Perhaps the issue of whether a snubber can be 'designed' for a typical valve amp HT winding is a little clouded by how non-linear a HT winding is at frequencies above mains.
To get a feel, I checked a 385V half-winding of a 385V-0-385V secondary, for a 125mA rated HT supply from circa 1960.
Initially, all windings open and independent except that the 385-0-385 has a common CT connection. The 50Hz inductance of the 385-CT winding was about 6H at 2mA excitation. The ESR was 138 ohm. The SRF was down at 2.3kHz, indicating a bulk capacitance of 800pF at the frequency. The next anti-resonance was about 50kHz , and then resonance about 75kHz. Connecting the ES to HT CT, reduced the SRF to about 1.8kHz, and higher resonances were lower too.
Shorting all other windings, gave 370mH leakage inductance, with SRF at 70kHz. Connecting the ES to the CT gave about 380mH leakage inductance but lowered the SRF to 40kHz, indicating about 42pF capacitance.
Configuring the windings to be somewhat equivalent to an operating circuit at the point of diode turn-off (ie. mains and heater windings shorted, but both HT windings open, and CT connected to ES), gave about 380mH with SRF at 40kHz as well.
So one query would be whether the commonly used snubber configurations and values would perform as expected for this application.
To get a feel, I checked a 385V half-winding of a 385V-0-385V secondary, for a 125mA rated HT supply from circa 1960.
Initially, all windings open and independent except that the 385-0-385 has a common CT connection. The 50Hz inductance of the 385-CT winding was about 6H at 2mA excitation. The ESR was 138 ohm. The SRF was down at 2.3kHz, indicating a bulk capacitance of 800pF at the frequency. The next anti-resonance was about 50kHz , and then resonance about 75kHz. Connecting the ES to HT CT, reduced the SRF to about 1.8kHz, and higher resonances were lower too.
Shorting all other windings, gave 370mH leakage inductance, with SRF at 70kHz. Connecting the ES to the CT gave about 380mH leakage inductance but lowered the SRF to 40kHz, indicating about 42pF capacitance.
Configuring the windings to be somewhat equivalent to an operating circuit at the point of diode turn-off (ie. mains and heater windings shorted, but both HT windings open, and CT connected to ES), gave about 380mH with SRF at 40kHz as well.
So one query would be whether the commonly used snubber configurations and values would perform as expected for this application.
Trobbins,
You stated that the secondary self capacitance determined by your method, with all windings open, was 800 pF. Form that and your SRF values, ES earthed and not earthed, that the self capaciatnace of the secondary with ES earthed is about 1300 pF - a not unreasonable figure.
First, this is about 100 to 130 times the likely rectifier tube capacitance, so the tube capacitance can be ignored.
Secondly, for a leakage inductance of 370 mH, and capacitance of 1300 pF, and an ESR of 138 ohms, the resonance is about 7300 Hz, both indicating a reactance of 17 kohm. The ESR of 138 ohms thus indicates a Q of 122 - sounds a bit scarey for some perhaps.
However, as stated before, the charge storage time in vacuum tube rectifier is of the order of a few nanoseconds. The conduction period is rather long - typically 30 degrees - a result of limits on the first filter capacitor size. Let's say the HT drain is 100 mA, then a maximum of 0.1 A x duty factor flows for 10 nSec (say) each time a tube anode turns off.
We can then estimate the maximum possible transformer voltage rise by assuming all this current goes in the transformer self capacitance, i.e., from V = I x t / C,
i.e 0.1 x180/30 x 10x10^-9 / 1300x10-12, i.e., 4.6 volts.
4.6 V is clearly negligible compared to the secondary voltage of 350 V.
I've used some simplifying assumptions, here, but I've done it in such a way as to grossly overestimate this voltage.
Clearly, a snubber network can be designed to bring down the Q, but even more clearly there is NO need for a snubber network.
You stated that the secondary self capacitance determined by your method, with all windings open, was 800 pF. Form that and your SRF values, ES earthed and not earthed, that the self capaciatnace of the secondary with ES earthed is about 1300 pF - a not unreasonable figure.
First, this is about 100 to 130 times the likely rectifier tube capacitance, so the tube capacitance can be ignored.
Secondly, for a leakage inductance of 370 mH, and capacitance of 1300 pF, and an ESR of 138 ohms, the resonance is about 7300 Hz, both indicating a reactance of 17 kohm. The ESR of 138 ohms thus indicates a Q of 122 - sounds a bit scarey for some perhaps.
