So, to find the interwinding capacitance for a simple two-winding transformer model, could I apply a known amplitude and frequency of RF to the primary of a real transformer (to be modeled), measure how much of that RF appears across the secondary, and then put an RF source (with the same amplitude and frequency) on the model's primary and adjust the model's interwinding C value until I get the same RF level across the model's secondary?
Just a wild guess. Otherwise, how?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
Just a wild guess. Otherwise, how?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
Or how about varying the primary's excitation frequency and looking for a resonant peak across the secondary? Would it then be possible to estimate the interwinding capacitance, if the primary and secondary (and leakage) inductances, and the resistances, are known?
By the way, does anyone have a feel for whether or not the INTRA-widing capacitances might be important to model? i.e. the capacitance seen between the two primary leads, and the capacitance seen between the two secondary leads?
I don't yet have much of a concept of the relative magnitudes of any of these capacitances. But it seems like they might be important, when looking at any types of high-frequency effects in any transformer-related circuits. But, if not, usually, then that would be very nice to know, right about now.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
By the way, does anyone have a feel for whether or not the INTRA-widing capacitances might be important to model? i.e. the capacitance seen between the two primary leads, and the capacitance seen between the two secondary leads?
I don't yet have much of a concept of the relative magnitudes of any of these capacitances. But it seems like they might be important, when looking at any types of high-frequency effects in any transformer-related circuits. But, if not, usually, then that would be very nice to know, right about now.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
Firstly, I am quite envious of the learned work done regarding equivalent models of UL output transformers. To my mind getting a fairly exact model, especially regarding distributed capacitances, must be quite a task. Up to now I have chickened out and done the reponse with tubes and transformer in situ. The hf side is not usually a problem and one can take phase shifts into account by tailoring the final response. But just my laziness!
Tom,
I have found that quite often the h.f. limit is governed more by the capacitance (interwinding plus tube) than the leakage reactance. There is an unbalance in a design where the leakage is low enough to take one to quite >100KHz, but the capacitance starts cutting after 30KHz. In this sense I am limiting my sectionalising to 3 - 4 secondaries and/or using some extra insulation thickness between pri-sec. sections. In a 100W C-core transformer I tried to do as good a job as possible of guestimating the equivalent internal C, and got a figure of around 1,2nF equivalent at the tube anodes. This was with 4 secondaries, not a split bobbin.
Tom,
I have found that quite often the h.f. limit is governed more by the capacitance (interwinding plus tube) than the leakage reactance. There is an unbalance in a design where the leakage is low enough to take one to quite >100KHz, but the capacitance starts cutting after 30KHz. In this sense I am limiting my sectionalising to 3 - 4 secondaries and/or using some extra insulation thickness between pri-sec. sections. In a 100W C-core transformer I tried to do as good a job as possible of guestimating the equivalent internal C, and got a figure of around 1,2nF equivalent at the tube anodes. This was with 4 secondaries, not a split bobbin.
I have done a little more research and have found that, for a simple two-winding transformer, the interwinding capacitance can be measured by shorting both windings and measuring the capacitance between them. Then, each winding can be shorted, one at a time, and the intrawinding capacitance of each one in parallel with the interwinding capacitance can be measured. Knowing the interwinding capacitance, each intrawinding capacitance can then be calculated.
But I'm still not sure what the best capacitance measurement methods would be, in those cases.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
But I'm still not sure what the best capacitance measurement methods would be, in those cases.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
Maybe to model those UL (UltraLinear) transformers well, one would need to resort to using the techniques similar to those described on these pages:
http://www.d4magnetics.com/page8/page8.html
The models are derived from the geometric and materials properties of the core and windings.
This looks quite interesting, too:
http://www.upmdie.upm.es/investigacion/ficheros/APEC_1997.pdf
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
http://www.d4magnetics.com/page8/page8.html
The models are derived from the geometric and materials properties of the core and windings.
This looks quite interesting, too:
http://www.upmdie.upm.es/investigacion/ficheros/APEC_1997.pdf
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
This author posted a new and improved spreadsheet at:
http://www.diyaudio.com/forums/tubes-valves/181578-spice-transformer-model-spreadsheet.html
All you always wanted to know about OPTs is in RDH4 chapter 5 !
In 5.3(v) you'll find how to calculate winding capacitances.
