AC coupling caps...

[Phil Allison]

Tabby wrote:
------------------

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.

** Not when you know all the facts involved.

Wild mismatch is routine with audio.
------------------------------------------------------------------

** Nothing \"wild\" about having low impedance sources feeding higher impedance loads.
From my earlier post here:
\" The aim is always to transfer signals from the source into the load with as little loss of voltage as possible.
Equal value matching would degrade the s/n ratio or cause active sources to distort their output signals trying to drive too low an impedance.\"

> Knowing all the facts involved doesn\'t change that.

** But JL does not know them and so continues to post anti audio drivel.

> Such mismatch is mostly not a problem.

** So there is no \"mismatch\" at all.
The \"max power transfer theorem\" does not apply nor is transmission line impedance matching necessary.


...... Phil

 

[Ricky]

On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:

On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.

??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

--

Rick C.

--+ Get 1,000 miles of free Supercharging
--+ Tesla referral code - https://ts.la/richard11209

 

[whit3rd]

On Friday, April 28, 2023 at 4:32:44 PM UTC-7, Ricky wrote:

> ... I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

Imagine, then, the problem of IC design in the days before multilayer
metallization; examine the old TTL books, the circuit diagrams don\'t
ever show wires crossing, except for wires crossing a resistor (buried layer)
or multiple connections to a transistor base (also a buried conductive layer).

 

[Lasse Langwadt Christensen]

lørdag den 29. april 2023 kl. 01.32.44 UTC+2 skrev Ricky:

On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

you can get blind and buried vias and you remove the annular ring on unconnected layers

 

[Ricky]

On Saturday, April 29, 2023 at 5:46:12 AM UTC-4, Lasse Langwadt Christensen wrote:

lørdag den 29. april 2023 kl. 01.32.44 UTC+2 skrev Ricky:
On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained.. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

you can get blind and buried vias and you remove the annular ring on unconnected layers

Blind and buried vias are expensive, $$$. Even removing the annular ring on unconnected layers still leaves a large keep out zone around the via hole.. The point is, this impacts layers that have no routing for that trace, just the via. Adding vias can approach being counter productive, using nearly as much space than they free up by using other layers. It\'s much better to route on a single layer, where possible.

--

Rick C.

-+- Get 1,000 miles of free Supercharging
-+- Tesla referral code - https://ts.la/richard11209

 

[John Larkin]

On Sat, 29 Apr 2023 02:46:08 -0700 (PDT), Lasse Langwadt Christensen
<langwadt@fonz.dk> wrote:

lørdag den 29. april 2023 kl. 01.32.44 UTC+2 skrev Ricky:
On Friday, April 28, 2023 at 6:36:26?PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35?PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

If I have a fixed format, like VME or PXI or a 1U rack, there may be
lots of room for parts, so a board can be neatly arranged. Sometimes,
in a limited area, we want to pack as many channels as possible, and
do that using both sides for parts. Often thermals dominate density.

https://www.dropbox.com/s/hgityabkaaord12/P500E_pcb_fab.jpg?dl=0

That\'s 10 layers, parts on both sides, lots of matched-impedance
minimum-delay picosecond stuff. Trace-trace crosstalk and low-noise
power distribution were important too.

But all schematics and board layouts should be beautiful, because
that\'s satisfying and because beautiful things work better.

1 uF and 1 Mohm is a time constant of 1 second, with LF corner freq
0.16 Hz.

you can get blind and buried vias and you remove the annular ring on unconnected layers

 

[John Larkin]

On Fri, 28 Apr 2023 16:00:09 -0700 (PDT), Phil Allison
<pallison49@gmail.com> wrote:

Tabby wrote:
------------------

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.

** Not when you know all the facts involved.

Wild mismatch is routine with audio.
------------------------------------------------------------------

** Nothing \"wild\" about having low impedance sources feeding higher impedance loads.
From my earlier post here:
\" The aim is always to transfer signals from the source into the load with as little loss of voltage as possible.
Equal value matching would degrade the s/n ratio or cause active sources to distort their output signals trying to drive too low an impedance.\"

Knowing all the facts involved doesn\'t change that.

