Understanding the state of the DSP market

[gnarly]

I began using wavelets for sub-to-main alignment after starting this thread on DIY 4 years ago...

Neat way to find delay, phase, and polarity at xover

Wavelets....as per this article.. Signal Aligning Using Wavelets Surprisingly repeatable measurements for low frequencies, that normally drive us crazy trying to nail down timing/phase at xover. Works great with Arta...just like the article says :)

www.diyaudio.com

At that time the SynAudCon article linked was showing how to use wavelets with ARTA, which was great and worked great.
Pat Brown's newer wavelet series JohnPM linked, that uses the REW scope, is even clearer and better imo. Nice job SynAudCon (as usual !!!
(Hey John, if you're reading....back at the time of the old thread above, I asked you on the REW forum if saving scope traces might become an option...and I would like to repeat that request )

I still haven't found anything more repeatable & precise than wavelets....

Anyway, as we've talked about the difficulty of measurement subs indoors, I see in my old thread I ran a comparison of Smaart's delay finder, REW's delay estimator, and wavelets using Aarta. Idea was to check for the variability in each, over successive measurements.
Here's the chart of measured times over 12 trials each:

The Post link for any interested....https://www.diyaudio.com/community/threads/neat-way-to-find-delay-phase-and-polarity-at-xover.370287/#post-6599413

 

[gnarly]

It will not be the IR of the driver.

For multi-ways (particularly using linear-phase xovers)

Do I need the full IR of the raw sub driver? Can't I just use the IR of the intended passband, or slightly extended passband?
I mean, after I put xover and filters in place to fully establish the pass-band, why am I interested in anything other than the final processed pass-band impulse?
thx

 

[LionIT]

For multi-ways (particularly using linear-phase xovers)

Do I need the full IR of the raw sub driver? Can't I just use the IR of the intended passband, or slightly extended passband?
I mean, after I put xover and filters in place to fully establish the pass-band, why am I interested in anything other than the final processed pass-band impulse?
thx

From my experience, I believe that there is no uniquely valid answer. To make the time alignment you need to detect the correct difference between wavefront arrival time (as well as equalizing the phase response at that point). Higher frequency content determines a clearer wavefront arrival, so it can be more accurate in time domain. But, in broader bandwidth there will also be more risk of including phase deviation due to the room, which can distort the measurement.
IMO use the cross-correlation with broad band measurements is a reliable way to estimate the relative delay in this case.
As a rule of thumb, the greater the overlap of bandwidth, the greater the accuracy of the time alignment.
Some even apply bass prefilters to tame the room and have an easier alignment job.
To always remember, is that the time alignment is valid on an infinitesimal point, that is, not exactly our way of being in a room or on a sofa / chair. So too much time precision can be useless, especially at higher wave lenght.
There is no absolutely right or absolutely wrong therefore, it depends on the goal, environment and compromise.
It is the uncertainty principle in the end.

 

[gnarly]

There is no absolutely right or absolutely wrong therefore, it depends on the goal, environment and compromise.
It is the uncertainty principle in the end.

All you said makes sense to me.
I think the uncertainty principle, the time vs frequency trade that exists in the analog domain, has that foundational measurement difficulty compounded by digital sampling which is linear vs frequency.

My entire focus on time alignment is that of a speaker builder. I only measure indoors to try to improve my measurement skills; never using indoor measurement for anything.
I will say this, time alignment with linear-phase speakers is an absolute breeze compared to minimum phase speakers. It's the biggest "it depends" factor I know of, when it comes to ease and accuracy of implementation.

 

[Keith_W]

My entire focus on time alignment is that of a speaker builder. I only measure indoors to try to improve my measurement skills; never using indoor measurement for anything.
I will say this, time alignment with linear-phase speakers is an absolute breeze compared to minimum phase speakers. It's the biggest "it depends" factor I know of, when it comes to ease and accuracy of implementation.

In general, doing anything in linear phase is an absolute breeze.

