Simple, no-math transformer snubber using Quasimodo test-jig

Hi Mark, I do not care to much about the 'HF-wiggles' as long as the resulting snubber values are valid. If these 'wiggles' are an artifact of the measurement setup, then I will just ignore them, the snubber resulting from the measurement is the only thing that counts (for me :)). So (I guess) the question is, is the result obtained valid (or do you think it is valid) (in this case using an 8 ohm resistor).

P.s. I did use the 33R to make adjustment a bit easier, maybe it is a good idea to advise a 250R potentiometer for people (mostly) snubbering large power transformers. In my case, I will change (the now 1K internal potentiometer) to a 100R internal potentiometer and the 'external' resistor will be placed in series with this.
 
Last edited:
P.s. (to the above) the question is, although I did dampen one frequency (in the middle picture), there is the other (higher) frequency. Any suggestions?

I count 6.5 periods of wiggles on a 200ns scope division, so the wiggle is at ~ 30 ns period or ~ 32 Mhz. That's a bit too slow for reflections on the scope cable (unless it's 3m long), but given the setup (large flying wire loops, etc), I'd say don't worry about it, anything above say 1 MHz is going to be seriously corrupted anyway, it is likely to be an artifact from measurement parasitics. Just like trying to measure any digital signal with the 20cm long ground clip...

Last time I adjusted a snubber, it was on already built gear, I just put the 1/10 scope probe on the secondary through a highpass (a few pF and some resistor to get rid of the 50 Hz AC). I could see the damped oscillations on the scope and suppress them by adjusting a RC snubber, that's about it, it works well...
 
Member
Joined 2011
Paid Member
FdW, I like to smoothly dial the snubber from very-underdamped to critically-damped, watching the oscilloscope the whole time. I've attached 5 photos showing the time sequencing that I prefer. I took these measurements today on a little 5 volt-ampere transformer that happened to be laying on the bench.

Picture #1 has the snubber resistor Rs removed from the board ("infinity ohms") so the secondary is extremely underdamped. Then I installed Rs and sloooooowly dialed it lower, lower, lower. This gave pictures #2, #3, #4. Picture #4 is critical damping (zeta = 1.0); the waveform "trough" at time=2.5 divisions, has been completely flattened.

Then, for the sake of this informational post, but NOT RECOMMENDED in real life circuits, I continued to dial Rs even lower, giving OVERdamping. When I dialled Rs down to ridiculously small values, producing ridiculous OVERdamping, I got picture #5. You'll see that it exhibits oscillations, at a lower frequency (scope sweep 5X slower). This is because Rs is now so small that the impedance of Cs is no longer negligible compared to the impedance of Rs. With a CRC snubber, it is possible to have too much of a good thing (damping). Picture #5 shows what happens when you try to set zeta to approximately nine. Don't do that!
 

Attachments

  • t1.png
    t1.png
    78.7 KB · Views: 3,095
  • t2.png
    t2.png
    65.6 KB · Views: 2,506
  • t3.png
    t3.png
    62 KB · Views: 2,527
  • t4.png
    t4.png
    61.8 KB · Views: 2,512
  • t5.png
    t5.png
    65.9 KB · Views: 1,693
Member
Joined 2011
Paid Member
I forgot to say in my above post, that I cannot believe that adding the .1uf film in you photo did not change the waveform. Photo 2 looks like a copy of photo 1. The cap had to do something, good or bad.
Hi Rick, please have a look at the horizontal sweep rate, which appears at the bottom of those scope photos. 20 usec/div for photo 1, 50usec/div for photo 2. Adding the cap reduced the oscillation frequency. But it didn't change the damping.
 
I see a bit of confusion here. Diode switch on in a real power supply occurs when the secondary voltage minus diode loss is greater than the filter capacitors charge (Voltage.) This is is much greater current than switch off noise. This will induce ringing in the transformer if not handled.

Switch off noise radiates as the energy has no clear discharge path.

I find on 1N4007 diodes 220 pf of capacitance lowers the switch off noise to about the level of line noise. Putting any capacitance across a diode increase the line noise that passes the rectifier stage. So too large or too small becomes an issue.

