Re: Well, THAT's Not Gonna Fucking Work!
Well ain't I just butchering the English language. D'oh. Typing two fast.
Well ain't I just butchering the English language. D'oh. Typing two fast.
Well, THAT's Not Gonna Fucking Work!
- Thread starter Bennett Prescott
- Start date
Mark DeArman
Junior
Re: Well, THAT's Not Gonna Fucking Work!
Here are the loopback results for the Lock-In amplifier. Granted I've never tried to measure frequencies this low with it, but it's pretty spot on. I also included images of the simple test setup.
(A) The Driverack 480 sitting on carts. It is plugged into a isolated power supply.
(B) NI Chassis with Lock-In amplifier setup.
(C) Closeup of panel configuration. PXI-6653 is connected internally with equal length paths to both cards. Signal output from the generator is T'd to AI0 of the PXI-4462. Software performs the Hilbert stuff.
(D) Disconnected signal conditioner normally used for acoustic measurements. But not in the signal chain.
Here are the loopback results for the Lock-In amplifier. Granted I've never tried to measure frequencies this low with it, but it's pretty spot on. I also included images of the simple test setup.
(A) The Driverack 480 sitting on carts. It is plugged into a isolated power supply.
(B) NI Chassis with Lock-In amplifier setup.
(C) Closeup of panel configuration. PXI-6653 is connected internally with equal length paths to both cards. Signal output from the generator is T'd to AI0 of the PXI-4462. Software performs the Hilbert stuff.
(D) Disconnected signal conditioner normally used for acoustic measurements. But not in the signal chain.
Attachments
Re: Well, THAT's Not Gonna Fucking Work!
Hi Mark,
FWIW, for some reason I think that some the DBX stuff is similar to the BSS … anyway this is what the BSS omini drive compact manual says regarding phase....
A classic 2-Way crossover using just a high-pass and a low-pass filter will
always meet the criteria predicted for a given alignment. A 2-Way Linkwitz-
Riley crossover ,for example, will produce its two outputs 'everywhere in
phase', and they will acoustically combine to a flat amplitude response. The
main lobe of the polar response will also be on-axis.
When more filters are added to the crossover for more than two drivers,
however, the crossover begins to depart from the mathematical perfection of
the 2-Way case. In a 3-Way crossover, the high-pass characteristics of the mid
band will be disturbed by the phase response of the low-pass in the mid band,
and vice-versa. This results in drivers of adjacent bands being driven out of
phase, producing irregularities in the amplitude response, and pushing the
main lobe of the polar response off-axis, further aggravating amplitude
response problems in some listening positions. Although these effects may be
subtle when the crossover frequencies are well separated, 4 and 5-Way
systems can produce significant errors.
The phase compensation scheme employed in the FDS-355 analyses these
phase anomalies whenever adjustments are made, and introduces phase
adjustment into certain bands such that the phase difference between all
adjacent bands is always close to zero degrees. It will of course allow the user
to introduce intentional phase differences, using the phase and delay
parameters. The FDS-355 will not attempt to apply phase compensation if the
high and low frequencies or shapes of the adjacent bands do not match, on
the assumption that the user does not expect to produce a standard alignment.
Peter
Hi Mark,
FWIW, for some reason I think that some the DBX stuff is similar to the BSS … anyway this is what the BSS omini drive compact manual says regarding phase....
A classic 2-Way crossover using just a high-pass and a low-pass filter will
always meet the criteria predicted for a given alignment. A 2-Way Linkwitz-
Riley crossover ,for example, will produce its two outputs 'everywhere in
phase', and they will acoustically combine to a flat amplitude response. The
main lobe of the polar response will also be on-axis.
When more filters are added to the crossover for more than two drivers,
however, the crossover begins to depart from the mathematical perfection of
the 2-Way case. In a 3-Way crossover, the high-pass characteristics of the mid
band will be disturbed by the phase response of the low-pass in the mid band,
and vice-versa. This results in drivers of adjacent bands being driven out of
phase, producing irregularities in the amplitude response, and pushing the
main lobe of the polar response off-axis, further aggravating amplitude
response problems in some listening positions. Although these effects may be
subtle when the crossover frequencies are well separated, 4 and 5-Way
systems can produce significant errors.
