How about we create a special chopper loading module to test the buffering capabilities of an amp?
I'd construct it this way: An idle high impedance of say 10K ~ 100K load on the supply output, that, via a control signal can be switched to say 10 Ohm/10W using a MOSFET for a period of say 10mS.
Then, just record the transient response at the supply output closest to the load and see what effect all kinds of bufferings and different caps, inserted wires to get your parasitic L etc, have on the transient response.
The goal is to work towards a way of buffering that alters the voltage at the load node the least. Reduce spiking, reduce sagging. This will be a relatively easy way to find out how to create the best buffer distribution/arrangements.
I'd construct it this way: An idle high impedance of say 10K ~ 100K load on the supply output, that, via a control signal can be switched to say 10 Ohm/10W using a MOSFET for a period of say 10mS.
Then, just record the transient response at the supply output closest to the load and see what effect all kinds of bufferings and different caps, inserted wires to get your parasitic L etc, have on the transient response.
The goal is to work towards a way of buffering that alters the voltage at the load node the least. Reduce spiking, reduce sagging. This will be a relatively easy way to find out how to create the best buffer distribution/arrangements.
The capacitors suggested in post #54 have an ESR specification of 16 mohm typical, 31 mohm maximum at 100 Hz / 20 degrees C. The impedance at 10 kHz = 30 mohm maximum so they are pretty low impedance at audio frequencies. It's going to require a large film capacitor to reach that low an impedance at 10 kHz.
Well yes, but look at the cap you linked. Is that a power supply main buffer cap, or would that be your local decoupling caps on the board both at board entrance and close to the output devices?
That's what I mean, the local electrolytics won't be kingsize like a main smoothing cap, they are in the region of 100uF ~ 2200uF, and their ESRs are higher than that of an accompanying 100nF MKT.
I think that it would be the power supply main buffer cap. I would expect local decoupling on the amplifier PCB as well. At 10 kHz the reactance of a 100 nF capacitor is 159 ohms so the audio current is going to come from the power supply main buffer capacitor.
I don't know if it would be neglible when talking the size of currents in a decent amp. The L could be just high enough to 'choke' the T=0 current and prohibit initial demand. I was about to put a simulation together of a full rectified supply with buffercaps and ESR simulation, along with the chopper circuit to view the transient response. I wanted to model the supply wire in there too.
The reactance (X) doesn't matter, it's not in series with the load, you have to view it as an ideal supply with an internal source resistance equalling the ESR paralleled to the main supply. Think 'shortcutting' a charged cap, that's a HUGE T=0 current and has nothing to do with its frequency dependant impedance.
Amps(T0) = Vcap/ESR
Additional info
Last edited:
PChi,
I've done this experience with different caps (I checked them first with a tan-delta meter), transformers, rectifiers and connections of the probes.
Each time, the phenomenon happened for high values of capacitor.
It has been repeatable for me but I do not exclude an error in my procedure.
Maybe somebody could do the same test and show his findings,
I've done this experience with different caps (I checked them first with a tan-delta meter), transformers, rectifiers and connections of the probes.
Each time, the phenomenon happened for high values of capacitor.
It has been repeatable for me but I do not exclude an error in my procedure.
Maybe somebody could do the same test and show his findings,
i doubt a 200nH inductance would have any significance
current of what size are we talking about here?
In the 4A 48-0-48 PSU that I described previously, I used 2mH and concentrated on the ability of the chokes to handle 4A continuously. (Air cored to prevent saturation)
As I said, It was a CLCC, CLCC, CLCC, CLCC supply. 2 x 600VA split secondary transformers feeding 4 x 50A Standard Bridges then 33000uF 2mH 33000uF 33000uF four times over for four separate supplies. The 33000uF were Cornellier Dubillier.
I also added 4.7uF Audio Grade Polyprops across each bank but it didn't seem to make too much difference.
The Pass Aleph 4 does not have any decouplers on the PCB board so all the decoupling is done at the PSU end.
The PSU cabling and PCB tracks are monsterous though (32A Auto Cable to the PCB).
As I said, It was a CLCC, CLCC, CLCC, CLCC supply. 2 x 600VA split secondary transformers feeding 4 x 50A Standard Bridges then 33000uF 2mH 33000uF 33000uF four times over for four separate supplies. The 33000uF were Cornellier Dubillier.
I also added 4.7uF Audio Grade Polyprops across each bank but it didn't seem to make too much difference.
The Pass Aleph 4 does not have any decouplers on the PCB board so all the decoupling is done at the PSU end.
The PSU cabling and PCB tracks are monsterous though (32A Auto Cable to the PCB).
Last edited:
Stray inductance resonating with the other cap? One cap means inductance is in series with the cap. Two caps means each inductance is almost in parallel with the other cap. The inductance is not so much caused by the length of wire but the area of the loop formed by the caps and the wire.forr said:
Alternative explanation: more stored charge in the rectifier junction? Bigger caps means shorter but larger charging pulse, hence more stored charge? Stored charge means the rectifier stays switched on when it should have switched off so having charged the caps up it can then drag them down again for a brief while until all stored charge has been removed from the junction. Having said that, I'm not sure I believe it myself but who knows?
I'm leaning towards rectifier drop as well, but when I was simming a PSU buffer chain and playing with values I recreated the same sort of distortion by giving the buffercaps a noticable ESR. I doubt that in reality ESR would go up with more caps in parallel so I'm considering the rectifier drop the purpetrator: At the peak the supply has to recharge both the cap and keep supplying the load as well. Less ESR means more current together with that what the load is demanding; depending on the bridge I could see this be a cause for the momentary drop at the peaks. Or, what DF96 says. A bigger, badder rectifier could possibly give an answer to that hypothesis.
I still think that the distortion is due to high ESR capacitors or measurement technique.
Is the power supply '0V' connection connected to mains earth?
The Oscilloscope ground is conected to mains earth. It might be worth probing the '0V' connection with the probe tip.
The following test was done with a single rail power supply, loaded by a 39 Ohm power resistor. The transformer came from an old B&O 2*20 W receiver where the reservoir cap was 3300 µF.
The scope probe was connected at the diode bridge + and - terminals.
The ground reference was given by the shield of the probe cable, the scope mains earth terminal was not connected to a real earth.
4400 µF
6600 µF
17000 µF
The scope probe was connected at the diode bridge + and - terminals.
The ground reference was given by the shield of the probe cable, the scope mains earth terminal was not connected to a real earth.
An externally hosted image should be here but it was not working when we last tested it.
4400 µF
An externally hosted image should be here but it was not working when we last tested it.
6600 µF
An externally hosted image should be here but it was not working when we last tested it.
17000 µF
4400 µF : two 2200 µF 63 V CMF-FP SIC SAFCO in parallel.
6600 µF : three 2200 µF 63 V CMF-FP SIC SAFCO in parallel.
17000 µF : one 2200 µF 63 V CMF-FP SIC SAFCO in parallel with one 15000 µF 50 V MARCON.
Given the caps mentioned and looking at the sensitivity of the last two scope images they are both 200mV/Div, only the first image is 500mV/Div.
The difference between B and C is the 3 small caps vs 1 small and one big one. Going from 3 to 2 caps increases ESR relatively, while the much increased capacitance increases the charge current notably for the big cap.
Adding up all of this I think it's safe to say it's an ESR issue, where the ESR performance of the 15mF cap seems quite bad in relation to the smaller caps.
The difference between B and C is the 3 small caps vs 1 small and one big one. Going from 3 to 2 caps increases ESR relatively, while the much increased capacitance increases the charge current notably for the big cap.
Adding up all of this I think it's safe to say it's an ESR issue, where the ESR performance of the 15mF cap seems quite bad in relation to the smaller caps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.