Now placing an RC network across the secondary can reduce ringing, and differential line noise pass through. I prefer a parallel RC network as I want the resistor to actually draw some power. This loads the secondary just a bit and makes sure the transformer voltage doesn't float as much as when unloaded. When a transformer is unloaded you also get a contribution to the output voltage from the capacitive coupling between the windings. Also just placing a load resistor drops the transformer noise by as much as 20 db at 20k from the unloaded state.

Now if you want to look at diode switch off noise I suggest an RF spectrum analyzer set for DC to 5 Mhz. This will show the spikes. Narrower bandwidth analyzers may entirely miss them.
 
I find on 1N4007 diodes 220 pf of capacitance lowers the switch off noise to about the level of line noise.

Where is this cap placed?


Now placing an RC network across the secondary can reduce ringing, and differential line noise pass through. I prefer a parallel RC network as I want the resistor to actually draw some power. This loads the secondary just a bit and makes sure the transformer voltage doesn't float as much as when unloaded. When a transformer is unloaded you also get a contribution to the output voltage from the capacitive coupling between the windings. Also just placing a load resistor drops the transformer noise by as much as 20 db at 20k from the unloaded state.

Robert Grodinsky advocated liberal use of damping resistors across several components, including the transformer windings.

https://docs.google.com/viewer?url=patentimages.storage.googleapis.com/pdfs/US4594561.pdf
 
Hi Rick, please have a look at the horizontal sweep rate, which appears at the bottom of those scope photos. 20 usec/div for photo 1, 50usec/div for photo 2. Adding the cap reduced the oscillation frequency. But it didn't change the damping.

Thanks for pointing out the time base change Mark. It makes sense that the cap would change the freq., but I would think that it would lower the level too.
 
Now placing an RC network across the secondary can reduce ringing, and differential line noise pass through. I prefer a parallel RC network as I want the resistor to actually draw some power. This loads the secondary just a bit and makes sure the transformer voltage doesn't float as much as when unloaded. When a transformer is unloaded you also get a contribution to the output voltage from the capacitive coupling between the windings. Also just placing a load resistor drops the transformer noise by as much as 20 db at 20k from the unloaded state.

Now if you want to look at diode switch off noise I suggest an RF spectrum analyzer set for DC to 5 Mhz. This will show the spikes. Narrower bandwidth analyzers may entirely miss them.

Thanks for suggesting the resistor across the secondary, Ed. I have tried this but found it impacted the dynamics of the music. I do think it is a good idea to always keep current flowing in the transformer and not let it "run dry" as we do with chokes. I will have to keep trying load current values of the R to find the right value. Do you have any suggested current values for this resistor?

Also I think that we should keep the C of the diodes as low a possible, thus no cap should be across them or the RF flow thru it.
 
Thanks for suggesting the resistor across the secondary, Ed. I have tried this but found it impacted the dynamics of the music. I do think it is a good idea to always keep current flowing in the transformer and not let it "run dry" as we do with chokes. I will have to keep trying load current values of the R to find the right value. Do you have any suggested current values for this resistor?

Also I think that we should keep the C of the diodes as low a possible, thus no cap should be across them or the RF flow thru it.

I size the resistor across the secondary at about 5% of rated current. I then pick the capacitor to match.

The diodes right across the diodes do let more noise pass but the switching spikes as seen on an rf spectrum analzer are significant. So the results will depend on how your circuit handles the noise. As I am beginning to use dual bridges with out of phase secondaries most of my line noise is gone.
 
Last edited:
Mark, I built your CRC snubber and used a .1 for Cx, and another .1 for Cs. While looking at the waveform with my Tektronix TDS 340 scope I dialed in 100 ohms with a 1k pot for the least ringing. The Cx removed the buzz from the radio, and the Cs Rs removed the ringing on the scope. How important it is to remove this last ringing is something I do not know. I have not listened to it with, and without Cs Rs in the same listening session. I would think that less waveform is best. It seemed that the exact value of Rs was not critical, 80 ohms to 110 ohms seemed to give the same waveform. Thanks for starting this thread.

Rick
 
I count 6.5 periods of wiggles on a 200ns scope division, so the wiggle is at ~ 30 ns period or ~ 32 Mhz. That's a bit too slow for reflections on the scope cable (unless it's 3m long), but given the setup (large flying wire loops, etc), I'd say don't worry about it, anything above say 1 MHz is going to be seriously corrupted anyway, it is likely to be an artifact from measurement parasitics. Just like trying to measure any digital signal with the 20cm long ground clip...

