Unless you have a choke input PSU the secondary conducts in short pulses, not for the whole cycle. My comment above about the voltage sweeping through zero is misleading; perhaps I should have said that the secondary voltage sweeps down past the standing DC voltage on the reservoir cap.
When the diode stops conducting the transformer attempts to maintain the same voltage across itself, before eventually failing. This leaves you with an almighty sharp corner in the transformer voltage waveform, and the high frequency energy in that corner can be coupled into the audio circuit via stray capacitance. There may also be some ringing in the tens of kilohertz range, which makes things worse. What you actually see in the audio circuit is then a pulse train that can be easily confused with true diode switching spikes.
It's starting to look like there are 2 different definitions of "switching noise" being discussed here.
1. My original posts refer to the engineering definition of "noise" created by a diode because of it design characteristics. Those are well known uV. switching spikes (if you can call a uV. a spike) in the Mhz range. Totally benign for any practical audio PS design consideration.
2. The ripple and harmonics noise that a complex circuit might create by not handling the diode on/off point well. Poor filtering.
1. My original posts refer to the engineering definition of "noise" created by a diode because of it design characteristics. Those are well known uV. switching spikes (if you can call a uV. a spike) in the Mhz range. Totally benign for any practical audio PS design consideration.
2. The ripple and harmonics noise that a complex circuit might create by not handling the diode on/off point well. Poor filtering.
Last edited:
There are effectively two regions of dI/dt to note. A good pictorial is shown in figures 27 & 28 of:
http://www.ieeta.pt/~alex/docs/ApplicationNotes/Rectifier Applications Handbook.pdf
The first region shows the relatively slow dI/dt as the current commutates from the secondary winding. With a valve rectifier this dI/dt is slower, as it takes longer for the current in the diode to change from load current level to 0A. With a diode rectifier it is a lot faster, and as Merlin says causes 'an almighty sharp corner in the transformer voltage'. This is the normal capacitive coupled splatter that Michael shows in his spectrum. Merlin has shown how it can also easily couple in to heater supplies and get injected to the input stage if heater layout is not that good.
The second region is the actual 'reverse recovery' influence. The dI/dt is typically much higher, and dependant on the diode rather than the circuit per se. It is the effect that causes the MHz ringing talked about - I can't find a good link to a frequency spectrum but it can show up as a distinct peak, and has caused many a smps to not pass radiated compliance.
http://www.ieeta.pt/~alex/docs/ApplicationNotes/Rectifier Applications Handbook.pdf
The first region shows the relatively slow dI/dt as the current commutates from the secondary winding. With a valve rectifier this dI/dt is slower, as it takes longer for the current in the diode to change from load current level to 0A. With a diode rectifier it is a lot faster, and as Merlin says causes 'an almighty sharp corner in the transformer voltage'. This is the normal capacitive coupled splatter that Michael shows in his spectrum. Merlin has shown how it can also easily couple in to heater supplies and get injected to the input stage if heater layout is not that good.
The second region is the actual 'reverse recovery' influence. The dI/dt is typically much higher, and dependant on the diode rather than the circuit per se. It is the effect that causes the MHz ringing talked about - I can't find a good link to a frequency spectrum but it can show up as a distinct peak, and has caused many a smps to not pass radiated compliance.
Last edited:
Awesome. Informed discussion.
Here are the switching pulse energy mechanisms that have been discussed so far:
commutation:
A wide rectangular current pulse produced as the rectifier switches on and off to charge the reservoir cap if there is less than 100% conduction angle (capacitor input filter e.g.). This is common to vacuum and solid state rectifiers. Increasing DCR reduces this by increasing the conduction angle for any given output current. Vacuum diodes have significant DCR... This would be the only source of energy beyond 120 Hz in an ideal system. With a critical inductance choke input filter there is no commutation.
