"Efficient" audio amp selection

[Don Y]

Hi,

I've been designing a "network audio client" (aka "network loudspeaker")
and now have to select a suitable amplifier to use in it.

1) The amplifier's presence is "optional" (e.g., when interfacing
to external equipment via a "line out", the amplifier is
superfluous)

2) The design is intended to support one or two audio channels
(e.g., drive a single speaker -- BTL? -- or a pair WHICH MIGHT
NOT BE IN CLOSE PHYSICAL PROXIMITY!)

3) Everything is PoE powered (though I will support PoE+ as well)
so the power budget is extremely critical.

4) Size is important. E.g., a few cu in for the whole device.
(note the implications on heat dissipation!)

5) "Louder is better" but LOUD CRAP isn't worth the effort
(i.e., trying to get the most GOOD sound -- not the MOST
sound)

6) Any controls need to be available thru software (gain, etc.)

7) Components need to be available in small quantities for
hobbyists, etc. -- though I suspect I could find a middleman
willing to make large buys and sell in smaller quantities
(hardware and software designs will be released as "Open Source")


I've accommodated #1 by putting all of this on a "daughter card".
Currently, the interface to that card is digital (i.e., for a
"line out" version, another card with DACs and buffers would
take its place). Of course, someone can always opt to change
this later...

#2 Could be accommodated with two different daughter cards
("mono" and "stereo"). Or, operating a stereo amp in bridge
mode (with suitable changes in load). A set of "mono"
devices might be used, for example, staggered around a listening
area (e.g., a 7.1 theater system) giving more power to each
device and/or making wiring more practical (run network drops
to each location instead of having to run drops to half of the
stations and "string wire" along a baseboard to the "other"
speaker in each pair).

Besides power budget, #3 also places constraints on how well
behaved the device is when powering up/down -- since power
can/will be applied/removed remotely. Clicks and pops need not
apply! :>

The size requirement in #4 *could* be flexible but impacts the
types of deployments possible. E.g., here, I plan to mount most
of them in 1G Jboxes (possibly using the box itself to help
dissipate heat). This eliminates all of the cosmetic "packaging"
issues.

Sound level and quality (#5) is, of course, a subjective assessment.
For high quality deployments, I imagine the amplifier will be
removed and replaced by something (external?) of much better
caliber. But, when present, it should be "usable" in normal
(i.e., non-teenager!) environments. A two channel device might
take the place of a "table radio". A set of mono devices (e.g.,
the 7.1 setup) could provide a rich home theater experience.
The impact headroom has on that quality becomes a significant
design issue (esp due to #3)

The controls (#6) need to be "remotable". I.e., a "network loudspeaker"
mounted in a ceiling would be impractical to adjust if the controls
were collocated with the speaker. Being able to "tell" the device
what you want and having the device perform that action itself seems
the only practical solution.

Personally, I could skip #7 (<grin>) but that would be a bit selfish
in light of the open nature intended. (building something out of
unobtainium is hardly a good way to promote its use!)

So far, my searches keep bringing me to TI's class D offerings -- though
none seems to be the perfect cherry. And, class D leaves me worrying
about sound quality and RFI (generated) -- esp when the loads aren't
close to the amp (e.g., the two channel case)

Pointers?

Thanks!
--don

 

[Tim Williams]

3) Everything is PoE powered (though I will support PoE+ as well)
so the power budget is extremely critical.

I'm not familiar with PoE, what V, I, P is available?

4) Size is important. E.g., a few cu in for the whole device.
(note the implications on heat dissipation!)

If the PoE only does a watt, of course, this isn't a problem, but if
you're allowed a bit more, like 10W, it'll start to get important.

Needless to say, efficient speakers will be a top priority -- 90dB
speakers on a class A amplifier (~20% efficient, assuming continuous full
volume of course) are *equal* to 83dB speakers on a lossless amplifier!

High efficiency speakers will tend to clash with your requirement of "high
quality" audio, particularly if they are as small as this device.
Speakers with high efficiency tend to be very resonant with poor frequency
response. The very good ones have high efficiency AND flat response, but
they are also large and expensive. If they have to be small, you might
ask Apple or Bose about their devices (...or marketing..).

These are all, of course, solved problems available on the market, but
you're going to pay for them.

