Delay compensation with hardware inserts

[scottgreiner]

There are two issues regarding ADC and hardware inserts that have not been addressed here (I think) and will affect parallel-type compression etc:

1) the inter-sample differences between a hardware inserted subgroup and it's ITB duplicate will always be about 1/2 sample off in either direction. ADC can only compensate down to the sample. The solution is to send your non-compressed subgroup on a round trip through the converters so it ends up with the same inter-sample delay as your compressed bus.

2) Aux sends and assigned outputs have different delays that can not be accounted for once the signal has left TDM. For example, a snare output assigned to output 1 and aux sent to output 2 will not line up outside the TDM environment (4 sample difference?) One solution using drum subgroups as an example is to "dummy output assign" all the drums (bus 63/64 for example); make post fader stereo aux sends 1/2 at unity gain on all drum tracks for your main drum subgroup, returned on an aux fader; make another post (or pre) fader stereo aux send 3/4 for your compressed parallel subgroup, returned on another aux fader. Now both subgroups share the same path, plus the added advantage of being able to automate what goes into the compressor. Also, if you sum OTB these subgroups will line up as well.

[d.f.]

Hi Scott.

Regarding your point 2 above: the problem with this "dummy output assign" is that you cannot have the individual drums going to their own outputs when summing OTB (ie kick to summing channel 1, snare to summing channel 2, etc etc). They will always be summing ITB.

What I want to do is have the drums individually assigned to OTB summing as above, make post fader aux sends on all these tracks that return on an aux fader that is my compressed parallel subgroup which then goes to OTB summing along with the "dry" signals. Is it possible to do this properly?

[ev33]

This is an issue I have wrestled with a lot in pro tools, as I like to use parallel compression blending for drum mixing. First of all, you will never get complete cancellation once the analog domain is in the chain. If you split a signal to 2 channels of an analog console and try to get them to cancel by flipping phase on one of the channels, you would notice that it is impossible to get 100% cancellation the way you can in a digital mix environment. There will always be some degree of a hi frequency artifact left. This is due to the tolerances of the components involved with the overall low pass filter in the audio circuit that is usually necessary to reduce RF interference or improve stability in the amplifiers.

I am using 192's on my HD rig so I don't have to deal with the additional complication of the H/W insert delay. When delay compensation is "ON" the system will try to compensate for the latency inherent in the 192's to keep everything in phase if an I/O insert is added to a channel. DIGI TAKE NOTE. Here is the incredibly pathetic part. Who ever was in charge of establishing that amount of compensation GOT IT WRONG!!!!!!!!!!!! It is off by 1 sample. I have re checked many times and always the same result. Here is the test method. create a source channel and send it to 2 aux faders via buss routing. put short time adjuster on both aux faders. put I/O insert on one of the aux faders. Patch the output directly to the input of the 192 for that channel, flip phase on one of the time adjuster plugins and voila!! excessive high frequency artifacts due to over compensation. now open the time adjuster plugin on the aux fader that has the I/O insert and set the delay amount to 5 samples instead of the default 4. You will hear a significant improvement in the degree of cancellation.

There is the other method of putting I/O inserts on both so the latency is forced to be the same on both signal paths. The phasing is a bit tighter with this method but the down side is that all of the signals that don't need any outboard processing suffer an additional generation of D/A then A/D conversion. I have not decided yet which method does the least harm. I would be curious to know if others get the same result or any other methods for dealing with this.

[Bob Olhsson]

I'd recommend the technique we used to use to set up time-code offsets.

You make up a 1K, -20 dBfs. tone region that is exactly one frame long to the sample. (this is known in the film world as a "2-Pop") This is real easy to do by zooming in. Then you simply run it through your outboard gear and record the results.

It is really easy to figure out exactly how many samples of latency are involved.

And yes, lots of software gets it wrong. This is why I think automagic latency compensation is way overrated.

[scottgreiner]

Originally Posted by d.f.

Hi Scott.

