In an conventional, bridge rectified capacitor input supply, where is the best place to RC snub the resonant circuit formed by secondary distributed capacitance and leakage inductance of the winding? I have seen it suggested to place RC snubbers across each diode. Wouldn't RC across the secondary be sufficient? Are there any advantages to doing it one way vs. the other?
Jim Hagerman has written the best audio related treatise on the subject to which I'm aware.
http://hagtech.com/pdf/snubber.pdf
http://hagtech.com/pdf/snubber.pdf
I had read the Hagerman paper, which is what prompted the question. In the opening of the paper he states that the problem is usually dealt with using RC snubbers across the diodes. At the end of the paper, the simulated circuit shows RC across the secondary.
I have followed Hagerman's method with good results. I made some regulated supplies that didn't attenuate the switching noise that much. Snubbers made the output very quiet.
I was mostly just wondering if there was a compelling reason to put RC snubbers across the diodes. It would require more parts so is there an advantage?
I have followed Hagerman's method with good results. I made some regulated supplies that didn't attenuate the switching noise that much. Snubbers made the output very quiet.
I was mostly just wondering if there was a compelling reason to put RC snubbers across the diodes. It would require more parts so is there an advantage?
Hagerman's article is interesting, but I think it contains an error. He correctly calculates the required resistance to damp an LC circuit. He then assumes that he can simply connect this resistor with a big capacitor in series with it. The snag with this is that the capacitor will lower the resonant frequency and the characteristic resistance of the tuned circuit. The result is that while significant damping will take place it will not be the critical damping he was aiming at. The result can be seen in Fig 10 - still a little ringing but at a much lower frequency.
He needs to do the full maths of the original LC with the CR snubber across it. I haven't done the sums, but I would guess you need a C about twice that of the LC (i.e. about a third of his value), and an R of about half his calculated value.
He also says that is important to use non-inductive resistors. Fine for damping UHF sproggies in a transmitter, but for damping sub-MHz ringing in a 50Hz power transformer any resistor will do. Film resistors are essentially resistive up to about 100MHz, so no need for carbon composition. Wirewound would do in most cases!
He needs to do the full maths of the original LC with the CR snubber across it. I haven't done the sums, but I would guess you need a C about twice that of the LC (i.e. about a third of his value), and an R of about half his calculated value.
He also says that is important to use non-inductive resistors. Fine for damping UHF sproggies in a transmitter, but for damping sub-MHz ringing in a 50Hz power transformer any resistor will do. Film resistors are essentially resistive up to about 100MHz, so no need for carbon composition. Wirewound would do in most cases!
My recollection is that Hagerman includes all first order reactive parasitics in his analysis, including rectifier diode capacitive reactance. They are treated as part of a single reactive network. So, my take is that rectifier diode commutation acts like the 'hammer' striking the resonant 'bell' formed by the reactive parasitics. I view rectifier diode commutation as providing an impulse to this resonant system, similar to what a hammer does every time it strikes a bell.
It would seem, although I have not done the analysis to confirm, that one could either damp the reasonant 'bell', or soften the impulse produced by the 'hammer' strike, as it were. So, my conclusion is that damping the effective tank circuit (per Hagerman), or reducing the effective dV/dT (or dI/dT) of the rectifier diode commutation would both work towards reducing high-frequency ringing by this network. Perhaps, the greatest effect would be had by doing both.
Last edited:
There could be two issues here. Diode switch-off can be very sharp - it was even used as a microwave source although using diodes developed for the purpose. That is why in RF situations there are sometimes small caps directly across the diodes. As an audio amp should not be sensitive to RF (although they sometimes are) there is less need. The transformer ringing then becomes the issue, and a snubber across the secondary deals with this. Note that a cap across the diodes simply shifts the frequency of the transfomer resonance but does not damp it.
So to stop transformer ringing when the diodes switch off put a snubber across the transformer. To stop RF radiation (or modulation hum) from the diode transition put a cap across each diode - but this might not be necessary in audio equipment. A snubber across each diode is an expensive way of stopping transformer ringing (4 snubbers instead of 1), and probably won't be so good at stopping RF as a simple cap.
So to stop transformer ringing when the diodes switch off put a snubber across the transformer. To stop RF radiation (or modulation hum) from the diode transition put a cap across each diode - but this might not be necessary in audio equipment. A snubber across each diode is an expensive way of stopping transformer ringing (4 snubbers instead of 1), and probably won't be so good at stopping RF as a simple cap.
I agree with the snubber across the secondary. Don't quite understand the use of caps across each diode. It would seem that they would bypass the diodes at RF, coupling any RF in the supply directly into the power entry to the amp. Could you elaborate on their use for stopping modulation hum?
Yes, caps across the diodes might increase the transfer of any RF in the supply, but there shouldn't be much RF there and a filter will stop it.
Modulation hum occurs when RF (probably from an oscillator in the equipment) is chopped at mains frequency by the rectifiers. This creates wide sidebands, which the RF equipment then picks up. It can be a particular problem in AC/DC radio receivers where the rectifier connects straight to the incoming mains. A mains transformer will not pass much RF, so AC-only sets are better in this respect. Putting caps across the diodes means that at RF the diodes no longer switch anything, so no modulation hum. This should not be a problem in audio equipment (apart from tuners and possibly some digital stuff) because there should be no RF oscillators inside. Note that a snubber will not stop modulation hum as the diode is still changing the RF impedance of the circuit.
Modulation hum occurs when RF (probably from an oscillator in the equipment) is chopped at mains frequency by the rectifiers. This creates wide sidebands, which the RF equipment then picks up. It can be a particular problem in AC/DC radio receivers where the rectifier connects straight to the incoming mains. A mains transformer will not pass much RF, so AC-only sets are better in this respect. Putting caps across the diodes means that at RF the diodes no longer switch anything, so no modulation hum. This should not be a problem in audio equipment (apart from tuners and possibly some digital stuff) because there should be no RF oscillators inside. Note that a snubber will not stop modulation hum as the diode is still changing the RF impedance of the circuit.
position of the snubber circuit
Hello,
first post in this great forum, by the way! I am planning on using such snubber circuits for the secondary of a toroidal transformer that is going to be used for powering a guitar amplifier. I am going to use full wave bridge rectification. And, having read the Hagerman article, I am a bit confused on where I should build the snubber circuits. Looking at the equivalent circuit, I firstly thought that I would need two such snubber circuits, one placed between the "+" of the secondary and audio ground, and the second symmetricallly placed between the "-" of the secondary and audio ground.
Is the above argument right, or is it adequate to put only one snubber circuit between the terminals of the transformer's secondary?
Hello,
first post in this great forum, by the way! I am planning on using such snubber circuits for the secondary of a toroidal transformer that is going to be used for powering a guitar amplifier. I am going to use full wave bridge rectification. And, having read the Hagerman article, I am a bit confused on where I should build the snubber circuits. Looking at the equivalent circuit, I firstly thought that I would need two such snubber circuits, one placed between the "+" of the secondary and audio ground, and the second symmetricallly placed between the "-" of the secondary and audio ground.
Is the above argument right, or is it adequate to put only one snubber circuit between the terminals of the transformer's secondary?
After reading the Hageman article a single but with a damping closer to .707 rather than the under damped .5 will generate a smoother transfer function . Google Zobel circuit which is the name for this snubber circuit from the 1930 after the man who came up with it . Zobel did work on speaker response but that what this, a circuit for to damp resonance .
I found that the transformer secondary could be significantly detuned with a variable R and variable C, while watching the switching transient on an oscilloscope. Works best with small transformers. At the null the period and amplitude both went down, a lot, but the observed ringing was never very great and what I could see was mostly low RF, certainly not VHF. Maybe a couple of volts P-P?
I think the best I can do is use soft recovery rectifiers, a Zobel on the secondary, and some generic RC snubbers across the rectifiers to help mop up some of the remaining noise. I think a plain C just doesn't work well to kill the actual resonance that a RC can.
We need a RF spectrum analyzer to see what's really going on, and I know I won't be getting one for Christmas.
I think the best I can do is use soft recovery rectifiers, a Zobel on the secondary, and some generic RC snubbers across the rectifiers to help mop up some of the remaining noise. I think a plain C just doesn't work well to kill the actual resonance that a RC can.
We need a RF spectrum analyzer to see what's really going on, and I know I won't be getting one for Christmas.
Last edited:
I found that the transformer secondary could be significantly detuned with a variable R and variable C, while watching the switching transient on an oscilloscope. Works best with small transformers. At the null the period and amplitude both went down, a lot, but the observed ringing was never very great and what I could see was mostly low RF, certainly not VHF. Maybe a couple of volts P-P?
I think the best I can do is use soft recovery rectifiers, a Zobel on the secondary, and some generic RC snubbers across the rectifiers to help mop up some of the remaining noise. I think a plain C just doesn't work well to kill the actual resonance that a RC can.
We need a RF spectrum analyzer to see what's really going on, and I know I won't be getting one for Christmas.
Well, such a gift would be a whole life's investment!
This is what I was thinking. I was planning to use four UF5408 diodes, and implement a zobel network on the secondary, as described in Hagerman's article. But I did not plan to use anything in parallel with the diodes, out of simplicity. In the worst case, some 100nF capacitors, as suggested by Morgan Jones.
For a guitar amp? Tube? I jump to thinking tube when you say guitar amp, even though you mention a torroidial transformer. Well, I can't add to the engineering theory or math, but looking at the tube guitar amps on my bench at the moment for a purely pragmatic baseline, LOL:
The Peavey Classic 50 and classic 100 are built to a price, and use...nothing before the main rail filter caps. But there is a small .01 ceramic cap across the primary (not a MOV).
The more powerful 180-watt Fender Super Twin with 6 6L6's uses some teensie .002 1KV ceramic disc caps, one across each of the 4 diodes in the bridge, then the main 220mfd filter caps are not directly across the 500 volt output, they are rated at 265v and go from each rail to the center tap of the secondary, and there's a 39K 2 watt resistor across each of those main filter caps. So that's like a 78K 4 watt resistor would be across the 500v rails. I always thought the resistors were to damp such resonances, but a note says they're mostly bleeders to drain the filters within 10 seconds as a safety precaution. That double-duty might be a practical consideration for you too.
I imagine the practical downside of a damping network across the secondary is of course the waste of a watt and generation of a little heat.
The .002mfd caps Fender uses across each diode are a lot bigger than the 100nF you propose.
If yours is solid-state or cost-no-object...then nevermind and Merry Christmas.
The Peavey Classic 50 and classic 100 are built to a price, and use...nothing before the main rail filter caps. But there is a small .01 ceramic cap across the primary (not a MOV).
The more powerful 180-watt Fender Super Twin with 6 6L6's uses some teensie .002 1KV ceramic disc caps, one across each of the 4 diodes in the bridge, then the main 220mfd filter caps are not directly across the 500 volt output, they are rated at 265v and go from each rail to the center tap of the secondary, and there's a 39K 2 watt resistor across each of those main filter caps. So that's like a 78K 4 watt resistor would be across the 500v rails. I always thought the resistors were to damp such resonances, but a note says they're mostly bleeders to drain the filters within 10 seconds as a safety precaution. That double-duty might be a practical consideration for you too.
I imagine the practical downside of a damping network across the secondary is of course the waste of a watt and generation of a little heat.
The .002mfd caps Fender uses across each diode are a lot bigger than the 100nF you propose.
If yours is solid-state or cost-no-object...then nevermind and Merry Christmas.
The uf5408 generate a lot less switching noise than the old 1n5408 so 100nf is good rather than .002mfd (2nf) . so it 50 times bigger then what fender used and it corner is thus much lower. Here a reference chart West Florida Components Capacitor Conversion Chart
We need a RF spectrum analyzer to see what's really going on, and I know I won't be getting one for Christmas.
That's what work is for, BTW they are quite handy for testing power supplies for noise. Power a single ended RF amplifier from the power supply to test, inject a fairly pure RF signal into the amplifier, compare output to input, look for AM sidebands. FWIW getting rid of 500 kHz noise is fairly easy with small passive filters, 50Hz and 100Hz are much tougher
Just to clarify things a bit, yes it will indeed be a tube amp. And saying "will", I am suggesting that this is a project that I am going to design and construct, for personal use. Thus, I want to do it the right way from all aspects, even though I may not be able to hear or feel any difference. I am going to use all this planning as a lesson for future builds, that could be hi-end - or simply demanding, in general - amplifiers.
Plus, this amp is going to be a Hiwatt replica, using slight amendments - which has already some hi-fi roots!
Plus, this amp is going to be a Hiwatt replica, using slight amendments - which has already some hi-fi roots!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.