choke input power supply filter

[Nelson Pass]

Originally posted by AndrewT
there is only one ClassA topology that has constant current draw.

All the other ClassA amplifiers whether single ended or push pull have modulated current on the supply rails.

You can used bridged amps off common rails to achieve constant draw while in ClassA mode.

For clarification, the average draw is constant, the instantaneous
draw may not be. If the time constants of the supply are longer
than the audio frequencies, this is not much of an issue.

[GRollins]

Yet another reason to apply e-flippin'-normous amounts of capacitance (and perhaps inductance) to every power supply. Get that time constance way, way down...

Grey

[audionut]

I just wanted to report the results of some experiments with the choke input supply versus a CLC supply...feeding two channels of brian gt's wonderful aleph 30 boards using 6 mosfets per channel.

Transformer voltage is 28 volts per rail. Current draw is 5 amps per rail...both channels working. Chokes are air core, 14 gauge, 2.5mh and .5 ohms dc resistance.

On the CLC filter, rail voltage was 35 volts per rail. the minus rail had 40 mv of ripple, and displayed an ugly wave form. The positive rail was 5 mv and displalyed a kind of regular looking wave form.

I conducted the electric drill test on the CLC supply...by plugging in an electric drill to the same socket as the Amp/power supply and observing the power supply trace on the oscilliscope. When running the drill, the wave form became thick with a lot of scattering.

On the choke input filter...rails came in at 26 volts. (just right for an Aleph 30). the minus rail had 100mv of ripple and displayed a nice regular wave form. The positive rail was 60 mv of ripple and also displayed a nice regular sine wave.

I conducted the same electric drill test. When running the drill, the wave form stayed about as it was before running the drill...no scattering or thickening of the trace.

Benefits...remember, I was trying to use these big 28 volt signal transformers that cost $5.00 ea.

voltage is now in range for an Aleph 30
wave form is better looking, might cancel better in A-30
regulation might be better
less strain on the bridge rectifiers
might start easier

detractions are:
the Aleph 30 with LC filter supply now draws 400 watts vs 300
watts for Aleph 30 with CLC supply.

I haven't heard how these aleph 30's sound compared to the Aleph 3's that I have had for last 3 years. When I am finished, it will be interesting to hear if a Choke Input supply sounds better than the CLC supply that I currently have in the Aleph 3's

I welcome any comments...

[AndrewT]

Hi,
nice experiment that electric drill test.
How severe is it compared to real life gunge that comes in from the utility?
BTW what size of bit do I need to make the test valid?

[wuffwaff]

Hi Audionut,

nice experiment. I always though those chokes had to be a bit bigger than 2.5mH to work properly.

Can you explain why the pos and neg rails display different ripple? The current should be the same in both shouldn´t it?

William

[MRupp]

Hi audionut,

Since the 2,5 mH are too low to have a continuous current draw you might get a lot of ringing in your circuit when the diodes are not conducting. If you haven't done so already you could put RC snubbers across the diode bridges, and a resistor across each choke might also be worth trying. I am not sure about the optimal value but 100 Ohms would be a good start?

[audionut]

thanks for your replys

I don't know why the ripple would be larger one rail versus the other...Its the same way on all of the alephs that I have built.

As far as ringing goes...It seems to me that I can see ringing when the CLC filter is in use-- the trace is jagged.

When the choke input filter is in use, with the O-scope set at 20mv, the trace makes a nice regular sinus curve.

When I do the calculation as set forth in one of the links above...it looks like the inductance should be 5mh...

[Nelson Pass]

Originally posted by audionut
I don't know why the ripple would be larger one rail versus the other...Its the same way on all of the alephs that I have built.

Something's wrong. The ripple should be the same.

[Kashmire]

My amps (operating in constant-current mode) use a choke-regulated power supply. I have used numerous chokes, larger and smaller than the critical inductance.

Larger chokes lower the system voltage, but reduce peak currents. Smaller chokes raise system voltage, but increase peak currents (through the transformer, choke, and filter caps).

The reason I used a choke-regulated PSU was to reduce transformer peak currents. High peak currents mean poor power factor (reducing the fidelity of the power mains?), high heat in the transformer, and lots of work for the filter caps.

HOWEVER, the diodes don't take nicely to choke regulation. The choke drags current from the diodes after the input voltage reduces from peak. This means when the diodes turn off, there is a big voltage difference. The PN junction slams shut, allowing a momentary (2 to 10 nanoseconds) reverse current peak to close the junction. I've used a dozen different diode types, and they all do this.

(Note: Tube diodes (valves) work great. They let the current die nicely. That's why tube amps can use choke-regulated PSUs. Also, since tubes are limited in current-carrying capability, low peak currents are very desirable. However, most valve diodes drop about 20-40 Volts at 100mA, which make them useful for PSUs of 300-400 Volts.)

The solid-state generated nanosecond spike is impossible to filter. PSU capacitors (even itty-bitty expensive ones) have too much inductance to shunt it. Inductors have too much interwinding capacitance to filter it. It jumps right across the capacitance-multiplier MOSFET on the Zen v4, and right into the loudspeaker.

I can hear a 120Hz buzz from my Fostex horns. Measured on an o-scope, I can see the nanosecond spikes occuring at 120Hz intervals. It is a direct result of the choke regulation causing a forcable PN junction collapse.

For my present setup, I have 3mH chokes right on the diodes (below critical inductance), feeding about 50,000uF capacitance. After that, I have a second LC stage of 120mH and 100,000uF. I have also added 0.01uF high-frequency caps at various locations. The nanosecond spike still leaks through.

Is there a way around this, or I am doomed to using a big transformer with lousy power factor? This just seems so ... unfriendly to my utility grid ... (and possibly against regulations in the EU)

[AndrewT]

Hi Kashmire,
have you tried an RC snubber across each rectifier diode?

Also 1nF to 10nF ceramic on very short leads and/or wires across the diode output and again after each choke might help.

Nice feedback on your achievements.