Jocko Homo wrote:
Maybe I should make everyone who reads this write a 25 word essay on everything that they know more about diodes in power supplies than John Curl does.
I haven't read many of John Curl's power supply ramblings, but here's my 10 line essay:
"You want the longest diode conduction time you can achieve."
And the best way to get started on this path is to download a copy of LTspice. Yes, there's plenty of people who say "oh, but SPICE is entirely theoretical simulation, building and listening is way more important", please remain quiet for now while I explain.
Some people think "oh, I'll build the best power supply I can for my amp" and use a toroid transformer with next to no leakage current, ultrafast recovery diodes designed for SMPS, and put half a farad of low ESR capacitance (ie, the aforementioned Panasonic FC) on each rail. In theory, this should give you an extremely low impedance power supply, right?
Well, lets build this power supply up in LTspice. Lay it out something like:
- AC voltage source (25VAC or whatever)
- Series inductor, equivalent to the secondary side referred leakage inductance of the toroid.
- Ultrafast rectifier, there's plenty of stock models in LTspice, or get a model from Fairchild's site.
- Low value (1mOhm) resistor in series with the rectifier. You want this for measuring rectifier current.
- Bulk capacitor, with a series ESR resistor and ESL inductor lifted from the capacitor datasheet.
- Load resistor.
Simulate this, and plot the current in the 1mOhm resistor. Voila, that's your conduction angle. Don't be surprised if the conduction angle is in the single digit degrees, and despite the 1A load that you have on the supply, the diode current is hitting 30A peaks.
Secondly, plot the voltage at the node where the leakage inductor meets the diode. Ugly, right? And thirdly, plot the voltage across the capacitor bank. During the diode conduction interval, do you see a bunch of hash on the power supply? Thought so. What's the frequency content of this? Pretty high, right? It probably travels directly through your linear regulators since they have limited loop bandwidth. After that, the PSRR of your amplifier design, op-amps and other chips you're using probably suck at that frequency also.
Now, time for some enlightenment.
- The 1mOhm resistor, increase its value. What happens to the conduction angle?
- How about snubbing the transformer output with a small C, or R+C? How about the diode?
- Are you sure you're using the right kind of diode in the first place? Directly reading the SPICE models for different types of diodes may provide further enlightenment.
- How about replacing your bulk capacitance bank with a fairly small C, a series L, and a big bank?
Anyway, play around for a while. Change things and see what happens. For that last idea, try replacing the small C with a diode so the thing looks like a buck regulator. Make notes.
Once you get some ideas in your head, start modifying your equipment with some of the things you've discovered. Make some notes on how its sounds while you're at it. After all, SPICE is entirely theoretical simulation, and building and listening is way more important.
