What if I wanted CMRR in the 80 or 90 dB range? What tolerance matching would I need then?
One of the Bryston engineers here says that the amps usually go into mid to high 70s dB CMRR for their SE output amps (so 3B and 4B in my case) and in the 80 dB range for the 7B which has balanced output. It would suck to limit expensive electronics by not spending a little more time matching some cheap parts.
BTW, not sure about the adapters - the Bryston amps are already very deep and I barely have any space at all behind them, so I might have to build the pads into the cables. But I don't have massive amounts of XLR cables, and those are staying with the hifi, I'm good about keeping things single-purpose like that.
You can find the post here:
to balance or not to balance.
I tried reading this page:
TLV6001: Calculate resistor tolerance to meet the minimum CMRR - Amplifiers forum - Amplifiers - TI E2E support forums
but I don't really know how it relates to the amplifier. Is the differential gain (G) going to be unity? I guess it is for a single ended amp - the balanced input stage will merely turn things into unity. But what about the 7B? That's basically two separate amps. Is the differential gain going to be 2 there?
One of the Bryston engineers here says that the amps usually go into mid to high 70s dB CMRR for their SE output amps (so 3B and 4B in my case) and in the 80 dB range for the 7B which has balanced output. It would suck to limit expensive electronics by not spending a little more time matching some cheap parts.
BTW, not sure about the adapters - the Bryston amps are already very deep and I barely have any space at all behind them, so I might have to build the pads into the cables. But I don't have massive amounts of XLR cables, and those are staying with the hifi, I'm good about keeping things single-purpose like that.
You can find the post here:
to balance or not to balance.
rmurray said:Hello Bryston fans. I just read the info explaining that the 4,3,2 sst squared amps are not 'fully' balanced. Is it really worth using this connection over my rca connects? Would there be sonic advantages ? thanks,......... RMURRAY.
James Tanner said:
I tried reading this page:
TLV6001: Calculate resistor tolerance to meet the minimum CMRR - Amplifiers forum - Amplifiers - TI E2E support forums
but I don't really know how it relates to the amplifier. Is the differential gain (G) going to be unity? I guess it is for a single ended amp - the balanced input stage will merely turn things into unity. But what about the 7B? That's basically two separate amps. Is the differential gain going to be 2 there?
Last edited:
The CMRR is not relevant because common mode noise will be way, way, way below the noise floor of your system unless your cables a just wet pieces of string, or really, really shonky rubbish.
CMRR is relevant for microphone inputs where µVs of signal transverse 10s or 100s of metres of cable in extreme EMI environments like television studios or stages with megawatts of dimmed lighting, or with nominally +4dBµ signal at transmission line receivers where cables are 10s or 100s of miles long.
In you case the CMRR is irrelevant. The benefit of the balanced inputs of the 10B and the Brystons comes from the signal not needing to be referenced to ground.
In fact the 'balanced' output from the Fiio is likely to have significant common mode signal in its output because it isn't truely balanced, but again that isn't relevant because the common mode signal is not noise or distortion in this case.
CMRR is relevant for microphone inputs where µVs of signal transverse 10s or 100s of metres of cable in extreme EMI environments like television studios or stages with megawatts of dimmed lighting, or with nominally +4dBµ signal at transmission line receivers where cables are 10s or 100s of miles long.
In you case the CMRR is irrelevant. The benefit of the balanced inputs of the 10B and the Brystons comes from the signal not needing to be referenced to ground.
In fact the 'balanced' output from the Fiio is likely to have significant common mode signal in its output because it isn't truely balanced, but again that isn't relevant because the common mode signal is not noise or distortion in this case.
The differential gain of your Bryston amplifiers is 29dB, the differential gain of the 10B is 0dB.
Another question I missed: the value or tolerance of the shunt resistor in the pad has no impact whatsoever on the balancing accuracy of the pad, only the accuracy of the attenuation. It also has zero effect on the CMRR of the following balanced or differential input. The matching of the values of the two series resistors is critical to maintaining a balanced signal, though.
That's why even if you buy 1% resistors, you should still select matching pairs on test for each attenuator, however as per my previous post accuracy has diddly squat* impact in your system. (*technical term roughly equivalent to F. A.) It would be quite different if the attenuator was being used on the end of a long microphone cable, or a long transmission line, for the reasons above.
