevinK,
just set a high latency and send a 3-4 bar drumbeat out and back into sonar. I don't think that it would take to much time when you are in front of your audiocomputer.
|
KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/21 18:49:57
(permalink)
It may seem like I have nothing better to do all day than to post
here, but other matters *do* intrude!
Anyway, I just ran a couple more experiments, and I am seeing
some odd stuff, though nothing musically nasty. If I just do audio
"pinging" via analog loopback, I can see that the copy is actually being
recorded some 60 samples *before* the original, regardless of the
buffer latency (from 1.5 ms to 46ms). As far as softsynth bounce
timing is concerned, depending on the soft synth, I would sometimes
see the bounced audio ahead of the recorded audio, sometimes after,
depending on the source, and the skew would vary from -30 to +150
samples.
The measurements were made looking at the second upward zero
crossing of the recorded signals, which were low frequency sine waves
from the cheezy Dreamstation DXi. I started out looking at percussion
samples from DR-008 (which is where I saw the negative skew on
"bounce" rendering), but those waveforms are complex and rather hard
to track reliably. With the sine waves, the initial attack and decay are
treacherous to track, due to the impedence of the wires on the attack
and a residual "ringing" on the decay, so I measured the distance
between the second upward zero crossing of a note.
|
billkath
Max Output Level: -69 dBFS
- Total Posts : 1076
- Joined: 2003/11/27 08:16:29
- Location: Ireland
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/21 18:59:50
(permalink)
I've put a little hep up on my site. It's called testlatency.zip http://www.heartbeatstudios.net/testlatency.zip This comprises a Cpw panel and a sample of a drum hit- total size less that 18Kb. (not too big.) The panel is at 24/44.1 track one contains a drum hit. Make sure input for this track is set to none. Set output to 1+2 analog out on whatever card you have. Track 2 is record readied. Set input to 1+2 analog in (or just 1 left analog in if you want). Set output to none. Simply loop a balanced patch lead (supposing you use balanced in's and out's) between the first out and first in. Hit arm, then hit record for a second. You will now be able toeasily measure exactly your latency. Simply mouse over the beginning of the wave on track one-the first up sample is easy to see. Write it down (mine was sample 5.) Then go to the same place on track 2 and mouse over (mine was 140) 140-5 = 135 135 samples latency. Hope that's helpful.
Billy E HeartBeat Studios
|
Mr. Ease
Max Output Level: -71 dBFS
- Total Posts : 960
- Joined: 2003/11/24 18:44:01
- Location: West Sussex, UK
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/21 20:41:17
(permalink)
In reply to Billy's question regarding delays on digital outputs that he did'nt expect. There are effectively three elements to the latency delays every soundcard has. The first is the latency of the software drivers - I guess we have all seen variations in the performance of different drivers as well as different DAW software. The second element is the time it takes to send a single sample. Each sample is sent in serial form (usually nowadays 24 bits) there are also some timing bits and most usually two channels are sent at once (stereo!). When the soundcard "knows" the data, it starts to send it - the receiver does'nt know what that data is until it is complete. OK this is done fairly quickly but it still takes time... The third element is when you add the D-A-D chain as again this takes a little more time. In essence we cannot get to discriminate between the first two and we see this as a fixed delay say 70 samples or so - this is mainly the software dirvers and interfaces but does include the serial delay. This goes up when we add the D-A-D process to say 130 or so samples. The bottom line is that the D-A-D process and the drivers will always take time and the ONLY way to remove it is to compensate for the delay. This is not so easy and could lead to errors. How is the software to know whether you are doing a digital transfer or using D-A-D process? Whichever compensation is used could be wrong. Remember that many soundcards mixer software does'nt even know what route your audio is taking and it is easily possible to run digital and analogue at the same time so what compensation should you use. I cannot see any way in which Cubase or Sonar can rectify this properly although things could be made tighter by using basic compensation (certain data is known when you run wave profiler) This basic compensation should also be able to work with differing soundcards as they are seperately profiled and the software DOES know which card you are using. Although better I cannot currently see how this could be made perfect. Maybe future soundcard hardware design could address this issue by recognising which I/O is active and knowing their own inherent delays to be passed to the DAW software. I am not aware of any soundcards that are doing this as yet. Still, basic compensation would be the first step and maybe Cubase has done this. Anyone know??