Regarding your point 2 above: the problem with this "dummy output assign" is that you cannot have the individual drums going to their own outputs when summing OTB (ie kick to summing channel 1, snare to summing channel 2, etc etc). They will always be summing ITB.

What I want to do is have the drums individually assigned to OTB summing as above, make post fader aux sends on all these tracks that return on an aux fader that is my compressed parallel subgroup which then goes to OTB summing along with the "dry" signals. Is it possible to do this properly?

Sure you can! Just Aux-send to a physical output. I do it all the time on the SSL. This way all the outputs come from aux sends and thus have the same amount of latency on the console (output assignments and aux sends have different delay).

Kick track aux send at unity to output 1; Snare track aux send at unity to output 2... etc. Then each track gets another stereo aux send to a pair of outputs which creates my parallel submix to be compressed.

Works the same ITB too.

[scottgreiner]

Originally Posted by ev33

I am using 192's on my HD rig so I don't have to deal with the additional complication of the H/W insert delay. When delay compensation is "ON" the system will try to compensate for the latency inherent in the 192's to keep everything in phase if an I/O insert is added to a channel. DIGI TAKE NOTE. Here is the incredibly pathetic part. Who ever was in charge of establishing that amount of compensation GOT IT WRONG!!!!!!!!!!!! It is off by 1 sample. I have re checked many times and always the same result. Here is the test method. create a source channel and send it to 2 aux faders via buss routing. put short time adjuster on both aux faders. put I/O insert on one of the aux faders. Patch the output directly to the input of the 192 for that channel, flip phase on one of the time adjuster plugins and voila!! excessive high frequency artifacts due to over compensation. now open the time adjuster plugin on the aux fader that has the I/O insert and set the delay amount to 5 samples instead of the default 4. You will hear a significant improvement in the degree of cancellation.

There is the other method of putting I/O inserts on both so the latency is forced to be the same on both signal paths. The phasing is a bit tighter with this method but the down side is that all of the signals that don't need any outboard processing suffer an additional generation of D/A then A/D conversion. I have not decided yet which method does the least harm. I would be curious to know if others get the same result or any other methods for dealing with this.

Yes, I believe Rail was the first to post about this 1 sample issue when ADC first came out. Not sure if it's even still an issue or if it's been fixed, as I do the "converter round trip" method for my uncompressed drum subgroup when mixing ITB. This way it solves the inter-sample phase issue as well.

Originally Posted by ev33

If you split a signal to 2 channels of an analog console and try to get them to cancel by flipping phase on one of the channels, you would notice that it is impossible to get 100% cancellation the way you can in a digital mix environment

I don't have this problem when I mix on a console, as long as all channels are leaving Protools the same way (via aux sends in my case) and I'm using the same kind of converter. Complete phase cancelation every time.

I have heard about a slight shift between separate 192 boxes (not within their respective groups of 8 outputs), but haven't confirmed it myself.

[dodlum]

Hey there,

Just to add - that I too use this 'dummy aux routing' method to compensate for the difference between routing out through physical 192 outputs and a parallel aux out.

Also - on my system I can verify that I run into trouble when I route out of a second 192, and the same goes for an expansion d/a card in the same 192.

David

[d.f.]

Originally Posted by scottgreiner

Sure you can! Just Aux-send to a physical output. I do it all the time on the SSL. This way all the outputs come from aux sends and thus have the same amount of latency on the console (output assignments and aux sends have different delay).

Kick track aux send at unity to output 1; Snare track aux send at unity to output 2... etc. Then each track gets another stereo aux send to a pair of outputs which creates my parallel submix to be compressed.

Works the same ITB too.

Many thanks for this Scott. I'll try this today.

[scottgreiner]

Originally Posted by d.f.

Many thanks for this Scott. I'll try this today.

Most welcome!

Originally Posted by dodlum

Also - on my system I can verify that I run into trouble when I route out of a second 192, and the same goes for an expansion d/a card in the same 192.

David

Good to know. I'm going to check my Aurora 16 between channels 1 and 16 and see what I get as soon as I get back to my studio.