Another question I missed: the value or tolerance of the shunt resistor in the pad has no impact whatsoever on the balancing accuracy of the pad, only the accuracy of the attenuation. It also has zero effect on the CMRR of the following balanced or differential input. The matching of the values of the two series resistors is critical to maintaining a balanced signal, though.
That's why even if you buy 1% resistors, you should still select matching pairs on test for each attenuator, however as per my previous post accuracy has diddly squat* impact in your system. (*technical term roughly equivalent to F. A.) It would be quite different if the attenuator was being used on the end of a long microphone cable, or a long transmission line, for the reasons above.
Last edited:
Haha, I see, yeah, that makes sense. I'll do things that way, then. The microphone cable thing put it into perspective - yeah, the interconnects are really short.
Just out of curiosity - how would one calculate the noise introduced by the shunt resistor? Or the series resistors? I doubt that they would matter here but I'm just really curious here.
Just out of curiosity - how would one calculate the noise introduced by the shunt resistor? Or the series resistors? I doubt that they would matter here but I'm just really curious here.
In the pad, only the shunt resistor is implicated in noise generation. (More correctly the resistance measured across the shunt resistor which also includes the parallel network network formed by the series resistors and output resistance of the source (10B), and the parallel input impedance of the load; for practical purposes all those can be ignored and the Rs noise is the dominant factor, and the actual noise level noise will be reduced by the parallel networks.)
There is a thermal noise calculator here:
Thermal noise of resistor – Calculator | Audio Perfection
If you take my suggestion of 1200 ohms for Rs, the noise calculates to -124dBv, which is quite a way below the noise floor of the 10B.
Another way of thinking about it is that when you plug the attenuator into the amplifier, it will potentially drop the noise level at the output of the amplifier (provided that input noise is the dominant source, which it should be for a competently designed amplifier). This is because the resistance component of the input impedance of the amplifier will be paralleled by the shunt, so the effective resistance is lowered, along with thermal noise at the input. This is why the noise level at the output of an RIAA preamp drops when a phono cartridge is plugged in, compare to open circuit inputs, and drops even more with shorted inputs, for example.
There is a thermal noise calculator here:
Thermal noise of resistor – Calculator | Audio Perfection
If you take my suggestion of 1200 ohms for Rs, the noise calculates to -124dBv, which is quite a way below the noise floor of the 10B.
Another way of thinking about it is that when you plug the attenuator into the amplifier, it will potentially drop the noise level at the output of the amplifier (provided that input noise is the dominant source, which it should be for a competently designed amplifier). This is because the resistance component of the input impedance of the amplifier will be paralleled by the shunt, so the effective resistance is lowered, along with thermal noise at the input. This is why the noise level at the output of an RIAA preamp drops when a phono cartridge is plugged in, compare to open circuit inputs, and drops even more with shorted inputs, for example.
Last edited:
Great link, thanks!
Thanks for the explanation. I'm not sure I follow. I agree that the shunt resistor is presented to the 10B's output in parallel to the input resistance of the amplifier. And I agree that effective resistance is lowered. But I don't know why thermal noise is being lowered at the input? Thermal noise of what? Do you mean the 10B's output noise? And why is it being lowered? Is it because a portion of it goes through the shunt resistor?
Sorry I'm so slow -- started writing this in response to your post #142. 
Some still relevant ..
What is the sound quality improvement you expect from 80 to 90dB CMRR, vs 50 to 60dB CMRR?
Sorry if I'm out of line here, but I'm starting to worry a little about you .. As somebody with first-hand experience with Numerical OCD Tendencies myself, I can attest that this rabbit hole that keeps drawing you in, is not serving you well. .. at least not as well as the investment you're making in it at times.
Cheers
Some still relevant ..What is the sound quality improvement you expect from 80 to 90dB CMRR, vs 50 to 60dB CMRR?
Sorry if I'm out of line here, but I'm starting to worry a little about you .. As somebody with first-hand experience with Numerical OCD Tendencies myself, I can attest that this rabbit hole that keeps drawing you in, is not serving you well.
.. at least not as well as the investment you're making in it at times.Cheers
Last edited:
Sorry I'm so slow -- started writing this in response to your post #142.
What is the sound quality improvement you expect from 80 to 90dB CMRR, vs 50 to 60dB CMRR?
Sorry if I'm out of line here, but I'm starting to worry a little about you .. As somebody with first-hand experience with Numerical OCD Tendencies myself, I can attest that this rabbit hole that keeps drawing you in, is not serving you well.