They have multiple sources (inter winding and inter layers) and always produce dips or peaks accompanied by suddent phase shifts that are highly undesirable !
The larger the OPT, the higher the parasitic capacitances
Yves.
Interesting application note on determining the intra winding capacitance
http://www.omicron-lab.com/fileadmin/assets/application_notes/AP_Note_Transformer_V_1_0c.pdf
Going to repeat here what I posted on the vacuum tube spice page. This is cobbled from Christophe Basso's book on switch mode power supplies. Add the parasitic elements to your heart's content. Multiple sections can be added in series and parallel to derive a UL transformer of your choosing by changing the parameter "ratio" in the spice model:
Attachments
32 section Hammond 1628SEA's UL model
I have attached my implementation of a 32 section distributed lumped element LTSpice model of the Hammond 1628SEA OPT.
This post is not meant to be complete or scientific, but please let me know if I am on the right or wrong track.
I was interested in attempting to model the primary-secondary capacitance and in particular the amplitude and phase relationship between the plate and UL taps of the primary at higher frequencies.
I had bench built a simple SE 6CA7 UL stage using the Hammond but had seen oscillations when connected in UL mode, but not in triode or pentode mode. Testing the OPT showed that around 88kHz there was a 180 deg phase shift between UL and plate connections, with V(UL) becoming larger than V(plate) (and leading to positive feedback to g2 I guess) so I wanted to see if this could be modelled in Spice.
The model is simple in that the primary and secondary are split into 32 mutually coupled inductors with the primary-secondary capacitance distributed as 33 discrete capacitors between the respective primary and secondary inductor nodes. In addition I have included dc resistance, primary winding capacitance, leakage reactance (manifesting as the coupling coefficient) and a series cap-resistor to attempt to model HF hysteresis losses….. (not very scientific I know…)
I found that the 32 section was an improvement on a 16 and 8 section model because the model performance in simulation was less sensitive to the input capacitance values, otherwise you end up with oscillations in the simulated circuit that are not there in real life.
To get the parameters not listed by the manufacturer I had only a LRC meter, oscilloscope and signal gen at hand, so used 3 small test circuits to measure parameters, then simulate the test circuit and compare results.
Extract of the sub circuit model:
************************************************************************************************************
* Generic UL SE OPT
* Input parameters:
* LP = manufacturer large signal primary inductance
* RP = manufacturer primary load impedance
* RS = manufacturer secondary load impedance
* RSP = measured primary series winding resistance
* RSS = measured secondary series winding resistance
* LPL = measured leakage inductance at primary with secondary short circuit. Used to calculate coupling coeffcient.
* CPS = measured primary/secondary inter-winding capacitance
* CP = estimated parallel primary winding capacitance - this is distributed accross the primary inductances - can create instability or too low resonant frequency;
* CPP = estimated primary winding capacitance Chose with CPS to get as close a match to resonant freq. Used to try simulate high freq core losses
* RHL = load resistor used in series with CPP for hysterisis losses simulation; for max loss RHL = output tube plate resistance;
* RPP = primary parallel resistance chose to match input level with open circuit secondary at resonance
* k = UL factor (winding fraction from B+)
.SUBCKT 1628SEA 1 2 3 4 5; P B+ S0 S1 UL;
*
X1 1 2 3 4 5 OPT LP=45 RP=5000 RS = 8 RSP = 178 RSS = 0.5 LPL = 25.8m CPS = 3.5n CP=0.5p CPP=1n RHL= 1000 k = 0.4 RPP=546k
.ends
This model does simulate the OPT performance reasonably well and shows the UL plate relative amplitude and phase change in a satisfactory way, although the frequencies at which this resonance occurs differs slightly in real life. Installing the model as a subcircuit in Spice correctly predicts the oscillation I observed at the bench, and allowed me to eliminate this with a 470pF capacitor between g2 and plate.
I have attached the simulated transformer frequency response for the plate connection (green) and the UL connection (blue) which shows the amplitude and phase change, very similar to what I measured.
It would be interesting to try a few other OPT’s (I only have the Hammond) to see how useful/adaptable this method is.
Any comments are appreciated.
I have attached my implementation of a 32 section distributed lumped element LTSpice model of the Hammond 1628SEA OPT.
