According to Morgan Jones's "Valve Amplifiers" book, there is need to bypass diodes connected to a transformer output to absorb field energy while all diodes are in cut-off ( +/- 0.7V around zero crossing). He suggests using 10nF film capacitors. I have a bunch of 10nF/1000V ceramic disc capacitors I use as switch bypasses, could these be okay? A film capacitor seems overkill across a diode, and aren't ceramics better suited to absorb fast transients?
Thank you for your reply. Are UF4007 considered old or new? They are "ultra-fast", whatever that means, but still they've been around for decades.
And, thinking of it, being fast doesn't eliminate the 0.7V threshold, the same problem subsides e.g. the transformer winding is still "disconnected" for a moment no?
And, thinking of it, being fast doesn't eliminate the 0.7V threshold, the same problem subsides e.g. the transformer winding is still "disconnected" for a moment no?
I use Schottky's only: they're orders of magnitude faster than "ultra-fast" junction diodes.
Disk ceramic caps belong to landfill.
Disk ceramic caps belong to landfill.
For 50/60Hz rectification, bog-standard silicon diodes with adequate PIV and current handling capabilities will work just fine. The fast reverse recovery types are necessary in HF switch mode supplies, though. The Schottky ones will work just fine in a 50/60Hz setting, but are really just unnecessary there. When using a snubber capacitor, it's a good idea to put a small resistor, say 10 ohms in series with the capacitor. The resistor dissipates the unwanted noise energy and eliminates any current spikes caused by the switching transients conducted through the capacitor. If you don't dissipate the unwanted noise energy, it may well end up somewhere else in the circuit.
In my book , yes, windings become disconnected below 0.7V ... but at the same time it means such disconnection happens at very low voltage, current and energy level, making such disconnection quite irrelevant.
It is not as if, say, disconnection happened at 20 - 30 - 40V
Personally do not lose sleep about it.
The day somebody shows me "horrible 50-60Hz artifacts" in signal output I might begin to worry ..... until then ....
Would also love to see side by side waveforms showing "with-without" capacitor situations.
Until then .....
It is not as if, say, disconnection happened at 20 - 30 - 40V
Personally do not lose sleep about it.
The day somebody shows me "horrible 50-60Hz artifacts" in signal output I might begin to worry ..... until then ....
Would also love to see side by side waveforms showing "with-without" capacitor situations.
Until then .....
It's not the forward voltage drop that does it, it's the delay in the diode switching off after the voltage reverses with the bog standard 1N400x rectifier and its "slow reverse recovery." When it does stop conducting, it switches off fast, and the stray inductance and stray capacitance rings at an RF frequency which gets picked up by sensitive audio circuitry - a short pulse of RF interference 100/120 times per second. There's at least one DIYA thread with some good descriptions of this, and surely oscilloscope traces. If not, I suppose I can do it and post some pics. I'm a little belt-and-suspenders here, I use a capacitor AND a series r-c snubber across each rectifier, and that's for a +/-20V supply to be regulated to +/-15V.
Spending a little more on fast-recovery rectifiers is another way to do this (I'd still do a cap and series r-c). As the OP seems to be making a supply for tube equipment, Schottkys may not be appropriate:
"Schottky rectifiers seldom exceed 100 volts in their working peak reverse voltage..."
https://www.microsemi.com/document-portal/doc_view/14630-introduction-to-schottky-rectifiers
TL;DR the 10nF caps should work okay.
Spending a little more on fast-recovery rectifiers is another way to do this (I'd still do a cap and series r-c). As the OP seems to be making a supply for tube equipment, Schottkys may not be appropriate:
"Schottky rectifiers seldom exceed 100 volts in their working peak reverse voltage..."
https://www.microsemi.com/document-portal/doc_view/14630-introduction-to-schottky-rectifiers
TL;DR the 10nF caps should work okay.
The diodes to use are soft-recovery, rather than fast recovery. Reducing loop area around the the rectifiers is a major win for any interference BTW, always the starting point for quiet rectification I'd have thought.