Transformer leakage inductance:
Stored energy in the leakage inductance of the transformer secondary winding releases when the rectifier stops conducting. can be absorbed by connecting a snubber network across the transformer secondary (leakage inductance is not common between transformer windings so don't try snubbing the primary for this) common to both vacuum and SS rectifiers but SS have the next 2 additional mechanisms which exacerbate this
Rectifier forward recovery:
When a solid state rectifier turns on, the junction charge needs to develop. This causes an abrupt turn-on current pulse when the charge reaches sufficient level to reach the conduction voltage threshold.
Rectifier reverse recovery:
When a solid state rectifier turns off, the stored charge releases promptly somewhat after the junction voltage crosses the switching threshold. This causes an abrupt turn-off pulse.
Forward and reverse recovery pulses happen regardless of leakage inductance being present (e.g. on the test stand) but can exacerbate the effect of leakage inductance. Vacuum diodes do have some capacitance but I've never encountered recovery pulses with vacuum diodes that I can recall.
Diodes also generate random noise as do any semiconductor junction. In my experience this is never referred to as switching noise and has no practical impact on power rectifier design.
My doctor says I have zero cholesterol issues so I'm going to go now and have a nice bacon steak and eggs breakfast 😛
Here are the switching pulse energy mechanisms that have been discussed so far:
commutation:
A wide rectangular current pulse produced as the rectifier switches on and off to charge the reservoir cap if there is less than 100% conduction angle (capacitor input filter e.g.). This is common to vacuum and solid state rectifiers. Increasing DCR reduces this by increasing the conduction angle for any given output current. Vacuum diodes have significant DCR... This would be the only source of energy beyond 120 Hz in an ideal system. With a critical inductance choke input filter there is no commutation.
Transformer leakage inductance:
Stored energy in the leakage inductance of the transformer secondary winding releases when the rectifier stops conducting. can be absorbed by connecting a snubber network across the transformer secondary (leakage inductance is not common between transformer windings so don't try snubbing the primary for this) common to both vacuum and SS rectifiers but SS have the next 2 additional mechanisms which exacerbate this
Rectifier forward recovery:
When a solid state rectifier turns on, the junction charge needs to develop. This causes an abrupt turn-on current pulse when the charge reaches sufficient level to reach the conduction voltage threshold.
Rectifier reverse recovery:
When a solid state rectifier turns off, the stored charge releases promptly somewhat after the junction voltage crosses the switching threshold. This causes an abrupt turn-off pulse.
Forward and reverse recovery pulses happen regardless of leakage inductance being present (e.g. on the test stand) but can exacerbate the effect of leakage inductance. Vacuum diodes do have some capacitance but I've never encountered recovery pulses with vacuum diodes that I can recall.
Diodes also generate random noise as do any semiconductor junction. In my experience this is never referred to as switching noise and has no practical impact on power rectifier design.
My doctor says I have zero cholesterol issues so I'm going to go now and have a nice bacon steak and eggs breakfast 😛
Last edited:
Baby spinach salad with ripe tomatoes, micro green thai basil and leaf lettuce, gorgonzola. Steak on the side, bacon crumbles and a hard boiled egg on the salad yummy 😀 and healthy
Attachments
The remarks about SS diodes are undoubtedly true for all PN junction types. High PIV Schottkys have changed the game. The PN diodes exhibit minority carrier injection, which leads to the switching noise. Like vacuum diodes, Schottkys are majority carrier only. No switching noise is the result.
It's important to note that the phenomenon of minority carrier injection is not always deleterious. BJTs depend on it for their operation. On an occasional basis, BJTs can be useful. 😉
Some nuances to the discussion are imho:
- the level of reverse recovery experienced is significantly modified by the dI/dt level of the current as the pn diode commutates (ie. the first region of current change). Many photos of diode reverse recovery, and forward recovery for that matter, are taken from smps circuits where the inherent switch transition is forced to be extremely rapid when compared to mains voltage rectification. Mains voltage rectifiers operate with much lower dI/dt, and the level of reverse recovery current is much less - which is why smps designers have to manage the effect much more.
