ProTools HD+ Lynx Aurora, expect total cancellation?

[SimonN]

Setting up ProTools HD hardware insert, and am not getting total phase cancellation coming back.
This is how I do it, please chip in to tell me wether it´s my method or my expectations that I should change!

ProTools HD 7.4
Aurora16, rev23, in 32channel mode
Digilinkcables less than 50 feet
External clock derived from UA2196

ADC on.
No extra time added in ProTools I/O-setup.
I create 2 tracks, 1 audio and 1 aux mono.
Both assigned to the same output.
Signal generator on audio track, buss send to aux track at 0VU.
With DigidesignEQ1 ONLY on the aux insert and flipping the phace I get total silence. Ie total cancellation.
Putting a hardware insert (patchcord just running straight from Output to Input on the interfaces) before the DigEQ and flipping the phase gives me a big leveldrop, but the tone is still there. Ie not total cancellation.
Same result over any channels used as insert.

Any ideas?

[theaero]

Why would it cancel completely? It's a different sound. Everytime you go through the ad/da process, the sound will be altered. Therefore, it won't null completely. It'll get really close (big level drop) but it will never null completely. Not until we have miracle converters that don't change the sound at all :)

[prophonic]

Hi Simon,

They will not cancel out because of the minor differences in trim settings and AD/DA non-linearities.

However, regardless of that there is a bug in 7.x to 8.x (confirmed by Digi now) which introduces phase shifts even when you do a digital loop back.

My original (long) post on this is below - sorry for the repeat.

Osman

-----
We use a lot of hardware inserts and having read the comments in this thread I did not want to believe them at first. So decided to do a scientific test (I am an Electronics Engineer by training).

The summary is that the problem is real and repeatable. Now read on...

1. We used a sweep file between 22Hz and 32Khz (the same one we use to generate convolution responses). We ran our test at 96K to minimise sampling errors.
2. Frequency sweep is useful as you can be in phase in some frequencies and not in others. We wanted to establish whether the phase coherency (or lack of it) was frequency dependent.
3. We generated two stereo loop-back connections, one analog and one digital. This was done so that we could isolate whether the reported problem was an AD/DA issue or an internal issue. For analog loopback we used A3-4 (output patched to input) on one 192. For digital we used AES at C9-10 on another 192, output patched to input.
4. We ran the sweep first with the analog insert (NullAnalog) and recorded the waveform post-insert. Repeated the same with the digital insert (NullDigital) and recorded the output to another audio track. Did the same for both mono and stereo (the uploaded file only shows the stereo.
5. Did another set of tests - this time instead of the insert, we used send and return, and recorded two more sets of audio tracks, one with digital, one with analog send-return.

The net net of it is that the concerns expressed in this thread are real and there is a real phase issue which gets worse at higher frequencies.

Here are our observations.

1. In the insert case, both the digital and the analog recorded audio tracks behave similarly. As the frequency increases, the recorded signal gets delayed further. We tried to null the results by phase inversion and you will see the master increasing in volume as the frequency increases. When doing this test remember to deactivate the insert!!
2. The send and return behaves differently and still has phase problems. Listen to the null test and you will see the phase shift oscilating up and down. You get the best results with the digital send/return where the phase shift is the least.

The test session can be found at http://www.songphonic.com/Private/InsertPhaseTest.zip

I would recommend running the same in your system and see what results you get.

Our system is 7.4cs10 on 10.4.9 Dual 2Mhz G5. We have an HD6 system with 6 192s.

We tested with or without ADC (and sample aligning by hand) - it is not an ADC problem. The behaviour is the same.

Try it yourself - if you sample align a random section, you will see that the rest of the waveform at some other point will become unaligned. The behaviour is like that of a frequency dependent but repeatable jitter (in the software). The original sweep and the recorded waveforms are NOT sample alignable however hard you try (and it is not a subsampling issue).

Pure analog always had frequency dependent phase shift (often adds character). Similarly DA/AD process will sometimes cause small phase shifts.

But I would have expected digital loop back to produce perfect null results. The fact that it doesn't makes me think that there is a fundamental internal timing issue lurking somewhere.

Sorry to be the bearer of bad news...

[SimonN]

Thankyou, both.
Especially prophonic for the scientific research done.
Of course, you guys are right.
I have been using hdw-insert in other studios that has claimed that it is totally phase coherrent, and I have thought of it as OK soundwise.
So, I´ll just stick to trusting my ears. Knowing that they lie a litle, there´s no such thing as total phase coherency with hdw-inserts.
And my system is behaving like it should, within the laws of physics.

Thanks for that.

/ Simon