• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Final REW measurements, opinions?

Keith_W

Major Contributor
Joined
Jun 26, 2016
Messages
2,881
Likes
6,642
Location
Melbourne, Australia
Looking at the measurement variations inherent with a USB Mic https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/post-99941 and then taking multiple measurements all with random timing errors it is no surprise DIRAC struggles.

Look at the scale of the graph you linked. That's about a 100-200 microsecond variance. You report that Dirac has your sub timing off by 125ms. That is 1000x the variance of the USB mic, so I don't think the USB mic is at fault. More likely faulty correction by Dirac. 125ms is HUGE and easily audible!
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,464
Likes
3,642
Location
San Diego
Look at the scale of the graph you linked. That's about a 100-200 microsecond variance. You report that Dirac has your sub timing off by 125ms. That is 1000x the variance of the USB mic, so I don't think the USB mic is at fault. More likely faulty correction by Dirac. 125ms is HUGE and easily audible!
While the measured variations are small they look like enough that they could easily "confuse" an algorithm that is looking to figure out where the LP is in relationship to the main speakers and the sub(s) based on a handful of measurements. Add in unknown DSP delays, Pre Ringing, room reflections, etc. I could see how the algorithm could behave unexpectedly in some circumstances. Of course I really don't know for sure how it all works under the hood.

My take away, after spending money and time buying and learning to use a loopback Mic, is that a reliable way to set/ check the sub timing is with a tape measure. Once I took into account the 6 ms DSP delay of my sub I would have come up with the exact same delays with a tape measure as I did with a lot of careful acoustic measurements. Of course you need to know about any DSP delays (SVS makes no secret about it and the information is readily available) in your system but physical measurements are a very good sanity check for any delay set up whether automatically or manually calculated from acoustic measurements. I did notice that you can manually set the Sub delay in DIRAC but I have not played around with that yet.
 
Last edited:
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
Looking at the measurement variations inherent with a USB Mic https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/post-99941 and then taking multiple measurements all with random timing errors it is no surprise DIRAC struggles.
It's not my case though. I use a professional microphone (Beyerdynamic MM-1) connected to the microphone input of the Ultralite MK5, so zero clock drift involved.
However, from the way Dirac emits the sweeps I believe it calculates the timing of the measurements with the acoustic signal (in fact it also works USB microphones like UMIK), so there may be some variable here.
Although, by displaying the pulses in Dirac they correspond with those measured in REW.
I don't know ... I would like to try adding a 100 ms delay to the mains and redo the measurements, but I need a DAW with plugins and Dirac works very badly in this case... I have never been able to complete a measurement without problems. I always have to use the standalone processor.
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
I have the PC based DIRAC, not sure on any other version, see attached.
Thanks. I had forgotten the presence of those settings.
I will check how that parameter works. If that delay is applied at the measurement stage it can be a great way to verify how the algorithm calculates the correction and if high GD is still present.
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,464
Likes
3,642
Location
San Diego
Thanks. I had forgotten the presence of those settings.
I will check how that parameter works. If that delay is applied at the measurement stage it can be a great way to verify how the algorithm calculates the correction and if high GD is still present.
Let me know what you find out, I am going to try again with DIRAC when I find some time.
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
Let me know what you find out, I am going to try again with DIRAC when I find some time.
I just tried... all unchanged...
That parameter doesn't seem to have any effect.
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
Look at this.
I measured Dirac filters using Blackhole for loopback, making sure to keep the correct time reference.
As you can see the phase response at low frequencies is consistent between mains and subs, however in terms of GD it corresponds to an increasing delay as the frequency decreases.
Note also the GD at the crossover point, which in this case is 120Hz. Strangely, the main left has a delay of about 20 ms compared to the others. Perhaps this is to compensate for the room effect (?!).
 

Attachments

  • Magnitude.jpg
    Magnitude.jpg
    302.5 KB · Views: 30
  • Phase.jpg
    Phase.jpg
    287.6 KB · Views: 29
  • IR.jpg
    IR.jpg
    289.1 KB · Views: 29
  • GD.jpg
    GD.jpg
    306 KB · Views: 28
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
And this is the excess phase of raw system vs Dirac applied.
It's about 50 ms of delay introduced al low end.
I'll definitely have to try to introduce 50 ms of delay to the mains and repeat measurements / filter creation to see the result.

EXCESS PHASE.jpg
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,464
Likes
3,642
Location
San Diego
And this is the excess phase of raw system vs Dirac applied.
It's about 50 ms of delay introduced al low end.
I'll definitely have to try to introduce 50 ms of delay to the mains and repeat measurements / filter creation to see the result.

