Stereo Tool

Hm, I think I know what that can be - I got some exceptions too when I was checking the logging. I'll start a new build now (exceptions should be fixed).

I hope this one doesn't crash: http://www.stereotool.com/download/radi ... ETA026.exe

Warning: Not an installer, this is the executable itself!

Granularity: 256samples - Maya44USB - 48kHz - minimal buffer size 5.4ms
After Forcing Card to "High Speed" - Granularity: 48samples - 48kHz - minimal buffer size 3.0ms

Actually i found that for stable output at 192kHz should be used 256samples and more, at soundcard, and at 48kHz 64samples and more.

I'll check with julia@ .... ..

hans and bojcha => a "asio config" button will be very usefull for stereo tool standalone , but i have already make a request for this

lot of software have this (bbp asio ,breakaway live ,sonos 4 , minihost(a vst host).... )

A question in between. Do you come to the Soundprocessing FreakDay in the Netherlands? http://www.processingfreakday.com/
Maybe to let some people hear there's more than orban, omnia, bw, bbp, vorsis.

Bojcha: Based on your reported 3 ms buffer size and 48 sample granularity I would expect:

- 512 / 48 = 10.666 -> 11 grains for audio to arrive
- 3 ms @ 44.1 kHz -> 3 grains for audio to be processed (could be 4 due to rounding issues if you're using 48 kHz...)
- 1 extra grain for output
Total 15 grains x 48 samples = 720 samples = 16.3 ms latency @ 44.1 kHz...

This assumes NO extra sound card latency.

beta26-standalone:Normal Output, No Phase Rotator, 0-84 web preset, 512 mode.


FM Out:
There are 2 strange things there.

- I can set lower buffer size then in Normal out. Normal out 4.0ms for 192kHz 256samples at soundcard.
For FM out it can be 2.8ms. So something is maybe still wrong in Normal out.

- And second, FM out is 16ms latency, no phase rotator, Composite limiter ON, RDS ON,
But when i just enable "Transmitter Calibration" (witch have some random adjustments) latency is 27ms

Great! That's much better than before

I'll repeat my calculation with your settings:
- 512 / 64 = 8 -> 8 grains for audio to arrive
- 3 ms @ 48 kHz -> 3 (2.06!) grains for audio to be processed
- 1 extra grain for output
Total 12 grains x 64 samples = 768 samples = 16.0 ms latency @ 48 kHz...

According to the ASIO spec there's apparently one extra grain needed: That adds another 1.33 ms giving 17.33 ms total. Which perfectly matches your findings.

In my processing I'm throwing away one grain that I could have used - so it must be possible to reduce the ASIO latency by 1 grain (1.33 ms at your settings), to 16 ms.
Fixing this is possible, but a little tricky (will need a lot of testing), so I'll postpone that for a later version.

[Note to self: It's possible to send data to an audio buffer AFTER I've returned control to the driver if I call ASIOOutputReady() when I'm ready].

@bojcha
How do you know that a buffer of 2.8ms isn't creating clicks/pops @192khz. Do you use some kind of measurement to get to the correct value?

@ Bojcha,

After looking at the code - and the difference between FM and normal output - I'm wondering if it's possible that you get clicks of a single sample every 256 samples in the FM output? This might be almost unnoticeable...

Why I'm asking:
- I send 256 samples to my processing code
- For each sample that I send, immediately another sample is sent to the FM chain. BUT for the LAST sample on the RIGHT channel only, I perform the send AFTER I do the processing - which takes a few ms.
- Normal output has to wait until the data is actually there.

This fully explains why FM latency can be set lower than normal latency.

My questions:
- If you slightly lower the FM latency, do you get this type of very short repetitive clicks, or does it start to click and hiccup badly? If it's just short repetitive clicks, I can easily fix it by outputting the samples BEFORE starting the processing of the next block - enabling another latency reduction of 1.33 ms...

I think I should do that anyway, and also for normal output. To be done!

On top of this, I think fixing this will also increase the maximum throughput - in other words, run better on older PCs.