** But JL does not know them and so continues to post anti audio drivel.

Audio is pretty simple, given objective, measurable goals and not
measurement by golden ears. Most cable drivers are low Z, receivers
are hi-Z, and there is no concern for the characteristic impedance of
the cables. The 600 ohm thing is an artifact of lossy open wires on
telephone poles from a couple centuries ago.

Hey, let\'s sell a special DVM just for audiphiles. (Audiophils?) It
would be very expensive and gold of course.

 

[Ricky]

On Saturday, April 29, 2023 at 10:37:38 AM UTC-4, John Larkin wrote:

On Sat, 29 Apr 2023 02:46:08 -0700 (PDT), Lasse Langwadt Christensen
lang...@fonz.dk> wrote:

lørdag den 29. april 2023 kl. 01.32.44 UTC+2 skrev Ricky:
On Friday, April 28, 2023 at 6:36:26?PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35?PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.
If I have a fixed format, like VME or PXI or a 1U rack, there may be
lots of room for parts, so a board can be neatly arranged. Sometimes,
in a limited area, we want to pack as many channels as possible, and
do that using both sides for parts. Often thermals dominate density.

https://www.dropbox.com/s/hgityabkaaord12/P500E_pcb_fab.jpg?dl=0

That\'s 10 layers, parts on both sides, lots of matched-impedance
minimum-delay picosecond stuff. Trace-trace crosstalk and low-noise
power distribution were important too.

But all schematics and board layouts should be beautiful, because
that\'s satisfying and because beautiful things work better.

1 uF and 1 Mohm is a time constant of 1 second, with LF corner freq
0.16 Hz.

This board probably is half the density of mine. I see gobs and gobs of room for more components if you needed to add them. With my board, it\'s hard to add anything, without removing something else.

Yes, and 1 Mohm in an op amp circuit can add high frequency roll off from parasitics, as well as DC offset from the input bias current. I had to lower the FB resistor from 200k to 100k to get that under control. So now the input is 36k. But that\'s not the point. The input impedance is 50 ohms when used single ended and 600 ohms differential. That requires much larger input caps than 1 uF.

--

Rick C.

++ Get 1,000 miles of free Supercharging
++ Tesla referral code - https://ts.la/richard11209

 

[Cursitor Doom]

On Wed, 26 Apr 2023 17:23:26 -0700 (PDT), Ricky
<gnuarm.deletethisbit@gmail.com> wrote:

On Wednesday, April 26, 2023 at 6:23:13?PM UTC-4, Tabby wrote:
On Tuesday, 25 April 2023 at 03:17:48 UTC+1, Ricky wrote:
On Monday, April 24, 2023 at 5:56:16?PM UTC-4, Tabby wrote:
On Monday, 24 April 2023 at 18:51:16 UTC+1, Ricky wrote:
Typically, an AC coupling cap is driven from a low impedance source and only the load needs to be factored into the frequency equation. But when the cap is on the input, is it typical to factor in the expected source impedance, or assume a worse case of low driving impedance?

I have a design with 600 ohm differential audio inputs and I keep seeing a different response on the input and output. I finally realized this was because my input simulation used a low impedance drive, but the output included the output impedance and the load impedance. Duh!

So, do I need to beef up the input caps to handle a low impedance driver? Or, I guess I\'m asking, what is typically expected of the frequency response at the input? Is it typically for a rated driving impedance or for all source impedances?

Not much point in asking my customer. He has indicated that the various test gear this will be connected to is all over the map on this. That\'s one reason why setting a dBm level on the test gear output does not give consistent voltages measured at the interface. I don\'t expect he will know much more about the user equipment this will be connected to.

The 47 uF input caps required for worse case are a bit on the large size and the voltage rating is limited. I\'d prefer to use 33 uF caps if practical.

I guess this caught me by surprise. I thought I had completed this design, but I moved the input caps to a slightly different point in the circuit, and things changed a lot.
The design choice of whether it stays in spec for off impedances is yours alone. If the customer is using it in all sorts of situations, keeping it in spec under a wide range seems wise. Indeed it seems wise even if the customer is not. But only you choose your specs/price tradeoff. For rock bottom price consumer products that\'s more likely to be \'sod it, this is cheaper, to hell with the customer\' not so much for pro gear.
With \"pro gear\", why would the source impedance not match the load?
IME it mostly doesn\'t. Why would it really?
It\'s app dependant of course.

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

Assuming you know or can measure the inter-stage impedances, why not
just Spice it?

 

[Phil Allison]

Ricky wrote:
------------------

But that\'s not the point.
The input impedance is 50 ohms when used single ended and 600 ohms differential.
That requires much larger input caps than 1 uF.

** 12V single supply rail - right?

So here is a 6V reference too, for the op-amps.
Make that the common ground and the need for DC isolation of ins and outs disappears.


....... Phil

 

[Ricky]

On Saturday, April 29, 2023 at 4:13:29 PM UTC-4, Cursitor Doom wrote:

On Wed, 26 Apr 2023 17:23:26 -0700 (PDT), Ricky
gnuarm.del...@gmail.com> wrote:
On Wednesday, April 26, 2023 at 6:23:13?PM UTC-4, Tabby wrote:
On Tuesday, 25 April 2023 at 03:17:48 UTC+1, Ricky wrote:
On Monday, April 24, 2023 at 5:56:16?PM UTC-4, Tabby wrote:
On Monday, 24 April 2023 at 18:51:16 UTC+1, Ricky wrote:
Typically, an AC coupling cap is driven from a low impedance source and only the load needs to be factored into the frequency equation. But when the cap is on the input, is it typical to factor in the expected source impedance, or assume a worse case of low driving impedance?

I have a design with 600 ohm differential audio inputs and I keep seeing a different response on the input and output. I finally realized this was because my input simulation used a low impedance drive, but the output included the output impedance and the load impedance. Duh!

So, do I need to beef up the input caps to handle a low impedance driver? Or, I guess I\'m asking, what is typically expected of the frequency response at the input? Is it typically for a rated driving impedance or for all source impedances?

Not much point in asking my customer. He has indicated that the various test gear this will be connected to is all over the map on this. That\'s one reason why setting a dBm level on the test gear output does not give consistent voltages measured at the interface. I don\'t expect he will know much more about the user equipment this will be connected to.

The 47 uF input caps required for worse case are a bit on the large size and the voltage rating is limited. I\'d prefer to use 33 uF caps if practical.

I guess this caught me by surprise. I thought I had completed this design, but I moved the input caps to a slightly different point in the circuit, and things changed a lot.
The design choice of whether it stays in spec for off impedances is yours alone. If the customer is using it in all sorts of situations, keeping it in spec under a wide range seems wise. Indeed it seems wise even if the customer is not. But only you choose your specs/price tradeoff. For rock bottom price consumer products that\'s more likely to be \'sod it, this is cheaper, to hell with the customer\' not so much for pro gear.
With \"pro gear\", why would the source impedance not match the load?
IME it mostly doesn\'t. Why would it really?
It\'s app dependant of course.

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.
Assuming you know or can measure the inter-stage impedances, why not
just Spice it?

Sorry, I\'m not sure what you are getting at??? I have a spice simulation. What are you trying to say about it???

--

Rick C.

--- Get 1,000 miles of free Supercharging
--- Tesla referral code - https://ts.la/richard11209

 

[Ricky]

On Saturday, April 29, 2023 at 11:16:59 PM UTC-4, Phil Allison wrote:

Ricky wrote:
------------------

But that\'s not the point.
The input impedance is 50 ohms when used single ended and 600 ohms differential.
That requires much larger input caps than 1 uF.

** 12V single supply rail - right?

So here is a 6V reference too, for the op-amps.
Make that the common ground and the need for DC isolation of ins and outs disappears.


...... Phil

Common ground for what? The input? Sorry, that won\'t work. The rest of the system is using the ground as ground. Even though the inputs are differential, there is a single ended signal mode. That has to be ground referenced. Also, the CODEC can\'t be referenced to the 6V virtual ground since the ground is shared between analog and digital. That\'s the second DC block required. The output side has two DC blocks, codec to op amp and op amp to output.

It\'s only the DC block on the 50 ohm terminator I\'m concerned about. I was thinking about this on the flight today and am thinking of removing the caps in front of the terminators, and just using the ones after the terminator (3.3 uF). I will beef up the power handling on the 50 ohm terminator. The 600 ohm termination should be ok. It takes a lot more voltage to kill it.
(0.5W * 600)^0.5 ~= 17 Vrms
(0.5W * 50)^0.5 ~= 5 Vrms

So I\'ll bump the 50 ohm termination to 1W which should allow short term voltage inputs of up to 10 Vrms. Eliminating the pair of 1206 caps will more than make up the increased size of the resistor.

The existing design has 1 uF into a 51 kohm input impedance. This design will use a 39 kohm input impedance at the op amp, so a 1 uF will work ok, as will 470 nF.

--

Rick C.

--+ Get 1,000 miles of free Supercharging
--+ Tesla referral code - https://ts.la/richard11209

 

[Cursitor Doom]

On Sun, 30 Apr 2023 01:14:50 -0700 (PDT), Ricky
<gnuarm.deletethisbit@gmail.com> wrote:

On Saturday, April 29, 2023 at 4:13:29?PM UTC-4, Cursitor Doom wrote:
On Wed, 26 Apr 2023 17:23:26 -0700 (PDT), Ricky
gnuarm.del...@gmail.com> wrote:
On Wednesday, April 26, 2023 at 6:23:13?PM UTC-4, Tabby wrote:
On Tuesday, 25 April 2023 at 03:17:48 UTC+1, Ricky wrote:
On Monday, April 24, 2023 at 5:56:16?PM UTC-4, Tabby wrote:
On Monday, 24 April 2023 at 18:51:16 UTC+1, Ricky wrote:
Typically, an AC coupling cap is driven from a low impedance source and only the load needs to be factored into the frequency equation. But when the cap is on the input, is it typical to factor in the expected source impedance, or assume a worse case of low driving impedance?

I have a design with 600 ohm differential audio inputs and I keep seeing a different response on the input and output. I finally realized this was because my input simulation used a low impedance drive, but the output included the output impedance and the load impedance. Duh!

So, do I need to beef up the input caps to handle a low impedance driver? Or, I guess I\'m asking, what is typically expected of the frequency response at the input? Is it typically for a rated driving impedance or for all source impedances?

Not much point in asking my customer. He has indicated that the various test gear this will be connected to is all over the map on this. That\'s one reason why setting a dBm level on the test gear output does not give consistent voltages measured at the interface. I don\'t expect he will know much more about the user equipment this will be connected to.

The 47 uF input caps required for worse case are a bit on the large size and the voltage rating is limited. I\'d prefer to use 33 uF caps if practical.

I guess this caught me by surprise. I thought I had completed this design, but I moved the input caps to a slightly different point in the circuit, and things changed a lot.
The design choice of whether it stays in spec for off impedances is yours alone. If the customer is using it in all sorts of situations, keeping it in spec under a wide range seems wise. Indeed it seems wise even if the customer is not. But only you choose your specs/price tradeoff. For rock bottom price consumer products that\'s more likely to be \'sod it, this is cheaper, to hell with the customer\' not so much for pro gear.
With \"pro gear\", why would the source impedance not match the load?
IME it mostly doesn\'t. Why would it really?
It\'s app dependant of course.

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.
Assuming you know or can measure the inter-stage impedances, why not
just Spice it?

Sorry, I\'m not sure what you are getting at??? I have a spice simulation. What are you trying to say about it???

Mea culpa. I didn\'t read your original post properly soz about that.

 

[Lasse Langwadt Christensen]

lørdag den 29. april 2023 kl. 15.25.56 UTC+2 skrev Ricky:

On Saturday, April 29, 2023 at 5:46:12 AM UTC-4, Lasse Langwadt Christensen wrote:
lørdag den 29. april 2023 kl. 01.32.44 UTC+2 skrev Ricky:
On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600..
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end. WTF are you talking about???