Last time I adjusted a snubber, it was on already built gear, I just put the 1/10 scope probe on the secondary through a highpass (a few pF and some resistor to get rid of the 50 Hz AC). I could see the damped oscillations on the scope and suppress them by adjusting a RC snubber, that's about it, it works well...

Thanks, I will ignore the HF wiggles.
 
Member
Joined 2011
Paid Member
The Cx removed the buzz from the radio, and the Cs Rs removed the ringing on the scope. How important it is to remove this last ringing is something I do not know. I have not listened to it with, and without Cs Rs in the same listening session. I would think that less waveform is best.
I think the same way.

Wouldn't it be a pretty big surprise if you (or listeners in general) preferred the sound without Cs+Rs? What if oscillatory ringing in the transformer's secondary is euphonic? Ringing that you can plainly see on an oscilloscope, increases listening pleasure?? Yikes that would be a scary outcome.

Thanks, -MarkJ
 
I think the same way.

Wouldn't it be a pretty big surprise if you (or listeners in general) preferred the sound without Cs+Rs? What if oscillatory ringing in the transformer's secondary is euphonic? Ringing that you can plainly see on an oscilloscope, increases listening pleasure?? Yikes that would be a scary outcome.

Thanks, -MarkJ

A number of well regard vacuum tube amplifiers have a power supply resonance that compliments the LF roll off in the output transformer!

So a ring at the LF cutoff of your loudspeakers may sound pleasing in some cases, as would 125 -150 hz for many recordings.

However accurate is better in the general case.
 
  • Like
Reactions: 1 user
Member
Joined 2011
Paid Member
... on already built gear, I just put the 1/10 scope probe on the secondary through a highpass (a few pF and some resistor to get rid of the 50 Hz AC). I could see the damped oscillations on the scope and suppress them by adjusting a RC snubber, that's about it, it works well...
What a nice idea! In fact, for DIY gear, why not install this highpass RC circuit right on the PCB itself? Design it into the PCB and provide a testpoint allowing future You to verify that the secondary is well-damped and free of oscillations. At a PCB cost of one square centimeter per transformer winding, it seems quite reasonable.

Thanks, -MarkJ
 
Very exiting reading hole thread, i have the DATS unit for speakers which think could work with math, but this unit i like to build for easy results. Think i will copy member FdW's nice universal unit. Thanks Mark for posting/sharing.

P.s. (to the above) the question is, although I did dampen one frequency (in the middle picture), there is the other (higher) frequency. Any suggestions?

I will build a PSU with snubber CRC's as indicated by the Quasimodo-test-jig and then test the result. When done I will report on this, but don't hold your breath :)
Any suggestions?
Looking at your test setup picture suggest easy trials to mayby cure the HF.
1 - CX 10nF try put leg from blue connector directly to yellow connector by small wire.
2 - CX 10nF is very beautiful but a old fabric, try some newer and mayby best the from Mark recommended EPCOS B32529 series.
 
Last edited:
Member
Joined 2006
Paid Member
Quasi V4 kit in action

I put my v4 Quasimodo together today, and measured my transformer, intended for an Aleph J build. I had gotten the rectifier bridges, but had to try out this method with figuring the snubber values for the DIYAudio PSU kit.

Transformer - Antek AN-5218
Cx (C2) - 0.01
Cs (C3) - 0.15

These are the caps that come with the kit, and worked fine.

Rs = 170R for my transformer.

Thanks for the wonderful board Mark. First use of my scope. Worked like a charm.

An externally hosted image should be here but it was not working when we last tested it.
 
Last edited:
Member
Joined 2011
Paid Member
Congratulations, Bones13! Your first use of a scope was a big success.

I think you may want to double-check or triple-check the parts placement in your Quasimodo capacitor sockets; what you typed above is not quite the recommended "try this first" starting point for further experimentation.

The recommended starting point is

  • Cx (C2) = 0.01uF (10nF)
  • Cs (C3) = 0.15 uF (150nF)
If you've accidentally swapped the caps, I suggest you may want to put them back the other way, the recommended way. Fire up Quasimodo again, and twirl the optimization dial once more.