5) "Louder is better" but LOUD CRAP isn't worth the effort
(i.e., trying to get the most GOOD sound -- not the MOST
sound)

That shouldn't be a problem. Even a simple design like this,
http://myweb.msoe.edu/williamstm/ClassD2_Schematic.pdf
has boom-box level quality. It would be greatly improved with a current
source in the triangle generator, instead of the exponential slope, which
produces a corresponding distortion in the output, visible on large
signal, medium frequency outputs (i.e., between feedback and filter cutoff
frequencies). Even as shown, this circuit is an excellent DC to LF
amplifier for motor, driver or subwoofer applications, because the
feedback loop pushes down the LF distortion.
http://myweb.msoe.edu/williamstm/Images/ClassD_4.jpg
http://myweb.msoe.edu/williamstm/Images/ClassD_5.jpg
As built, RFI is quite low. Differential and common mode chokes, and
ceramic caps, do a great job. >20MHz hash is invisible (a spec might
argue differently of course), ripple is visible (~120kHz) but not
substantial.

I can't imagine any popular monolithic solution would get any traction if
it didn't have the distortion to sell it. The only thing I'd watch out
for is style -- open loop, closed loop, voltage mode, current mode,
carrier frequency, etc. These aspects will dominate performance and fault
behavior. For instance, the above circuit is voltage mode, with no fault
protection (well, I think I tested it into a 1 ohm load, but that yanked
down the power supply instead..).

I've accommodated #1 by putting all of this on a "daughter card".
Currently, the interface to that card is digital (i.e., for a
"line out" version, another card with DACs and buffers would
take its place). Of course, someone can always opt to change
this later...

Hmm, only problem that comes to mind about this might be RFI. Mind where
your currents are flowing, keep RF out of the board-to-board connection.
I suppose the outputs will get filtering, which will help too. You
probably already know all this.

Tim

 

[Phil Allison]

"Tim Williams"

on a class A amplifier (~20% efficient, assuming continuous full volume of
course)

** A class A amplifier is up to 50% efficient at full sine wave power.

This is true for single ended stages using an output transformer and push
pull stages with or without a transformer.

Class B push-pull can go up to 78.5% efficiency at full level.


.... Phil

 

[[email protected]]

Try an 8 inch high quality speaker in an infinite boffle; you will be
amazed at the result.

I'm baffled by your boffle, seen that word twice recently.

Infinite baffle, aka acoustic suspension, nice. I don't like
the boomy vented enclosures common these days. At least, not
the ones in my price range.

Grant.

 

[[email protected]]

Try an 8 inch high quality speaker in an infinite boffle; you will be
amazed at the result.

I'm baffled by your boffle, seen that word twice recently.

Infinite baffle, aka acoustic suspension, nice. I don't like
the boomy vented enclosures common these days. At least, not
the ones in my price range.

Grant.

 

[Phil Allison]

[email protected]>

Infinite baffle, aka acoustic suspension, nice.

** No really.

Infinite baffle implies a rather big enclosure while " acoustic suspension"
implies a rather small one.

In the former, the bass resonance frequency is set mainly by the driver
itself and in the latter mainly by the enclosure volume.

I don't like the boomy vented enclosures common these days.

** Vented ( ie tuned port) boxes are not boomy.

Not if Messer's Thiele and Small have anything to do with it.



..... Phil

 

[Tim Williams]

** A class A amplifier is up to 50% efficient at full sine wave power.

This is true for single ended stages using an output transformer and
push pull stages with or without a transformer.

25% being the case for single ended, resistive loaded stages, and in
practice, much lower being common (although I've had tubes up around
15-20% plate efficiency in this mode before, though the distortion was not
what one would call high-fidelity).

Tim

 

[Phil Allison]

"Tim Williams = Wanker "

"Phil Allison"


25% being the case for single ended, resistive loaded stages,

** Totally irrelevant to an amplifier driving a speaker.

You pathetic, context shifting, over snipping bullshitter.





.... Phil

 

[Don Y]

Hi Bill,

Put some high-frequency filtering at the output of the amplifier.

I'm concerned with the *bulk* that would involve (note the entire
device wants to be just a couple of cubic inches). It's alarming
how quickly that volume gets eaten up (electronics, connectors, etc.)

My original vision was for a single channel device bolted directly
to the driver. I.e., it's bulk hides in the speaker's envelope;
no long wires leading *to* the voice coil from the device, etc.

I.e., the two channel optimization may prove NOT to be a net
improvement due to the other consequences it introduces.

 

[Don Y]

Hi Tim,

I'm not familiar with PoE, what V, I, P is available?

The biggest constraint is available *power* (~12W for PoE, about
double that for PoE+) as you can create whatever supplies you need
from the input supply (at various efficiencies).