Schottky's can be tricky to use here as they might need to be conservatively rated or heat-sinked to avoid thermal runaway if used at high voltages due to the massive reverse-leakage when they get hot. For +/-15V supply a Schottky rectifier is fine, but say +/-80V rails and you might be at risk of this.
Schottky's can be tricky to use here as they might need to be conservatively rated or heat-sinked to avoid thermal runaway if used at high voltages due to the massive reverse-leakage when they get hot. For +/-15V supply a Schottky rectifier is fine, but say +/-80V rails and you might be at risk of this.
Capacitors across rectifier diodes are known as ratelcondensatoren in Dutch, rattle capacitors, and their purpose is to prevent modulating RF signals with mains harmonics, which would cause rattling sounds in nearby radios.
As a bad example, suppose you have some piece of (double-insulated) equipment without protective earth, with a mains transformer without a shield between the primary and secondary windings, with a bridge rectifier and with long signal cables connected to it.
At radio frequencies, the capacitance between the primary and secondary of the transformer couples the mains wiring to the AC side of the rectifier, while the case and the signal wires are connected to the other side. Together, this forms a kind of dipole antenna that gets opened and shorted 100 or 120 times per second. Any reflections from this structure will therefore be modulated with a rattle.
In any case, they can be ceramic. The only disadvantage I can think of is that ceramic class 2 capacitors are piezoelectric and might emit some humming sound of their own when used like this, but I doubt if that will be strong enough to be noticable.
As a bad example, suppose you have some piece of (double-insulated) equipment without protective earth, with a mains transformer without a shield between the primary and secondary windings, with a bridge rectifier and with long signal cables connected to it.
At radio frequencies, the capacitance between the primary and secondary of the transformer couples the mains wiring to the AC side of the rectifier, while the case and the signal wires are connected to the other side. Together, this forms a kind of dipole antenna that gets opened and shorted 100 or 120 times per second. Any reflections from this structure will therefore be modulated with a rattle.
In any case, they can be ceramic. The only disadvantage I can think of is that ceramic class 2 capacitors are piezoelectric and might emit some humming sound of their own when used like this, but I doubt if that will be strong enough to be noticable.
Last edited:
look at The Quasimodo thread and stop using old methods of rectifier snubbing.
https://www.diyaudio.com/community/...-using-quasimodo-test-jig.243100/post-7677299
https://www.diyaudio.com/community/...-using-quasimodo-test-jig.243100/post-7677299
So, a Cx and Cs+Rs neglects the need for additional caps/resistors across the diodes?
I haven't looked into it too much yet but I remember reading an article about different diodes and how a Cx and Cs+Rs made caps across the diodes unnecessary. I might be wrong though so don't quote me on that.
Couple of good references on the subject:
https://www.hagtech.com/pdf/snubber.pdf
https://sound-au.com/articles/psu-snubber.htm
Hal
https://www.hagtech.com/pdf/snubber.pdf
https://sound-au.com/articles/psu-snubber.htm
Hal
Absolutely.
Capacitors across diodes are a very bad solution considering common mode noise. They connect transformer secondary to the rectified output and transfer common mode HF noise. When rectifier diode switch off, there will be RLC ringing. However, at that moment as all diodes are switched off, secondary is disconnected from the first reservoir capacitor. Ringing produces only radiated common mode noise and has direct path only through small diodes' capacitance. Here is an example with EI transformer, active rectifiers (LT4320) and no snubbers:
How much of common mode noise will be converted to the differential noise/signal depends on circuit design, wiring and PCB.
With snubbers, at the switch off point, there would be only one smooth step down and no ringing at all.
That Cx/Cs+Rs snubbing method does look better. Cures the problem at the source so to speak and avoids the coupling of the transformer to the input filter through the diode bypass capacitors. The ESP article is quite extensive, complete with a test setup to experiment with. Looks like I'll take this route.
Thanks a lot!
I'm in the design phase - Nothing to hear yet
But this thread has been very instructive in deconstructing what looks like a myth about the need to bypass diodes and point out how it could even worsen the problem.
Last edited: