Principles: Time of Arrival Delays, vs Phase tuning ?

[Sigberg Audio]

@john61ct Then then the next step I suspect is to differentiate between the issue of building a single speaker, and the issue of making the whole system work coherently together. Since you are going multichannel, the latter is probably easiest to do by simply purchasing an AVR that have all of this built-in.

If you are going the active speaker route, simply using the same amplification/DSP modules for all channels will ensure similar latency for each channel (not that it really matters that much if you go the AVR route), and then you are only left with time aligning the individual drivers in each speaker, to the extent that you choose to care about that.

 

[john61ct]

@john61ct Then then the next step I suspect is to differentiate between the issue of building a single speaker

Well that gets into spinorama and anechoic, I'm not asking about that here.

The specific topic of this thread, is the phase / timing domain issues of crossovers aspect.

Not multichannel gear issues and (FYI) I def want to avoid AV gear for now. Same with what are called active speakers - active system yes, but with passive single-driver boxen.


> If you are going the active speaker route

each box gets its own channel, with DSP before the amp if needed. However some of the FR boxen are multi-way and passive. Even the single driver boxen are passive, in the sense that the amp is externally located - for example, I've seen powerful big-ch count amps on eBay going for under $30/ch.

I am not concerned about delays through such purely analog components, different wire lengths etc.

The two timing issues I'm trying to deal with (grok) here are

phase at crossover region between multi-way vs between single driver boxen? are those contexts so radically different?

and preventing D/A & A/D converters' clock drift, ideally without having to invest in a single interface with 32+ analog I/O.

> time aligning the individual drivers in each speaker, to the extent that you choose to care about that.

That is what I'm here to learn from the hive mind, at least to the extent that is possible before being equipped to test for myself.

Part of the challenge is, my own ears are old, but I'd like the system to be well optimised for the youngsters when they are also present.

...

> the issue of making the whole system work coherently together

I prefer to keep DRC or EQ in general for separate threads

...

relevant

SMSL SU-2: a dumb, wasted opportunity

I have to believe the network based AoIP protocols that ensure clock sync are used for good reason in pro environments SMPTE ST 2110, Ravenna, AES67, Dante, Ravenna, Q-LAN My quandary (besides cost) is that I want RPi being fed 6 (or more) **different** analog signals to stay in sync. I'm...

www.audiosciencereview.com

 

[john61ct]

Oooh ooh ooh this is exciting direct from an RME guy!

is this a valid way of synchronizing two USB DACs?

It has been discussed here millions of times: you can't just use various USB DACs simultaneously to have a multichannel system because they are not synchronized and will drift apart. But yesterday watching the last video at eevblog, where he talks about crystal oscillators, he mentions what...

www.audiosciencereview.com

@MC_RME

Not just for AES, but SPDIF, the sender (dumb splitter hub say) controls the master clock, slave DACs are actually controlled by the timing signals inherent in the music content stream?

 

[Sigberg Audio]

Well that gets into spinorama and anechoic, I'm not asking about that here.

The specific topic of this thread, is the phase / timing domain issues of crossovers aspect.

Not multichannel gear issues and (FYI) I def want to avoid AV gear for now. Same with what are called active speakers - active system yes, but with passive single-driver boxen.


> If you are going the active speaker route

each box gets its own channel, with DSP before the amp if needed. However some of the FR boxen are multi-way and passive. Even the single driver boxen are passive, in the sense that the amp is externally located - for example, I've seen powerful big-ch count amps on eBay going for under $30/ch.

I am not concerned about delays through such purely analog components, different wire lengths etc.

The two timing issues I'm trying to deal with (grok) here are

phase at crossover region between multi-way vs between single driver boxen? are those contexts so radically different?

and preventing D/A & A/D converters' clock drift, ideally without having to invest in a single interface with 32+ analog I/O.

> time aligning the individual drivers in each speaker, to the extent that you choose to care about that.

That is what I'm here to learn from the hive mind, at least to the extent that is possible before being equipped to test for myself.

Part of the challenge is, my own ears are old, but I'd like the system to be well optimised for the youngsters when they are also present.

...

> the issue of making the whole system work coherently together

I prefer to keep DRC or EQ in general for separate threads

...

relevant

SMSL SU-2: a dumb, wasted opportunity

I have to believe the network based AoIP protocols that ensure clock sync are used for good reason in pro environments SMPTE ST 2110, Ravenna, AES67, Dante, Ravenna, Q-LAN My quandary (besides cost) is that I want RPi being fed 6 (or more) **different** analog signals to stay in sync. I'm...

www.audiosciencereview.com

From this post I'm still slightly confused about what your concerns are, so I think I will politely step out of the conversation. I wish you the best of luck with both the project and your theoretical enlightenment!

