Understanding the state of the DSP market

[dsnyder0cnn]

In almost a decade using Acourate, I've never encountered a need to build active crossovers. However, I am aware that Acourate can be used to do this.

I guess I still don't care about timing beyond what's required to take a valid end-to-end impulse response using REW.

 

[Music1969]

I guess I still don't care about timing beyond what's required to take a valid end-to-end impulse response using REW.

Well you originally wrote "we don't care about timing"

we don't care about the timing.

That was the entire reason for me explaining why some of us want to verify Acourate's simulation.

For the final time - REW's acoustic timing reference works great in this application, 15Hz to 24kHz verification measurement.

Hopefully you now understand.

 

[OCA]

Could you share some links to content where i can learn more about this method? Thank you.

Unfortunately I never made a video on that and haven't seen it shown anywhere either. It's a bit complicated but a quick and dirty trick you can try is to generate minimum phase version of the sub response, move that new response (sub-MP) on top of the original sub response in the left column (so that sub will need to shift towards its minimum phase version). MP version impulse will inevitably start at time zero and then cross corr align them. Cumulative IR shift value will show you precise IR peak time of the sub no matter how difficult the sub response is:

Now, is syncing IR peaks the best possible alignment for a sub and speaker? Often not, brute force optimization on total sound energy usually results in a little beyond IR start alignment in my experience but it's a good starting point.

 

[LionIT]

Unfortunately I never made a video on that and haven't seen it shown anywhere either. It's a bit complicated but a quick and dirty trick you can try is to generate minimum phase version of the sub response, move that (sub-MP) on top of the sub response (so that sub will need to shift towards its minimum phase version which inevitably will start at time zero) and cross corr align them. Cumulative IR shift value will show you precise IR peak time of the sub no matter how difficult the sub response is:
View attachment 447784
Now, is syncing IR peaks the best possible alignment for a sub and speaker? Often not, brute force optimization on total sound energy usually results in a little beyond IR start alignment in my experience but it's a good starting point.

Wouldn't it be more appropriate to apply an FDW before estimating the minimum phase?
In my experiments an FDW is highly effective in isolating the quasi anechoic response

 

[OCA]

Wouldn't it be more appropriate to apply an FDW before estimating the minimum phase?
In my experiments an FDW is highly effective in isolating the quasi anechoic response

FDW changes response in frequency domain, will have no effect on time domain and impulse delays.

 

[LionIT]

FDW changes response in frequency domain, will have no effect on time domain and impulse delays.

You're right, but quoting REW guide:

The minimum phase response is generated by using the measurement amplitude and calculating the corresponding minimum phase from it, using a mathematical relationship between the two that holds for minimum phase systems.

Generate minimum phase will produce a minimum phase version of the measurement using the current IR window settings. The minimum phase impulse then shows the response of a system having the same frequency response as the measurement but with the lowest phase shift such a system could have. Note that it is best to make full range measurements if the minimum phase response is to be generated as a good result relies on measuring beyond the bandwidth of the system being measured. This control also activates minimum and excess phase and group delay traces on the SPL & Phase and GD graphs respectively.

Note that the IR window settings are important as the minimum phase response is derived from the frequency (magnitude) response of the measurement, which in turn is affected by the IR window settings. If the window settings are subsequently changed Generate minimum phase should be used again to reflect the new settings. Note also that the shape of the left side window (the window applied before the peak) affects the minimum phase result, a rectangular window will produce a response with lower phase shift than, for example, a Hann window.

If the system being measured was inherently minimum phase (as most crossovers are, for example) the minimum phase response is the same as removing any time delay from the measurement. Room measurements are typically not minimum phase except in some regions, mainly at low frequencies. For more about minimum and excess phase and group delay see Minimum Phase.

So, windowing is applied before, and applying an FDW should help calculate the minimum phase on a cleaner IR.
Then the excess phase have to be calculated with respect to the full IR windowing.
And the excess phase is particularly indicative of the alignment between drivers, at least where it has a linear trend.