The phase compensation scheme employed in the FDS-355 analyses these
phase anomalies whenever adjustments are made, and introduces phase
adjustment into certain bands such that the phase difference between all
adjacent bands is always close to zero degrees. It will of course allow the user
to introduce intentional phase differences, using the phase and delay
parameters. The FDS-355 will not attempt to apply phase compensation if the
high and low frequencies or shapes of the adjacent bands do not match, on
the assumption that the user does not expect to produce a standard alignment.
Peter
Last edited:
Re: Well, THAT's Not Gonna Fucking Work!
I thought that study sounded familiar... from the Wiki page on the study / paper:
"Awards
Dunning and Kruger were awarded the 2000 Ig Nobel Prize in Psychology for their report, "Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments".[11]"
Just last week I handled house audio for the 21st First Annual Ig Nobel Awards at Harvard's Sanders Theater.
Here is a summary: http://www.youtube.com/watch?v=g6L6gUqdb4I
The whole thing should be up on YouTube soon. Hysterically fun annual gig. One of my favorites.
I thought that study sounded familiar... from the Wiki page on the study / paper:
"Awards
Dunning and Kruger were awarded the 2000 Ig Nobel Prize in Psychology for their report, "Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments".[11]"
Just last week I handled house audio for the 21st First Annual Ig Nobel Awards at Harvard's Sanders Theater.
Here is a summary: http://www.youtube.com/watch?v=g6L6gUqdb4I
The whole thing should be up on YouTube soon. Hysterically fun annual gig. One of my favorites.
Peter,
The problem with all of thy is, while it may be very clever, with modern DSP there is absolutely no reason to use a standard alignment, and freely available delay removes most of the concern regarding non-L/R or unmatched filter types and slopes. Very few loudspeakers have acoustic response that works well with a standard alignment.
The problem with all of thy is, while it may be very clever, with modern DSP there is absolutely no reason to use a standard alignment, and freely available delay removes most of the concern regarding non-L/R or unmatched filter types and slopes. Very few loudspeakers have acoustic response that works well with a standard alignment.
Re: Well, THAT's Not Gonna Fucking Work!
Yes I understand all of that …and some … the reason for the post was to note that BSS deliberately modifies the phase response in some situations, not every DSP/ crossover will do that.
Peter
Yes I understand all of that …and some … the reason for the post was to note that BSS deliberately modifies the phase response in some situations, not every DSP/ crossover will do that.
Peter
Re: Well, THAT's Not Gonna Fucking Work!
Peter,
What the BSS is doing is NOTHING like the effects seen here in the DBX. For a LR, in the stopband (i.e. well away from the XO knee) the phase asymptotically approaches a given value that depends on the order of the LR filter. The slope of this asymptote is essentially constant, which infers that the group delay in the stopband is also constant. If the group delay in the stopband is constant, then the phase difference induced by the XO filter that causes the constant group delay can easily be accounted for by a fixed delay on the next lower passband of the loudspeaker.
BSS is speaking of circumstances where the next XO point occurs before the crossover is fully in the stopband. Here the group delay is not constant, as the phase rotation is still in process. That is where their phase compensation kicks in. I've included a picture representing this below. The filter response is lifted from Charlie Hughes' excellent website.
Peter,
What the BSS is doing is NOTHING like the effects seen here in the DBX. For a LR, in the stopband (i.e. well away from the XO knee) the phase asymptotically approaches a given value that depends on the order of the LR filter. The slope of this asymptote is essentially constant, which infers that the group delay in the stopband is also constant. If the group delay in the stopband is constant, then the phase difference induced by the XO filter that causes the constant group delay can easily be accounted for by a fixed delay on the next lower passband of the loudspeaker.
BSS is speaking of circumstances where the next XO point occurs before the crossover is fully in the stopband. Here the group delay is not constant, as the phase rotation is still in process. That is where their phase compensation kicks in. I've included a picture representing this below. The filter response is lifted from Charlie Hughes' excellent website.
Last edited:
Re: Well, THAT's Not Gonna Fucking Work!
My post said - For What It’s Worth (FWIW), no more – I didn’t say this was the cause. I just pointed out to Mark something that was a little different.
