Help understanding group delay

[David_M]

Think of a race where there are fat people and fit skinny people. They all start off at the same time. But when they reach the finish line, all the fat people are grouped together and they finish later. It's the same with sound - low frequencies have a longer period, so they naturally take longer to be reproduced if they all start at the same time.

How linear-phase DSP fixes this: imagine that fat people are allowed to start first. After a certain amount of time, skinnier people are allowed to start, and the skinniest and fittest people start the last. At the finish line, the fat people and skinny people all arrive together.
.

.
.

Ported vs. sealed: Sealed means airtight. Anything that is not airtight is ported. It doesn't matter if the port is opened or partially closed by stuffing it, it's simply a ported loudspeaker with altered port tuning. And, as you can imagine, you can have an unintentionally ported subwoofer if you construct it very poorly!

Good explanation, though a few corrections are needed:

1. Low, medium, or high frequencies will arrive at a stationary detector at the same time, because the speed of sound is independent of frequency or wavelength. A similar analog applies to RF waves; they will arrive at the same time as optical lightwaves to a stationary detector. It's the characteristics of the medium that determine wave speed (temperature, density, etc) and not its frequency.

2. ... unless you are talking about open baffle subwoofers .

 

[Chris A]

My understanding is UMIK-1 cannot ?

Not correct.

This is why I present a very "conventional" view whilst privately holding different beliefs.

A politician in our midst.

My 'audio politics' are a bit lower on the PC scale (...as you might have noticed...).

Chris

 

[Chris A]

How do I get from that to the version you posted? We're obviously using different settings. Mine was just the REW default. Can you clarify? Cheers,

I've placed my "Controls" dialog in the figure below so you can see the differences...

Chris

 

[levimax]

Not correct.

Because a UMIC-1 does not have it's own clock and relies on USB it is not 100% reliable for timing measurements. It is fine for FR measurements and some basic sanity checks on timing and phase but not really good enough for "speaker design". Sub integration is getting pretty close to speaker design if you take it beyond the basics so if you plan to do a lot of things that require accurate timing measurements then a UMIC-2 or a "regular MIC" and interface are a better option. If you have a regular MIC and a 2 channel interface you can do "loop back" measurements which are by far the most reliable way to get accurate timing measurements (Although recent REW versions are getting quite good using acoustic reference) See post about UMIC-1 timing issues. https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/#post-99941

 

[thcdru2k]

UMIK-2 has higher sample rates and I've repeatedly heard in this forum that UMIK-2 is the minimum necessary for accurate timing work but I am wondering if that is necessarily true if one is using REW with acoustic timing reference. UMIK-1 at 48khz so 48,000 samples / sec means 1 sample is 1/48,000 sec = .0000208s = .0208 ms

I'm not sure how to define timing repeatability but let's just say its a range from 1-3 samples. So could be .0208ms or .0625ms.

Let's assume standard subwoofer integration so work around 80hz crossover. 1 cycle = 1/80sec = 12.5ms . If we take the upper range of timing repeatability that is .0625ms / 12.5ms which is .5% of a cycle .

I would say UMIK-1 is audibly probably good enough and than there is the issue of implementing those changes. I have trouble figuring out what is published but what is typical AVR internal processing rate? What is typical of Audyssey, YPAO, ARC, Dirac Live/ART, Trinnov optimizer etc.

I know for minidsp Flex HTx it is 48khz so doesn't that mean even if you have advantage of UMIK-2 ultra-fine measurement timing, can those measurements actually be applied properly so there is an audible benefit? I have UMIK-2 myself but I have trouble understanding if I got caught up in the upgrade bug or not.

 

[levimax]

UMIK-2 has higher sample rates and I've repeatedly heard in this forum that UMIK-2 is the minimum necessary for accurate timing work but I am wondering if that is necessarily true if one is using REW with acoustic timing reference. UMIK-1 at 48khz so 48,000 samples / sec means 1 sample is 1/48,000 sec = .0000208s = .0208 ms

I'm not sure how to define timing repeatability but let's just say its a range from 1-3 samples. So could be .0208ms or .0625ms.

Let's assume standard subwoofer integration so work around 80hz crossover. 1 cycle = 1/80sec = 12.5ms . If we take the upper range of timing repeatability that is .0625ms / 12.5ms which is .5% of a cycle .

