Dear All,
After bread-boarding my a design I had been simulating for months rather successfully the time has arrived to start the PCB layout. To minimize PSU impedance I opted to place the PSU components right by the OPS on the same board with a ground plane as return path for reservoir and capacitor caps.
This arrangement makes sense intuitively: (load current from speaker to caps indicated by green arrows and Charging current by Red) Is this okay?
However I struggle to understand how to arrange the rectifiers, reservoir caps and negative speaker terminal. My understanding is that the charging currents between the rectifier and capacitors cause an AC voltage between them which may be imposed on the negative speaker terminal (speaker ground), causing audible hum. The speaker terminal draws its current from the reservoir caps which in turn draw current from the rectifier. Does this mean that when the speaker terminal and capacitor share a path to the rectifier hum may be created?
Next there's the matter of where to connect (small)signal ground. My plan is to connect it right at the negative speaker terminal. My reasoning being that when there is an AC voltage between signal ground and speaker ground hum is created.
Simulated PSRR is over 140dB so hum shouldn't be injected through the voltage rails. PCB layout will make or break it, I feel my reasoning above might be incorrect. Please feel free to correct me but most of all do tell me how the PSU/OPS layout may be improved.
Long story short, is this layout fine or should it be reworked?
Much thanks and Cheers!
Ruben
After bread-boarding my a design I had been simulating for months rather successfully the time has arrived to start the PCB layout. To minimize PSU impedance I opted to place the PSU components right by the OPS on the same board with a ground plane as return path for reservoir and capacitor caps.
This arrangement makes sense intuitively: (load current from speaker to caps indicated by green arrows and Charging current by Red) Is this okay?
However I struggle to understand how to arrange the rectifiers, reservoir caps and negative speaker terminal. My understanding is that the charging currents between the rectifier and capacitors cause an AC voltage between them which may be imposed on the negative speaker terminal (speaker ground), causing audible hum. The speaker terminal draws its current from the reservoir caps which in turn draw current from the rectifier. Does this mean that when the speaker terminal and capacitor share a path to the rectifier hum may be created?
Next there's the matter of where to connect (small)signal ground. My plan is to connect it right at the negative speaker terminal. My reasoning being that when there is an AC voltage between signal ground and speaker ground hum is created.
Simulated PSRR is over 140dB so hum shouldn't be injected through the voltage rails. PCB layout will make or break it, I feel my reasoning above might be incorrect. Please feel free to correct me but most of all do tell me how the PSU/OPS layout may be improved.
Long story short, is this layout fine or should it be reworked?
Much thanks and Cheers!
Ruben
Why would you place outputs next to the filter caps? The entire circuit between rectifiers and filter caps should be away from the amp, and have low loop-area, so that you dont induce noise voltages directly on the output circuit.
Using twisted pair (or triple) between rectifiers and caps is a good way to reduce this area, or on a PCB run the high current traces on top of each other on either side of the PCB. With two rails and one ground the PCB solution isn't so straightforward, discrete wiring may be better.
The big charging current pulses create magnetic fields that can induce voltages on nearby circuits, dependent on the mutual inductance (loop areas, proximity). The lower the loop area the less distance the magnetic field extends. The further the noisy circuit is from the amplifier the less coupling of magnetic fields to it.
Given a load of 8 ohms, keeping the PSU impedance below, say, 1% of this (80 milliohms) is all you need to do. Its easy to find wire thick enough for < 80 milliohms over a few tens of cm run. Having the PSU droop an extra 1% due to wiring losses is less than 0.1dB reduction on the output maximum level. Or you can specify even smaller droop if you want - the losses in the wire to the speakers will be much greater in practice as they have to run long distance.
Each amplifier should have a single ground reference point used by the speaker return and the input and feedback networks, then the amp is oblivious to whether that ground reference has some extra voltage on it. There are issues if powering more than one amp from the same power supply if the amp inputs are single-ended and share signal ground - a ground loop exists. So usually you want to pay close attention to ground routing and any loop area and resistance in the ground system. You should also be aware of which wires carry linear currents and which carry non-linear currents or noise currents.
The speaker wires carry linear currents and their magnetic fields are unlikely to cause issues (other than crosstalk between channels), where as the rail supply wires carry non-linear currents (half-cycles of the output current), and the wires between rectifier and filter caps carry noise currents.