Of course, I have a low voltage supply available for the logic
but that takes minimal power. A second supply for the amplifier
(and any analog processing). Due to the overall power limitations,
I can't imagine that second supply being more than 15-18V
(single ended).

Since some instances may opt to power the amplifier locally
(more power available, less demands on the PoE suppy, etc.),
it would be good to pick a "nice number" to make the choice
of local power supply easier to satisfy.

If the PoE only does a watt, of course, this isn't a problem, but if
you're allowed a bit more, like 10W, it'll start to get important.
Exactly.

Needless to say, efficient speakers will be a top priority -- 90dB
speakers on a class A amplifier (~20% efficient, assuming continuous full
volume of course) are *equal* to 83dB speakers on a lossless amplifier!

High efficiency speakers will tend to clash with your requirement of "high
quality" audio, particularly if they are as small as this device.
Speakers with high efficiency tend to be very resonant with poor frequency
response. The very good ones have high efficiency AND flat response, but
they are also large and expensive. If they have to be small, you might
ask Apple or Bose about their devices (...or marketing..).

These are all, of course, solved problems available on the market, but
you're going to pay for them.

I'm not concerned with the choice of speaker. I.e., all of the value
added, here, is in the hardware and software designs. If someone opts
to tack a good/bad speaker onto one... <shrug> as long as it fits their
requirements, what do *I* care? :>

My personal needs vary in fidelity, durability, etc. E.g., the speaker
used "for announcements" (someone is at the door; the garage door was
just opened; time to check the roast; etc.) can be of low fidelity
but must endure a harsh physical environment (i.e., without cone drying
out). OTOH, those with multimedia application will tend to want to
be of higher quality...

That shouldn't be a problem. Even a simple design like this,
http://myweb.msoe.edu/williamstm/ClassD2_Schematic.pdf
has boom-box level quality. It would be greatly improved with a current
source in the triangle generator, instead of the exponential slope, which
produces a corresponding distortion in the output, visible on large
signal, medium frequency outputs (i.e., between feedback and filter cutoff
frequencies). Even as shown, this circuit is an excellent DC to LF
amplifier for motor, driver or subwoofer applications, because the
feedback loop pushes down the LF distortion.
http://myweb.msoe.edu/williamstm/Images/ClassD_4.jpg
http://myweb.msoe.edu/williamstm/Images/ClassD_5.jpg
As built, RFI is quite low. Differential and common mode chokes, and
ceramic caps, do a great job.>20MHz hash is invisible (a spec might
argue differently of course), ripple is visible (~120kHz) but not
substantial.

Ouch! Physically too large. I was hoping to tap into the market
created by battery powered, hand-held devices -- though my power
requirements tend to be an order of magnitude higher than most of
those (and the voltage/power available accordingly)

I can't imagine any popular monolithic solution would get any traction if
it didn't have the distortion to sell it. The only thing I'd watch out
for is style -- open loop, closed loop, voltage mode, current mode,
carrier frequency, etc. These aspects will dominate performance and fault
behavior. For instance, the above circuit is voltage mode, with no fault
protection (well, I think I tested it into a 1 ohm load, but that yanked
down the power supply instead..).

What are the major tradeoffs with each? Remember the environment:
aside from assembly, there should problem with "shorted outputs"
unless a voice coil overheats/shorts -- in which case, the cost
of the speaker and labor to replace it might be high enough to
justify replacing the "amplifier" as well.

Hmm, only problem that comes to mind about this might be RFI. Mind where
your currents are flowing, keep RF out of the board-to-board connection.
I suppose the outputs will get filtering, which will help too. You
probably already know all this.

I'm concerned with the efficacy and size of filters on the output.
Since space is *really* tight, even high frequency devices can
be (relatively) large.

 

[Tim Williams]

I'm not concerned with the choice of speaker. I.e., all of the value
added, here, is in the hardware and software designs. If someone opts
to tack a good/bad speaker onto one... <shrug> as long as it fits their
requirements, what do *I* care? :>

My personal needs vary in fidelity, durability, etc. E.g., the speaker
used "for announcements" (someone is at the door; the garage door was
just opened; time to check the roast; etc.) can be of low fidelity
but must endure a harsh physical environment (i.e., without cone drying
out). OTOH, those with multimedia application will tend to want to
be of higher quality...

Ok, so it could be kind of all over, implementation dependent. That'll
make loudness difficult to tack down, but maybe that's not such a big deal
after all. The PA speaker might be a cheap, tinny, relatively efficient
type, or outdoor compatible; the hi-fi types might be low efficiency, but
that might not be a problem for the casual iPod user who wants something
for quiet listening without those horrible earbuds.

