Diode bypass caps

ssssly · 14691

0 Members and 1 Guest are viewing this topic.

Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
on: November 17, 2010, 02:34:18 AM
Have been reading up on some power supply theory and was wondering if putting a small pf bypass cap across rectifier diodes would improve the linearity or power transformer operation. In theory I would think it should smooth ripple by limiting voltage swings related to the switching of the diodes.

However since I have never seen any of the big brains around here mention it, it also crossed my mind that I might simply misunderstand something.

Anyone ever try this? Any thoughts on if it is practical or would even work?



Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #1 on: November 17, 2010, 02:56:56 AM
Those caps snub noise caused by the diode.

It wouldn't hurt but gets messy.



Offline Paul Joppa

  • Global Moderator
  • Hero Member
  • *****
    • Posts: 5828
Reply #2 on: November 17, 2010, 11:15:34 AM
Primarily, those caps snub the reverse recovery spike which is characteristic of standard silicon diodes. In the past, we made extensive use of a filter between the diodes and the power transformer (it is still a part of the Seduction design) and so-called "ultrafast" silicon rectifiers which have an especially small and easily attenuated spike. More and more we are now using silicon carbide Schottky diodes for high voltage rectification. Schottky rectifiers (like tube rectifiers) don't have any reverse recovery effect.

An interesting technical bit is that the preferred capacitors for this application are low-grade mylar caps. Those caps have a lot of nonlinearity especially at high frequencies, which shows up as an equivalent series resistance. If you used good caps, you would have to add a series resistor to get good snubbing.

Paul Joppa


Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
Reply #3 on: November 18, 2010, 03:10:12 AM
Hmmm.. not sure if we are talking about the same thing. As I understand it a reverse recovery spike is actual voltage leak backwards through the diode caused by the speed in which it switches off.

Also am I correct in saying that the snubber network that you are describing consists of a reservoir capacitor after the rectifier and then connected to the center tap on the transformer?

I was thinking of putting a ~10pf low esr cap across each diode. Primarily to keep the transformer from seeing the diodes pass back through a 0V state. For when all the diodes are switched off wouldn't that induce mechanical vibration in the transformer at approx twice line voltage AC waveform?

Or are we all talking about the same thing and splitting hairs on the definitions?



Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #4 on: November 18, 2010, 03:56:36 AM
ssssly,

Going from the general to the specific, what piece are you thinking of putting snubbing the caps in?

If it is Bottlehead and newer than the Seduction it might be superfluous.  Newer Bottlehead kits include soft recovery diodes for the high voltage. 

The Seduction has what is called the RRSF on the heaters to keep diode noise out of the power supply.  That reverse leakage comes out of the capacitors too, not just from the diodes.  So it notches the power supply.

VoltSecond has described both the RRSF and a cheap way to eliminate that noise on his site.  The quick and dirty way is to put a soft recovery diode in series with the DC.  That is, after the bridge and before the first capacitor.  It doesn't allow the reverse voltage to get out of the supply.

Here is VoltSecond's site:

http://www.siteswithstyle.com/VoltSecond/index.html



Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
Reply #5 on: November 18, 2010, 04:47:29 AM
My primary goal is actually to reduce physical vibration and resultant hum in the transformer itself. With a secondary goal of smoothing the AC in my PMour 1 heater circuit.

Don't know if it is typical of the PMour transformers, or just the ones I have, but they have an audible hum to them when operating. 



Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #6 on: November 18, 2010, 05:12:39 AM
Then the series Cree/ShotTkey diode in the B+ would make a difference.  I don't think the bridge can notch without a connection to the capacitor bank.

Have you tried snugging up the screws that hold the transformer?  The laminations should be "varnished" together but the transformer still has torque on it.  That can cause it to vibrate on the top plate.

It is worth a try.

I just checked.  I have a tiny bit of hum from my transformers.  I have to get down on the floor and almost put my eye into the driver tube to hear it.  I assume yours is more.
« Last Edit: November 18, 2010, 10:43:15 AM by Grainger49 »



Offline Paul Joppa

  • Global Moderator
  • Hero Member
  • *****
    • Posts: 5828
Reply #7 on: November 18, 2010, 09:37:05 AM
Just to clarify a few confusing issues:

When we have DC filament power, it has always come from Schottky rectification. That includes the Seduction, of course.

The Seduction uses UF4007 diodes, which are of the so-called "ultrafast soft recovery" type. That means the reverse recovery spike is very short and relatively small. These diodes can be easily replaced by the 600-volt IXYS silicon carbide Schottky diodes that we use in other products; they are quite inexpensive.

