diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   Ripple in unregulated supply for class-A amp (http://www.diyaudio.com/forums/power-supplies/221234-ripple-unregulated-supply-class-amp.html)

IG81 9th October 2012 01:42 PM

Ripple in unregulated supply for class-A amp
 
Hello,

I have a little DIY project for a class-A amplifier, 4-channels @ ~10W each. It consists in a simple common-emitter gain stage followed by an emitter-follower output, biased with a CCS.

I use a large Hammond transformer, 36VAC with center tap (unused-floating) for 0V-46VDC supply rails, rated 432VA. Rectification is via a large bridge that mounts on the heatsink. This is followed by 8x10000F Panasonic caps, 63V/105C.

Each channel draws ~1.91A at idle, so total current is 7.64A. The formula I/(2*F*C) gives me ~800mV of ripple. I measure ~600mV.

The problem is that I hear this ripple as a hum in the output, as it modulates the signal. I guess I'm just stuck with this ripple figure, as the math predicts it. Increasing capacitance would quickly become ridiculous IMO and not get me to un-noticable levels. Is there anything that can be done outside of adding regulation?

I don't think it's a grounding problem either. I have the AC mains ground connected to the all-metal chassis. DC and signal ground is separate from chassis and uses short runs of thick gauge wire. Any resistance there will be under 10mili-ohm and never cause such a voltage (600mV) to develop at the worst-case 1.91A current.

Thanks,

IG

agdr 10th October 2012 11:02 AM

4 Attachment(s)
A single-ended amplifier (0-46Vdc vs. dual-ended +/-23Vdc) will have very little power supply rejection. It will be more sensitive to ripple on the supply.

If you can sacrirfice 2-3 volts of supply voltage, try adding some 0.1R resistors between the caps to form a CRCRCRC filter. 20W rating on the resistor between the leftmost two caps, 10W on the other two. Put the load ground where shown at the rightmost cap, below, to keep the filter loop currents before the load. Star ground your load(s), but don't star ground the filter sections. Wire the CRC filter sections as a ladder as shown to keep the filter ripple currents within each loop. The leftmost 2 caps that are fed from the bridge need to have a minimum ripple current rating of around 8A each or so, 2A minimum each for the middle 2 caps and 0.5A for the rightmost 2. So if your 8 10MF caps have a ripple current rating of at least 8A or so they are fine to use in all 8 cap positions.

Headbanger 10th October 2012 11:18 AM

An alternative is to use capacitance multiplier or simple mosfet reg, a pair of mosfets should be good (one for 2 ch) if heatsink is big enough. Cons: more power dissipation, and voltage drop.

IG81 10th October 2012 12:55 PM

A capacitance multiplier is what I had come down to after doing a bit more research. This project had been shelved for a couple of years and I actually intended to use one back then, but just forgot since. It won't hurt that I might have to reduce my bias levels also, the heatsinks get hotter than I'm comfortable with.

IG

demeterart 10th October 2012 01:54 PM

my guess is, with the mentioned audible hum, that the problem lies mostly with the biasing of the transistors. make sure you use capacitors for bypassing the resistive biasing to the bases even if it means splitting a single resistor into two (usually unequal parts) to insert a capacitor for hum filtering.

IG81 10th October 2012 02:13 PM

Quote:

Originally Posted by agdr (Post 3196245)
A single-ended amplifier (0-46Vdc vs. dual-ended +/-23Vdc) will have very little power supply rejection. It will be more sensitive to ripple on the supply.

If you can sacrirfice 2-3 volts of supply voltage, try adding some 0.1R resistors between the caps to form a CRCRCRC filter. 20W rating on the resistor between the leftmost two caps, 10W on the other two. Put the load ground where shown at the rightmost cap, below, to keep the filter loop currents before the load. Star ground your load(s), but don't star ground the filter sections. Wire the CRC filter sections as a ladder as shown to keep the filter ripple currents within each loop. The leftmost 2 caps that are fed from the bridge need to have a minimum ripple current rating of around 8A each or so, 2A minimum each for the middle 2 caps and 0.5A for the rightmost 2. So if your 8 10MF caps have a ripple current rating of at least 8A or so they are fine to use in all 8 cap positions.

I could sacrifice a few volts. Thanks for doing the sim. I'll input that into AIMspice to see what it gives me.

Quote:

Originally Posted by demeterart (Post 3196431)
my guess is, with the mentioned audible hum, that the problem lies mostly with the biasing of the transistors. make sure you use capacitors for bypassing the resistive biasing to the bases even if it means splitting a single resistor into two (usually unequal parts) to insert a capacitor for hum filtering.

You mean bypassing the ground to base resistors? I'm not certain I follow, or at least can't see how this'd be done without affecting the signal.

IG

demeterart 10th October 2012 04:18 PM

1 Attachment(s)
i mean the following:

IG81 10th October 2012 08:08 PM

Seems like the simplest thing to do at this point in my build might be to split my 8x10000F capacitor bank into 2x40000F and insert a choke in-between. Common-mode ferrite-donuts are pretty cheap and 4x7.3mH would reduce the ripple to a simulated 12mV, low enough IMO. It would also reduce buzz by smoothing-out the sawtooth ripple into a quasi-sinusoidal waveform.

IG

agdr 10th October 2012 10:33 PM

2 Attachment(s)
That should work, but I'm not seeing any place the ripple would appear as common mode. Those CM chokes are most useful for conducted EMI. May need to be a standard choke like a Hammond 159ZL:

159ZL - HAMMOND - CHOKE | Newark

http://www.hammondmfg.com/153.htm

IG81 11th October 2012 02:08 AM

I would not be using it in common mode, just the two coils in series, this can be done right? I'm gonna try some 2x7.3mH, 9A units, four coils in series for 29.2mH. My simulation has ripple down to roughly 10mV with this inserted between the two 40000uF caps.

IG


All times are GMT. The time now is 03:31 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2