And for FWIW here is an example of my work understanding phase and that stuff … It’s a 3 way mid high box
Yes Phil I understand all of that … but still no one has managed to explain what happening… and its unusual. The options are – firmware fault, hardware fault, measurement error or a design problem.
My post said - For What It’s Worth (FWIW), no more – I didn’t say this was the cause. I just pointed out to Mark something that was a little different.
And for FWIW here is an example of my work understanding phase and that stuff … It’s a 3 way mid high box
Attachments
Re: Well, THAT's Not Gonna Fucking Work!
Peter,
My apologies, your posting was far enough out of the topic at hand, that I jumped on it. I think this a design issue with the DBX product(s), or perhaps their choice of biquad implementation, but I'm not really sure.
PS, I know that you grok phase, but a real man wouldn't use 9pt smoothing when he displays his measurement
~:-D~:grin:
Peter,
My apologies, your posting was far enough out of the topic at hand, that I jumped on it. I think this a design issue with the DBX product(s), or perhaps their choice of biquad implementation, but I'm not really sure.
PS, I know that you grok phase, but a real man wouldn't use 9pt smoothing when he displays his measurement
~:-D~:grin:Re: Well, THAT's Not Gonna Fucking Work!
How this then :razz:
Peter,
My apologies, your posting was far enough out of the topic at hand, that I jumped on it. I think this a design issue with the DBX product(s), or perhaps their choice of biquad implementation, but I'm not really sure.
PS, I know that you grok phase, but a real man wouldn't use 9pt smoothing when he displays his measurement
How this then :razz:
Attachments
Re: Well, THAT's Not Gonna Fucking Work!
Ha!
Its great for reminding me that Renkus never called me back to fix my copy of Systune, which has forgotten its licensing information, again.
I'll have to call them when they open over in Cali.
Ha!
Its great for reminding me that Renkus never called me back to fix my copy of Systune, which has forgotten its licensing information, again.
I'll have to call them when they open over in Cali.
Re: Well, THAT's Not Gonna Fucking Work!
Another senseless flat-view tragedy.....
If I hadn't followed this chronologically, I'd have zero idea of who/what you replied to...
Another senseless flat-view tragedy.....
If I hadn't followed this chronologically, I'd have zero idea of who/what you replied to...
Re: Well, THAT's Not Gonna Fucking Work!
Peter - did you make, ahem, scale for those measurements?
Anyway...to augment Mark's measurements I measured my 480 using Smaart 7 again. The results are interesting:
Changing the high pass and low pass topologies only changes the phase response of the former. The 260 was whacked on both ends if you recall. The two polarity inversions are for 12 dB and 36 dB Linkwitz-Riley filters. I can't say that I've seen this happen with any other processor. You might also note that the 480 - at least my sample - is 3 to 6 dB down at 20 Hz.
Peter - did you make, ahem, scale for those measurements?
Anyway...to augment Mark's measurements I measured my 480 using Smaart 7 again. The results are interesting:
Changing the high pass and low pass topologies only changes the phase response of the former. The 260 was whacked on both ends if you recall. The two polarity inversions are for 12 dB and 36 dB Linkwitz-Riley filters. I can't say that I've seen this happen with any other processor. You might also note that the 480 - at least my sample - is 3 to 6 dB down at 20 Hz.
Mark DeArman
Junior
Re: Well, THAT's Not Gonna Fucking Work!
I guess the point I was trying to make with all the measurements was this behavior is not unexplained. I also measured the Xilica, Ashley, and BSS and they all do the same thing in the LF. This is because they are all AC coupled DSPs. The difference with the others compared to the Drive Rack 480 is that the HP filter phase response crosses zero deg before 20 Hz. All of my AC coupled Endevco Microphone preamps have a cutoff of 0.5Hz. For the dbx products they obviously chose 10Hz.
Now the fact that the phase response changes with the filter sections is obvious. The slope should increase as the order gets larger, even when it is marked as OUT. This seems to be true from the other posts but I did in fact not measure it myself.
The only thing I have not been able to produce is Bennett's original data showing the HF phase rising as if there was a notch filter near by. I haven't looked into this very much. Like I said previously it could be the RF choke, part of the AA filter at 24kHz, etc.