 

[levimax]

I will say this, time alignment with linear-phase speakers is an absolute breeze compared to minimum phase speakers. It's the biggest "it depends" factor I know of, when it comes to ease and accuracy of implementation.

In general, doing anything in linear phase is an absolute breeze.

Maybe you can help me understand the best way to think about this. The instructions for my new sub say to use a 2nd order high pass filter at 17 Hz to protect the driver below the port tuning. So do I use a Linear phase filter with pre-ringing or a Minimum phase filter with phase shift? I am not sure I did this right but I created a LP and MP filter in rephase and then imported it into REW and looked at the impulse and step responses. I also looked at the phase shift of the MP filter in Rephase. When I see this I don't want to use either filter at all but the Linear Phase filter step response looks scary to me and the phase shift less scary. Am I looking at this right? For this application what is better LP or MP and if it depends what are the other things to think about? If I am doing everything wrong I would like to know that as well. Thank you.

Linear Phase Impulse and Step Response

Minimum Phase Impulse and Step Response

Minimum phase Phase Shift

 

[gnarly]

Am I looking at this right? For this application what is better LP or MP and if it depends what are the other things to think about?

That's what I look at...either we are both right or both wrong

I do feel confident that the system high-pass when one is needed like for our subs, should be minimum phase. If linear-phase high pass is used, here is no way to offset the step response dive before t=0 your graph correctly shows.

I try to use the minimum order IIR sub high-pass needed for excursion protection.
Sometimes I play with phase linearizing about 1/2 the IIR order, saying to myself I'll take a 50/50% poison potion of group delay vs pre-ring potential. Just to see. When doing that I use the shortest FIR filter possible, looking whatever rePhase shows as an acceptable amount of slippage.

The only fully safe use of linear-phase imo, is with acoustic complementary crossovers, and we get full pre-ring potential cancelled.
This in theory requires complementary response holds up throughout polars. My experience is that it's needless concern, especially with high-order lin-phase xovers that minimize summation regions.
It's worth remembering that a linear-phase crossover between sub and main only works without prering potential, when they are co-located, preferably within 1/4 wave length summation distance at xover.

 

[ernestcarl]

Linear Phase Impulse and Step Response

Perhaps it looks worse than it really is in that plot view... potential audible issues are more likely to occur when one just blindly removes the excess phase contribution from the room -- if there's a lot of it! -- i.e. phase inversion part of "room correction".


Alternate ETC view:

Still, I would not want to exceed levels beyond the set cursor -- or in general, keep it as low as possible.


Speakers (COUCH setup) before and after phase EQ (no subwoofer added here yet)


BM speakers (i.e. now with sub) -- min phase and "mixed phase" filtering


Filters (tested more than one variant):

*I did say previously speakers are not so phase compliant other than corresponding L-R pairs.


Given the room dictates the sub's response at the listening positions, EQ is needed regardless. Fortunately, the bass freq. response of the sub does not vary too much and so the "mixed" phase equalization also carries out over a wider area:



Main speakers are spread far apart from each other across the room and there's more variation with LP change -- thus I didn't bother cutting any excess GD peaks -- and, because they occur above 100 Hz.

 

[levimax]

If you measure a driver over a range that is less than the driver's bandwidth there is in effect a frequency domain truncation of the driver's response, which will result in ringing in the time domain in the IR corresponding to that measurement and a broadening of the IR commensurate with the bandwidth reduction. It will not be the IR of the driver.

There are other ways to align speakers/subs or individual drivers, Pat Brown has a set of videos illustrating an alternative approach here.

The link in this post is great. I was trying to set this up as they showed it but I don't have a way to adjust phase or delay "on the fly". I believe he was using a "separate box" for DSP but not sure.

Does anyone know of Windows software for crossovers and DSP that allows "on the fly" adjustments like shown in the video? Thanks.

 

[DWPress]

Does anyone know of Windows software for crossovers and DSP that allows "on the fly" adjustments like shown in the video?