The RRSF (reverse recovery spike filter) in the Seduction is a CRC filter between the transformer and the diodes, which reduces or eliminates ringing in the transformer caused by the reverse recovery spike. We tried several approaches back in the day, and this one was the most effective.

A low-ESR capacitor will not reduce the reverse recovery spike energy, because is has no energy dissipation mechanism. Snubbers are usually designed with a capacitor in series with a resistor; the capacitor makes the spike longer and shallower, and the resistor absorbs its energy. It's mostly the total energy that causes the transformer to ring. The cheap mylar capacitors have significant ESR at the very high frequencies characteristic of the spike, so they don't need the separate resistor.

Technically speaking, Schottky diodes are NOT "soft recovery" - they have no recovery at all, soft or hard.

Paul Joppa


Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
Reply #8 on: November 19, 2010, 03:29:15 AM
Alright, now I know we are not talking abuot the same thing.

I am not trying to attenuate the reverse recovery spike.

I am trying to keep the transformer from switching off while allowing the diodes to switch on and off.



Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #9 on: November 19, 2010, 03:53:14 AM
I'm trying to follow you.  Getting onto the reverse recovery spikes was my fault.

Alright, now I know we are not talking about the same thing.

I am not trying to attenuate the reverse recovery spike.

I am trying to keep the transformer from switching off while allowing the diodes to switch on and off.

If the amp is on the power transformer is also on.  That is true even f the load isn't drawing current on the load side of the transformer which doesn't happen in an operating amp. 

At start up the power supply is drained so there is an inrush of current charging the power supply caps.  After that there is still a steady but much lower current that supplies the B+ and heater current at idle.  With music playing there is a higher draw of current from the power supply and the transformer.

Jumping back to your original post:

Have been reading up on some power supply theory and was wondering if putting a small pf bypass cap across rectifier diodes would improve the linearity or power transformer operation. In theory I would think it should smooth ripple by limiting voltage swings related to the switching of the diodes.  .  .  .   

I read that you want to improve the linearity of the power transformer operation/smooth the ripple in the power supply.  There is some ripple after the two 47uF (IIRC) caps.  You might be able to reduce that by using lower ESR and higher value capacitors.  But that gets crowded.  I know!  The caps across the diodes do nothing for ripple.

And more:

My primary goal is actually to reduce physical vibration and resultant hum in the transformer itself. With a secondary goal of smoothing the AC in my PMour 1 heater circuit.  .  .  .   

Are your transformers snugly attached?  Can you get longer screws and put a rubber grommet under the transformer?  Do keep the transformer grounded but maybe physical isolation will help the hum.

I don't think you can smooth the AC for the heaters.  The way to do that is go DC on the heaters. 



Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
Reply #10 on: November 19, 2010, 07:01:00 AM
Describing technical topics tends to end in "term creep". Particularly when one only has enough knowledge about a technical topic to confuse himself (IE me). How many different terms can be used to label a "coupling capacitor"?

If I understand the math and the circuit correctly, with the circuit under load (amplifying any signal within the audio bandwidth), there is a periodic occurrence, which is approximately twice line voltage, where none of the diodes are forward biased. In this state the rest of the circuit would see the transformer as off (or non existent) and the transformer would be acting as if there was no load on it (a space heater). This would not present a linear load to the transformer and would also induce non sinusoidal artifacts in the output AC. Not to mention it would explain why the tranny was humming at about 240hz.

From what I calculated, an approximate 10pf bypass capacitor across each individual diode alleviated these issues. The recharge time of the cap kept the transformer from ever seeing a state of all non forward biased diodes.

Or I could just completely misunderstand the math and the circuit.

Further physically isolating the transformer from the top plate would probably reduce the audible hum. I'm sure the top plate amplifies the vibration quite well. But I would rather stop the hum in the transformer if it is possible to do so, or at least to the extent that it is possible.



Offline JC

  • Sr. Member
  • ****
    • Posts: 485
Reply #11 on: November 19, 2010, 07:28:12 AM
As a starting point, I think I would try to decide if the sheet metal "bell-end(s)" or cover(s) are contributing to the problem.  Instinctively, it occurs to me that these lighter-mass parts might be prone to acting as little diaphragms and could be treated with some deadening materials.

Isolation of the entire unit from the chassis plate may also help for approximately the same reason, but do make sure to keep a safety ground on the transformer frame.