Using twisted pair (or triple) between rectifiers and caps is a good way to reduce this area, or on a PCB run the high current traces on top of each other on either side of the PCB. With two rails and one ground the PCB solution isn't so straightforward, discrete wiring may be better.
The big charging current pulses create magnetic fields that can induce voltages on nearby circuits, dependent on the mutual inductance (loop areas, proximity). The lower the loop area the less distance the magnetic field extends. The further the noisy circuit is from the amplifier the less coupling of magnetic fields to it.
Given a load of 8 ohms, keeping the PSU impedance below, say, 1% of this (80 milliohms) is all you need to do. Its easy to find wire thick enough for < 80 milliohms over a few tens of cm run. Having the PSU droop an extra 1% due to wiring losses is less than 0.1dB reduction on the output maximum level. Or you can specify even smaller droop if you want - the losses in the wire to the speakers will be much greater in practice as they have to run long distance.
Each amplifier should have a single ground reference point used by the speaker return and the input and feedback networks, then the amp is oblivious to whether that ground reference has some extra voltage on it. There are issues if powering more than one amp from the same power supply if the amp inputs are single-ended and share signal ground - a ground loop exists. So usually you want to pay close attention to ground routing and any loop area and resistance in the ground system. You should also be aware of which wires carry linear currents and which carry non-linear currents or noise currents.
The speaker wires carry linear currents and their magnetic fields are unlikely to cause issues (other than crosstalk between channels), where as the rail supply wires carry non-linear currents (half-cycles of the output current), and the wires between rectifier and filter caps carry noise currents.
Thank you Mr. Tillotson for the elaborate reply!
To summarize you suggest minimizing loop area of the non-linear charging currents and increase proximity to the OPS and that these goals are better achievable using wiring oposed to using PCB.
So why would I place OPS and PSU together?
An amplifier with high UGF needs a really (fast) low impedance supply. I got this feedback from Grimm Audio local to my town.
Placing the PSU on the OPS board seems to me the way to achieve the lowest possible impedance PSU lines. This is after I noticed how much Gain and phase margin goes lost to supply inductance. The remedy I found was placing the big reservoir caps right by the output. Indeed al other grounds should connect to the negative speaker terminal where it meets with the reservoir caps.
I guess my remaining question is whether the run between reservoir caps and rectifier (non-linear high current path) should be considered when aiming for low high frequency PSU impedance.
Or should only the linear high current path between amp and reservoir caps be considered when aiming for a very low high frequency impedance PSU?
Much thanks for your thoughts.
Cheers,
Ruben
To summarize you suggest minimizing loop area of the non-linear charging currents and increase proximity to the OPS and that these goals are better achievable using wiring oposed to using PCB.
So why would I place OPS and PSU together?
An amplifier with high UGF needs a really (fast) low impedance supply. I got this feedback from Grimm Audio local to my town.
Placing the PSU on the OPS board seems to me the way to achieve the lowest possible impedance PSU lines. This is after I noticed how much Gain and phase margin goes lost to supply inductance. The remedy I found was placing the big reservoir caps right by the output. Indeed al other grounds should connect to the negative speaker terminal where it meets with the reservoir caps.
I guess my remaining question is whether the run between reservoir caps and rectifier (non-linear high current path) should be considered when aiming for low high frequency PSU impedance.
Or should only the linear high current path between amp and reservoir caps be considered when aiming for a very low high frequency impedance PSU?
Much thanks for your thoughts.
Cheers,
Ruben
With some simulation I could verify that the linear loop is important for stability. On breadboard I moved the reservoid caps right to the output stage with recififier about 12cm down a tightly braided wire. This seems a very quiet configuration. Ill likely design the board such that the rectifier can be mounted remotely too.
Im thinking of opening a new thread for the rest of the layout. The simulated results are about 1000 times better than on breadboard so I think this design might be knitpicky about layout. On breadboard DUT is approaching the performce of the Scarlet focusrite im using as source and measurement device.
Cheers,
Ruben
Im thinking of opening a new thread for the rest of the layout. The simulated results are about 1000 times better than on breadboard so I think this design might be knitpicky about layout. On breadboard DUT is approaching the performce of the Scarlet focusrite im using as source and measurement device.
Cheers,
Ruben