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

you can get blind and buried vias and you remove the annular ring on unconnected layers
Blind and buried vias are expensive, $$$.

it adds a few more production steps but since your board is probably small it might not add that much per board

 

[Phil Allison]

Ricky wrote:
-------------------

But that\'s not the point.
The input impedance is 50 ohms when used single ended and 600 ohms differential.
That requires much larger input caps than 1 uF.

** 12V single supply rail - right?

So there is a 6V reference too, for the op-amps.
Make that the common ground and the need for DC isolation of ins and outs disappears.


...... Phil
Common ground for what?

** Having split +/- supplies for the input stage/s eliminates your problem with input caps.
Create a -12V one if need be, standard practice with digital audio devices that have analog ins and outs.

Enjoy that corner you have painted yourself into.

...... Phil

 

[John S]

On 4/28/2023 6:00 PM, Phil Allison wrote:

Tabby wrote:
------------------

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.

** Not when you know all the facts involved.

Wild mismatch is routine with audio.
------------------------------------------------------------------

** Nothing \"wild\" about having low impedance sources feeding higher impedance loads.
From my earlier post here:
\" The aim is always to transfer signals from the source into the load with as little loss of voltage as possible.
Equal value matching would degrade the s/n ratio or cause active sources to distort their output signals trying to drive too low an impedance.\"

Knowing all the facts involved doesn\'t change that.

** But JL does not know them and so continues to post anti audio drivel.

Such mismatch is mostly not a problem.

** So there is no \"mismatch\" at all.
The \"max power transfer theorem\" does not apply nor is transmission line impedance matching necessary.


..... Phil

+1 (or more)

 

[Ricky]

On Sunday, April 30, 2023 at 8:58:21 PM UTC-4, Phil Allison wrote:

Ricky wrote:
-------------------

But that\'s not the point.
The input impedance is 50 ohms when used single ended and 600 ohms differential.
That requires much larger input caps than 1 uF.

** 12V single supply rail - right?

So there is a 6V reference too, for the op-amps.
Make that the common ground and the need for DC isolation of ins and outs disappears.


...... Phil
Common ground for what?
** Having split +/- supplies for the input stage/s eliminates your problem with input caps.
Create a -12V one if need be, standard practice with digital audio devices that have analog ins and outs.

Except that it doesn\'t. One of my concerns is that applying something like 5V or 12V to the input will exceed the rating of the 50 ohm input impedance resistor. The 600 ohm input impedance resistor requires a much larger input to over drive it. The design is very, very tight, so I was focusing on keeping the resistors small. But the caps are much larger. It\'s probably best to increase the wattage of the 50 ohm resistors to suit. I need to check to see at what input the switches are over current though.


> Enjoy that corner you have painted yourself into.

It\'s nice when you have total control over a design. This design is a 4 square inch daughter card, with an FPGA, RS-422 I/O, and two channels of audio input and output with selectable impedances. I suppose I could construct a daughter card for the daughter card. Otherwise, there is no place possible to add a -12V supply.

Actually, -12V is supplied to the daughter card, but it\'s only a few mA, originally intended for biasing serial port drivers. These analog I/Os have up to ~50 mA each and there are two differential outputs. If you can design a circuit for this, that will fit in 0.25 inches square (single sided components), I\'d be happy to use it. Oh, wait, it would probably need to use the +12V line and that might be close to maxing out.

--

Rick C.

-+- Get 1,000 miles of free Supercharging
-+- Tesla referral code - https://ts.la/richard11209

 

[Tabby]

On Saturday, 29 April 2023 at 00:00:13 UTC+1, Phil Allison wrote:

Tabby wrote:
------------------

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.

** Not when you know all the facts involved.

Wild mismatch is routine with audio.
------------------------------------------------------------------

** Nothing \"wild\" about having low impedance sources feeding higher impedance loads.

it\'s just a mismatch. And the ratio can be large.

From my earlier post here:
\" The aim is always to transfer signals from the source into the load with as little loss of voltage as possible.

