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: David Sturzenbecher on August 18, 2013, 10:18:53 AM
-
Hello,
I am in the process of pulling some JBL presets from a DBX4800 and creating ItechHD presets from the transfer functions. When I took the TF of the DBX4800 I get alot of ripple in the HF. My resolution is quite fine, and when smoothed even a little it goes away, but I am wondering if anyone can take a TF of the 4800 and see if you see anything similar. Setup was just a Octapre sending pink noise into the 4800, and measuring the output from the individual channels.
-
Hello,
I am in the process of pulling some JBL presets from a DBX4800 and creating ItechHD presets from the transfer functions. When I took the TF of the DBX4800 I get alot of ripple in the HF. My resolution is quite fine, and when smoothed even a little it goes away, but I am wondering if anyone can take a TF of the 4800 and see if you see anything similar. Setup was just a Octapre sending pink noise into the 4800, and measuring the output from the individual channels.
Are the two signals time aligned? Since the 4800 probably adds some propagation delay the result in the TF will be comb filtering, which I suspect is what your measurement shows.
-
I don't know why the two signals would need to be time aligned. They don't ever get combined, so there would be no way for them to comb. I took the exact setup over to the ITechHD with a VI box and 8ohm resistor and there was no ripple.
-
I don't know why the two signals would need to be time aligned. They don't ever get combined, so there would be no way for them to comb. I took the exact setup over to the ITechHD with a VI box and 8ohm resistor and there was no ripple.
If you don't sync your measuresignal to your referencesignal there is no way for the software to display a accurate line on screen. Since there is no ripple in LF or MF the delay to be set is probabbly pretty small (0.xxms)
-
Can verify if it's a comb filter by looking at the phase. A comb filter should also be visible with an RTA, albeit perhaps not quite so obvious. A comb filter is created when there are two arrivals that differ in time. How deep the troughs are will vary based on how much the signals vary in level.
I'd first verify that the routing and processor settings are correct.
-
Can verify if it's a comb filter by looking at the phase. A comb filter should also be visible with an RTA, albeit perhaps not quite so obvious. A comb filter is created when there are two arrivals that differ in time. How deep the troughs are will vary based on how much the signals vary in level.
I'd first verify that the routing and processor settings are correct.
Attached is a picture with phase. I suppose it could be crosstalk or bleed in the Octacapture, but I am not sure why it wouldn't be visible when the octacapture was moved to the ItechHD.
-
Attached is a picture with phase. I suppose it could be crosstalk or bleed in the Octacapture, but I am not sure why it wouldn't be visible when the octacapture was moved to the ItechHD.
Doesn't look like a comb filter to me. The Octacapture is quite configurable, I'd first make sure it's setup correctly. If you have another device, I'd test the amp using it as well. Anything to rule out pieces of the signal chain.
-
The phase display shows a time-difference, so you might need to check the measurement delay.
-
Attached is a picture with phase. I suppose it could be crosstalk or bleed in the Octacapture, but I am not sure why it wouldn't be visible when the octacapture was moved to the ItechHD.
As Malcolm said, there is definitely a time offset between the reference signal and the DUT signal. When you are measuring a piece of electronics the phase plot should be mostly flat across the whole band. With no filters the phase plot should be completely flat, but with the band pass and parametric filters you have you will not have a flat phase curve, but it should not look like what you have.
Mac
-
As Malcolm said, there is definitely a time offset between the reference signal and the DUT signal. When you are measuring a piece of electronics the phase plot should be mostly flat across the whole band. With no filters the phase plot should be completely flat, but with the band pass and parametric filters you have you will not have a flat phase curve, but it should not look like what you have.
Mac
The idea was not to measure the 4800, but rather the preset inside the 4800, so i really wanted to capture the phase slope to make sure my signal delay would be consistent in the new device. My measurements are not time windowed so I was under the impression that there would be no loss of data.
My concern is that if I add delay for each bandpass individually for the reference measurement to get a flatter phase it changes the time relationship between the band passes. If i remember correctly from when I took the measurements I did set the delay to the Low frequency transfer function, then left it in place for the HF transfer functions.
Thanks for the input everyone, sometime in the near future I should have access to this 4800 as well as some others, plus I also use a USBPre2 that I can retake the measurements again.
-
My concern is that if I add delay for each bandpass individually for the reference measurement to get a flatter phase it changes the time relationship between the band passes. If i remember correctly from when I took the measurements I did set the delay to the Low frequency transfer function, then left it in place for the HF transfer functions.
No it doesn't. It is only changing the measurement time so you get the correct phase display. You will then be able to see where the 2 phase plots are through the crossover, and if you need to change it, you add or subtract delay in the device without resetting the time in your measurement software.
Mac
-
No it doesn't. It is only changing the measurement time so you get the correct phase display. You will then be able to see where the 2 phase plots are through the crossover, and if you need to change it, you add or subtract delay in the device without resetting the time in your measurement software.
Mac
What Mac said. Also:
As someone else said, the phase response of an electronic measurement should be dead flat, except in the HF due to the sampling rate not offering a delay value that will completely flatten it. You can get closer by using the highest sample rate available.
And, it doesn't look like a comb filter to me. There is no complete cancellation at 1/2 f, with subsequent notches every f Hz above the first cancellation.
Edit: of course the "complete cancellation" occurs only when the misaligned signals are equal in level. Anything else changes the way it looks in a measurement. Still, it doesn't look like a comb filter to me.
Edit 2: and since it is not combining with anything, how could it be a comb filter? It may be a jiggy 4800.
-
Are the two signals time aligned? Since the 4800 probably adds some propagation delay the result in the TF will be comb filtering, which I suspect is what your measurement shows.
Not accounting for device latency would obviously affect the phase response, but I would not expect it to affect the magnitude response, at least in SysTune.
-
Not accounting for device latency would obviously affect the phase response, but I would not expect it to affect the magnitude response, at least in SysTune.
It should not change the magnitude response in Smaart either.
Mac
-
If i understand correcty you want to copy the presets into your own unlocked preset file? first of i would start with measuring the device.
1. Use a flat preset (no changes to gain/delay/polarity/eq/phase/x-overs) and store a snapshot of it as reference
2. Open the preset you want to copy and write down the propogational delay for each of the outputs (from input to output of your fft I/O)
3. Now use the transfer function to store snapshots of each output and make sure you set the measurement delay correctly.
4. Create a new preset and start changing stuff output by output until they matchup with what you see in the previously recorded snapshots. (i usually start with gain, then delay, then x-overs, then eq, then allpass and tweak the bunch untill it's good.
For a transfer function to be of any use you really need to align the measurement signal to the reference signal, otherwise the phase display just isn't usable and you are missing vital info.
I hope this helps :)