< Message edited by Mr. Ease -- 7/22/2004 1:46:29 AM >
|
KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 03:32:16
(permalink)
ORIGINAL: billkath
I've put a little hep up on my site. It's called testlatency.zip
http://www.heartbeatstudios.net/testlatency.zip
Thanks a lot, Bill. Different experiments turn up different results, as I've observed looking at this phenomenon off and on over the past few days. And even for the same experiment (yours), in the case of my RME 9632 rig, there seems to be some kind of synchronization or calibration that occurs at power-up or reset, because I get very consistent results, down to the sample, across a range of latencies and driver models on a given set of runs, but if I reboot my machine, it will shift a bit. The values I observed this morning with your test project were 51, 60, and 54 samples. I'm pretty darned certain that I wasn't hallucinating last night when I repeatedly observed negative skew, or when I observed essentially zero skew (the signal didn't have as good a signature as your test wav, so there could have been +/- 10 samples and I would not have been certain) two days ago, but I don't have time to do the set of 20 or so reboots that it might take to see those again.
|
billkath
Max Output Level: -69 dBFS
- Total Posts : 1076
- Joined: 2003/11/27 08:16:29
- Location: Ireland
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 07:27:23
(permalink)
Very wierd Kev,
Here was me hoping that this was going to be consistent samples. I've tried it dozens of times with Layla24 and got exactly the same results. I had hoped that this was going to be the same (not the numbers, but the consistency of the numbers) on every sound card, and thus the fix would have been easy. My own method of a fix is very easy.
The first track is fine- I leave that where it is.
I just turn the time bar to samples and zoom in.
For subsequent tracks that are looped through an effect I slip edit the tracks by 135 samples, apply trimming, then drag the track over by 135 samples.
If its digital in and out to effects i do it by 67 samples.
It only takes seconds.
What I'm going to try on my next project is also doing this for tracks that are not on a loop (in other words for normally recorded digital and analog tracks.)
Also I'm going to work out, according to what my mixing latency is set at, samples/ms.
and then for any soft synths that need rendering I'll slip edit and slide them as well.
Hopefully this is the right way to go.
Billy E HeartBeat Studios
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 07:40:58
(permalink)
I cannot see any way in which Cubase or Sonar can rectify this properly although things could be made tighter by using basic compensation (certain data is known when you run wave profiler) This basic compensation should also be able to work with differing soundcards as they are seperately profiled and the software DOES know which card you are using.
Still, basic compensation would be the first step and maybe Cubase has done this. Anyone know??
I'm not a technician, I'm a musician. I can accept what is possible and what is not (yet) possible. Read my posts again: I tested Cubase (which I own) and demoversions of adobe audition, sony vegas/acid and only Sonar has this problem! So the conclusion!?
If I remember right I saw the same problem with Sonar2 which I demoed, but I thought it was a problem of my soundcard. I'm wondering that there are not more people active in this topic. I'd like to see if this applies only to some people or if this is a general sonar problem. I stil don't have words to express my feelings about this situation. I mean, whose computer was able to play and record at 2-4 ms latency 2-3 years ago? I have a P4 at 1.8 gHz and I'm able to record and add some effects in Cubase without any problem. If I don't record virtual instruments, I don't need a low latency setting. Cubase conpensates everything very fine. In Sonar it is like adding an echo to individual waves that are recorded in!
|
KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 08:44:18
(permalink)
ORIGINAL: sani
I'm not a technician, I'm a musician. I can accept what is possible and what is not (yet) possible. Read my posts again: I tested Cubase (which I own) and demoversions of adobe audition, sony vegas/acid and only Sonar has this problem! So the conclusion!?
I don't have Cubase, nor Vegas, but I do have Adobe Audition, so I ran Bill's experiment there (or a slight variant). I see the same 50-odd sample skew in Audition as I do in Sonar. Are you *sure* about what you just said?
|
UnderTow
Max Output Level: -37 dBFS
- Total Posts : 3848
- Joined: 2004/01/06 12:13:49
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 09:12:43
(permalink)
ORIGINAL: sani
Thanks Kevin for trying to help.