If that were true the load impedance would always be as high as practical. That is clearly not the one aim or the practice.

Equal value matching would degrade the s/n ratio or cause active sources to distort their output signals trying to drive too low an impedance.\"
Knowing all the facts involved doesn\'t change that.
** But JL does not know them and so continues to post anti audio drivel.
Such mismatch is mostly not a problem.
** So there is no \"mismatch\" at all.

low Z driving high Z is by definition a mismatch. And is mostly no problem with audio.

> The \"max power transfer theorem\" does not apply nor is transmission line impedance matching necessary.

indeed.

 

[Tabby]

On Saturday, 29 April 2023 at 00:32:44 UTC+1, Ricky wrote:

On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end.

ok

> WTF are you talking about???

the reduction of coupling caps to trim off the bottom end to enable speakers & amps to perform better. It\'s common practice.

This design is not so cost constrained, but it is very size constrained. Higher value caps can be larger. Also, with the issues of supply, locking into a higher value cap in the smallest possible footprint, means you have no alternates if supply gets tight.

No, this is not a trivial design decision for some cases. In designs like Larkin\'s, where he uses a board some three times the necessary size and loves to marvel at the beauty of the regular arraignments of the long traces, you can use any part you want, and even provide multiple footprints for alternates. I don\'t have that luxury. I barely have room for vias. They are actually a PITA, taking up room on all layers. Too bad I can\'t just eliminate them.

 

[Ricky]

On Monday, May 1, 2023 at 6:49:13 PM UTC-4, Tabby wrote:

On Saturday, 29 April 2023 at 00:32:44 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 6:36:26 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 22:24:52 UTC+1, Ricky wrote:
On Friday, April 28, 2023 at 1:26:35 PM UTC-4, Tabby wrote:
On Friday, 28 April 2023 at 04:14:05 UTC+1, Phil Allison wrote:
John Larkin wrote:
-----------------------------

My understanding is that impedance mismatch can cause issues of frequency response, such as the one I\'m seeing.

In a distribution system, with long wires, you match the source and/or the load with
the characteristic line impedance, because too LOW resistance makes the
wire an inductor, and too HIGH resistance makes the wire a capacitor.
High-Z you lose high frequency response, low-Z you lose low frequency.
Matched, the response is flat.

in some cases. In many there is no real world problem. It\'s app dependant.

Microphone cords are the only really long audio wires where it matters, in most
cases.

The differential impedance of a typical shielded twisted pair is
around 100 ohms... nothing like 600, much less 1200.
** No consequence in the audio band with the usual short (< 50m) runs involved.
Cable C is only 100 to 250 pF per meter.

FYI by far the most common mic impedance is 200ohms - not 600.
Audio systems are, usually, wildly mismatched.
** Not when you know all the facts involved.
Wild mismatch is routine with audio. Knowing all the facts involved doesn\'t change that. Such mismatch is mostly not a problem.
Except that it does impact the low end frequency response. That is the filter that, in general, uses a larger and a bit more expensive capacitor, so this would want to be minimized in consumer electronics. If the output impedance of the signal source would impact this response, they would need to make a decision as to how much they want to spend to accommodate the various signal sources. It\'s only pennies, but they count pennies in commercial gear.
Ie it\'s a design decision that\'s trivial to make a nonproblem.
Or you can use it as a plus, trimming down on frequencies a nonideal speaker can do without.
??? I\'m supposed to provide 20 Hz at the low end.
ok
WTF are you talking about???
the reduction of coupling caps to trim off the bottom end to enable speakers & amps to perform better. It\'s common practice.

Do you mean setting the value of the coupling cap to provide a high pass filter? That\'s rather a DUH. There are two reasons to reduce the value of the coupling cap. One is to make it cheaper, the other is to tune the filter to the frequencies of interest. But setting the corner frequency of the filter has to take into account all other aspects of the use, which is where this thread started. Perhaps you might want to read the first few posts again.

--

Rick C.

-++ Get 1,000 miles of free Supercharging
-++ Tesla referral code - https://ts.la/richard11209