However, as stated before, the charge storage time in vacuum tube rectifier is of the order of a few nanoseconds. The conduction period is rather long - typically 30 degrees - a result of limits on the first filter capacitor size. Let's say the HT drain is 100 mA, then a maximum of 0.1 A x duty factor flows for 10 nSec (say) each time a tube anode turns off.
We can then estimate the maximum possible transformer voltage rise by assuming all this current goes in the transformer self capacitance, i.e., from V = I x t / C,
i.e 0.1 x180/30 x 10x10^-9 / 1300x10-12, i.e., 4.6 volts.
4.6 V is clearly negligible compared to the secondary voltage of 350 V.
I've used some simplifying assumptions, here, but I've done it in such a way as to grossly overestimate this voltage.
Clearly, a snubber network can be designed to bring down the Q, but even more clearly there is NO need for a snubber network.
Last edited:
You postulate a slowly decaying (Q>120!) sinewave at 7.3 kHz whose initial amplitude is 4.6 volts. This is superimposed on a 50 Hz sinewave whose amplitude is 500 volts (350 rms).
Seems to me the quantities of interest are the power supply's line rejection at 7.3 kHz and the active electronics' PSRR at 7.3 kHz. If the output signal amplitude is 1V, and if you seek a signal-to-supply-noise ratio of 80dB at 7.3 kHz, then you want
Seems to me the quantities of interest are the power supply's line rejection at 7.3 kHz and the active electronics' PSRR at 7.3 kHz. If the output signal amplitude is 1V, and if you seek a signal-to-supply-noise ratio of 80dB at 7.3 kHz, then you want
- line rejection dB @7.3kHz + PSRR dB @7.3kHz < (1V / 4.6V) * -80dB
- line rejection dB @7.3kHz + PSRR dB @7.3kHz < -93dB
Last edited:
How about putting a small air core inductor, on the order of a few times the power xfmr leakage L, in series on the primary side. Should attenuate any HF spikes coming in. Could be used to prevent a primary side MOV from blowing out too.
Then there are the common mode inductors for the CM issue as well. One could just wrap the power cord around (through) a big ferrite toroid a few times for CM protection. (need to check for series resonance with the xfmr common mode C, pri to sec, though, AM radio or something get through)
Then there are the common mode inductors for the CM issue as well. One could just wrap the power cord around (through) a big ferrite toroid a few times for CM protection. (need to check for series resonance with the xfmr common mode C, pri to sec, though, AM radio or something get through)
Last edited:
In my calculation I established an upper limit. The actual voltage will be a lot less. the Q of the transformer self capacitance will be very low for a start.
This 7.3 KHz is at the transformer, and completely isolated from the amplifier power supply because the rectifier tube is non-conducting at the time the 7.3 kHz is present. It was the tube cutting off that started it remember! There is no need for Mark J to worry about sufficient PSRR. And 7.3 kHz is way too low for any radiative coupling to signal circuits in any reasonable layout. There's hundreds of volts of signal at the output transformer.
This 7.3 KHz is at the transformer, and completely isolated from the amplifier power supply because the rectifier tube is non-conducting at the time the 7.3 kHz is present. It was the tube cutting off that started it remember! There is no need for Mark J to worry about sufficient PSRR. And 7.3 kHz is way too low for any radiative coupling to signal circuits in any reasonable layout. There's hundreds of volts of signal at the output transformer.
If it's a multi secondary transformer then the 7.3 kHz is coupled into the other windings where it may do some harm... different diodes in different supplies having different conduction angles and all. If the transformer includes a 6.3VAC secondary for heaters, the 7.3 kHz also gets into that. Although I have no idea whether the transfer function from the heater to other electrodes is -200dB at 7.3kHz or -0.2dB at 7.3kHz ... I'm a valve virgin.
A non-existent problem.
The worst possible case is the 7.3 kHz at 4.6 V on the HT windings gets 100% coupled into a lower voltage winding - in which case the design step down ration applies.
In fact, remember that this 7.3 kHz arises in the leakage inductance of the HT winding. By definition, leakage inductance is not coupled into other windings. So, no coupling at all.
You can look at it another way: The 7.3 kHz arises due to a supposed high Q circuit. If there IS significant coupling, either by another rectifier on a different conduction angle, or there's no rectifier at all (as in heater windings), then there is a low impedance load on it. A very low impedance load in fact - either filter parts of heaters. This low impedance will dramatically reduce the working Q and heavily damp out the oscillation.
So, we have a situation where if there is interwinding coupling at 7.3 kHz there cannot be a problem, and if there is no coupling there cannot be a problem. So we don't have to wrorry about interwinding coupling
Yep. Definitately a non problem.
Ask yourself this: Noting that snubbers in solid state equipment is routine, but was never seen in tube based equipment, domestic or professional, during the tue era, could there be a reason?
Of course there is a reason. In tube based equipment there is nothing needing to be snubbed.
Hi,
I've been designing with separate heater xformers since 1986. All true dual mono to the extreme so what else is new?
Cheers,
Having worked for a transformer manufacturer decades ago as their engineer, I've been designing and winding my own power and output transformers for my own projects at home ever since. So all windings go on the same core as it's cheaper and takes up less room. And if required I can put in a screen (which it never is in tube equipment for this reason, though as screen between the primary and all secondaries certainly cuts down AC mains-borne noise).
But if I was restricted to commercial transformers, it can often be the case where an extra 6.3 V winding, 6.3V CT winding, or other low voltage winding is needed and is in nobody'd catalogue.
If you put the 6.3 volt heater winding just atop the primary, would that act as a screen (shield), assuming that this 6.3V winding is grounded?
Then above this, the high voltage B+ secondary would be wound.
There is a significant leakage inductance (0.37H) in the example winding that would have a current waveform that has a discontinuity occurring at diode turn off. Prior to turn-off, the diode current would be ramping down at some level of dI/dt. Switching a valve diode device is akin to zero-current-switching.
A 'snubber' is a means of managing any energy in the winding that may want to continue to circulate via the winding terminals, rather than find other paths to take.
The noise that the screen stops is in part switching transients that are certainly reduced by earthing the heater run, but sometimes enough energy still remains on the heater wiring run to couple into grid circuits.
It isn't usually a big problem - most tube radios were built without screens.
A good measure of shielding can be had by having the HT secondaries wound as two groups. The first group has its earthy end next to the primary - so that its' first layer acts partly as an earthed screen. This is usually how it is done.
From the point of view of transformer construction it is better to have the heater windings the outer most windings. The first layer you wind on has nice neat square turns closely fitting around the former. As you put more layers on, the corners become smooth quarter turns of radius equal to teh distance out from the former. The straight runs between corners bow outwards and you end up needing considerable tension in the wire to minimise the bowing outward.
The bowing outward, which gets worse and worse as you put more layers on wastes space, forcing you to use a smaller wire guage, and uses up more wire length, both of which increase electrical resistance and reduces transformer efficiency. It also increases leakage flux a little.
So the heavy gauge heater windings, which resist bending the most, are best put on last, so that bowing is manageable.
The design of transfomers is part engineering and part art, and there are sometimes other reasons for putting particular windings on in a particular order. Sometimes heavy current windings are done in bifilar or trifilar style so that a small gauge wire can be used. For a particular mix of design load current on each winding, this gives better use of the winding later width. A transformer engineer tries to avoid winding layers that occupy less than the full width.
I have several 1950's textbooks on the design & engineering of transformers, and none of them cover the full range of design aspects of a fairly typical tube amp power transformer or tube amp output transformer.
Last edited:
I forgot to mention another reason for putting the heater winding(s) on last.
Mains borne noise that gets on the the HT windings generally doesn't matter much because most of the time it is blocked by the rectifier during the rectifier's off time, and what gets through is considerably attenuated by the ripple filter.
But the heater winding goes to every tube, and normally isn't filtered. In input stages there may be sufficient capacitive coupling to transfer the noise into the grid circuit.
Some hobbyists put filtering and even rectification and electronic regulation in heater runs, with reducing noise as the justification. But, having considerable commercial experience, I don't recommend that. It just isn't needed if you do everthing else right - such as transformer windings correctly sequenced, good layout, and wiring practices, etc.
Mains borne noise that gets on the the HT windings generally doesn't matter much because most of the time it is blocked by the rectifier during the rectifier's off time, and what gets through is considerably attenuated by the ripple filter.
But the heater winding goes to every tube, and normally isn't filtered. In input stages there may be sufficient capacitive coupling to transfer the noise into the grid circuit.
Some hobbyists put filtering and even rectification and electronic regulation in heater runs, with reducing noise as the justification. But, having considerable commercial experience, I don't recommend that. It just isn't needed if you do everthing else right - such as transformer windings correctly sequenced, good layout, and wiring practices, etc.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.