While trying both programs to do the same thing, I didn't change the asio settings for the audiocards. I'm glad to hear that your system behaves as it should. My both programs are set to use the clock of the audiocard. In Cubase it doesn't matter how high or low I set the latency. I could set the latency for hundreds of miliseconds and the record would still be sample accurate, but not in Sonar.
But at the end, if cubase works well with the asiodrivers of the card and sonar not, it still means to me, that sonar is not quite asio compatible as it should be.
I did some tests with my Terratec EWS88MT and I have a fixed loop-back latency when using WDM drivers (66 samples) and a much larger variable one when using ASIO. I have never liked the way this sound card interacts with Sonar in ASIO mode. Things get out of sync. Especially as I increase the latency settings to compensate for the CPU demands of large projects. I have always stuck to WDM. UnderTow
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 09:16:53
(permalink)
Are you *sure* about what you just said? I didn't run Bill's experiment on audition, a took some guitar part that I recorded on one of my previous sessions. I also didn't zoom in up to the individual samples, I can do it later this day. To be honest, I have to say that I'm not sure if the new recording in Audition was sampleaccurate. I don't use audition, so maybe I didn't set the apropriate drivers (I guess it does not work with asio). On Cubase, I have to say: yes, it was sample accurate. I zoomed in and the samples were (slightly differen, I guess due to the a/d conversion) but the sample points were on the same place. I didn't look in audition if the recording was exactly sample accurate, because visualy the recording seems to be on the right place. The same goes for vegas/acid. I zoomed in to see the beats a little better, but not to the sample level. The difference to Sonar is that in sonar I saw immediately that the record is late, without the need to zoom in. Another thing: when I play it: in Cubase you don't hear any change. The mix output is just louder until the recorded track ends. In audition and acid/vegas it sounded chorused, but in sonar it was a mess: a delayed sound, like two guitarplayers playing out of sync.
|
KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 09:44:06
(permalink)
ORIGINAL: sani
Are you *sure* about what you just said?
I didn't run Bill's experiment on audition, a took some guitar part that I recorded on one of my previous sessions. I also didn't zoom in up to the individual samples, I can do it later this day. To be honest, I have to say that I'm not sure if the new recording in Audition was sampleaccurate. I don't use audition, so maybe I didn't set the apropriate drivers (I guess it does not work with asio).
On Cubase, I have to say: yes, it was sample accurate. I zoomed in and the samples were (slightly differen, I guess due to the a/d conversion) but the sample points were on the same place.
I didn't look in audition if the recording was exactly sample accurate, because visualy the recording seems to be on the right place. The same goes for vegas/acid. I zoomed in to see the beats a little better, but not to the sample level. The difference to Sonar is that in sonar I saw immediately that the record is late, without the need to zoom in. Another thing: when I play it: in Cubase you don't hear any change. The mix output is just louder until the recorded track ends. In audition and acid/vegas it sounded chorused, but in sonar it was a mess: a delayed sound, like two guitarplayers playing out of sync.
OK, but note that if the delay is less than 60 or so samples, you do have to zoom in pretty close to notice the skew. At 44.1KHz, 60 samples is about 1.3ms, or the difference between moving a musican or monotor 4 feet further away - measurable, but not musically problematic, and certaily not something that the ear percieves as a delay. The observed skew with Audition was invariant with whether or not I set the "sample accurate sync" option, by the way.
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 09:57:46
(permalink)
60 samples is about 1.3ms, or the difference between moving
a musican or monotor 4 feet further away - measurable, but not
musically problematic, and certaily not something that the ear
percieves as a delay True, 1.3 ms is either musically problematic, nor can a singer or a live recorded drumplayer play tighter. But I hear a delay. Well, I'm at my office computer so I cannot tell you how large delay is in ms or samples, but it is intolerable. You will definitely not hear 1.3 ms latency as delay, would you? UnderTow is right: the terratec as a fixed latency in sonar with wdm drivers, but with the asio drivers selected the latency depends on the setting for the audiocard. UnderTow didn't say how he calibrates the card and what is his latency setting. My terratec can work only with 10 ms for the dma buffer. If I try a higher or lower setting, the sound is distorted. And finally: if Sonar is asiocompatible it must not have problems with asiodrivers that are proven to work good with other sequencers. What happens if someone only has asiodrivers?
|
billkath
Max Output Level: -69 dBFS
- Total Posts : 1076
- Joined: 2003/11/27 08:16:29
- Location: Ireland
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 09:58:42
(permalink)
Very interesting.
