When using silicon rectifier diodes like 1N4007 for tube amp plate supplies, people put film capacitors from the anode to the cathode of each diode, to suppress noise from switching spikes.
I've read that contrary to the common wisdom, tube rectifiers (like the 5Y3GT I'm using) also create those switching spikes. If that's the case, then it should be helpful to put something like 10nF to 100nF from each diode to the cathode of the 5Y3. Has anyone tried this? Any results to report?
Some questions:
- What capacitor type is best for this? I would think polypropylene film and foil construction would be best, since that's the type that's rated for the highest ripple currents. Can metalized polyethylene or similar also be used for this? Ceramic dielectric 1kV parts?
- What voltage rating do the caps have to have? If for a +400VDC supply, would a 630WVDC rating be enough? Or is 1kVDC necessary?
____________
I've read that contrary to the common wisdom, tube rectifiers (like the 5Y3GT I'm using) also create those switching spikes. If that's the case, then it should be helpful to put something like 10nF to 100nF from each diode to the cathode of the 5Y3. Has anyone tried this? Any results to report?
Some questions:
- What capacitor type is best for this? I would think polypropylene film and foil construction would be best, since that's the type that's rated for the highest ripple currents. Can metalized polyethylene or similar also be used for this? Ceramic dielectric 1kV parts?
- What voltage rating do the caps have to have? If for a +400VDC supply, would a 630WVDC rating be enough? Or is 1kVDC necessary?
____________
It is possible the snubbing you have heard recommended for tube rectifiers is actually the result of transformer ringing, rather than reverse recovery switching noise which occurs in semiconductor diodes. These are different issues, and are addressed with snubbing in different locations. I would suggest snubbing any transformer that is used with rectification regardless of rectifier type.
The snubber consists of a series RC network placed in parallel with the transformer secondary winding. There are alternative locations with varying degrees of success, but I have found this to be the ideal location.
The capacitors you are referring to (directly across the diode) are to address reverse recovery.
Metallized polypropylene is my suggestion, with an AC voltage rating appropriate for the secondary winding. Resistor is either carbon comp or some other low inductance technology.
The snubber consists of a series RC network placed in parallel with the transformer secondary winding. There are alternative locations with varying degrees of success, but I have found this to be the ideal location.
The capacitors you are referring to (directly across the diode) are to address reverse recovery.
Metallized polypropylene is my suggestion, with an AC voltage rating appropriate for the secondary winding. Resistor is either carbon comp or some other low inductance technology.
As zigzag says, any rectifier can suffer from transformer ringing, but it will be much less with a valve rectifier because its internal resistance inherently damps the process. But an RC network across the secondary will do not harm. The component type does not matter. Normally you would use a ceramic cap because they're readily available with kilovolt ratings. You can use a plastic cap if you want to but don't expect it to create a holographic soundstage or anything. Use about 10nF, then add a resistor (doesn't matter what type, the inductance is negligible) to tune out the ringing. If you don't have the means to do this, just use a value between about 1k and 5k, which is likely to do most of what you need.
Any switch can create switching spikes. Only certain solid-state diodes produce reverse-recovery spikes.
Given that valve rectifiers turn off gradually I would not expect a problem with them. However, adding a cap with constant AC stress may reduce PSU reliability. Best to do nothing unless you can identify a real problem which then needs solving.
Given that valve rectifiers turn off gradually I would not expect a problem with them. However, adding a cap with constant AC stress may reduce PSU reliability. Best to do nothing unless you can identify a real problem which then needs solving.
Thanks for the replies and clarifications. Much appreciated.
So, if I see ringing in the power supply feed, that would be from transformer ringing. That can be tamed by a series RC between the transformer secondaries. Most likely will use 10nF (1kV rated) and resistor between 1k and 5k.
However, putting a capacitor between the secondaries means that part will always be seeing high voltage AC pulses. That adds a possible failure mode. So if it ain't broke, don't fix it! (Good point, DF96.)
From reading the old RDH4 and a couple of other old books, I found that resistors are often put in series with the plates of the rectifier tube to decrease current pulses to the tube. This is useful for ensuring longevity of the tube, especially if using a fairly large value of reservoir capacitor (I'm using 47uF 450V). I was thinking of putting 100R in series before each plate (pins 4 and 6 on the 5Y3GT).
