Further to your proposed circuit showing that I'm not exaggerating, below the simulation of the generated output noise and the worst case CMRR.
Noise with 3.4uV is quite a bit higher as most modern main amps.
Output is 471mV for a 4Volt pk-pk differential input signal.
Worst case output is 1.92mV for the same 4Volt pk-pk CM input signal when using 0.1% resistors.
CMRR is 20*log(471/1.92) = 47.8dB, worse as the differential main amp.
Hans
Last edited:
Well, I'll be... 30 years servicing electronics and I'm still learning new things... I re-ran my sims here and indeed breaking that ground connection makes all the difference. I stand corrected.
But I would wonder how often real world power amps give access to both sides of the differential inputs. If you are designing for this, I get your point now, but bridging a pair of amplifiers already out in the wild (where I would most likely see them) is a different story.
Last edited:
Well, in my own defence that was a quick sketch up to demonstrate a principle... I doubt I would ever actually use that circuit in a real project.
Last edited:
I am not wrong, the common-mode signal sails straight through to be a common mode signal on the speaker wires. The speaker can only react to the differential component of course, but the power amps see the common-mode signal and push it out on the speaker wires. That's what I meant by "Your approach pushes the common mode signal right through the amp and out the other side"
The differential signal is amplified by perhaps x20 to x50 or so, and the common-mode is amplified by unity (ideally). IE you can view the common-mode signal as being injected (inverted) onto the power rails (from the point of view of the differential signal). Most power amps are not designed to handle high frequency interference on the power rails (remember this is not something the decoupling caps can address because from the POV of the speaker signal the ground carries the inverted common mode signal).
[/quote]
But on top of that you get the suppression at the LS output between the 2 channels that will add at least another 20dB to 40dB resulting in 60 to 80dB in total.
While I now see what they're doing and how it works, I'm with you in viewing this as a specific-case hack. When you're looking at a couple of RCA jacks on the back of an amplifier, splitting the XLR balanced signal to use as a cheap bridging technique is indeed fraught with problems.
A CM signal leeds to no current flowing into the speaker.
This can be compared to driving an amp without load.
How can this possibly lead to an injection of whatever anomaly in the power rails.
Maybe you can clarify this with a simulation.
That's right, but whe have a 2 step rejection mechanism. The first one caused by the used topology that gives a 34dB rejection for the OP's case, practically independent of the above mentioned CMRR and matching of feedback resistors.
The second one being the rejection at the LS output is dependent on the matching of feedback resistors and CMRR of both amplifiers.
Even when using 0.1% feedback resistors, the difference at the LS output will most likely stil be way above the added effect of the CMRR of the amplifiers.
That's why I mentioned a total CMRR for the diff amp of above 60dB and yes this is frequency dependant, diminishing at ultrasonic frequencies.
In the situation you describe, both inputs at 10V, no current is flowing into the load, so what's the problem ?
Apart from this, a CM signal of 10V is ridiculous high and can only be caused by a sharp peak of very short duration, causing clipping and possibly oscillations in a SE amp, much, much worse as what happens in this diff amp.
Have you ever encountered a stability problem or a fried Zobel network problem with two bridged amps when driven with opposite polarity signals, I've never heard of this happening.
To conclude, I don't share your concerns, too my opinion it's all too far fetched from a real world situation.
When putting a 10Khz input signal on an amp at full power, you can blow any Zobel network, but that never happens with music polluted with ultrasonic peaks.
I would like to be convinced in a simulation where an SE amp can handle large ultrasonic peaks better that a diff amp with this topology.
Hans
Last edited:
Any conversion with a reasonable, well designed circuit will NOT add any sonic imprint!
Hi
XLR input to balanced amps: Bryston uses a pair of differential opamp blocks with their inputs cross-wired to generate balanced drive signals to the two power amps. I believe this is the best solution for the OP, since he was nominally starting with existing stereo PAs. This overall scheme requires no modification of the PAs, nor cross-connecting them, or any other trick that might be used in an integrated balanced design.
In the OP's original drawing, his mistake was to show a signal into pin-1 of the XLR. Although this could represent a ground difference, it does make it look nonintuitive to tie pin-1 to chassis as it should be. The independent PAs cannot really do anything about common-mode signals, where the cross-diff arrangement above can.
XLR input to balanced amps: Bryston uses a pair of differential opamp blocks with their inputs cross-wired to generate balanced drive signals to the two power amps. I believe this is the best solution for the OP, since he was nominally starting with existing stereo PAs. This overall scheme requires no modification of the PAs, nor cross-connecting them, or any other trick that might be used in an integrated balanced design.
In the OP's original drawing, his mistake was to show a signal into pin-1 of the XLR. Although this could represent a ground difference, it does make it look nonintuitive to tie pin-1 to chassis as it should be. The independent PAs cannot really do anything about common-mode signals, where the cross-diff arrangement above can.
Last edited:
Good to see that you keep going on and on.
Since your Spkr output is floating, how did you connect the probe of your scope, because signal gnd is no longer a reliable reference.
Or do you have a floating scope on batteries connected directly to the Spkr ?
In that case it would also be nice to record the CMRR by connecting the same input signal between signal gnd and both inputs connected together.
Hans
Would it be a lot of work for you to measure the CMRR with the same input level (25mV ?) as in your first THD measurement ?
Hans
The input of the power amplifier is at -30.1dBV. The closed loop gain of the amplifier is 32dB. First pic is a "diff" and second one "CM".
Attachments
Last edited: