Thanks Scott!
Yes, a stereo amp, as two mono amps in a single enclosure. In my case, there is just one transformer with multiple secondaries and all else but safety earth is duplicated.
In which case, there is no reason to bond the RCA shields, just add 1nF-2nF caps from shield to case for each channel.
Yes, a stereo amp, as two mono amps in a single enclosure. In my case, there is just one transformer with multiple secondaries and all else but safety earth is duplicated.
In which case, there is no reason to bond the RCA shields, just add 1nF-2nF caps from shield to case for each channel.
If you build a dual mono amp and everything is separate then you do not have the common impedance problem you have with standard single transformer stereo designs that we have been discussing here.
However, thats an expensive solution!
Either use a HBR or go balanced for a more cost effective solution.
However, thats an expensive solution!
Either use a HBR or go balanced for a more cost effective solution.
Thanks Bonsai.
I was a bit concerned that if the transformer in this "dual mono" amp has multiple secondaries and one set is used for Left channel, and the other for right, then since there is magnetic coupling between the secondaries within the transformer, and a common safety earth, the cross channel issues will still be present.
I was a bit concerned that if the transformer in this "dual mono" amp has multiple secondaries and one set is used for Left channel, and the other for right, then since there is magnetic coupling between the secondaries within the transformer, and a common safety earth, the cross channel issues will still be present.
Last edited:
There are a lot of comments to the effect that the local decoupling on an amp PCB prevents PSU return currents from going through the power umbilical. This logic makes the mistake of thinking that a capacitor has less impedance than a twisted pair (only at RF, and even then not necessarily as much as you think). In reality, even if you load the PCB with capacitance, you will usually not reduce the umbilical currents to less than half.
Try out this twisted pair calculator and see what the impedance of that umbilical really is:
https://www.eeweb.com/tools/twisted-pair
The reason this doesn't end up being a big issue is not so much because power return currents don't flow through the umbilical (they do), but because the impedance of the umbilical is just extremely low. Furthermore the haversine currents from the output stage are tightly coupled through the umbilical back to the power supply, so the magnetic fields combine back into a continuous sine wave long before they reach any susceptible PCB traces or routing.
Try out this twisted pair calculator and see what the impedance of that umbilical really is:
https://www.eeweb.com/tools/twisted-pair
The reason this doesn't end up being a big issue is not so much because power return currents don't flow through the umbilical (they do), but because the impedance of the umbilical is just extremely low. Furthermore the haversine currents from the output stage are tightly coupled through the umbilical back to the power supply, so the magnetic fields combine back into a continuous sine wave long before they reach any susceptible PCB traces or routing.
Why don't you sim it and put it up so we can discuss it?
I think you are correct at LF, but the situation changes quickly once you get above 1kHz and up at 10 kHz, with decent on-board decoupling, most of the speaker current will be getting supplied by the local decoupling loops.
This simply reinforces the case for good layout, good decoupling, twisting wires and making sure you terminate the speaker return back at the amp module and not the PSU 0V in order to control the loop areas.
If you are trying to build amplifiers >> -100 dB noise I don't think you can assume anything.
I don't need to simulate to know that the 1KHz capacitive reactance of a 10mF capacitor without ESR is 15.9mohms, and the resistance of a foot of 14ga wire is 2.5mohms. And that calculator gives 6.23nH per inch for a 14ga twisted pair.
With 10mF 12nH 10mohm ESR on both sides of the umbilical, the current wants to split evenly between the two capacitors. 6 inches of 14ga twisted pair reduces this by a whopping 1db.
With 10mF 12nH 10mohm ESR on both sides of the umbilical, the current wants to split evenly between the two capacitors. 6 inches of 14ga twisted pair reduces this by a whopping 1db.
Attachments
Last edited:
I don’t get the sim results you do but then my model is a bit more complex. Above 1 kHz, on board decoupling provides benefits by localizing the loop and that improves with frequency.
My comments about good layout, good, wideband local decoupling, twisting wires etc still stand.
I am hung up on taking your results to their logical conclusion: don’t bother about local on board decoupling. If you can explain how to get away with that in a practical amplifier where the PSau and amp modules are separated, I’m all ears.
My comments about good layout, good, wideband local decoupling, twisting wires etc still stand.
I am hung up on taking your results to their logical conclusion: don’t bother about local on board decoupling. If you can explain how to get away with that in a practical amplifier where the PSau and amp modules are separated, I’m all ears.
If the capacitors are large enough to do the job properly, given the high speaker currents at low frequencies, then the decoupling becomes more or less an extension of the main power supply. Also, capacitors that large in value take up a lot of space.
Typically on a big amp your main reservoirs would be somewhere between 22mfd and say 50 mfd per rail. There are plenty of designs that take it to extremes (100’s of thousands of mfd) but beyond a certain point I don’t see the benefit. I have gone over to using 1000 mfd to 2000 mfd per rail local decoupling around the OPS.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.