I figure 1W rated parts will do. My thinking:
I figure the 1W rated parts give lots of headroom.
Advisable?
So, if I see ringing in the power supply feed, that would be from transformer ringing. That can be tamed by a series RC between the transformer secondaries. Most likely will use 10nF (1kV rated) and resistor between 1k and 5k.
However, putting a capacitor between the secondaries means that part will always be seeing high voltage AC pulses. That adds a possible failure mode. So if it ain't broke, don't fix it! (Good point, DF96.)
From reading the old RDH4 and a couple of other old books, I found that resistors are often put in series with the plates of the rectifier tube to decrease current pulses to the tube. This is useful for ensuring longevity of the tube, especially if using a fairly large value of reservoir capacitor (I'm using 47uF 450V). I was thinking of putting 100R in series before each plate (pins 4 and 6 on the 5Y3GT).
I figure 1W rated parts will do. My thinking:
Plate supply draws about 40mA total.
100R * 0.04A = 4V (dropped across each 100R resistor)
4V * 0.04A = 0.16W (power dissipated in each 100R resistor)
100R * 0.04A = 4V (dropped across each 100R resistor)
4V * 0.04A = 0.16W (power dissipated in each 100R resistor)
I figure the 1W rated parts give lots of headroom.
Advisable?
Some diyAudio members are especially conservative when designing circuits that connect to power transformers in vacuum tube equipment. They feel it necessary to assume that a bunch of worst case scenarios eventually will occur, simultaneously
Me, personally, I think this is a bit too conservative for my tastes. While I agree that following these guidelines will make the equipment more robust, I think I prefer to spend my "ridiculous overkill" budget on other areas of the circuit design. No 1.6 kilovolt snubber caps for me, thank you. Flameproof resistors, sure, why not; the cost increase is negligible.
By the way, readers may enjoy independently analyzing the old wives' tale that you should use only carbon composition "low inductance" resistors for snubbers. Here's the data you need to carry out the analysis: (i) maximum frequency of resonant oscillation is 3 MHz and often much much lower, especially in vacuum tube HT windings; (ii) typical values of optimum snubber resistance (Q<0.5) for high voltage secondaries, range from 200 ohms to a maximum of 2000 ohms.
There you go. That's everything you need to calculate a value of parasitic series inductance which becomes non-negligible (more than 10% of the resistor value) at the resonant frequency. It's a straightforward piece of arithmetic and if you use the values above you'll get a number of microhenries. Not nanohenries, microhenries. Metal film resistors, carbon film resistors, metal foil resistors, surface mount resitors, just about any kind of resistor except wirewound resistors, have series inductance measured in nanohenries. Three orders of magnitude lower. Conclusion: in a snubber feel free to use whatever resistor type you like, as long as it's not wirewound. Let grandpa stay with his humidity-absorbing carbon composition resistors if he likes; YOU don't need to.
- IT IS ASSUMED the equipment contains no surge protectors, MOVs, gas discharge tubes, transZorbs, or other countermeasures. The entire upstream chain of power strips and power conditioners and plugs and jacks, is also devoid of surge protectors and lightning suppressors, all the way to the utility company's transformer on the pole. The whole chain does absolutely nothing to limit voltage surges on the mains
- IT IS ASSUMED a surge of at least 1000 extra volts WILL occur on the mains, while the equipment is turned on and playing music through the speakers
- IT IS ASSUMED the surge WILL pass from the transformer primary (times the turns ratio) to the secondary; you cannot depend on the transformer to act as a lowpass filter when the manure hits the fan
Me, personally, I think this is a bit too conservative for my tastes. While I agree that following these guidelines will make the equipment more robust, I think I prefer to spend my "ridiculous overkill" budget on other areas of the circuit design. No 1.6 kilovolt snubber caps for me, thank you. Flameproof resistors, sure, why not; the cost increase is negligible.
