Stereo Tool

much better noisegate ...

my final measurement is multiband+singlecomressor 8ms
+ noisegate 11ms
+ hardlimit 12ms

This is awesome .. for microphone processing only, i think there is not need for other processors. Maybe Band-pass.

Hans, Great job!

I know - still, if I could get the other steps with as little latency as these, then the entire thing would work very well for live broadcasts. So I'll give the other filters a shot as well.

(I have already checked the final limiter, and IF there are no loud peaks limiting can be done in about 12 ms. But I want it to be able to handle loud peaks as well, so I'll keep looking).

Hi Hans,
Can you add a ASIO control panel button and "realtime priority" checkbox?
i know that for ESI cards there is not "asio panel" but for other cards there is.
I think this is important for using different soundcards with ASIO and lowlatency processing.

A few updates:
- Bandpass filter added. I'm now able to filter at frequencies near 14-18 kHz with a latency of only 1.6 ms - unfortunately in some cases I need 3 of those filters in a row.

Filters that are now working in very low latency mode:
- Noisegate filter, latency of 1.6 ms
- Stereo Boost, no latency. Only works if Multiband is enabled.
- Dualband, no latency (0.7 ms latency if 'remove remaining peaks' is enabled)
- Multiband, no latency.
- Multiband clipping for frequencies above 800 Hz (band 5+). Clipping causes a latency of 4.3 ms!
- Bandbass filter. Latency is 1.6 ms, and it's far less steep than before. When a highpass filter is used (works only when Multiband is enabled!), the phase of the bass sounds changes. When filtering at 14 kHz, the signal at 18 kHz is already near -40/-50 dB - not as good as the normal latency version, but good enough.
- Singleband compressor never had any latency.
- Preemphasis (and deemphasis), no latency.
- Hard Limit. If no FM processing is used, the latency is 1.1 ms. If FM processing IS used, Stereo Tool will perform some extra processing steps to make sure that the signal - when played over an analog device - won't cause too high spikes. This causes an extra delay of about 4 ms (so 5 ms in total).

With everything turned on, the processing delay is now 12 ms.

Not yet working in low latency mode:
- MPX signal generation: FM Overdrive filter (probably another 4-6 ms, 0 without overdrive protection).
- Stereo Boost excessive reverb protection (I think it is possible in about 1.6 ms).
- Final Limiter / Loudness - that's going to be the most difficult part if I want to avoid both distortion caused by deep bass sounds and the "bad bass" sounds from the initial 4.00 beta. (Each bass protection step causes a latency of AT LEAST 23 ms...)

New version should be online in about an hour.

It's online now.

During testing, I found 2 bugs.

1. Pre Limiter: "Remove remaining loud peaks" causes huge amounts of distortion, and makes the sound far less punchy. So for now, turn it off!
2. Multiband compressor: Set "limit output" (on the right) to 100%, otherwise loud peaks are being clipped WITHOUT ANY filtering afterwards, which again leads to horrible distortion.

Kewl

Tonight I managed to do something I thought was impossible: I created a phase linear lowpass filter that filters at 60 Hz (1 waveform is 1/60 sec = 16 ms long) - with a latency of only 2.2 ms!

Unfortunately the CPU load is currently about 16 (!) times higher than that of the normal lowpass filter... Still, the fact that it is possible at all (which is not at all what I expected...) might open opportunities for very low latency 'bass boost' filtering (both the actual 'bass boost' and the similar filter that I'm currently using in normal latency mode to reduce the bass distortion).

The higher CPU load might not be as bad as it sounds - it's roughly equal to the CPU load of the Multiband filter in high quality, HQ on, clipping on and post filter on. And I can decrease the CPU load by increasing the latency, or decreasing the filter accuracy. Both may be needed to reach acceptable levels. The bad news is that I need to run this filter at least 3 times (which, among others, leads to 6.6 ms latency) - and some other filters as well. On the other side, I think I can use the same trick that I used for the lowpass filter also for the Loudness filter, to reach roughly the same quality with far less latency (unfortunately also with a much higher CPU load).

I might have just found a way to do it in 0.7 ms with a very low CPU load....

But it's getting way too late, I'm off to bed, maybe tomorrow I'll discover that what I'm thinking now doesn't make any sense at all...

The 0.7 ms (and probably even less) latency lowpass filter for low bass sounds with very low CPU load is working! Unfortunately I cannot use it for higher frequencies because the CPU load would get far too high. But for low freqs it works perfectly. And I have also found a method to reduce the existing latency of the normal lowpass filters by half.

i see you'r deeeep in it
seem you'll find many other solutiions for this.

I am curently waiting to come another Juli@ (not maya) to try how microphone processing works with ST in real life.
Quick test in radio sounds promising .. using multiband, noisegate and single sompressor.
Why not Maya? Because that card seems can't give good phantom power supply.
So i'll do it with Juli@ and to use "insert" jacks on mixer .. maybe this is even better.