What I think I'll do is have a scootch around to find my old copy of cubase. As I said- I don't use ASIO-WDM seems to behave a lot better with Layla24 and Gina24 than ASIO.
For everyone- I do think that if we are going to test, then we should all test the same way-otherwise we may all get different results. What do you think? Can anyone devise a better, more accurate test, or do you think that little file i put up is OK for now?
Billy E HeartBeat Studios
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 10:12:46
(permalink)
It is difficult to make an identic enviroment for all users. We are using different comps, audiocards, some have better drivers, some worse, some cards work better with wdm drivers, other better asio drivers. I do not expect absolute synchronisation but anything above 15-20 miliseconds is IMO not acceptable. Thank you Billy for making a downloadable test, but generally any wave will do it.
|
Patrice Brousseau
Max Output Level: -83 dBFS
- Total Posts : 391
- Joined: 2003/11/05 20:12:10
- Location: Montréal, Québec, Canada
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 10:14:24
(permalink)
I ran a similar test a month ago and I discovered that with WDM on my Echo Mia, the result was the same, whatever latency settings in Audio Options. For the record, it was 129 samples. There was this topic on the subject: http://www.cakewalk.com/forum/tm.asp?m=130201
|
sani
Max Output Level: -85 dBFS
- Total Posts : 300
- Joined: 2003/11/06 14:25:54
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 10:25:45
(permalink)
Thank you Patrice Brousseau again for the link. The conclusion would be: wdm drivers have some sort of "standard" latency, depending how you set up the card; in most cases it can be done in a suitable way to work without any remarkable latency. Asio drivers are, how Ron told it "a gray area", but I guess mostly for the Cakewalk stuff. I preffered asio drivers because with wdm drivers for the terratec card, sonar didn't show me all inputs/outputs and I had to choose them from the audiocards mixer, so I wanted to work with asiodrivers but this is obviously not possible as UnderTow made the test with his terratec card: ASIO drivers with 1024 buffer (effective 23.2 ms) ==> offset 2110 samples.
ASIO drivers with 2048 buffer (effective 46.4 ms) ==> offset 4158 samples.
|
KevinK
Max Output Level: -73 dBFS
- Total Posts : 863
- Joined: 2003/11/06 06:30:43
- Location: Le Bar sur Loup, France
- Status: offline
RE: Does Sonar compensate for latency recording?
2004/07/22 10:33:43
(permalink)
ORIGINAL: sani
Thank you Patrice Brousseau again for the link.
The conclusion would be: wdm drivers have some sort of "standard" latency, depending how you set up the card; in most cases it can be done in a suitable way to work without any remarkable latency.
Asio drivers are, how Ron told it "a gray area", but I guess mostly for the Cakewalk stuff. I preffered asio drivers because with wdm drivers for the terratec card, sonar didn't show me all inputs/outputs and I had to choose them from the audiocards mixer, so I wanted to work with asiodrivers but this is obviously not possible as UnderTow made the test with his terratec card:
ASIO drivers with 1024 buffer (effective 23.2 ms) ==> offset 2110 samples.
ASIO drivers with 2048 buffer (effective 46.4 ms) ==> offset 4158 samples.
There must be a lot of variation in ASIO driver implementations, as I saw the same number of samples of skew regardless of what latency I selected with the RME drivers, and regardless of whether I sued ASIO or MME drivers (RME doesn't do WDM/KS). Now, as it happens, RME sells cards/converters to Steinberg that are resold under Steinberg's name, which is one reason they're well tweaked. But there's nothing about ASIO, nor even about Sonar's ASIO support, that *necessarily* causes long delays, nor delays equivalent to the buffer latency (which is what UnderTow was reporting for the Terrtec).
|