By the way, readers may enjoy independently analyzing the old wives' tale that you should use only carbon composition "low inductance" resistors for snubbers. Here's the data you need to carry out the analysis: (i) maximum frequency of resonant oscillation is 3 MHz and often much much lower, especially in vacuum tube HT windings; (ii) typical values of optimum snubber resistance (Q<0.5) for high voltage secondaries, range from 200 ohms to a maximum of 2000 ohms.
There you go. That's everything you need to calculate a value of parasitic series inductance which becomes non-negligible (more than 10% of the resistor value) at the resonant frequency. It's a straightforward piece of arithmetic and if you use the values above you'll get a number of microhenries. Not nanohenries, microhenries. Metal film resistors, carbon film resistors, metal foil resistors, surface mount resitors, just about any kind of resistor except wirewound resistors, have series inductance measured in nanohenries. Three orders of magnitude lower. Conclusion: in a snubber feel free to use whatever resistor type you like, as long as it's not wirewound. Let grandpa stay with his humidity-absorbing carbon composition resistors if he likes; YOU don't need to.
You are equating DC current draw with AC RMS current. There is a factor of 2-3 to be used. This means 4-9 in power terms. As a minimum, I would assume 5 times as much power dissipation in the added resistors as a naive calculation gives.rongon said:
Thanks once again, DF96.
I need to find someplace that explains how to predict power dissipated in a load with AC voltage across it. This could be useful for speaker power and all sorts of things.
I'll use 5W resistors, for sure. Fortunately, I have some 100R 5W 'cement box' types.
I need to find someplace that explains how to predict power dissipated in a load with AC voltage across it. This could be useful for speaker power and all sorts of things.
I'll use 5W resistors, for sure. Fortunately, I have some 100R 5W 'cement box' types.
That is easy: it is V^2 /R, just as for DC, but you have to use the RMS value.rongon said:
Similarly, it is I^2 R for current - but again you use the RMS value.
For sinewaves RMS = 0.707 x peak. For square waves RMS = peak. For other AC waveforms you have to do a somewhat messy calculation, or get a computer to do it for you!
Keep a couple of formulas in mind.
V = IZ The AC form of Ohm's Law.
Z = (R2 + {XL - XC}2)1/2 Where XL is the inductive reactance and XC is the capacitive reactance. Impedance is a complex, not real, number.
AC power = (V) (I) (cos[n]) Where n is the phase angle between the current and voltage vectors. This is the scalar (dot) product of the vectors.
In a pure reactance, the phase angle is 90o and the power is zero. Power is dissipated only in a resistance.
V = IZ The AC form of Ohm's Law.
Z = (R2 + {XL - XC}2)1/2 Where XL is the inductive reactance and XC is the capacitive reactance. Impedance is a complex, not real, number.
AC power = (V) (I) (cos[n]) Where n is the phase angle between the current and voltage vectors. This is the scalar (dot) product of the vectors.
In a pure reactance, the phase angle is 90o and the power is zero. Power is dissipated only in a resistance.
Well, I'm embarrassed to have to say that those formulas are beyond my meager understanding. I'll have to infer safe operating parameters from the data sheets.
The 5Y3GT data sheet shows in the chart "Operation Characteristics - Full Wave Rectifier with Capacitor-Input Filter" that a series resistance of 135 ohms per plate is used for DC output voltages of 400V and higher.
Since my supply is putting out almost exactly 400V DC, I'll look for the secondary winding to have a plate-to-plate DCR of 270 ohms or so. If it is lower than that, I'll add a series resistor just before each plate on the 5Y3 socket to get to 135 ohms series resistance per plate (secondary winding DCR to CT plus series resistor per plate = 135 ohms, or 270 ohms plate-to-plate).
Yes?
--
The 5Y3GT data sheet shows in the chart "Operation Characteristics - Full Wave Rectifier with Capacitor-Input Filter" that a series resistance of 135 ohms per plate is used for DC output voltages of 400V and higher.
Since my supply is putting out almost exactly 400V DC, I'll look for the secondary winding to have a plate-to-plate DCR of 270 ohms or so. If it is lower than that, I'll add a series resistor just before each plate on the 5Y3 socket to get to 135 ohms series resistance per plate (secondary winding DCR to CT plus series resistor per plate = 135 ohms, or 270 ohms plate-to-plate).