As an aside, the excess group delay plot also clearly shows there is a time offset between the subwoofer and the main speaker, the sub being about 25ms delayed, which is not so obvious from the overall group delay plot. Excess group delay is a useful plot for time aligning speakers.

 

[OCA]

You're right, but quoting REW guide:


So, windowing is applied before, and applying an FDW should help calculate the minimum phase on a cleaner IR.
Then the excess phase have to be calculated with respect to the full IR windowing.
And the excess phase is particularly indicative of the alignment between drivers.

You are right, MTW response produces even cleaner MP response. I hadn't tried that before. Thanks.

 

[LionIT]

You are right, MTW response produces even cleaner MP response. I hadn't tried that before. Thanks.

View attachment 447860

Thanks to John Mulcahy for writing REW guide

 

[LionIT]

I think I have the settings right so lets try and see:

Distances while measuring left speaker and mic positioned as follows:

- Mic 1.235 m from Right tweeter
- Mic 100 cm from left mid upper ring
- Mic 125 cm from left port

Results:

View attachment 447609
other speaker ac. ref


View attachment 447607
same speaker ac ref


View attachment 447608
both

Now, REW reports -1.340 m for the "other" speaker ac ref which is close to the 1.235 m. No EQ or any DSP so no delay there, the electronics should be some negligible some.
Is there a chance to calculate by triangulation?

What's interesting is that both generated minimum phase are referring to 0 despite all other traces.

In reality the post wasn't directly referring to you. I was actually suggesting to @Keith_W some ideas regarding the anomalies he detected, since he exposed his doubts about REW.
In my experience REW is pretty accurate, even if it requires the appropriate precautions and knowledge to obtain accurate results.

Anyway, for me, the calculation of excess phase with respect to the minimum phase generated with appropriate FDW (I use the spectrogram to establish it) is an extremely effective method to set/verify the alignment. Clearly then the precise alignment is to be set/verified with unwrapped phase graph.
This obviously does not totally make up for the non-linearity induced by the room (very early reflections) and the measurement system (the microphone is not perfectly linear phase, unless it is calibrated and compensated for that) ... but it is already good enough (if not perhaps the best that can be done at home).
To be kept in mind that alignment is valid for only one infinitesimal point, and it is not even certain that the signal (music) has extreme fidelity ... so too much precision doesn't make much sense.

 

[levimax]

Anyway, for me, the calculation of excess phase with respect to the minimum phase generated with appropriate FDW (I use the spectrogram to establish it) is an extremely effective method to set/verify the alignment. Clearly then the precise alignment is to be set/verified with unwrapped phase graph.
This obviously does not totally make up for the non-linearity induced by the room (very early reflections) and the measurement system (the microphone is not perfectly linear phase, unless it is calibrated and compensated for that) ... but it is already good enough (if not perhaps the best that can be done at home).
To be kept in mind that alignment is valid for only one infinitesimal point, and it is not even certain that the signal (music) has extreme fidelity ... so too much precision doesn't make much sense.

Rather than trying to measure at the LP and apply windows to an already relatively "low res" measurements at LF why not measure NF to get the system delays and then use a measuring tape (or laser pointer) to determine the distance difference to the LP and combine the two to calculate the relative delays? I am not saying this is right rather asking you your opinion. Thanks.

 

[LionIT]

Rather than trying to measure at the LP and apply windows to an already relatively "low res" measurements at LF why not measure NF to get the system delays and then use a measuring tape (or laser pointer) to determine the distance difference to the LP and combine the two to calculate the relative delays? I am not saying this is right rather asking you your opinion. Thanks.