I would say UMIK-1 is audibly probably good enough and than there is the issue of implementing those changes. I have trouble figuring out what is published but what is typical AVR internal processing rate? What is typical of Audyssey, YPAO, ARC, Dirac Live/ART, Trinnov optimizer etc.

I know for minidsp Flex HTx it is 48khz so doesn't that mean even if you have advantage of UMIK-2 ultra-fine measurement timing, can those measurements actually be applied properly so there is an audible benefit? I have UMIK-2 myself but I have trouble understanding if I got caught up in the upgrade bug or not.

See the link I posted. It is not about sampling rate, it is about the inherent issues of the USB bus on a computer. The UMIK-2 has it's own clock, the UMIK-1 does not and relies on the USB bus for timing. https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/#post-99941

 

[thcdru2k]

See the link I posted. It is not about sampling rate, it is about the inherent issues of the USB bus on a computer. The UMIK-2 has it's own clock, the UMIK-1 does not and relies on the USB bus for timing. https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/#post-99941

Thank you; sorry I didn't read that in first place. So sounds like the issue is USB-clock derived sampling vs internal mic clock derived sampling if that is the case UMIK-2 wins hands down. Whether or not it's audible is probably up for discussion

 

[Salt]

For measuring frequency range (amplitude) and impulse response (all you need), Umik 1 is all you need ( don't ask why, but I own Umik 2, in addition).

 

[levimax]

Thank you; sorry I didn't read that in first place. So sounds like the issue is USB-clock derived sampling vs internal mic clock derived sampling if that is the case UMIK-2 wins hands down. Whether or not it's audible is probably up for discussion

Yes that is right. If you are doing speaker design and trying to get the tweeter to mid-range timing right the UMIK-1 is not going to be accurate enough. For subwoofer timing it may be OK but sub integration and trying to get useful in room measurements is hard under the best of circumstances. If you get variable results (not repeatable) it makes things even more confusing. Measurements under 100 Hz in room have many confounding variables so the more you can eliminate the better.

 

[levimax]

For measuring frequency range (amplitude) and impulse response (all you need), Umik 1 is all you need ( don't ask why, but I own Umik 2, in addition).

If you look at the link I posted it would appear the UMIK 1 is not reliable for impulse response measurements. This will vary based on the specific computer and USB implementation.

 

[Salt]

If you look at the link I posted it would appear the UMIK 1 is not reliable for impulse response measurements. This will vary based on the specific computer and USB implementation.

That is why I commented "but I own Umik 2" ... for daily use (and loudspeaker measurement at home) it does no difference, the membranes of the transducers are the slave of physics, not the microphone's (yes, a little) problem, and USB the least.

 

[levimax]

That is why I commented "but I own Umik 2" ... for daily use (and loudspeaker measurement at home) it does no difference, the membranes of the transducers are the slave of physics, not the microphone's (yes, a little) problem, and USB the least.

For non timing use the UMIK-1 is fine. It works well for FR amplitude and even for the "shape" of the impulse. Where I ran into issues was using the "delayed impulse" method of time aligning a subwoofer https://www.minidsp.com/applications/rew/measuring-time-delay . I could not get repeatable results and it took me a long time to figure out it was the USB bus issue. While overkill for most hobby use I ended up buying an Earthworks M23 MIC. It is considerably easier to use (no calibration file, much more headroom, much lower distortion) and is a beautifully crafted tool. For hobby use I think the UMIK-2 is probably the best mix of price, quality, ease of use, and performance.

 

[Chris A]

Guys, if you want an instrumentation grade measurement microphone, then buy one. But first, get out your checkbook ($1k-$4K USD, etc.)

UMIK-2 has higher sample rates and I've repeatedly heard in this forum that UMIK-2 is the minimum necessary for accurate timing work but I am wondering if that is necessarily true if one is using REW with acoustic timing reference.

Yes.

I have no complaints related to the UMIK-1--because I don't try to rely on its absolute amplitude response at the upper limit of human perception, nor do I use it as "a timing reference". The type of techniques where the home hi-fi buff is trying to use the microphone as a timing reference are just not recommended. [If you need more information on this, we can discuss how this can be--but probably not in this thread.]

Horses for courses, they say, and the UMIK-1 is just fine for what it does. That's why it was developed and has sold so many units over the years.

I have no use for a UMIK-2 for dialing in my loudspeakers. I recommend the UMIK-1 to all the guys I've helped dial-in their setups remotely using DSP (and that includes many dozens of instances)--because the price is good and the performance is more than sufficient.

 

[john61ct]

miniDSP Tide16 - Holy Grail with 16 Channel Atmos/DTS:X, high SINAD

I don’t anticipat any issues plugging such devices directly into my TV. this seems like an exclusively projector and bluray issue. it’s a bit much t its not for ANYONE with full bitrate Dolby Vision and TrueHD Atmos or DTSX Master audio. o call it a failed product when there is a narrow use...

www.audiosciencereview.com

UMIK-2 has higher sample rates and I've repeatedly heard in this forum that UMIK-2 is the minimum necessary for accurate timing work but I am wondering if that is necessarily true if one is using REW with acoustic timing reference. UMIK-1 at 48khz so 48,000 samples / sec means 1 sample is 1/48,000 sec = .0000208s = .0208 ms

I'm not sure how to define timing repeatability but let's just say its a range from 1-3 samples. So could be .0208ms or .0625ms.

Let's assume standard subwoofer integration so work around 80hz crossover. 1 cycle = 1/80sec = 12.5ms . If we take the upper range of timing repeatability that is .0625ms / 12.5ms which is .5% of a cycle .

I would say UMIK-1 is audibly probably good enough and than there is the issue of implementing those changes. I have trouble figuring out what is published but what is typical AVR internal processing rate? What is typical of Audyssey, YPAO, ARC, Dirac Live/ART, Trinnov optimizer etc.

I know for minidsp Flex HTx it is 48khz so doesn't that mean even if you have advantage of UMIK-2 ultra-fine measurement timing, can those measurements actually be applied properly so there is an audible benefit? I have UMIK-2 myself but I have trouble understanding if I got caught up in the upgrade bug or not.

 

[john61ct]

I am trying to design and build both stereo mid-bass couplers and trueSub boxen likely mono

And testing out proper integration of bandpass enclosure ideas, where phase rotation / group delay may become the limiting factor striving for ever-deeper, extreme bass extension with the latter, or wider bandwidths with the former, likely ~200Hz, or even up into the lower midranges

So "backing off" from when delays have gone too long is a big part of the trial and error.

Yes that is right. If you are doing speaker design and trying to get the tweeter to mid-range timing right the UMIK-1 is not going to be accurate enough. For subwoofer timing it may be OK but sub integration and trying to get useful in room measurements is hard under the best of circumstances.

 

[Arindal]

some effects that I've personally noticed after reducing all-pass phase growth in-room and significantly smoothing the group delay response,

No disagreement from my side, maybe you noticed that I had already mentioned 4 of the 5 aspects of sound quality deterioration you are listing, just using other words.

I also agree that in theory group delay distortion could create a deterioration of subjective transparency, what you name harshness. In practice, I just never encountered a situation in which *solely* group delay caused such issue, and thereby cannot be just disappearing the moment you linearize phase response without affecting amplitude response by any means.

Could you please give an example of a particular speaker or situation in which harshness was caused only by group delay distortion, while no apparent signs of amplitude peaks, cancellation, nonlinear distortion, edge diffraction were noticeable? If one of the latter is at play, I would rather suspect than solving these has greater positive effect for subjective transparency, and linearizing group delay is so to say a positive side-effect, but not the main root cause.

 

[Chris A]

Could you please give an example of a particular speaker or situation in which harshness was caused only by group delay distortion, while no apparent signs of amplitude peaks, cancellation, nonlinear distortion, edge diffraction were noticeable?

The first link that I posted on this subject has your answer (i.e., one further post down from the start of conversation of the effects of phase/group delay flattening).

Here is a direct link to that particular post to save you a little time: Elimination of Harshness. I invite you to read a little more into that thread, both before and after I discovered the PEQ/shelf filter-only crossover approach.

This effect suddenly appeared when I flattened the phase response of the entire setup, (i.e., well after flattening the amplitude response of the 5.1 array). This was as big of a surprise as anything that I had encountered before in my home hi-fi adventures.

This effect revealed a great deal to me in terms of understanding why some listeners experience apparent harshness while listening full-range horn setups that are not compensated for phase growth. My setup currently sounds like the entire front and front side walls are playing music like the musicians are in the room. There is no "horn sound" at all. You can also walk from left wall to right wall and not experience any gaps in the soundstage while playing multichannel (5.1) recordings. This was a difficult-to-achieve capability that took a few years of investigation and work on loudspeakers and acoustic treatments.

But note that you need three things to achieve all the effects that I talk about in that thread (i.e., simultaneously):

1. full-range controlled directivity--down to the room's Schroeder frequency (~100 Hz in my listening room)
2. suppression of early reflections (less than 5 ms from direct arrivals)
3. flattened amplitude and phase/group delay response (±1.5 dB on axis using psychoacoustic smoothing, less than 180 degrees of total phase growth from 20 kHz down to 100 Hz, less than 1 ms group delay spikes above 400 Hz using psychoacoustic smoothing, less than 1 period of sound at frequencies below 400 Hz)

 

[levimax]

The first link that I posted on this subject has your answer (i.e., one further post down from the start of conversation of the effects of phase/group delay flattening).

Here is a direct link to that particular post to save you a little time: Elimination of Harshness. I invite you to read a little more into that thread, both before and after I discovered the PEQ/shelf filter-only crossover approach.

This effect suddenly appeared when I flattened the phase response of the entire setup, (i.e., well after flattening the amplitude response of the 5.1 array). This was as big of a surprise as anything that I had encountered before in my home hi-fi adventures.

This effect revealed a great deal to me in terms of understanding why some listeners experience apparent harshness while listening full-range horn setups that are not compensated for phase growth. My setup currently sounds like the entire front and front side walls are playing music like the musicians are in the room. There is no "horn sound" at all. You can also walk from left wall to right wall and not experience any gaps in the soundstage while playing multichannel (5.1) recordings. This was a difficult-to-achieve capability that took a few years of investigation and work on loudspeakers and acoustic treatments.

But note that you need three things to achieve all the effects that I talk about in that thread (i.e., simultaneously):

1. full-range controlled directivity--down to the room's Schroeder frequency (~100 Hz in my listening room)
2. suppression of early reflections (less than 5 ms from direct arrivals)
3. flattened amplitude and phase/group delay response (±1.5 dB on axis using psychoacoustic smoothing, less than 180 degrees of total phase growth from 20 kHz down to 100 Hz, less than 1 ms group delay spikes above 400 Hz using psychoacoustic smoothing, less than 1 period of sound at frequencies below 400 Hz)

PEQ is just a type of "user interface" and doesn't have anything to do with the underlying filters so why would it make any difference when creating crossovers?

 

[Chris A]

First, and just as a courtesy to others--if you're going to quote a longer previous post, I wouldn't do it unless you can quote just the specific sentence that you're focusing on. If you're using a cell phone that doesn't easily allow you to quote less than a full previous post, then I wouldn't quote any longer posts (like I just posted above to address the deeper question that @Arindal posed--which is on a subject of some interest [at least to me] that's not often discussed, so I took a little time to answer his question fully).

Second, and to address your immediate question...

I would answer your question directly, but it seems that you might use this moment to continue your learning path in this subject area (passive IIR, active IIR crossover filters and the difference between those filters and PEQs/shelves used in DSP equalization).

I'll give you a hint: look at the phase shifts of both types of filters. I think you'll answer your own question.

You could also read the thread that I've linked to above (now three times). That will also answer your question.

Chris

 

[levimax]

I would answer your question directly, but it seems that you might use this moment to continue your learning path in this subject area (passive IIR, active IIR crossover filters and the difference between those filters and PEQs/shelves used in DSP equalization).

PEQ is not a type of filter, it is just a type of a human interface that helps visualize the filters you are creating which are IIR filters. There is no practical difference between passive IIR and active IIR. There are phase differences between IIR filters based on their steepness and other parameters. In addition to IIR filters there are also FIR filters where phase and amplitude can be adjusted separately. In addition to phase differences there are also differences in how each filter type "rings" and how much group delay they create. There is no magic or free lunch.

I did read the link you posted and was surprised you dismiss "any filter that contains someone's last name" for crossover use. The science behind Linkwitz-Riley crossovers is widely accepted. What I didn't understand about your "no harshness" PEQ/IIR crossovers was if you don't use one of the standard crossover curves what "curve" are you trying to match with you PEQ IIR filters when you create your crossover?