There are various considerations in design trade off. The input connected the way Mark's schematic shows does reduce interaction between the source device when it is connected with the source ground. We generally have not so much control of how the source is designed, and it is a good idea to minimize their effects.
This is certainly different from what I have tried to analyse in the past.
It appears that Gain set by the NFB is changed when HBRR><0 ohms.
For non-inverting amp: Gain = Rupper / Rlower
I have always stopped there. I did not consider that the resistance in the voltage reference link Signal Return to Power Ground as part of the negative feedback route.
If I take post954 and include HBRR as being inside the NFB loop,
then when a non zero HBRR is added in the gain becomes: Gain = Rupper / {Rlower +HBRR}
typically Rupper could be 9k (more typically for generallised 3886 it would be 27k) and Rlower could be 1k
and the new gain equation becomes marginally lower.
9000/1001 + 1 =9.991009 instead of 10. That amounts to a 0.09% reduction in gain. not subtracted from output, but a proportional reduction in output.
Is it distortion?
But I have to ask:
Why does the NFB loop include HBRR?
Why does it not stop at the Input Signal Return. The amplifier is reading the input signal, across R5, after all.
It appears that Gain set by the NFB is changed when HBRR><0 ohms.
For non-inverting amp: Gain = Rupper / Rlower
I have always stopped there. I did not consider that the resistance in the voltage reference link Signal Return to Power Ground as part of the negative feedback route.
If I take post954 and include HBRR as being inside the NFB loop,
then when a non zero HBRR is added in the gain becomes: Gain = Rupper / {Rlower +HBRR}
typically Rupper could be 9k (more typically for generallised 3886 it would be 27k) and Rlower could be 1k
and the new gain equation becomes marginally lower.
9000/1001 + 1 =9.991009 instead of 10. That amounts to a 0.09% reduction in gain. not subtracted from output, but a proportional reduction in output.
Is it distortion?
But I have to ask:
Why does the NFB loop include HBRR?
Why does it not stop at the Input Signal Return. The amplifier is reading the input signal, across R5, after all.
Last edited:
The feedback current flow is from OUTPUT to PGND but the negative input only sees the voltage over Ri. V(HBRR) is subtracted from the normal V(Ri).
Increase HBRR to 1k and the output increases by 1V.
Last edited:
What does the sim say about the THD with/without HBRR? I'd be very surprised if you find an improvement by adding HBRR.
Note: You need to run with a fairly small time step and ensure that you sample an integer number of periods for the THD to simulate reliably.
Tom
I apologize, my last comment was an unsuccessful attempt at humor.
For me it is about understanding the circuit. It seems that on all aspects of design there are always two contrasting methods, center taped or two secondary transformer, star ground or bus ground.
Adding a HBRR resistor to your design gives you the two options. Just replace the resistor with a wire and you still have a good layout.
For me it is about understanding the circuit. It seems that on all aspects of design there are always two contrasting methods, center taped or two secondary transformer, star ground or bus ground.
Adding a HBRR resistor to your design gives you the two options. Just replace the resistor with a wire and you still have a good layout.
I think the gain including HBRR should be:
Gain=1+Rfb/Ri+HBRR/Ri
HBRR increases the gain by HBRR/RI.
HBRR is not required for a monoblock amplifier.
HBRR/HBRL is only required where a stereo pair of amplifiers inside the same case/chassis incorporate a grounding loop with the incoming Signal Returns from the stereo source.
HBRR/HBRL is there to attenuate the ground loop current that otherwise afflicts stereo installation when unbalanced inputs have a commoned Signal Return back at the Source.
Since a standard Tomchr implementation is balanced input and since each PCB can be assembled as a monoblock, then the HBRR problem does not exist in that installation. HBRR would be zero ohms, or as close as possible to get to zero.
Last edited:
How about this one: THAT Driver: Differential line driver / preamp
I'll have boards available within a couple of days.
Tom
I'll have boards available within a couple of days.
Tom
Totally fair. Play with it in the simulator or write out some equations.
That's engineering for you. There's not one solution. There's a solution space. Where you land within that space depends on which variables you're optimizing for - or rather which weight you attach to each optimization variable.
Oh, no! You don't get off that easily. You could replace it with a capacitor or inductor. Or how about a diode? Or spark gap? Why limit yourself?
On a serious note, I (and many others) use two different schematic symbols for the power ground and the signal ground as this assigns different net names to the two grounds. This makes it possible to join them at one point only in the layout - usually via a 0 Ω resistor. In one pro design (Parasound A23), I did see a ferrite bead used instead of the 0 Ω resistor.
Tom
A good practice implemented by very few on this Forum.
Even better would be adopting three different symbols.
PE. Three legged fork.
Power Ground. Three bars starting with long and finishing with short.
Signal Return. Open black border triangle.
That leaves a fourth, black filled triangle, for special uses.
For simulation of everything except the ground network, that's sufficient. For simulating the ground network, you need to include the various ground trace impedances.
You're welcome. Fedex says bords by Tuesday. Gimme a day or two to assemble and test the board. - Or order now through my website (www.neurochrome.com) if you'd like to be an early adopter.
Tom