Yes?
--
Eli isn't … Z = √( R² + (XL - XC)² ) a solution for a particular configuration of R, L & C?
The conclusion “in a pure reactance, phase angle is 90°, and power is zero” is for either a parallel C || L bandstop or a C -- L series bandpass, at the resonant frequency. At any other frequency, phase angle isn't 90°. OR at least that's what I seem to remember.
GoatGuy
The conclusion “in a pure reactance, phase angle is 90°, and power is zero” is for either a parallel C || L bandstop or a C -- L series bandpass, at the resonant frequency. At any other frequency, phase angle isn't 90°. OR at least that's what I seem to remember.
GoatGuy
I snub secondaries (necessary or not being another matter) I use one snubber per half secondary which eases the needed rating.
Some transformers have one half of the secondary wound around the outside of the other. This makes the resistance of the outer winding a little larger (perhaps 10-20 ohms) and adding resistance gives an opportunity to balance this.
this is what i do in my builds.....but rc network is in the primary side, an MOV in parallel and an NTC in series completes the deal...
In a pure reactance, phase is 90 and power is zero. This is always true, however that pure reactance is produced: L, C, L+C in series (resonant or not), L and C in parallel (resonant or not).GoatGuy said:
In real life, you never get a pure reactance as there is always some resistance lurking - but at low frequencies you can get sufficiently close to pure reactance for all practical purposes.
I have a couple of questions as my freshly wound double C core transformer is showing still some buzzing...I am using 350-0-350v and 70mA in choke input, rectifier is a rgn1064. I added a 1uf cap before the first choke, which made everything more silent. But...still some buzzing in the main transformer, which are really huge guys, perfectly able to deliver 300mA or more.
So,
- Besides component max ratings: Does it matter if one snubber is added over the whole secondary from plate to plate or from each end only to CT, so basically to earth ?
- Besides using my ear to here if the transformer is buzzing or not: Can this stuff be seen on a scope and what is it to look for ... so is there a way to fine tune the snubber regarding capacity and reistance and to have some measurement equipment to support this process ?
I believe, I tried 100R+100nF from secondary ends to CT and the whole thing started to buzz even worse...
So,
- Besides component max ratings: Does it matter if one snubber is added over the whole secondary from plate to plate or from each end only to CT, so basically to earth ?
- Besides using my ear to here if the transformer is buzzing or not: Can this stuff be seen on a scope and what is it to look for ... so is there a way to fine tune the snubber regarding capacity and reistance and to have some measurement equipment to support this process ?
I believe, I tried 100R+100nF from secondary ends to CT and the whole thing started to buzz even worse...
hey Blitz, transformers will buzz when very close or at saturation....
so if your situation is like this, then derating the traffo is needed,
or else get a bigger one....
varnishing or dipping the whole traffo assembly in varnish might help...
so if your situation is like this, then derating the traffo is needed,
or else get a bigger one....
varnishing or dipping the whole traffo assembly in varnish might help...
This is a trafo wound by AE Europe especially for quietness...and is significantly oversized...so, I guess I need to find a fix elsewhere...like a snubber...
why not email them and tell them your issues, they should be able to come up with fixes.....
tbh, it is hard to troubleshoot in front of a pc monitor...
transformer size could be one thing, but the operating flux density is another, so if they designed the traffo to run on flux very near the knee of the magnetization curve, then buzzing is easily achieved....
i design and build my own and i make it a point to run flux densities way below the knee of the magnetization curve precisely to avoid such issues...regulation and power is a bit lower, but this is the compromise i take to get cool running and quiet traffos......
tbh, it is hard to troubleshoot in front of a pc monitor...
transformer size could be one thing, but the operating flux density is another, so if they designed the traffo to run on flux very near the knee of the magnetization curve, then buzzing is easily achieved....
i design and build my own and i make it a point to run flux densities way below the knee of the magnetization curve precisely to avoid such issues...regulation and power is a bit lower, but this is the compromise i take to get cool running and quiet traffos......
- Status
- Not open for further replies.