[002R] Hardware Inserts Latency: how much is it?

[HotRats_Reloded]

I've made a simple test with my 002R, PT LE 7.0 in a PC

trkA:1 audio track with a 4 bars region of recored click with hardware insert (in/out #5)
trkB:1 audio track with a 4 bars region of recored click without inserts.
trkC:1 audio track recording the TrkA output.

(the hardware connected to my insert was a joemeek VC1Qcs)

I've repeted the test with all the H/w buffer settings: 64, 128, 256, 512, 1024.

After the reconrding I've measured the delay between TrkB and TrkC due to the harware insert
and here are the results:

buffer size: 64 -> delay: 334
buffer size: 128 -> delay: 462
buffer size: 256 -> delay: 718
buffer size: 512 -> delay: 1230
buffer size: 1024 -> delay: 2254

the delay is alwayes 2 x buffer size + 206.


Is it right?

three questions:
1. why the delay indicator in on my TraA says always 6 samples?
2. is the latency of harware inserts a constant + 2 times the h/w buffer size?
3. do you kwon the value of this constant? (is it really 206?)

[MikeTrahearn]

Hi - I wonder if you could do the test but with the hardware insert looped back and bypassed, this would test if there is any latency in your outboard?
I bet the constant changes from 206 to some other value because it may be you outboard giving you the constant? Just a thought.

Mike

[HotRats_Reloded]

Yes, I know that outboard may add some latency and that the amount depends form the gear you're using.

In my test I was not testing latecy itself but the VC1Qcs as "effect" for "reamping" some previously recorded guitar and bass tracks.

What I need to know is if the 002R delay is really a constant + 2 x (h/w buffer size),
the value of this constant and why the delay indicator says 6 samples instead of something like
460 (with a 128 sampels buffer).

[daeron80]

The 002 latency is constant for a given sample rate and hardware buffer setting. I think it might be different for digital I/O than for analog, but I haven't been able to test that yet.

If you're connected by analog I/O (you said #5, so you must be), then I don't know how you were able to get the higher buffer settings to have such low latency. At 1024, for example, a simple loopback yields 2561 and 2586 @ 44.1 and 48 KHz, respectively. Did you use the null test method?

Please see my web page about 002 mix latencies. Full details for analog connections.

[HotRats_Reloded]

double post, sorry.

[HotRats_Reloded]

Thanx a lot, DAvid!

may be in may test I made some mistake beacuse my latency table is different form yours
except for the 44.1, 128 buffer. I've mesured 462 samples instead of 462.

where have you found the 2.34375*b + 161 forumla? that's what I was looking for.

[daeron80]

why the delay indicator in on my TraA says always 6 samples?

I've always wondered that myself. Most peculiar.

BTW, the formula for latency is different depending on sample rate, but it's not quite as simple as 2*buffer+x. At 44.1 KHz, it's 2.34375*b+161. At 48 KHz, it's 2.375*b+154. At the double rates, the multiplier is the same, but the number added is a bit larger. I'd love to know why.

I'm not familiar with the Meek piece you're using, but if it's an all analog path, it shouldn't add any measurable latency. If it uses any digital processing, it will.

There's a simpler and more certain way to measure latency than the one you described. This is known as the null test.
1. Duplicate a track that has no plug-ins on it (it can be a musical track instead of click, if you want)
2. Put an analog I/O insert on one track, and a "TimeAdjuster long" plug-in on the other.
3. Engage the invert ("phase") switch on the TimeAdjuster.
4. Connect the appropriate analog in and out with a short, good quality patch cable. If you're wanting to measure latency with a piece of gear inserted, make sure the gear is set so as to change the sound as little as possible, both level and timbre.
5. While playing back, change the TimeAdjuster Delay parameter until the two tracks cancel each other out as much as possible. They won't quite completely cancel because the converters and I/O line amps will color and distort the sound, and change its level slightly, as will any gear in the path.

[daeron80]

where have you found the 2.34375*b + 161 forumla? that's what I was looking for.

I found it by trial and error using an Excel spreadsheet. I think I want to know why it works that way. On the other hand, if I knew, it might scare me.

[HotRats_Reloded]

Great work man. I'm a statistian so I do appriciate your trial&error approach!
The null test with time adjuster is a preatty good idea.

see ya.