Cheers
I wasn't imagining any improvement at the moment - all I was looking for was that the expensive electronics aren't being limited by a bunch of cheap resistors. However, as John said, it doesn't really matter at all - even less so for the short cable runs I'm dealing with.
Yes, that's correct. An attenuator doesn't need to be exactly 20.00dB, usually 19.8 or 20.1 would be close enough for practical purposes.
However to preserve CMRR the matching of the two series resistors does have to be exact. 60dB requires matching to 1/1000 or 0.1%, 80dB requires 1/10,000 or 0.01%. 100dB 0.001% matching.
Looking at an output of a circuit e.g. a microphone, there is noise from the circuit, but even if the circuitry was perfectly quiet, there will be noise generated by the resistive component of the output stage. This is Johnson noise. The aim of good circuit design is to have the following electronics contributing as little noise as possible over and above the Johnson noise inherent in the source resistance, be that a pickup, microphone or another amplifier stage.
Lets say a moving coil microphone has a coil with a resistance of 150Ω; the Johnson noise at 25ºC is -133.1dBV. If a preamplifier was designed such that it's noise floor was -132.6dBv when the input terminal is connected to a 150Ω resistor, then that preamplifier's noise is 0.5dB above the theoretical best possible performance for that source. The only way to make it quieter is to lower the temperature everything is operating at.
I took you question to mean how much noise is being added by the shunt resistor in the attenuator. The shunt resistor is across the input terminals of the stage following it, and the Johnson noise it generates is what I calculated previously.
If you have, say, two 10kΩ resistors in parallel, the effective resistance is 5kΩ; the Johnson noise across two 10kΩ resistors in parallel is the same as the Johnson noise across a single 5kΩ resistor.
The output impedance of the previous stage added to the series resistors, and the input impedance of the next stage, are both in parallel with the the shunt resistance of the attenuator, so the actual Johnson noise generator is across the effective resistive part of all of these components paralleled, which will be less than the value of the shunt resistance on its own, so the noise will be less.
Hope this helps.
Input noise of the amplifier with attenuator in the present signal chain is the RMS sum of three components:
1. crossover output noise attenuated same as any other signal
2. effective source impedance provided by the attenuator and XO output (dominated by the former, being = (Rsource + R1) || R2)
3. Amplifier effective input noise itself
I actually created a little online calculator for just this kind of application at one point.
It be mentioned that, especially with resistors easily being hand-matched by the electronics enthusiast, the U-type attenuator with 4 resistors (R2 split in half and a shield connection established at midpoint) is generally preferable to the floating 3-resistor type, and more so the greater the attenuation required. This may seem counterintuitive in the light of potential common-mode to differential mode conversion brought about by resistor mismatch (and I certainly wouldn't advise using a pot), however the latter variant has a crucial weakness - it does not attenuate any common-mode components at all (just differential mode), while the 4-resistor one very much does. You can see how this may put considerable strain on input CMRR as the CM/DM ratio degrades as attenuation goes up.
So the crossover is suffering from Pin 1 Problem? Hmm. I would be inclined to inspect its construction for the feasibility of a mod to establish AES48-2005 compliance without making too much of a mess of serviceability. Should the unit feature panel-mounted XLR jacks connected with flying leads, it would be fairly easy to remove the existing pin 1 wire and establish a connection to chassis instead... PCB-mounted jobs would make things more difficult, then the scalpel may have to come out. Long-term such a mod strikes me as the more sensible option... having to hack up your cables is annoying, and I literally cannot think of any upsides to a device with such a defect left in. It is simply a construction problem that the designers were not sufficiently aware of at the time and would have avoided otherwise. It may still be overruled by concerns of assembly to this day, but that's mostly in the penny-pinching world of inexpensive active monitors and the like (5"/6" KRK Rokits come to mind).
1. crossover output noise attenuated same as any other signal
2. effective source impedance provided by the attenuator and XO output (dominated by the former, being = (Rsource + R1) || R2)
3. Amplifier effective input noise itself
I actually created a little online calculator for just this kind of application at one point.
It be mentioned that, especially with resistors easily being hand-matched by the electronics enthusiast, the U-type attenuator with 4 resistors (R2 split in half and a shield connection established at midpoint) is generally preferable to the floating 3-resistor type, and more so the greater the attenuation required. This may seem counterintuitive in the light of potential common-mode to differential mode conversion brought about by resistor mismatch (and I certainly wouldn't advise using a pot), however the latter variant has a crucial weakness - it does not attenuate any common-mode components at all (just differential mode), while the 4-resistor one very much does. You can see how this may put considerable strain on input CMRR as the CM/DM ratio degrades as attenuation goes up.