View attachment 368675
I have never looked at this type of analysis before but looking at this it appears to me that the sub(s) delay is about 50 ms "off"? How does this analysis tell the difference between "delay" and "group delay". Seems like if the sub timing was shifted 50 ms you would have what you wanted? I think using Qbase with your Motu you could delay the mains 50 ms as an experiment to see what happens. The real question is does DIRAC do this on purpose or is it a "bug"?
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
I have never looked at this type of analysis before but looking at this it appears to me that the sub(s) delay is about 50 ms "off"? How does this analysis tell the difference between "delay" and "group delay". Seems like if the sub timing was shifted 50 ms you would have what you wanted? I think using qbase with your Motu you could delay the mains 50 ms as an experiment to see what happens. The real question is does DIRAC do this on purpose or is it a "bug"?
No. If you look at the trend as the frequency increases, you see that the difference is reduced, so it is not a pure delay on the whole band.
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,464
Likes
3,642
Location
San Diego
No. If you look at the trend as the frequency increases, you see that the difference is reduced, so it is not a pure delay on the whole band.
What is the crossover point and slope for the subs? Is the crossover in place for boths sets of measurements?
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
What is the crossover point and slope for the subs? Is the crossover in place for boths sets of measurements?
Raw measurements are without crossover / filters. With DLBC the XO is set to 120Hz but don't know the slope. But if you look at my previous post there is digital loopback measurements of DLBC filters, so you can estimate it.
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
The real question is does DIRAC do this on purpose or is it a "bug"?
I believe that Dirac does not apply FIR at low frequencies, probably because when it runs on AVR it doesn't have enough taps, so it cannot correct the excess phase at low frequencies.
I do not know the mathematics of the IIR but perhaps the increase in delay is proportional to the increase in gain, which is what ultimately happens in my filter in that band.
However, by manually compensating for the delay of the mains I should get a similar response to the original one.
This does not mean having an optimal correction, because it would require a fir to compensate for the LF excess phase due to the room and reflections.
Dirac declares to use a mixed-phase correction so at some point I should find a delay value for mains that represents a good compromise.
I'll do some tests...
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,464
Likes
3,642
Location
San Diego
I believe that Dirac does not apply FIR at low frequencies, probably because when it runs on AVR it doesn't have enough taps, so it cannot correct the excess phase at low frequencies.
I do not know the mathematics of the IIR but perhaps the increase in delay is proportional to the increase in gain, which is what ultimately happens in my filter in that band.
However, by manually compensating for the delay of the mains I should get a similar response to the original one.
This does not mean having an optimal correction, because it would require a fir to compensate for the LF excess phase due to the room and reflections.
Dirac declares to use a mixed-phase correction so at some point I should find a delay value for mains that represents a good compromise.
I'll do some tests...
It does look a lot like what an IIR filter would do as far as Group Delay goes. I know you can use Rephase to create a FIR "Linerize" filter for a 120 Hz 24 dB per Octave Low Pass filter which should correct the sub Group Delay but not sure how it would work (or how you would apply it) if you applied it just to the sub output since FIR filters add latency. If you could figure out a way to both apply the FIR linearize filter to the subs and then re-adjust the sub to mains timing it may give you what you want.
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy

It does look a lot like what an IIR filter would do as far as Group Delay goes. I know you can use Rephase to create a FIR "Linerize" filter for a 120 Hz 24 dB per Octave Low Pass filter which should correct the sub Group Delay but not sure how it would work (or how you would apply it) if you applied it just to the sub output since FIR filters add latency. If you could figure out a way to both apply the FIR linearize filter to the subs and then re-adjust the sub to mains timing it may give you what you want.
I tried to manually apply a delay and redo the calibration.
I was able to solve the artifact problems using the plugin instead of the standalone processor fortunately.
So, I applied exactly 75ms of delay to mains, create the filter with same target as previous one, and then redo the REW measurements with appropriate timing ref.
The result was terrible in my opinion.
Dirac has applied 151ms of delay to subs, according to the processor indication (it is not clear to me why there is only one value if there are two subs ... each should have its own value).
Despite this, the subs were with the same GD as the old filter, while the mains showed over 220ms of delay.
Inexplicable honestly...

But not only that, I saw that by moving the crossover and putting it back to the exact same place, it is possible that a different filter is calculated.
I had reported this to technical support and they provided me with an explanation that is not very technical but according to which it is natural ...

I don't know, this bass control is starting to get tired.
I can't wait to see if ART will work properly. Its technology is unique and I hope it works.
 

RenPa

Member
Joined
Dec 7, 2022
Messages
37
Likes
11
This is a bit frustrating. In fact, every time I remake the measurements with Dirac, even with the same setup, I get a different sound. Sometimes the soundstage is wider, intelligible, other times it is sterile, collapsed. Sometimes the phantom center is present but backward, other times the sound seems to come out exactly from the cone of the speaker. Sometimes the image is shifted left or right Even if the levels are exactly the same (it will be some precedence effect). Other times the bass have a good slam, other times they are inconsistent.
I also noticed that pointing the microphone up or down slightly changes the high frequencies, especially in the reflections.
This is why I started experimenting with alternative measurement setups like the one I explained earlier.
This is my exctly same experience with DLBC and 3.4.4 firmware version.

Also here a dumb question: I use REW 5.30.8 version and can't find your IACC calculation button. Wher is it in REW?
 
OP
Davide

Davide

Addicted to Fun and Learning
Joined
Jul 6, 2020
Messages
560
Likes
215
Location
Milan, Italy
This is my exctly same experience with DLBC and 3.4.4 firmware version.

Also here a dumb question: I use REW 5.30.8 version and can't find your IACC calculation button. Wher is it in REW?
Go to ALL SPL graph, then Action button. In the window that appears there is IACC button. Choose L and R measurements and calculate.
 

RenPa

Member
Joined
Dec 7, 2022
Messages
37
Likes
11
Go to ALL SPL graph, then Action button. In the window that appears there is IACC button. Choose L and R measurements and calculate.
Sorry I don't see it. I will ask to John in an other forum if needed.
1715538500613.png
 
Top Bottom