ProSoundWeb Community
Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: Erik Jerde on March 04, 2019, 05:50:47 PM
-
I've never bothered to measure the flatness of my USB Pre2 till today when I didn't have a cable tester on hand and I figured that if I compared the suspect cable to a known good one it would become obvious if the suspect one was bad. It was immediately obvious. :)
I was somewhat surprised to see what I've understood to be a very good measurement interface wasn't phase flat. I'm thinking that since real systems aren't going to be phase flat and since this really isn't a huge deviance that it isn't an issue but I'd love the thoughts of more experienced folks.
-
That's a surprisingly large phase difference between the reference and measurement channels. Something is wrong. Or is it possible that you're using an "internal" reference through the software that doesn't use a hardware reference input to cancel the interface's frequency response?
--Frank
-
That's a surprisingly large phase difference between the reference and measurement channels. Something is wrong. Or is it possible that you're using an "internal" reference through the software that doesn't use a hardware reference input to cancel the interface's frequency response?
--Frank
Smaart v7.5.3.2 mic input 1 to generator.
-
Smaart v7.5.3.2 mic input 1 to generator.
Erik, set up Smaart to use one input of the interface for the reference and another for the measurement. Connect both inputs to the generator output with a Y-cord. You should see at most a few degrees difference at 20 kHz. --Frank
-
Smaart v7.5.3.2 mic input 1 to generator.
From the measurement you've done it's impossible to say whether the phase deviation you're seeing caused by the interface's input, output, or a combination of the two. One of the reasons to use a Y-cable for a transfer function instead of comparing to the generator internally is that it controls for these things: it doesn't matter what distortion the outputs have because transfer functions are source-independent, and it doesn't matter what distortion the inputs have (as long as it's the same distortion) since the TF will negate them out.
-Russ
-
Erik, set up Smaart to use one input of the interface for the reference and another for the measurement. Connect both inputs to the generator output with a Y-cord. You should see at most a few degrees difference at 20 kHz. --Frank
You were right on, thanks for the wisdom.
Now I’ve got to figure out why my Q-SYS has a smiley faced phase response when an input is wired directly to an output. It starts and ends 30 degrees out of phase from the middle of the trace.
-
Now I’ve got to figure out why my Q-SYS has a smiley faced phase response when an input is wired directly to an output. It starts and ends 30 degrees out of phase from the middle of the trace.
Hi Erik, here's what I got...input straight to output...
How does your trace compare?
-
Hi Erik, here's what I got...input straight to output...
How does your trace compare?
Here's what I've got. I mis-remembered when I posted and it's 60 degrees out on each end. It's +3 in the middle and +9 at 122 and 2880. I'm guessing your interface is responsible for the sharp drop off that you're seeing at the top. My straight through cable measurement is truly flat so this slope is all the qsys.
I did email qsc this AM and they are looking into it.
I'm wondering what the expected phase response of a system processor running straight through should be. I (perhaps naively) expected it would be flat except perhaps at the edges.
-
Here's what I've got. I mis-remembered when I posted and it's 60 degrees out on each end. It's +3 in the middle and +9 at 122 and 2880. I'm guessing your interface is responsible for the sharp drop off that you're seeing at the top. My straight through cable measurement is truly flat so this slope is all the qsys.
I did email qsc this AM and they are looking into it.
I'm wondering what the expected phase response of a system processor running straight through should be. I (perhaps naively) expected it would be flat except perhaps at the edges.
I look forward to what you hear from support...btw, support gave me great help understand the q-sys limiters...public thank you, QSC!
The screen I posted was at Smaart at 96kHz.
The one below includes the previous trace, and two taken at 48kHz, one sample apart.
The red phase trace is per Smaart's delay finder at 3.17ms, and I think matches what you got.
The darker green trace is manually ticking the delay finder down 1 sample to 3.15ms.
As you can see, the high end tail either fades up or down around zero.
Smart said +83degrees, and -65 at 20Khz. So a spread of almost 150 degrees for one sample timing change.
Which I think makes sense.
One wavelength at 20kHz has a period of 0.05ms. One sample is 0.021ms.
So one sample represents 360 * (.021/.05) degrees, or about 150 degrees as seen in the traces.
I think the "luck" of what the true timing is, places it right in the middle of samples at 48k,
and that is why 96k is more at zero...
That's just what I've come to on my own...looking forward to what QSC says...
-
I remember something like this being discussed in detail a few weeks/months ago, can't remember the explanations why it happens.
-
QSC got back to me today. Here's what they had to say:
We have confirmed a difference in the phase response of the 110f inputs vs. the phase response of other QSC input circuits such as the CIML4 card. Keep in mind the 110f was conceived to be used in the meeting room environment rather than the performance audio space. One requirement of that space is a smaller form factor, and we believe this is a result of the miniaturization of the input circuitry.
Note the measured phase shift only applies to 110f analog inputs, so this will not occur on any sources brought in through the network, SIP, USB, etc. If you have applications where you feel this phase response is unacceptable, you could use the analog inputs of a CXDQ amplifier or IO Frame in conjunction with a core 110f, or move to a core model with the CIML4 card such as the Core 510i.
I need to grab a BSS Blu-100 and measure it now to see if it has similar issues.
-
QSC got back to me today. Here's what they had to say:
We have confirmed a difference in the phase response of the 110f inputs vs. the phase response of other QSC input circuits such as the CIML4 card. Keep in mind the 110f was conceived to be used in the meeting room environment rather than the performance audio space. One requirement of that space is a smaller form factor, and we believe this is a result of the miniaturization of the input circuitry.
Note the measured phase shift only applies to 110f analog inputs, so this will not occur on any sources brought in through the network, SIP, USB, etc. If you have applications where you feel this phase response is unacceptable, you could use the analog inputs of a CXDQ amplifier or IO Frame in conjunction with a core 110f, or move to a core model with the CIML4 card such as the Core 510i.
I need to grab a BSS Blu-100 and measure it now to see if it has similar issues.
Thanks for that.
Below is a CXDQ input straight to a core110f output at 96k, as suggested by QSC.
Versus the analog core input like we were previously posting; I'd call the low freq end definitely flatter, and the high end less squirrel-y.
It's not the super smooth traces we usually see from a pure speaker management processor, but certainly more than acceptable IMO.
As far as the high end phase tail wagging either up or down with a one sample delay change, I was able to confirm at Smaart class this weekend that it is just the "luck" of pure timing being between samples at 48kHz....as was speculated...