I use External Insert to send/return a track to hardware EQ-Compressor, etc.  Everything is played in sync.  Now I try to record that by create a new track with the input set to the same port as the return port of the the External Insert.  What I observed on the Track View is the new re-recorded clip is recorded much earlier than the original clip.  I would estimate 4000 samples early (buffer set at 512, 48k sampling). 
Couldn't make sense out of this yet.  I'd thought that the re-recorded clip would be in sync as latency was compensated by the External Insert, or come in after the source signal by x latency factor.  But before, how is that even possible?

The real-time delay compensation for an External Insert is done by delaying the playback of other tracks to match the delayed signal returning from the external hardware. This delay compensation works just like Plugin Delay Compensation for plugins working inside the box.
 
When you record with PDC active, the signal is usually from a direct-monitored live performance that's expected to be in sync with the delayed ouput (including the metronome click), and will be compensated for that. The delay for the signal returning from the external hardware to another input has already been compensated by PDC, but SONAR doesn't 'know' that so the recorded audio will end up being over-compensated. But I think you must be mistaken about that 4000-sample discrepancy at a buffer of 512, unless maybe the external hardware is digital, and is introducing a lot of A/D/A conversion and DSP latency on top of the DAW interface latency.
 
In any case. the bottom line is that to render processed output of an EI with proper compensation, you need to do a real-time bounce (i.e. Fast Bounce disabled) to another track, and then bypass the EI and mute/archive the original track.

Would I achieve the same result if instead of using EI, send the track output to a soundcard 'out' port and the re-record track' input is set to the soundcard 'in' port.  That would eliminate the EI PDC and should be just like normal tracking, and latency is just ASIO and is handled automatically, right?  If that's the case then I have to re-think why using EI at all.

Hi micv,
Hmm...something seems funny here. If you designate a port as a return port in EI, Sonar shouldn't let you choose that input port as a source for other tracks....at least that's how it works here. So for example, if I set input 67 as a return in EI, I cannot then use input 67 as a source for an audio track.
 
That aside, the way to do this is probably either with Bounce, Freeze, or by routing the output of the track with EI to an aux and recording the aux. You'll get the same result as recording a hardware input and won't (or shouldn't) run into strange PDC behavior.
 
As far as routing to a hardware out and recording back in, delay compensation can't account for that because Sonar has no way of knowing that you're setting up a round trip through your A/D and a piece of hardware and then back in. That's the situation that EI is intended to fix.

Dean

Hmmm.... yeah, good point about re-using the return input. Actually, I've noticed SONAR will let you choose the mono sides of a stereo input pair that the EI is using, which is a little odd.
 
But mainly I wanted to note that recording to an Aux track will give the opposite problem: the return is under-compensated.
 
Non-fast Freeze or Bounce are really the only viable options. I had forgotten that Freeze had a Non-Fast option; thanks for reminding me.

Hunh....yeah Brundle, you're right. I just tried to record from a track with EI on it to an aux and the resultant track was delayed. I would have thought that PDC would handle a track with EI on it in the same way it handles any other plugin. Thanks for pointing that out.
 
OP - Yep, just freeze or bounce.
 
Dean

Thanks for the valuable info guys.
I have a Lynx Hilo usb that allows its 4 internal USB ports to map to any of its analog in/out.  So for example I route "usb play" to al 4 ports, the IE uses one and yes that port will not be visible to select but there are other 3 ports that were routed to the same source. But as you have pointed out record with IE just won't work no matter what.
Routing track to DA-HW-AD works for me though as I believe that all those ports visible to Sonar are asio so Sonar was able to compensate for the round trip delay.  Yes there might be some additional delay from the outboard gears but they are all analog so latency should be negligible.  I will run some more test just to be thorough but a quick test shows that everything lined up.

I spent the last couple evening looking into this carefully.  What I stated earlier about the re-track by routing to port is wrong.  By using a test signal, I was able to measure that at 48k, 512 samples buffer, the re-track was recorded late by 155 samples.  If I add this value to the Asio Manual Offset, the re-track synced up perfectly.
Now I just need to understand what I've observed.

Yes, you need to have your record latency compensation dialed in before complicating the picture with External Insert. I should have mentioned it but didn't because you were talking about errors of 4000 samples, and Manual Offset is usually on the order of 20-50 samples.
 
It sounds like you are now compensating for the some EI delay with Manual Offset, which you don't want to do, because Manual Offset will apply to compensation of 'normal' recordings, and they will end up over-compensated.
 
To set Manual Offset, use a tool like the free CEntrance Latency Tester to measure actual round trip latency, subtract whatever is reported in for Total Round-Trip in Preferences > Audio > Driver Settings (usually smaller), and enter that difference (usually positive) as the Manual Offset.
 
You can also measure the actual round trip using EI with nothing in the loopback path, and add one buffer to that number. It should come out the same as CEntrance.
 
Manual Offset should be fixed regardless of buffer size so you shouldn't have to touch it again.

brundlefly
It sounds like you are now compensating for the some EI delay with Manual Offset, which you don't want to do, because Manual Offset will apply to compensation of 'normal' recordings, and they will end up over-compensated.

Not sure about this.  I removed all instances of EI in Sonar and ran test on basically a hardware loopback (In connects to Out).  Nothing but a source track and a record track, so EI shouldn't be in the equation.  Since the Hilo has addition buffer settings (standard. low, etc.) I suspect that it has something to do with the high number.  I can collect more data on different settings.
 

brundlefly
To set Manual Offset, use a tool like the free CEntrance Latency Tester to measure actual round trip latency, subtract whatever is reported in for Total Round-Trip in Preferences > Audio > Driver Settings (usually smaller), and enter that difference (usually positive) as the Manual Offset.
You can also measure the actual round trip using EI with nothing in the loopback path, and add one buffer to that number. It should come out the same as CEntrance.

I did this and confirm that EI's report is one buffer less than CEntrance.
 

brundlefly
Manual Offset should be fixed regardless of buffer size so you shouldn't have to touch it again.

I need to run more test at different buffer but so far at 256 and 512 buffer the Offset is the same, but at 128, the offset is -102.  Just when I thought I got this figured out!!! 
 
A question I have is Is the reported asio latency in Sonar came from the soundcard driver? if so is the below correct?:
CEntrance = HW + asio/usb driver latency.
Sonar reported asio = asio/usb driver +  sonar input-output buffer latency
then
CE - Sonar = HW + sonar input-output buffer