Finally, I'm still not so sure that the "snubber" is too terribly far off the mark.  Part of the issue with the Reverse Recovery Spike, IIRC, was that it tended to react with the transformer in undesirable ways, and even couple to windings other than the one that the diodes are connected to.  Dissipating that energy before it ends up back in the transformer may be a worthwhile pursuit, particularly if it would help to keep some noise from finding its way onto windings used for filaments.

Of course, a 'scope may be of great use helping you to "see" what you're doing with snubbers and caps on the diodes.  Do you have access to one?

Jim C.


Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #12 on: November 19, 2010, 07:33:30 AM
JC posted before I got through but here it is...

I'll give it a go.  

Coupling Capacitors:

When connecting a tube or transistor amplifying stage and you just want to pass the AC audio signal there is a coupling capacitor or transformer to block the DC needed to power the tube or transistor.  When connecting between two stages, like between the two stages of the Seduction, it is often called an interstage coupling capacitor.  

In my Foreplay, Seduction and in the Crack there are output capacitors.  They are the last capacitor before the output.  In the case of the Crack it is feeding the speaker terminals.  The Seduction and FP feed another device downstream that doesn't want DC on the input.

Back to the transformer and capacitors and diodes:

Yes, at 120 times a second (twice line frequency) the AC incoming and the AC at the output of a power transformer goes to zero.  Assuming here in the US where we have 60 Hz AC.  The current goes to zero at that point then starts climbing again.  The voltages and current capability changes going through the transformer.  All circuits that I can think of now have zero voltage 120 times a second.  Maybe switching supplies don't.  And possibly PJ knows of a circuit that the transformer has a continuous load on it.  

The secondary, load side, of the transformer still goes to zero volts 120 times a second even if there is a load there.  But the voltage downstream of the first power supply capacitor and resistor are considerably smoothed and will have a minimum constant current after turn on.  

I think the problem is that the input to the four diodes producing the power supply voltage, still goes to zero 120 times a second.  So there is no voltage to keep the diode on.

I'm not  sure I'm making this clearer.  Sorry!

How bad is the hum you get from your transformer?  Do you hear it when a CD is in pause?  When music is playing?  

I have a pair of Paramours and my transformer doesn't hum or vibrate.  I checked yesterday.  The "zero crossing" may not be the cause of the problem.  A noisy line voltage, full of harmonics can cause a transformer to hum.  Also, heavy loading causes transformers to hum (at least power distribution transformers).

To measure the load you can measure the voltage dropped across the power supply resistor.  In mine that is a 270 ohm between the 2 47uF caps.  I (current) is equal to the voltage read divided by the measured resistance.  Turn it off before measuring resistance.  (You knew that but a new reader might not)
« Last Edit: November 19, 2010, 07:35:23 AM by Grainger49 »



Offline ssssly

  • Sr. Member
  • ****
    • Posts: 386
Reply #13 on: November 20, 2010, 07:06:18 AM
The hum isn't loud by any stretch of the imagination. Can't hear it over ambient noise from my listening position (about 5 meters depending on which speakers I am listening to). I can hear it when I am changing records (2 feet).

My PMour 1s, while I still love them (my first bottlehead kit), are now largely my experiment amps. I have a pair of PMour 2s that I listen to far more often.

So this isn't really a problem as much as a learning experiment. Never really bothered me to be honest. Have just been studying about power supplies and it struck me that with the amount of effort that people put into designing power supply filters down stream of the diode bridge that leaving the bridge itself as an ocilator inducing non linearity in the tranny seemed a little odd.

Then it struck me that it could be the reason that the PMour 1 trannies hum.

So other that the fact that I was only counting every full sine wave as 0V instead of 0v twice a cycle (on the up and down swing of the sine wave) I think I at least understand what I am talking about.

So then my next question would be what would be the result of placing 10pf caps bypassing each individual diode?

From what I understand, as one of the diodes approaches and passes through 0V, failing to be forward biased, the transformer would then charge the capacitor. The capacitor would then be used to forward bias the diode which would then continue to conduct. 

This should present a theoretic constant load to the power transformer and prevent the diode from pulling voltage from other places to forward bias (which can only be causing non linearity someplace else).

Whether this will stop the hum who knows? But it should prevent the possibility of the diodes creating 120hz oscillation in the power transformer correct?



Offline Grainger49

  • Hero Member
  • *****
    • Posts: 7175
Reply #14 on: November 20, 2010, 08:46:55 AM
  .  .  .   So then my next question would be what would be the result of placing 10pf caps bypassing each individual diode?  .  .  .   

Just to answer that question, like I said in the first reply:  "It wouldn't hurt but gets messy."  By that I mean that it gives you a lot of leads in one very tight space.  If it doesn't sound better you can clip off the caps.