Yes, very slick. I'd like to know as well. I went poking around the website briefly but couldn't find anything outside the offered courses paywall.

edit: after watching them again I think I could work that out using plugins in the DSP chain. Polarity and delay are easy enough, just need to find one that moves the all-pass filter smoothly through the graph. Looks like pro version of software similar to miniDSP with the XO blocks but obviously more capable.

The obvious disclaimer to doing it this way is it's only valid for that one point in space but probably still applicable to the surrounding area maybe better than impulse matching.

 

[Sokel]

Yes, very slick. I'd like to know as well. I went poking around the website briefly but couldn't find anything outside the offered courses paywall.

Judging by the buttons, colors, etc I would think it's biamp's own software.

Uber Filter

There's some shots at the video showing tesira, so...

Product Categories - products.biamp.com - WYDZZE4LM6SYVY9T_TST

products.biamp.com

 

[Robbo99999]

Hi folks, just a quick question for you, I thought it wasn't worth starting it's own thread, because I'm thinking it's a quick question & is related to DSP inasmuch as we're mostly talking about in it's relation to RoomEQ and/or setting up speakers ideally in a room. Does toe in of speakers affect frequency response below 200Hz at listening position? I have a 2.1 system with speakers crossed over to sub at 120Hz with the JBL 308p speakers port stuffed with socks to take the port out of the equation. I have 2 different listening positions for which I load up seperate Room EQ profiles depending on which one I'm at. One is on my sofa 3.8m away from the speakers, and the other one is at an ideal equilateral position about 1.5m away from the speakers - and historically I've only ever had the speakers set up in terms of toe in to be perfect at the nearer 1.5m listening position - so therefore as a byproduct of this when I'm on couch at the 3.8m listening position then the speakers are no longer at the ideal amount of toe in. I suppose I set it up like this because I couldn't be bothered to swivel the speakers each time (& was concerned it could affect RoomEQ), but they are easy to swivel because the stands swivel and to be honest I wouldn't have to do this more than once per day at the max really, so I'm thinking I'd like to swivel the speakers so they're always pointing at me when I switch between the two listening positions. Will I have to redo my RoomEQ if I only have RoomEQ up to 200Hz? I'm thinking the toe in position isn't going to be affecting frequency response at listening position below 200Hz, and therefore no need to redo RoomEQ? Seems to be OK bass wise with a quick listen. (I could remeasure but don't want to waste the time)

 

[JanRSmit]

It's worth remembering that a linear-phase crossover between sub and main only works without prering potential, when they are co-located, preferably within 1/4 wave length summation distance at xover.

Interesting remark, is this your findings? The distance, is that from listening point, so time of flight?

 

[JanRSmit]

I began using wavelets for sub-to-main alignment after starting this thread on DIY 4 years ago...

Neat way to find delay, phase, and polarity at xover

Wavelets....as per this article.. Signal Aligning Using Wavelets Surprisingly repeatable measurements for low frequencies, that normally drive us crazy trying to nail down timing/phase at xover. Works great with Arta...just like the article says :)

www.diyaudio.com

At that time the SynAudCon article linked was showing how to use wavelets with ARTA, which was great and worked great.
Pat Brown's newer wavelet series JohnPM linked, that uses the REW scope, is even clearer and better imo. Nice job SynAudCon (as usual !!!
(Hey John, if you're reading....back at the time of the old thread above, I asked you on the REW forum if saving scope traces might become an option...and I would like to repeat that request )

I still haven't found anything more repeatable & precise than wavelets....

This remembers me of posts in aktieves-hoeren.de about convolving wavelets in the filters and to measure responses, summed it should reproduce rhe wavelet correctly. This as one way of aligning time between drivers. All using Acourate.
Need to search those posts again.