 

[john61ct]

sad face. but thanks, sincerely

 

[3ll3d00d]

I also don't understand what the question, in an ideal world each way will be time aligned and the handover from one to the other will result in a perfect sum which typically means each way has an exactly on target roll off for whatever target you are aiming for.

 

[-Matt-]

Yes, I'm also puzzled. The OP has started numerous threads and seems to be generally interested in developing a multi way digital crossover using raspberry pi hardware.

They also may be interested in having more than two output channels but want to avoid true AV type multi channel. I.e. They are allergic to AVRs, but seem to want to play stereo over many speakers, (basic all channel stereo, rather than upmixing). They don't seem to be interested in using 5.1, 7.1 or Atmos to obtain more than 2 input channels (true multi-channel).

Apologies if the above summary is not exactly correct - this is the meaning I've managed to glean from your various messages.

I think it might be clearer if the OP attemped to tackle one part at a time, and I'd suggest that the active crossover part seems to be where the OP's primary focus lies.

With that in mind, I'd recommend the OP to follow the MiniDSP application notes. Even if you don't plan to buy such a device (which I suggest you do), it will help you to understanding of the normal workflow etc.

https://www.minidsp.com/applications/digital-crossovers/digital-crossover-basics

https://www.minidsp.com/applications/digital-crossovers/stereo-34way-xover

 

[john61ct]

I think maybe my sharing the current "bigger picture" use case may have been a bad idea.

On the theory side I guess my lack of understanding prevents re-phrasing into smarter questions.

Either way I am happy to answer any clarifying questions.

I think this is just repeating, but here goes:

I'm asking about the orders of magnitude between what people say is "acceptable" delay, as in acoustically perceptible as problematic.

Is this due to some inherent difference between "time delay only" vs Phase tuning of crossovers?

Phase tuning of crossovers within a multi-way box,

being somehow inherently different from

Phase tuning of crossovers between enclosures?

I am not aspiring to "perfection" here.

Maybe the answers posted have already answered as well as can be answered, and as I later parse them closely, or learn more - maybe after hands-on - then I will grok those answers better.

Meantime as always, if anyone has links to extisting learning resources, in this case structured explanations of these specific topics, posting those will be greatly appreciated, even if not noobish enough for me to grok at this stage.

Thanks again to all for whatever efforts you've expended.

 

[john61ct]

Thanks much!

There are many ways to skin a cat, this thread is not about gear or implementation methods, but I will respond anyway.

two output channels but want to avoid true AV type multi channel. I.e. They are allergic to AVRs, but seem to want to play stereo over many speakers, (basic all channel stereo) rather than upmixing.

Yes to the AV allergy, also to HDMI and proprietary multichannel codecs.

Upmixing to at least five bed channels is a yes for "not just stereo" mode.

Splitting several of those five into at least two speaker output channels each is a must, F-LR likely four ways, so a dozen total very likely before any summed mono subs

Thus could end up 20+ total by the time I'm done,

which would then drop to ~12 in "just stereo" mode, 3-channel stereo maybe in between.

...

> the active crossover part seems to be where the OP's primary focus lies.

That is certainly what this thread is about, specifically wrt DSP.

i do plan to start with analog, and maybe even test passive xover hardware.

Analog crossovers, only using DSP to treat issues the analog system cannot fix

If I have Main L/R comprised of a mix 2- 3-way speakers + midbass couplers + stereo trueSubs all co-located Those subs easily switched to mono if that turns out to test better, there are also other trueSubs already in mono thus freely placed, can be treated as a single sub-group in the DSP...

www.audiosciencereview.com

> MiniDSP application notes. Even if you don't plan to buy such a device (which I suggest you do), it will help you to understanding of the normal workflow etc.

Thanks much! I was under the impression that these were only relevant to their units.

I will have a 2x4 HD unit for learning perhaps to be deployed for a small part if the overall system.

I will also be building at least one PC, and am opportunistically collecting (cheaper per-port but high SQ) multichannel DAC / format converters / interfaces used.

not the topic here but this is one suited to continue if you like

'ADAT interface hosts, high SQ but cheaper than RME Digiface USB?' https://www.audiosciencereview.com/...gh-sq-but-cheaper-than-rme-digiface-usb.71592

 

[-Matt-]

I'm asking about the orders of magnitude between what people say is "acceptable" delay, as in acoustically perceptible as problematic.

The answer to this specific point is the same as always... it depends!

I'll try to give you a tool that you can use to make your own estimates.

The importance of phase at the crossover is that in the ideal case with both drivers in phase you will get a nice smooth and controlled transition from one driver to the other.

If a specific frequency is 180 degrees out of phase you can potentially get complete cancellation. I.e. An enormous dip in frequency response centred at the frequency where the out of phase cancellation occurs.