I guess the point I was trying to make with all the measurements was this behavior is not unexplained. I also measured the Xilica, Ashley, and BSS and they all do the same thing in the LF. This is because they are all AC coupled DSPs. The difference with the others compared to the Drive Rack 480 is that the HP filter phase response crosses zero deg before 20 Hz. All of my AC coupled Endevco Microphone preamps have a cutoff of 0.5Hz. For the dbx products they obviously chose 10Hz.
Now the fact that the phase response changes with the filter sections is obvious. The slope should increase as the order gets larger, even when it is marked as OUT. This seems to be true from the other posts but I did in fact not measure it myself.
The only thing I have not been able to produce is Bennett's original data showing the HF phase rising as if there was a notch filter near by. I haven't looked into this very much. Like I said previously it could be the RF choke, part of the AA filter at 24kHz, etc.
Re: Well, THAT's Not Gonna Fucking Work!
If you are looking at the HP output of a 12 or 36 dB LR filter, by definition the output is inverted
~~:smile:
But I’m still a little lost, maybe there is phase compensation like I mentioned above for the BSS and its playing a trick on us. Its set up so that you can flip between 12,24,36,&48 dB LR and not have to change anything other than the filter order … just a thought... maybe I'm just reading the colours on your graph wrong.
Peter
If you are looking at the HP output of a 12 or 36 dB LR filter, by definition the output is inverted
~~:smile:But I’m still a little lost, maybe there is phase compensation like I mentioned above for the BSS and its playing a trick on us. Its set up so that you can flip between 12,24,36,&48 dB LR and not have to change anything other than the filter order … just a thought... maybe I'm just reading the colours on your graph wrong.
Peter
Last edited:
A fault of Tapatalk's inability to let me reply to a post without quoting it. Trying to moderate and read on my phone on the left coast, so you'll have to just put up with it!
Mark DeArman
Junior
Re: Well, THAT's Not Gonna Fucking Work!
Well I reproduced the phase response in the HF now. The error comes from a one sample difference of the t0 to the impulse response of the HF LP filter type response of a DSP. I'll produce the results here in the forum later on. This is what I was wondering about originally. The t0 choice can dramatically effect results with SMAART or when other measurements which determine the frequency response from the FFT of the derived impulse response. I would like to also point out that I will make available all algorithms and programs which I used to make my measurements if requested. I can't pass out the swept sine programs but the algorithms I will give away for free. In fact they are on www.cascadeacoustic.com in the tutorial section. The Lock-In Amp program from Feldman I will pass on if anyone wants.
P.S. I'm not trying to sell any product to the sound reinforcement industry. I don't want anyone to think my website plug is against SMAART or SysTune. Course I'll bet on the SysTune measurements all day above others.
I'm still trying to get a SMAART system down here... duh this is taking like a week... too long. We have five licenses.
Well I reproduced the phase response in the HF now. The error comes from a one sample difference of the t0 to the impulse response of the HF LP filter type response of a DSP. I'll produce the results here in the forum later on. This is what I was wondering about originally. The t0 choice can dramatically effect results with SMAART or when other measurements which determine the frequency response from the FFT of the derived impulse response. I would like to also point out that I will make available all algorithms and programs which I used to make my measurements if requested. I can't pass out the swept sine programs but the algorithms I will give away for free. In fact they are on www.cascadeacoustic.com in the tutorial section. The Lock-In Amp program from Feldman I will pass on if anyone wants.
P.S. I'm not trying to sell any product to the sound reinforcement industry. I don't want anyone to think my website plug is against SMAART or SysTune. Course I'll bet on the SysTune measurements all day above others.
I'm still trying to get a SMAART system down here... duh this is taking like a week... too long. We have five licenses.
Last edited:
Re: Well, THAT's Not Gonna Fucking Work!
Mark,
I said this many posts back, this is not news. I even directly replied to this effect in my last reply to you. SMAART's reference delay granularity almost always make the phase at Nyquist lead or lag on either side of a single sample offset of the reference delay. :roll:
Mark,
I said this many posts back, this is not news. I even directly replied to this effect in my last reply to you. SMAART's reference delay granularity almost always make the phase at Nyquist lead or lag on either side of a single sample offset of the reference delay. :roll:
Last edited: