Hi, gang.
 
So I've been trying to figure out my recording latencies at different sampling rates via loopback recording tests.
 
Here are the settings I used to get solid results at the two given sample rates:
 
44100
Buffer Size: 3 msec / 132 samples
Effective Latency: 3 msec
Record Latency Adjustment Manual Offset: 2 samples
 
48000
Buffer Size: 2 msec / 96 samples
Effective Latency: 2 msec
Record Latency Adjustment Manual Offset: 46 samples
 
(All other applicable parameters the same in both modes -- Buffers in Playback Queue: 2, Read/Write Caching: Off, Playback/Record I/O Buffer Sizes: 128 KB)
 
Now if the only setting in question was the buffer size, the answer to my question would seem obvious. But there is also the latency adjustment offset, and I'm not clear on which way that factor swings the final latency amount...
 
Can someone fill me in on that point, thus making it clear which of these settings translates to lower latency?
 
Thank you!
 

All things being equal the higher the sample rate, you will always get lower latency.
The trade off is the higher you go the more room the files take, the more load is on your CPU and your hard drive because you have to move more data in the same time.

If your system is powerful enough then it's no problem.

The question I would ask is Why? Why do you want to record at 48khz instead of 44.1khz. Do you know of some improvement of one over the other or are you preparing audio for DVD media?

2ms is lower than 3ms.
 
Do you know what the record latency offset is? Have you actually tested your record latency to see if you have the correct record latency offset set under "sync and Caching" in the Unified Preferences view?

Thanks for the replies, guys.
 
 

mudgel
All things being equal the higher the sample rate, you will always get lower latency.
The trade off is the higher you go the more room the files take, the more load is on your CPU and your hard drive because you have to move more data in the same time.

 
Thanks, Mike. I do know all of that already, though. My main question was really about in what way the Record Latency Manual Offset factors in to my final latency result.
 
Obviously the 2 msec @ 48k is smaller than the 3 msec @ 44.1k, but then the 48k's recordings require a higher manual offset (46 samples) as compared to that of 44.1k (2 samples) for exact round-trip synchronization.
 
My assumption would be that, since the looped recordings were coming in 46 samples late in the 48k tests, the total latency would be 2 msec + 46 samples (or about 2.96 msec). And the same for the 44.1k tests, which came in 2 samples late, I would assume to be a final figure of 3 msec + 2 samples (about 3.05 msec).
 
But I don't want to just assume... For all I know, the actual latency might simply be the total offset amounts required to get things in sync. Or, I might be on the right track up there^ but, as I've read in prior threads, SONAR's reported effective latency figures may not be entirely trustworthy.
 
This is why I wanted to pick the brains of the experienced forum users first.
 

mudgel
The question I would ask is Why? Why do you want to record at 48khz instead of 44.1khz. Do you know of some improvement of one over the other or are you preparing audio for DVD media?

 
Well the main driving force was the matter of latency... if there was enough of a gain from 44.1 to 48 to make it worth using. I realize that the difference in that regard (especially if it's only around 1 msec or less) is miniscule, and almost certainly imperceptible. But on a more academic level, just "to know", I am curious how the manual offset factors in to the final figures.
 
There is also the matter of aliasing with certain plugins when using one sampling rate over another, but that's a whole other enchilada that's been covered in other threads.
 
 

thomasabarnes
Do you know what the record latency offset is? Have you actually tested your record latency to see if you have the correct record latency offset set under "sync and Caching" in the Unified Preferences view?

 
Yes, of course. Did you read all of my original post? I gave the offset parameters needed for proper synchronization under both sampling rates.
 

To compute the total round trip latency including the manual offset add the offset to the ASIO reported latency to the right of the offset and divide by the sample rate. See the "Record Latency Adjustment" section near the bottom of http://www.cakewalk.com/D...age=3&help=0x22B1A

dimelives1:
 
I did read all of your post, but it seemed like you didn't really understand what the Record Latency Offset is.
 
But to answer your question of which the title of this thread is: I think the 2ms is the lower latency of the two. That seems obvious to me, but maybe I'm missing something and you're looking for one of Jim Roseberry's magnificent analyses. 

scook
To compute the total round trip latency including the manual offset add the offset to the ASIO reported latency to the right of the offset and divide by the sample rate.

 
Thanks, scook. That's essentially what I was looking for.
 
But what if I didn't use ASIO (for instance, if I chose to use WDM) and therefore didn't have the ASIO reported latency? Would I just be left to guess at that point?

No with WDM the latency is listed next to the buffer size slider under Audio Driver Settings.

The only way to know true latency is to measure round trip latency or loop back as you've mentioned. But I don't understand your figures.

Simply do a recording of some external input. Take note of the sample offset that Sonar introduces,

Do the same recording but this time route the audio out of your device via a cable and back in via an input and record that signal.

The difference between the two tracks in samples is your round trip latency. Don't forget to factor in any sample offsets introduced by Sonar. If that's what you've done I'm sorry but I don't get that from you're figures.


So whether you're using ASIO or WDM won't matter in so far as being able to produce a result. It will tell you which driver and which sample rate is truly the smallest.

That gets away from what a driver tells you and internal safety buffers which are not reported to the driver.

thomasabarnes
But to answer your question of which the title of this thread is: I think the 2ms is the lower latency of the two. That seems obvious to me, but maybe I'm missing something and you're looking for one of Jim Roseberry's magnificent analyses. 

 
Yes, yes, obviously 2 msec is lower than 3, I'm not that hopeless What was making the grand total unclear to me was that the 2 msec tests required an offset of 46 samples to get things properly synced, while the 3 msec tests required an offset of only 2 samples. So the sampling rate that offered a lower base latency (48k) required a larger offset than the sampling rate that offered a slightly higher base frequency (44.1k). I just wasn't clear on which way the manual offset swung the results, i.e. if the fact that the 2 msec setup required a higher offset actually made it have a higher total latency than the other setup, which needed a smaller offset.
 
Hopefully that makes more sense and I don't sound too lost in the woods here.
 
 

John
No with WDM the latency is listed next to the buffer size slider under Audio Driver Settings.

 
Thanks, John. But it was my understanding that that given figure from SONAR may not be entirely trustworthy, and/or that it may not account for the other factors that the ASIO reported latency appears to (A/D conversion latency, etc.).
 
And if it is accurate, is that telling you the round-trip latency, or just one way?
 
 

mudgel
The only way to know true latency is to measure round trip latency or loop back as you've mentioned. But I don't understand your figures.

 
What seems unclear about my results, Mike? I simply gave the buffer sizes and manual offsets required to get stable, in-sync loopback recordings.
 
 

mudgel
Take note of the sample offset that Sonar introduces,

 
Like I mentioned to scook, if I'm using a driver mode other than ASIO, that figure is not present.
 
 

mudgel
Do the same recording but this time route the audio out of your device via a cable and back in via an input and record that signal.

The difference between the two tracks in samples is your round trip latency. Don't forget to factor in any sample offsets introduced by Sonar. If that's what you've done I'm sorry but I don't get that from you're figures.

 
That is precisely what I've done, my friend Only I did it without SONAR's automatic offset (since I wasn't using ASIO for these tests). I apologize for not making that more clear.