Ouch! Physically too large. I was hoping to tap into the market
created by battery powered, hand-held devices -- though my power
requirements tend to be an order of magnitude higher than most of
those (and the voltage/power available accordingly)

That circuit could be miniaturized quite a bit as-is; notice it's discrete
on a two layer board (the best I've ever made by hand, if I do say so
myself.. the layers lined up perfectly!). A couple drivers, or a whole
monolithic chip, should leave enough room for your receiver and DAC
business.

What are the major tradeoffs with each? Remember the environment:
aside from assembly, there should problem with "shorted outputs"
unless a voice coil overheats/shorts -- in which case, the cost
of the speaker and labor to replace it might be high enough to
justify replacing the "amplifier" as well.

Yeah, it's not quite consumer material, but if you have to use BGAs and
such to fit the package, a dead circuit is pretty much throwaway to
anyone.

As for amps, I don't know much about them beyond my own experience. Even
open loop can give reasonable fidelity, but protection circuitry is extra.
Closed loop cleans up the response, asymptotically, but usually leaves
distortion at higher frequencies, relating to the open-loop artifacts that
feedback isn't fast enough to clean up. (As with any audio, you get the
best results with the best of everything, and only then add NFB to make it
electrically perfect.)

Current mode is the best method for switching supplies in general, because
having first-order control over inductor current prevents any possible
fault problems. Unfortunately, speakers aren't current mode (well, most
of them*..), so a voltage-mode loop is required, which means the overall
response will be inductive at high frequencies, coming down to resistive
or voltage sourced at low frequencies. On the plus side, current control
is usually a first order system, so its dominant pole can be almost as
high as the clock frequency, impacting overall frequency response
minimally. With the inner loop keeping average current accurately at the
setpoint, linearity should be high, even if the PWM method is poor.

*A friend of mine built a variable impedance amplifier for the audiophile
circle. You can dial in the boominess at will. Some specially made
speakers respond quite well to high impedances, most are ridiculously
terrible!

I'm concerned with the efficacy and size of filters on the output.
Since space is *really* tight, even high frequency devices can
be (relatively) large.

If you can put the boards behind the speakers, this will be fine -- the
voice coils or magnets might get a bit toasty from the eddy currents,
which will hit you with quiescent current, but if it's not too much, no
big deal. This does rule out stereo amplifiers unless you're going to
have a panel connector for the other one.

I grilled a turkey last night, over charcoal of course (anything else is
wasted time!). Soaked in brine and all that. (I'd use the grill to claim
hardcore Wisconsinism, but it's been downright muggy this year.) This
turkey sandwich is *delicious*. Oven roasted, never again ;-)

Tim

 

[josephkk]

Hi Tim,



The biggest constraint is available *power* (~12W for PoE, about
double that for PoE+) as you can create whatever supplies you need
from the input supply (at various efficiencies).

Presuming so much as 6 watts to the speaker (normally plenty to way loud)
it is worthwhile to invest in quality 4 to 6 inch drivers, US $50 and up
each. No, i am not kidding, good drivers co$t plenty. With external
power you can get less expensive higher power drivers and compensate them
in electronics (al la Bose)

Of course, I have a low voltage supply available for the logic
but that takes minimal power. A second supply for the amplifier
(and any analog processing). Due to the overall power limitations,
I can't imagine that second supply being more than 15-18V
(single ended).

Since some instances may opt to power the amplifier locally
(more power available, less demands on the PoE suppy, etc.),
it would be good to pick a "nice number" to make the choice
of local power supply easier to satisfy.

Militating strongly for class D. Keep switch frequency above 200 kHz but
below 500 kHz, and place then in a metal (foil or even film) shielded box.

 

[josephkk]

Hi Bill,



I'm concerned with the *bulk* that would involve (note the entire
device wants to be just a couple of cubic inches). It's alarming
how quickly that volume gets eaten up (electronics, connectors, etc.)

My original vision was for a single channel device bolted directly
to the driver. I.e., it's bulk hides in the speaker's envelope;
no long wires leading *to* the voice coil from the device, etc.

I.e., the two channel optimization may prove NOT to be a net
improvement due to the other consequences it introduces.

I am reasonably sure that one channel devices is the way to go. Whether
it is mono from the doorbell, stereo, or 7.1. Too many problems, too
little advantage.