 

[OCA]

Yes, very slick. I'd like to know as well. I went poking around the website briefly but couldn't find anything outside the offered courses paywall.

edit: after watching them again I think I could work that out using plugins in the DSP chain. Polarity and delay are easy enough, just need to find one that moves the all-pass filter smoothly through the graph. Looks like pro version of software similar to miniDSP with the XO blocks but obviously more capable.

The obvious disclaimer to doing it this way is it's only valid for that one point in space but probably still applicable to the surrounding area maybe better than impulse matching.

REW's alignment tool already does most of that if you use "Filter IRs at cursor":

Maybe if we beg enough, we can get @JohnPM to add a sliding allpass filter to it

 

[gnarly]

Judging by the buttons, colors, etc I would think it's biamp's own software.

Yes, Biamp Tesira for sure....an open-architecture DSP.
(Personally, I would include open-architecture DSPs into state of the DSP market discussions.)
I use QSC's Q-Sys...there are a good handful of other brands out there.
It's quite amazing all the things open-architecture allows, from changing I/O counts and their routing easily, to embedding untold filters, functions, meters, etc.
You can even put dual-channel FFT analyzers into a schematic. (qsc calls theirs a Responsalyzer)
Here's a shot of the filter functions available in q-sys.

All the open-architect processors (OAP) I've seen so far are for pro-sound, without any auto room eq type functions, etc. But I do think we will see OAPs in home audio in the near future, if only for their easy of configuring and changing mult-channel audio I/O counts.

I don't know much about computer filter programs, but I have to think on-the-fly all-pass is available.
Most prosound hardware processors (which are intended for loudspeaker management) have on-the-fly filters including all-pass. The Linea ASC48 is an example.

 

[gnarly]

Interesting remark, is this your findings? The distance, is that from listening point, so time of flight?

Yep, my finding after soaking in other's teachings.
Just like the degree of off-axis comb filtering or lobing for separated drivers, that lobing is what creates the degree of pre-ring potential. Because only fully complementary acoustic coupling allows the two sides of a crossover to completely cancel pre-ring potential. (I keep saying potential because i think it's more theory than not when it comes to audibility.)

I'm taught, and have measured, that when two drivers are within 1/4 wavelength throughout xover summation, that they won't lobe off-axis.
So if I can co-locate sub and main, I don't have any direct sound lobing or pre-ring potential at any listening area. Nor reflected potential.

Separated beyond 1/4 WL, whatever direct lobing exists at any particular location, direct pre-ring potential exists. Seems to me, most all reflected sound has to have pre-ring potential too.
The safe listening place for direct sound would be when the arrival times flight for of the sub and main are within 1/4 WL (for regular combing too) .
But that doesn't do anything for reflections.

 

[gnarly]

REW's alignment tool already does most of that if you use Filter IRs at cursor:

Agreed. But there's something more elegant, more intuitive, that I feel I can trust more.......about the wavelet scope method.

REW's scope and generator are awesome tools ! imo....

 

[DWPress]

I may have found an appropriate plugin solution: Chameleon

 

[LionIT]

Unfortunately, most software that aims towards "ease of use" also sacrifice flexibility. For example, Dirac is a wizard-based program that walks you through all the steps required to generate filters. Time alignment is automated, no need to manually inspect the impulse response. But if it gets it wrong, there is no way to use Dirac itself to determine the correct delay, you have to use REW. And what if you don't want to time align, you want to phase align? Answer: REW. You might then be asking yourself why you need to use REW if you paid close to a thousand big ones for Dirac? Umm, that's a question for Dirac

Hi @Keith_W.
I was deepening the market situation of Dirac ART and given the title of the thread I'm wondering why you skipped a bit on it, although it seems to theoretically represent the most versatile and effective DRC technology (Trinnov is a bit more hardware related). I see that there are delays in rollout and some bugs, but I also see incredibly promising results, unrivalled by simple FDW + inversion filters, even with low count speakers.
Beyond reliability/flexibility, do you know any other aspects that manual software/methods are preferable for?
Do you have any comparison of results between Dirac and Acourate for example?