So the crossover is suffering from Pin 1 Problem? Hmm. I would be inclined to inspect its construction for the feasibility of a mod to establish AES48-2005 compliance without making too much of a mess of serviceability. Should the unit feature panel-mounted XLR jacks connected with flying leads, it would be fairly easy to remove the existing pin 1 wire and establish a connection to chassis instead... PCB-mounted jobs would make things more difficult, then the scalpel may have to come out. Long-term such a mod strikes me as the more sensible option... having to hack up your cables is annoying, and I literally cannot think of any upsides to a device with such a defect left in. It is simply a construction problem that the designers were not sufficiently aware of at the time and would have avoided otherwise. It may still be overruled by concerns of assembly to this day, but that's mostly in the penny-pinching world of inexpensive active monitors and the like (5"/6" KRK Rokits come to mind).
Thanks, that's a really interesting idea. How does the 4-resistor circuit attenuate common mode components?
Which shell should I connect the mid point to: the shell of the output (the crossover) or the shell of the input (the amp)?
Is it fine if instead of connecting to shell, I connect it to pin 1 of the crossover's output? It's not connected to pin 1 of the amp's input.
Which shell should I connect the mid point to: the shell of the output (the crossover) or the shell of the input (the amp)?
Is it fine if instead of connecting to shell, I connect it to pin 1 of the crossover's output? It's not connected to pin 1 of the amp's input.
Wow, thanks for that! I did not realise the lack of common mode reduction with a 3 resistor U attenuator; of course the feed through of common mode is obvious when you think about it.
This is fine (good, actually) as long as we're only talking line-level stuff. And maybe modern mic inputs are more tolerant of higher output mic's than 'back in the day'. But having once had to build pads into many cheap board inputs, I gotta mention to include a 50V capacitor at the split-R2-node ..
Aww, forget it -- these are not going to get exposed to Phantom outputs .. Dunno what I was thinkin' ..
And you may want to check AES48-2005 (as Stephan said) -- pretty sure you want to use pin 1, and leave the shell out of the circuit.
Cheers
) as long as we're only talking line-level stuff. And maybe modern mic inputs are more tolerant of higher output mic's than 'back in the day'. But having once had to build pads into many cheap board inputs, I gotta mention to include a 50V capacitor at the split-R2-node ..Aww, forget it
-- these are not going to get exposed to Phantom outputs .. Dunno what I was thinkin' ..And you may want to check AES48-2005 (as Stephan said) -- pretty sure you want to use pin 1, and leave the shell out of the circuit.
Cheers
Last edited:
This is fine (good, actually
Aww, forget it
And you may want to check AES48-2005 (as Stephan said) -- pretty sure you want to use pin 1, and leave the shell out of the circuit.
Cheers
Sorry, who's Stephan? I didn't see any mention of AES48-2005 or any posts by a person named Stephan, maybe I forgot?
At this stage I think I'll just build 20 dB pads into the cables and that'll be a great step.
I guess I should ask the next question then. It would be nice to be able to change the attenuation. Unfortunately that means attenuating 6 pads at once (8 pads once I get separate amps for the XBDs and a separate crossover or a digital crossover). This would mean throwing 16 poles, and this would then go to as many outputs as I would need levels. Could either build 3 levels (0 dB, -20 dB, -40 dB), or maybe I could have some sort of incremental switching happening: a -20 dB pad, -10 dB, -5 dB, -2.5 dB, each switched in or bypassed as needed.
The logical conclusion to this is to have a remotely controlled attenuation box. Operated from some sort of remote (even if wired). I imagine a bunch of relays switching different attenuation levels in and out. I don't really need to switch it that smoothly, but it would be great to be able to switch it quickly without having to crawl behind my tv into the tiny space the electronics are sitting. What would be the best way to set this sort of thing up? Relays? Reed switches? And which kinds? I would love it to be, at best, inaudible from 2m away. I guess I can pad whatever mechanical elements with acoustic foam, use suspension, etc. Hearing relays thonking would be pretty distracting during monitoring.
I'm not sure about the reliability: Once the convenience of a remote is in, I'm not sure if I would be changing them once a week several times a day.