Frequency f is given in Hz.
Period (the time for a complete cycling of phase from 0 degrees to 360 degrees )is 1/f (in seconds).

So at 20 Hz the period is 50 ms. The equivalent time to get 180 degrees out of phase (a gross phase error) would therefore be 25 ms.

At 10 kHz the period is 0.1 ms, so our rule of thumb for what constitutes a gross phase error at 10 kHz would be 0.05 ms (or 50 us).

The audability of this is another question altogether, which I won't try to tackle.

 

[john61ct]

Yes, so far I follow that, thanks again

I was hoping for at least some estimates of audibility, the point where "it matters" which of course "it depends"

If I do end up using DSP, and handling all the crossover timing / phase issues from a single clock then I guess lots of this gets inherently resolved for me. It's just that the extra few hundred bucks means longer time saving pennies

Which time will be well spent learning

 

[-Matt-]

I know you don't want it to be the main topic of the thread but I'm having a hard time understanding the opposition to using an AVR (perhaps as a pre-amp, using the analogue pre-outs as the inputs for your DSP system).

Option A) The opposition is based on a desire to use DACs with the highest possible SINAD. True you could get outboard DACs with 120dB SINAD instead of the 100dB which can be had in an AVR. However, this may be largely inaudible. Furthermore, you plan to use ADCs on the input. This conversion from analogue to digital likely represents a far more significant degredation of SINAD (i.e. down to 80dB before it even gets into the DSP). This is where an AVR with HDMI has an advantage because it can accept digital audio input without any signal degredation.

Option B) Cost. If your opposition to using an AVR is because these are expensive devices I can assure you that building a multi-way, multi-channel system by trial and error is going to end up costing a lot more. Furthermore there is a high probability that it may fail to work, or take years to get working and will achieve lower audio quality than could be obtained by a basic AVRs automated setup.

Option C) Something else. If so please explain what are the obstacles that you feel would prevent the use of a more standard Stereo HiFi or AVR type system?

What components do you currently have (e.g. sources, speakers (diy?), amps, subs, etc)?

 

[john61ct]

Let's just call it a preference. Same with big proprietary DSP boxen. If not RPi then little old quiet x86 units and Linux over windoze.

Remember the learning journey is all, the destination "success vs failure" less relevant.

I've spent so little on the boxes and piles so far, all easily resold for a profit, so easy to swap out if better options come along.

Wiim Ultra, Schiit SYN and OG LS50 are cornerstones though. Likely Outlaw ICBM-1 to the extent "mostly analog" holds for crossovers.

GFA-555 as a benchmark, to be left for the kids.

Loudness contours units to test

Pre-amps with Loudness - OSD Pre-1 ?

Most of my source-switchers and volume controls are analog-only, which is my preference. I am also frugal, some months actually poor 8-( I like having physical tone controls, and miss the old variable loudness curve feature, ideally automatically coupled with the volume level, but I guess that...

www.audiosciencereview.com

But honestly I really don't want to divert this thread further off track

 

[-Matt-]

Yes to the AV allergy, also to HDMI and proprietary multichannel codecs.

Upmixing to at least five bed channels is a yes for "not just stereo" mode.

Upmixing usually refers to proprietary algorithms such as Dolby Surround Upmixer, or Auro 3D.

How do you intend to upmix without these?

Simply splitting the signal to many speakers is not really upmixing, it could perhaps be termed "matrixing". I.e. If the input is 2 channel stereo you could build a matrix array of 2 or 3 output speakers for each of the Left and Right channels (4 or 6 speakers total). Note that this wouldn't strictly be a multichannel setup as there are only actually 2 channels of source information. In the AVR world it would be "All channel stereo" mode. (This is often used when the goal is simply multi-room stereo playback - or party mode).

Is it then correct to describe your design aim as a stereo DIY DSP system which allows active crossovers and multiple speaker outputs (some of which will have active crossovers, and where some may be subwoofers)?

(This is assuming you intend to send the input stereo signals to the subs, via DSP active crossovers). If not, how do you intend to get the .1 LFE signal?

 

[3ll3d00d]

Maybe the answers posted have already answered as well as can be answered, and as I later parse them closely, or learn more - maybe after hands-on - then I will grok those answers better.

All of it has been answered multiple times as far as I can see. Instead of asking people to repeatedly rewrite the same content (you can ask some llm to do that tbh), you need to do your own research, try to design some speakers on paper using manufacturer published measurements, play with the geometry and see what happens next. Ask specific Q's about the bits you don't understand. Requires no investment in anything other than time and brainpower.

 

[Waveform Fidelity]

I think maybe my sharing the current "bigger picture" use case may have been a bad idea.

On the theory side I guess my lack of understanding prevents re-phrasing into smarter questions.

Either way I am happy to answer any clarifying questions.

I think this is just repeating, but here goes:

I'm asking about the orders of magnitude between what people say is "acceptable" delay, as in acoustically perceptible as problematic.

Is this due to some inherent difference between "time delay only" vs Phase tuning of crossovers?

Phase tuning of crossovers within a multi-way box,

being somehow inherently different from

Phase tuning of crossovers between enclosures?

I am not aspiring to "perfection" here.

Maybe the answers posted have already answered as well as can be answered, and as I later parse them closely, or learn more - maybe after hands-on - then I will grok those answers better.

Meantime as always, if anyone has links to extisting learning resources, in this case structured explanations of these specific topics, posting those will be greatly appreciated, even if not noobish enough for me to grok at this stage.

Thanks again to all for whatever efforts you've expended.

On the topic of delay, the only practical problem is Lip Sync - when Audio and Video should be synchronous, otherwise why would you care if your system took, for example a second or more to create its sound?

On the phase/time topic, if you keep looking at sine-waves, it will be hard for you unlearn all this confusion about phase and time delay, that is, that phase and time are two separate things.

So, imagine you send a short rectangular pulse (whose voltage rises with time - signal goes positive) to a loudspeaker and receive it with a microphone, some distance away. On a time display you can see the pulse transmitted by the speaker and later when it was received by the microphone. Now change the phase by 180 degrees by swapping the polarity at the loudspeaker terminals and resend the pulse. The pulse now falls (goes negative) but it is received exactly at the same time as before except the pulse looks “upside-down”. So you changed the phase and nothing happened to time delay. Now move the microphone further away which increases the delay (the pulse has to travel further) but nothing happened to the phase of the signal either.

This is such a basic concept, which is so widely misunderstood, mostly in forums like these, that I felt it worth writing this comment.

 

[kemmler3D]

Yes, so far I follow that, thanks again

I was hoping for at least some estimates of audibility, the point where "it matters" which of course "it depends"

If I do end up using DSP, and handling all the crossover timing / phase issues from a single clock then I guess lots of this gets inherently resolved for me. It's just that the extra few hundred bucks means longer time saving pennies

Which time will be well spent learning

I will "bottom line" this for you - you should definitely not use different clocks for crossovers, i.e. different frequency ranges for the same channel of audio. I believe clock drift will ruin the crossover between mid and high frequencies, and low-to-mid is also questionable. Changes or inaccuracies on the order of fractions of a millisecond will cause the response of your speaker to go wrong.

If you can tolerate a wandering stereo image, multiple clocks for multichannel might be acceptable to you, but is technically not good.

Interaural time difference - Wikipedia

en.wikipedia.org

People can detect timing differences between different sounds of less than 1ms, sometimes MUCH less.

If you can keep everything on the same clock, as you note, the potential problems go away.

 

[john61ct]

ISO cable for OG RME Digiface HDSP CardBus/PCMCIA connection

My goal is not recording or mixing / producing music, but a high count of multichannel DSP and crossovers for recreational reproduction, cheap. If members have other similar gear sitting in closets, or alternative suggestions feel free to speak up. In this case I like the RME OG Hammerfall...

www.audiosciencereview.com

 

[john61ct]

On the topic of delay, the only practical problem is Lip Sync.

Wut?

No the problems being discussed here are not that, and that is no problem for me.

System delay in total is fine for me, this topic is just time/phase domain crossover issues, and discrepencies between the speaker enclosures.

time-of-arrival differences, vs

group delays

audible phase issues at crossovers, and

DSP-caused delays

> phase and time are two separate things

I am realising that I will likely not grok the acoustic effects of these very different time domain issues without hands-on testing.

 

[john61ct]

How do you intend to upmix without these?

Schiit SYN to start, as stated. I have a couple Hafler units somewhere also to compare.

> not really upmixing, it could perhaps be termed "matrixing"

I just meant analog simulated surround as opposed to decoding. Maybe one day I will want more sophisticated algo's, start with DD AC-3 - I hear DTS Neural X is pretty good. But not there yet so far does not interest me.

> Is it then correct to describe your design aim as a stereo DIY DSP system which allows active crossovers and multiple speaker outputs (some of which will have active crossovers, and where some may be subwoofers)?

Yes, the option of switching between Stereo-only and 5 bed channels is not at all relevant to the topic of this thread.

...

> how do you intend to get the .1 LFE signal?

I will test the differences between

The Wiim Ultra sub output - DSP configurable.

The Schiit SYN sub output, fixed at 80Hz LPF.

LP filtering from its Center channel.

Or, properly summing to mono from

stereo input to SYN vs its F-LR output

the stereo bass output from Outlaw ICBM-1 is already LP'd

maybe miniDSP 2x4 HD if I use that for say MSO, can use that ABX test a pair of the above at a time.