 

[Don Y]

Hi Joseph,

I am reasonably sure that one channel devices is the way to go. Whether
it is mono from the doorbell, stereo, or 7.1. Too many problems, too
little advantage.

The biggest problem comes with tabletop/countertop/bedside
deployments -- where the speakers tend to be *relatively*
lose together. You don't want to have to run *two* drops
to the same place.

Or, in our case, the pair of speakers above the kitchen sink
(separated by about 4 ft).

(The "boombox" option can typically be addressed by adding
components to the "enclosure" housing the speakers)

 

[Don Y]

Hi Tim,

Ok, so it could be kind of all over, implementation dependent. That'll

Yup. As it should be, IMO. Let the end user trade dollars for
performance. Folks with "tin ears" or who just listen to talk
radio will care little for the quality of the driver.

make loudness difficult to tack down, but maybe that's not such a big deal
after all. The PA speaker might be a cheap, tinny, relatively efficient
type, or outdoor compatible; the hi-fi types might be low efficiency, but
that might not be a problem for the casual iPod user who wants something
for quiet listening without those horrible earbuds.

Exactly. E.g., we often listen to "background music" in the
kitchen while preparing a meal. We don't need head banging
volumes but would prefer resonable *quality*.

That circuit could be miniaturized quite a bit as-is; notice it's discrete
on a two layer board (the best I've ever made by hand, if I do say so
myself.. the layers lined up perfectly!). A couple drivers, or a whole
monolithic chip, should leave enough room for your receiver and DAC
business.


Yeah, it's not quite consumer material, but if you have to use BGAs and
such to fit the package, a dead circuit is pretty much throwaway to
anyone.

Yup. And, my point was, you aren't dicking with those
connections (intentionally) often. Not like "time to move
the hifi as we rearrange furniture". Nor are you likely
to encounter UNintentional changes ("oops! I tripped on
the speaker wires and pulled them out of their bindings!")

I.e., you might stress the network connection but not the
*internal* ones!

As for amps, I don't know much about them beyond my own experience. Even
open loop can give reasonable fidelity, but protection circuitry is extra.
Closed loop cleans up the response, asymptotically, but usually leaves
distortion at higher frequencies, relating to the open-loop artifacts that
feedback isn't fast enough to clean up. (As with any audio, you get the
best results with the best of everything, and only then add NFB to make it
electrically perfect.)

Current mode is the best method for switching supplies in general, because
having first-order control over inductor current prevents any possible
fault problems. Unfortunately, speakers aren't current mode (well, most
of them*..), so a voltage-mode loop is required, which means the overall
response will be inductive at high frequencies, coming down to resistive
or voltage sourced at low frequencies. On the plus side, current control
is usually a first order system, so its dominant pole can be almost as
high as the clock frequency, impacting overall frequency response
minimally. With the inner loop keeping average current accurately at the
setpoint, linearity should be high, even if the PWM method is poor.

*A friend of mine built a variable impedance amplifier for the audiophile
circle. You can dial in the boominess at will. Some specially made
speakers respond quite well to high impedances, most are ridiculously
terrible!

Hmmm.. I thought you always wanted to keep the damping factor as
high as practical (of course, the impedance of the speaker also
varies)

If you can put the boards behind the speakers, this will be fine -- the
voice coils or magnets might get a bit toasty from the eddy currents,
Yes.

which will hit you with quiescent current, but if it's not too much, no
big deal. This does rule out stereo amplifiers unless you're going to
have a panel connector for the other one.

Even a connector implies a *cable*. Then, I think the filter
becomes necessary.

I grilled a turkey last night, over charcoal of course (anything else is
wasted time!). Soaked in brine and all that. (I'd use the grill to claim
hardcore Wisconsinism, but it's been downright muggy this year.) This
turkey sandwich is *delicious*. Oven roasted, never again ;-)

Some years ago, a friend roasted a pig (in his driveway). It
was, without a doubt, the best pork I'd ever eaten! THough
I admit some trepidation in taking those first bites: "Do I
*really* think this thing has been prepared and cooked properly???!"

Unfortunately, its not the sort of exercise you can repeat
without dozens of hungry mouths available!

 

[Don Y]

Hi Vladimir,

TPA2018

<frown> Kinda wimpy -- 1(3)W into 8(4) ohms.

That seems to be typical of the sorts of devices I encounter...
as if intended for a small, battery powered PMP, etc. (e.g.,
the 6V power supply)

I guess the rationale is that you will "tolerate" class-D
for those extreme low power applications where nothing else
is possible. I seem to be operating at a point a bit too
high for most of these devices -- but still constrained
on power so unable to use AB style monolithic amps.