I doubt you would notice any forward recovery with mains rectification, because of the relatively low dI/dt.
The DCR in the commutation region of a valve amp is dominated by the valve diode and secondary winding (with a bit of reflected primary) - eg. a 5Y3GT is nearly 1k at commutation, whereas the secondary winding may be up to about a hundred ohm. An ss circuit has effectively zero diode resistance.
The higher DCR, and lower dI/dt in a valve rectifier circuit, are the reasons why turn-off noise is much lower in a valve rectified supply, because this lowers the energy in the secondary winding leakage inductance. Placing an ss diode in series with the tube diode for protection then effectively has negligible reverse recovery impact, and so makes an excellent failsafe and stress reducer.
- the level of reverse recovery experienced is significantly modified by the dI/dt level of the current as the pn diode commutates (ie. the first region of current change). Many photos of diode reverse recovery, and forward recovery for that matter, are taken from smps circuits where the inherent switch transition is forced to be extremely rapid when compared to mains voltage rectification. Mains voltage rectifiers operate with much lower dI/dt, and the level of reverse recovery current is much less - which is why smps designers have to manage the effect much more.
I doubt you would notice any forward recovery with mains rectification, because of the relatively low dI/dt.
The DCR in the commutation region of a valve amp is dominated by the valve diode and secondary winding (with a bit of reflected primary) - eg. a 5Y3GT is nearly 1k at commutation, whereas the secondary winding may be up to about a hundred ohm. An ss circuit has effectively zero diode resistance.
The higher DCR, and lower dI/dt in a valve rectifier circuit, are the reasons why turn-off noise is much lower in a valve rectified supply, because this lowers the energy in the secondary winding leakage inductance. Placing an ss diode in series with the tube diode for protection then effectively has negligible reverse recovery impact, and so makes an excellent failsafe and stress reducer.
What about a semiconductor voltage doubler circuit? Are there any considerations as opposed to just a straight semiconductor full wave bridge?
BTW, "micro green thai basil" sounds awesome.
BTW, "micro green thai basil" sounds awesome.
Last edited:
Baby spinach salad with ripe tomatoes, micro green thai basil and leaf lettuce, gorgonzola. Steak on the side, bacon crumbles and a hard boiled egg on the salad yummy 😀 and healthy
Hmmm... I don't see that bacon on the picture. But diode spikes were visible on another picture. 😀
The Greinacher, AKA full wave, doubler is (in fact) a pair of 1/2 wave rectifiers wired back to back. As both halves of the AC cycle are used, it does not present "standing" DC to the rectifier winding, which is good. The conduction angle argument applies "in spades", as high value caps. are needed in the doubler stack to get decent regulation of the resulting rail. A LC section following the doubler stack is found in all good designs. Scan the archives for my posts on "hash" filters.
So far no mention has been made of synchronous rectification.
Might it be the one true path? Possibly avoid generating all those nasty transients in the first place....
I have a vague recollection from years back, of a circuit shown in EW+WW as it then was, of something referred to as a "semi-resonant synchronous rectifier" and being impressed with its elegance.
I can't recall the details, but I think it worked without any control logic ICs and was notable for its efficiency and the lack of transients it produced.
Are there any archivists of EW+WW who recall this at all - Mr Self, for instance?
The design was realized with mosfets, but might be adaptable to all valve construction.
Just a thought.
Might it be the one true path? Possibly avoid generating all those nasty transients in the first place....
I have a vague recollection from years back, of a circuit shown in EW+WW as it then was, of something referred to as a "semi-resonant synchronous rectifier" and being impressed with its elegance.
I can't recall the details, but I think it worked without any control logic ICs and was notable for its efficiency and the lack of transients it produced.
Are there any archivists of EW+WW who recall this at all - Mr Self, for instance?
The design was realized with mosfets, but might be adaptable to all valve construction.
Just a thought.
Member
Joined 2009
Paid Member
At the time I understood it well, though I can't remember it now, sadly. It was simply and elegantly done, that much I do remember.
The thoughts about synchronous rectifiers and mosfets with the body diodes don't come in to play here in general - they are very relavant in smps design. Synchronous is all about aiming for lower on-voltages of the 'diode' by using fets in their place - a smps efficiency improvement holy grail.
Yes, I am aware of their usual area of application, and the reasons for that, but thank you anyway.
As I did mention, a feature of the particular circuit that I was referring to was the lack of switching transients. It was also rather beautiful.
I'm sorry to be unable to lay hands on the actual schematic.
Then you'll also be aware that synchronous rectifiers lack switching "transients" because they switch when the rectifier voltage or current is zero.
They are synchronous with the SMPS controller.
Can you briefly explain how this concept might be applied to a 50 of 60 Hz line frequency rectifier? How will you avoid commutation, i.e having switched the rectifier in at a zero crossing, how will you avoid the rectangular current pulse when the rectifier voltage crosses the reservoir cap voltage?
I can think of two systems that avoid 120 Hz commutation 1. critical-inductance choke input filter 2. SMPS with power factor correction
Last edited:
You would want to switch the Mosfet on when it had zero voltage across it (input V to cap V) to avoid the transient (so zero current turn on), and switch it off again when it had zero current. An SCR could do that too with the same trigger. The resonant circuit might have been a way to derive the trigger voltage, which seems to need a little lead factor for the gate threshold. With a rectifier in series to block the internal back diode, the Mosfet could just be (prematurely) turned on at the input V zero crossing (when it gets above the gate threshold, some scaling and gate protection needed of course).
Hmm, does this buy you anything over just a plain Schottky rectifier though?
Maybe something like this. The transformer has some leakage inductance. So use a 1st cap after the rectifiers that is 50/60 Hz resonant with the the leakage inductance. I was considering this for a fixup to a PFC to keep it supplied during the power line zero crossing (so it could maintain output regulation full time).
Hmm, does this buy you anything over just a plain Schottky rectifier though?
Maybe something like this. The transformer has some leakage inductance. So use a 1st cap after the rectifiers that is 50/60 Hz resonant with the the leakage inductance. I was considering this for a fixup to a PFC to keep it supplied during the power line zero crossing (so it could maintain output regulation full time).
Last edited:
Smoking-amp,
I don't think it buys anything much over a schottky or even carefully designed 'ordinary diode' circuit, especially at these frequencies and power levels!
As far as I can remember, the circuit used small coupled inductors arranged to drive the rectifier mosfet gates without any connection to the smps controller, though it was indeed part of a smps.
It wasn't a circuit suitable for scrs, the gate control signals were ramped, albeit quickly, rather than stepped, at both turn-on and turn-off. The motivation for all this was of course efficiency, and the topology was such that it was increased by this, rather than otherwise.
In the original application, the comparative softness of the switching, ie max di/dt & dv/dt being low was a means to an end, not the end itself.
I don't think it buys anything much over a schottky or even carefully designed 'ordinary diode' circuit, especially at these frequencies and power levels!
As far as I can remember, the circuit used small coupled inductors arranged to drive the rectifier mosfet gates without any connection to the smps controller, though it was indeed part of a smps.
It wasn't a circuit suitable for scrs, the gate control signals were ramped, albeit quickly, rather than stepped, at both turn-on and turn-off. The motivation for all this was of course efficiency, and the topology was such that it was increased by this, rather than otherwise.
In the original application, the comparative softness of the switching, ie max di/dt & dv/dt being low was a means to an end, not the end itself.
Don't underestimate the bean counters!
A lot of changes in audio have been made due to cost, convenience ect'. At times at the expense of audio quality.