Geometrically you should put the mic on the bisector between LP and the center of the two drivers, but acoustically there are too much non-linearities at stake (directivity, shape of wave front, acoustic centroid) to detect something sufficiently valid to translate at LP.
Alternatively you should put the mic exactly on the axis between LP and the center of the driver (each) and measure exactly the distance between there and LP. But in practice you would not have a better accuracy in alignment than the LP measurement (with aforementioned calculations).
If there is any reason why it can be more advantageous, honestly it escapes me...
I understand that by measuring NF you have a better ratio between direct and indirect sound, but in any case if you are doing room correction together with the alignment you want to know the actual phase response at LP to correct the excess phase where possible / make sense (*) and calculate alignment on the corrected result (still talking about sub-main integration... for higher frequencies IR + phase alignment, without excess phase correction, should do the job).
With a single measure at LP you can do everything together then.
* Not a rule, someone disagrees, but in my experience the bass sounds much better.

 

[gnarly]

Wouldn't it be more appropriate to apply an FDW before estimating the minimum phase?

Definitely. It's often the only way to get meaningful data when strong reflections abound.

 

[gnarly]

I see. So once the acoustic centre of the subwoofer has been determined, the piezo tweeter becomes the substitute. Thanks for the explanation!

Glad it made sense !

From what I'm told, the piezo trick was one of the first things live sound guys thought of, when FFT measurements started being used to set up systems in the field.
Laser finders weren't around, and the piezo made for a precise distance finder. Beat the hell out of a surveyor's tape.

Some subs, horn-loaded for example, have horn paths almost 3m long from the front of the cabinet to the acoustic center. It's nice to be able to quickly assess distance to acoustic center...without a lot of on site sweeps or timed pink.

What's interesting to me, and testimony to the difficulty of locating subs....sub-to-main integration is still by far the #1 subject asked about in prosound measurement training classes I've attended..

 

[AlmaAtaKZ]

@Keith_W
Learning a lot about DSP in this thread. Thank you!

How would you evaluate the DSP capabilities of Hypex Fusion amps?

 

[Keith_W]

@Keith_W
Learning a lot about DSP in this thread. Thank you!

I don't deserve all the credit (apart from starting the thread), I think the ASR community also deserves thanks. Plenty of knowledgeable folk here.

How would you evaluate the DSP capabilities of Hypex Fusion amps?

According to the manual:

15 biquads per channel for a plate amp intended for subwoofers is plenty. You will use maybe 1-2 biquads for the low-pass, which will still leave you plenty of biquads. I did not do enough reading to find out how you program the biquads. Remember the warning I gave in my first post: biquads have poor portability. If you want to use REW (for example), then REW needs to be able to specifically export biquads for Hypex Fusion.

HOWEVER, (IMO!!!) it is not a good idea to implement DSP on subwoofers only. This is because:

1. Subs are usually delayed with respect to the mains. The correction involves delaying the mains to match the subs. However, DSP on the subs adds additional subwoofer latency, which worsens the alignment between mains and subs.

2. you want to use DSP to tie subs to the mains, which means that both the mains and subs should be DSP'ed together. Of course you can have one DSP unit for the sub, and another DSP unit for the mains, and then somehow program these two independent units to work together. But this isn't going to be easy, and it will drive you nuts trying to figure it out. Some people enjoy the challenge, but I am not that much of a masochist. It is better to have a single DSP upstream of both speakers that can control them together.

My personal opinion: I would rather have a "dumb" subwoofer. The only features that subwoofers need to have is a signal detecting auto on/off switch and a volume control. And of course, size, styling, max output, distortion, etc. + all the usual considerations. I don't need anything else, I will control it all from DSP.

Pros love this kind of solution though. There are many amps on the pro market that are DSP enabled. They distribute the DSP load over multiple amps and speakers, then use another layer of DSP to control it all. For example, Powersoft makes an 8 channel amp, and each channel has its own DSP with biquads. You could use this amp to correct a single speaker. You then deploy maybe a dozen of these things, each speaker corrected to be "perfect". Then you tie it in with another DSP layer that makes them play together.

 

[OCA]

Thanks to John Mulcahy for writing REW guide

Live now:

 

[SoySalsaPicante]

I don't deserve all the credit (apart from starting the thread), I think the ASR community also deserves thanks. Plenty of knowledgeable folk here....

...HOWEVER, (IMO!!!) it is not a good idea to implement DSP on subwoofers only. This is because:

1. Subs are usually delayed with respect to the mains. The correction involves delaying the mains to match the subs. However, DSP on the subs adds additional subwoofer latency, which worsens the alignment between mains and subs.

2. you want to use DSP to tie subs to the mains, which means that both the mains and subs should be DSP'ed together. Of course you can have one DSP unit for the sub, and another DSP unit for the mains, and then somehow program these two independent units to work together. But this isn't going to be easy, and it will drive you nuts trying to figure it out. Some people enjoy the challenge, but I am not that much of a masochist. It is better to have a single DSP upstream of both speakers that can control them together.

My personal opinion: I would rather have a "dumb" subwoofer. The only features that subwoofers need to have is a signal detecting auto on/off switch and a volume control. And of course, size, styling, max output, distortion, etc. + all the usual considerations. I don't need anything else, I will control it all from DSP.

Pros love this kind of solution though. There are many amps on the pro market that are DSP enabled. They distribute the DSP load over multiple amps and speakers, then use another layer of DSP to control it all. For example, Powersoft makes an 8 channel amp, and each channel has its own DSP with biquads. You could use this amp to correct a single speaker. You then deploy maybe a dozen of these things, each speaker corrected to be "perfect". Then you tie it in with another DSP layer that makes them play together.

Reading this now...

In my current arrangement, I have one DSPeaker Anti-Mode X2D correcting my speakers. I have a second DSPeaker Anti-Mode 8033S-II correcting my subwoofer. Finally I am using a WiiM Ultra streamer's bass management to time-align the output of the sub to the speakers. I am not using the PEQ/room correction of the WiiM at all. I've set the crossover point and the volume level of the sub versus the speaker in such a way as to closely approximate the overall frequency response of a Harman curve. Due to the placement of the subwoofer, there does not appear to be an ideal phase setting for the subwoofer (the sub's DSP can adjust phase in single-degree increments anywhere from 0-180, separately from a full polarity switch), because the phase relation of the sup to the left speaker is never in sync with the phase relation to the right speaker (sub is farther left in the room than the left speaker--I cannot place the sub dead between them, the media rack is there). Some advice on best way to determine sub phase would be helpful here.

Does this seem like a reasonable setup, or is there any major point of flaw you can see in this arrangement?

 

[PristineSound]

Dense content, not something you can just read on your phone while watching TV.

Dammit, another thing to add to the long list of reading. This place just pulls you into the rabbit hole deeper and deeper.

 

[ernestcarl]

Reading this now...

In my current arrangement, I have one DSPeaker Anti-Mode X2D correcting my speakers. I have a second DSPeaker Anti-Mode 8033S-II correcting my subwoofer. Finally I am using a WiiM Ultra streamer's bass management to time-align the output of the sub to the speakers. I am not using the PEQ/room correction of the WiiM at all. I've set the crossover point and the volume level of the sub versus the speaker in such a way as to closely approximate the overall frequency response of a Harman curve. Due to the placement of the subwoofer, there does not appear to be an ideal phase setting for the subwoofer (the sub's DSP can adjust phase in single-degree increments anywhere from 0-180, separately from a full polarity switch), because the phase relation of the sup to the left speaker is never in sync with the phase relation to the right speaker (sub is farther left in the room than the left speaker--I cannot place the sub dead between them, the media rack is there). Some advice on best way to determine sub phase would be helpful here.

Does this seem like a reasonable setup, or is there any major point of flaw you can see in this arrangement?

Measurements are needed to “see” the phase… it’s also better to have the ability to EQ each speaker and sub output channel individually.

 

[SoySalsaPicante]

Measurements are needed to “see” the phase… it’s also better to have the ability to EQ each speaker and sub output channel individually.

The Anti-Mode X2D is correcting the speakers in 2.0 mode, and the Anti-Mode 8033S-II is separately correcting the subwoofer, so effectively all channels are already being, “individually EQ’d.” I’m just not using the WiiM’s software to do it.