 

[Don Y]

Hi Joseph,

Understood. Small change in the module boundaries; one NIC, two (maybe
three) audio channels; for those (few?) locations. I had focused on the
distributed ceiling speakers. Oops.

It *really* feels "wrong" to support two channels. Where
do you stop? 2? 2+1? 4? etc.

And, when the amplifier is NOT required (e.g., when an external
amplifier is available and all you need is "line out"), then
you have to think about "is two actually *enough*"?

Depending on the device(s) chosen, I *could*, conceivably, stack
multiple "output cards" (since any package that has to present
"line out" jacks would need a larger enclosure). Likewise for
power amps.

Of course, the more channels you support, the tougher the
processing requirements (so you start needing more resources
ahead of the amp/DAC/etc.)

Of course, maybe you do not need stereo at boom box locations. Only place
for 7.1 is the home theater. (?)

Different requirements result in different system specifications and thus
modularization. Do what supports your needs.

Yup. I may have to resign myself to different hardware designs
to address the differing deployments. But, I should be able
to scale the software easily (which is the hardest part of such
a project, anyway!)

 

[[email protected]]

Hi Vladimir,



<frown> Kinda wimpy -- 1(3)W into 8(4) ohms.

That seems to be typical of the sorts of devices I encounter...
as if intended for a small, battery powered PMP, etc. (e.g.,
the 6V power supply)

I guess the rationale is that you will "tolerate" class-D
for those extreme low power applications where nothing else
is possible. I seem to be operating at a point a bit too
high for most of these devices -- but still constrained
on power so unable to use AB style monolithic amps.

There are fairly high-power AB power amps, used in the automotive industry.
Cooling won't be easy and they may be hard to find, onesy-twosy, though.

 

[josephkk]

Hi Joseph,



It *really* feels "wrong" to support two channels. Where
do you stop? 2? 2+1? 4? etc.

And, when the amplifier is NOT required (e.g., when an external
amplifier is available and all you need is "line out"), then
you have to think about "is two actually *enough*"?

Maybe the module boundary should be in terms of streams.

Depending on the device(s) chosen, I *could*, conceivably, stack
multiple "output cards" (since any package that has to present
"line out" jacks would need a larger enclosure). Likewise for
power amps.

And one of the streams could be 5.1 digital.

Of course, the more channels you support, the tougher the
processing requirements (so you start needing more resources
ahead of the amp/DAC/etc.)

<grin> But, that's what makes it all *fun*!

Bloody straight.

Yup. I may have to resign myself to different hardware designs
to address the differing deployments. But, I should be able
to scale the software easily (which is the hardest part of such
a project, anyway!)

If you overlay the various deployments and try to factor that for
modularization you may find your happy place.

?-)

 

[Don Y]

Hi Joseph,

Maybe the module boundary should be in terms of streams.

The problem inherently defies a simple partitioning.
E.g., if you want to drive speakers directly, you
partition the streams differently than if you want
to drive a "typical" home theater system (i.e., all
of the interfaces in a single location to connect
to external kit).

The degenerate case -- one channel, one speaker -- is the
obvious building block. But, doesn't lend itself well to
network deployment (esp wired networks).

And one of the streams could be 5.1 digital.

The CODEC I've designed really only wants to deal with two
audio "channels", at most. (Yes, I could change that)
Consider, the more "channels" you support in the stream,
the more resources you implicitly require on the ends of
that stream.

Eg., if you pack 5.1 into a single stream, then the
client device has to be able to unpack and DECODE
all of those streams in the same "unit time" that it
could otherwise have been responsible for a *single*
(or, perhaps *pair* of) "channel".

I.e., you can't just arbitrarily scale up the "width"
of each stream without consequences on the clients
(and the goal is to make the clients *really* simple
and inexpensive)

If you overlay the various deployments and try to factor that for
modularization you may find your happy place.

(sigh) Yes. The problem is trying to be "considerate"
(accommodating?) of future users' needs. I could, instead,
just focus on *my* needs and offer *my* implementations as
"reference designs" and let other folks "roll their own".
IME, this is too high a bar for many people to reach.
It seems far easier for people to repeat someone else's
success and make *minor* changes than to try to "grok"
the underlying truths and extrapolate those to another
implementation. Esp when there are lots of subtleties
in the design that they may not be qualified to understand
or accommodate (in another implementation).

<frown>