I'm not sure what the best way would be to do that. I assume up to something like 48Vpp, relays and reed switches will not affect the sound. Am I correct to think so? I heard that relays will fail if used too often or if not used enough; are reed switches a better solution here? I saw some at some point and they were essentially a metal leaf spring inside a glass envelope, operated by what seems to be an electromagnet. Could that put buzz in the audio?
Eventually I'd like to build up a more comprehensive monitor control box: route to different monitors (eg PMC electronics + speakers vs a set of small active speakers); amplify every band on the PMCs separately vs 2-way or 1-way, while re-routing the speaker bands to specific powered outputs; use Bryston amps vs another set of amps; switch in high pass and low pass filters or high and low shelves; turn on and off a monitoring bridge.
I guess I should ask the next question then. It would be nice to be able to change the attenuation. Unfortunately that means attenuating 6 pads at once (8 pads once I get separate amps for the XBDs and a separate crossover or a digital crossover). This would mean throwing 16 poles, and this would then go to as many outputs as I would need levels. Could either build 3 levels (0 dB, -20 dB, -40 dB), or maybe I could have some sort of incremental switching happening: a -20 dB pad, -10 dB, -5 dB, -2.5 dB, each switched in or bypassed as needed.
The logical conclusion to this is to have a remotely controlled attenuation box. Operated from some sort of remote (even if wired). I imagine a bunch of relays switching different attenuation levels in and out. I don't really need to switch it that smoothly, but it would be great to be able to switch it quickly without having to crawl behind my tv into the tiny space the electronics are sitting. What would be the best way to set this sort of thing up? Relays? Reed switches? And which kinds? I would love it to be, at best, inaudible from 2m away. I guess I can pad whatever mechanical elements with acoustic foam, use suspension, etc. Hearing relays thonking would be pretty distracting during monitoring.
I'm not sure about the reliability: Once the convenience of a remote is in, I'm not sure if I would be changing them once a week several times a day.
I'm not sure what the best way would be to do that. I assume up to something like 48Vpp, relays and reed switches will not affect the sound. Am I correct to think so? I heard that relays will fail if used too often or if not used enough; are reed switches a better solution here? I saw some at some point and they were essentially a metal leaf spring inside a glass envelope, operated by what seems to be an electromagnet. Could that put buzz in the audio?
Eventually I'd like to build up a more comprehensive monitor control box: route to different monitors (eg PMC electronics + speakers vs a set of small active speakers); amplify every band on the PMCs separately vs 2-way or 1-way, while re-routing the speaker bands to specific powered outputs; use Bryston amps vs another set of amps; switch in high pass and low pass filters or high and low shelves; turn on and off a monitoring bridge.
Agreed, so obvious once you see it, but I'm not sure how the 4-resistor pad attenuates the differential signal? I understood that the 3-resistor pad was a 3-element resistive divider that looked like this:
(in +)
|
| Rser/2
|
(out +)
|
| Rpar
|
(out -)
|
| Rser/2
|
(in -)
Is that correct?
If so, then the 4-resistor pad should work the same way:
(in +)
|
| Rser/2
|
(out +)
|
| Rpar/2
|
(gnd)
|
| Rpar/2
|
(out -)
|
| Rser/2
|
(in -)
Also, how exactly does this get rid of common mode voltages? Let's say we have a pad where Rpar/2 = Rser/2, and there's a 1V common mode voltage. Here's how I imagine it works. Is that correct?
Code:
1V 1V
| |
| |
| |
(out +) (out-) <-- 0.5V
| |
| |
| |
+----------+-- (gnd)So it gets attenuated.
However, what happens if gnd introduces noise? Let's say (in+) = 5V, (in-) = -5V, gnd=1V.
Code:
(output) (attenuator) (input)
__________________________________________________
+---(out+)------(in+)
| | |
| | | Rser/2
| | |
|R1 | R2 (out+)----------(in+)-----+
| | | | |
| | | Rpar/2 | | R4
| | | | |
| (gnd1)------(gnd1) | R3 (gnd2)
| | | | |
| | | Rpar/2 | | R4
| | | | |
| | R2 (out-)----------(out+)----+
| | |
| | | Rser/2
| | |
+---(out-)------(in-)
gnd1
|
| Rg
|
gnd2Rg will be roughly ~1 Ohm. I can imagine what's going to be in Rser and Rpar. But what's going to be in R1, R2, R3, and R4? I'll have to look at the schematics. Are there good values to assume here? Thanks.