This post is not meant to be complete or scientific, but please let me know if I am on the right or wrong track.
I was interested in attempting to model the primary-secondary capacitance and in particular the amplitude and phase relationship between the plate and UL taps of the primary at higher frequencies.
I had bench built a simple SE 6CA7 UL stage using the Hammond but had seen oscillations when connected in UL mode, but not in triode or pentode mode. Testing the OPT showed that around 88kHz there was a 180 deg phase shift between UL and plate connections, with V(UL) becoming larger than V(plate) (and leading to positive feedback to g2 I guess) so I wanted to see if this could be modelled in Spice.
The model is simple in that the primary and secondary are split into 32 mutually coupled inductors with the primary-secondary capacitance distributed as 33 discrete capacitors between the respective primary and secondary inductor nodes. In addition I have included dc resistance, primary winding capacitance, leakage reactance (manifesting as the coupling coefficient) and a series cap-resistor to attempt to model HF hysteresis losses….. (not very scientific I know…)
I found that the 32 section was an improvement on a 16 and 8 section model because the model performance in simulation was less sensitive to the input capacitance values, otherwise you end up with oscillations in the simulated circuit that are not there in real life.
To get the parameters not listed by the manufacturer I had only a LRC meter, oscilloscope and signal gen at hand, so used 3 small test circuits to measure parameters, then simulate the test circuit and compare results.
Extract of the sub circuit model:
************************************************************************************************************
* Generic UL SE OPT
* Input parameters:
* LP = manufacturer large signal primary inductance
* RP = manufacturer primary load impedance
* RS = manufacturer secondary load impedance
* RSP = measured primary series winding resistance
* RSS = measured secondary series winding resistance
* LPL = measured leakage inductance at primary with secondary short circuit. Used to calculate coupling coeffcient.
* CPS = measured primary/secondary inter-winding capacitance
* CP = estimated parallel primary winding capacitance - this is distributed accross the primary inductances - can create instability or too low resonant frequency;
* CPP = estimated primary winding capacitance Chose with CPS to get as close a match to resonant freq. Used to try simulate high freq core losses
* RHL = load resistor used in series with CPP for hysterisis losses simulation; for max loss RHL = output tube plate resistance;
* RPP = primary parallel resistance chose to match input level with open circuit secondary at resonance
* k = UL factor (winding fraction from B+)
.SUBCKT 1628SEA 1 2 3 4 5; P B+ S0 S1 UL;
*
X1 1 2 3 4 5 OPT LP=45 RP=5000 RS = 8 RSP = 178 RSS = 0.5 LPL = 25.8m CPS = 3.5n CP=0.5p CPP=1n RHL= 1000 k = 0.4 RPP=546k
.ends
This model does simulate the OPT performance reasonably well and shows the UL plate relative amplitude and phase change in a satisfactory way, although the frequencies at which this resonance occurs differs slightly in real life. Installing the model as a subcircuit in Spice correctly predicts the oscillation I observed at the bench, and allowed me to eliminate this with a 470pF capacitor between g2 and plate.
I have attached the simulated transformer frequency response for the plate connection (green) and the UL connection (blue) which shows the amplitude and phase change, very similar to what I measured.
It would be interesting to try a few other OPT’s (I only have the Hammond) to see how useful/adaptable this method is.
Any comments are appreciated.
Attachments
I agree it is a bit like taking a sledgehammer to a crack a nut and does slow down the simulation somewhat. It is not clever but simply an interpretation of the physical that lets Spice do the maths for you. For a small SE type circuit the wait is not too long, but long transient analysis will take time.
It has been a useful educational exercise for me and it did correctly predict the oscillation with UL connection,and confirmed the corrective measure would work without having to stick my hands into HV!
If you know the physical winding arrangement you can construct any model that represents the OPT. I have now added the (not connected) 16 Ohm tap and it is interesting how that changes the resonant frequencies.
All good fun. Time to now build the real thing.
It has been a useful educational exercise for me and it did correctly predict the oscillation with UL connection,and confirmed the corrective measure would work without having to stick my hands into HV!
If you know the physical winding arrangement you can construct any model that represents the OPT. I have now added the (not connected) 16 Ohm tap and it is interesting how that changes the resonant frequencies.
All good fun. Time to now build the real thing.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.