I hope this is the right place for this question. I'm running an LTSpice model of an EL84 PP amp. I built the OPT model using the Transformer Models.xls spredsheet that was published here. If the mutual inductance is set at 1, the model runs ok.
If it is set at 0.9995... the model blows out and shows massive oscillation. Is this due to a problem with my design, or is it some quirk of the OPT implementation?
If it is set at 0.9995... the model blows out and shows massive oscillation. Is this due to a problem with my design, or is it some quirk of the OPT implementation?
k=1 is an ideal case, which does not exist in real life, 0.9995 seems more realistic and if your sim oscillates then something is wrong with your circuit.
Mutual inductance is measured in henries. Do you mean that or do you mean the inductance coupling coefficient, k, for these 1 and 0.9995 values?
Do you have realistic series resistances? The matrix with mutual inductances included can get very close to being singular.
It boils down to leakage inductance.... You can leave the K=1 and add leakage inductance to the transformer.... or simply adjust K to account for the leakage inductance... Either way the results will be the same... Don't forget to add the DC winding resistance as well as winding capacitance ....both interwinding and self winding....11.1H seems extremely low for 1/4 the 8K plate load... Inductance is not a constant.... It varies greatly with signal level...this is usually modeled based on the BH loop of the steel and the way it is stacked...
You should post the whole amplifier circuit: if you take global feedback post-transformer, you could have these symptoms as soon as you have the slightest leakage inductance.
The OPTs are old partridges which I measured - the series resistances are 364 ohms plate-to-plates and 0.5 ohm on the speaker side. I haven’t put any mutual inductance as I misnamed it in
I've found that quite often you need to add in the compensation networks to the simulation in order for it to solve (e.g. RC across first stage anode resistor, or Zobels from B+ to either end of the primary winding).
Thanks all for the very informative and helpful replies. I have to admit that came came blundering into this. This is my first LTSpice model, and I spent quite a lot of time on the valve sections and then just bunged in an OPT model that I found on someone else's circuit, and adjusted without much insight. For example, I have no capacitance in there.
@jazbo8 that LTwiki page is pretty informative and describes exactly the issue I had.
@tikiroo as you can see below, I put the zobels in, not much thought went into values, just fiddled around until they made a difference. They all but stopped the HF oscillation. Empiricism instead of reason. In your amps, did you only have to put the zobels into the spice model, or did you need them in real life too? In other words, is the oscillation an LTSpice artifact or real? Also, do you have a model for your Partidges? (I remember you saying had a pair very similar to mine)
@elvee the circuit is below - all pretty standard. The things I haven't worked out too carefully yet are the coupling caps and the cathode bypass caps - still need to justify the values I have in there. Also the nfb RC is just common values - I have to give that more thought too.
But anyway, just because this model doesn't have to be massively correct I'm setting k = 1 just so I can finish the design and build the amp. If any of these issues occur in the real world, then I'll come back to LTSpice and correct the design. It's just components and soldering after all.
PS: fiddling around, I noticed that the circuit is quite sensitive to the value of C in the NFB loop - the wrong value also causes oscillation, so a bit of work needed there anyway.
@jazbo8 that LTwiki page is pretty informative and describes exactly the issue I had.
@tikiroo as you can see below, I put the zobels in, not much thought went into values, just fiddled around until they made a difference. They all but stopped the HF oscillation. Empiricism instead of reason. In your amps, did you only have to put the zobels into the spice model, or did you need them in real life too? In other words, is the oscillation an LTSpice artifact or real? Also, do you have a model for your Partidges? (I remember you saying had a pair very similar to mine)
@elvee the circuit is below - all pretty standard. The things I haven't worked out too carefully yet are the coupling caps and the cathode bypass caps - still need to justify the values I have in there. Also the nfb RC is just common values - I have to give that more thought too.
But anyway, just because this model doesn't have to be massively correct I'm setting k = 1 just so I can finish the design and build the amp. If any of these issues occur in the real world, then I'll come back to LTSpice and correct the design. It's just components and soldering after all.
PS: fiddling around, I noticed that the circuit is quite sensitive to the value of C in the NFB loop - the wrong value also causes oscillation, so a bit of work needed there anyway.
Last edited:
The transformer numbers are off.....they seem to be representing TURNS ratio which is not what they should be... They need to be an actual INDUCTANCE value in Henries...
These numbers are too small and introduce significant Phase Shift that may be causing the oscillations... For an 8K plate load, your best to start about 200H this will place your low freq -3dB POLE at 6.3Hz...this POLE is also where you have your 45 degree phase shift... This will make L5 and L6 to be 1/4 at 50H each....thus making L4 to be .411H
These numbers are too small and introduce significant Phase Shift that may be causing the oscillations... For an 8K plate load, your best to start about 200H this will place your low freq -3dB POLE at 6.3Hz...this POLE is also where you have your 45 degree phase shift... This will make L5 and L6 to be 1/4 at 50H each....thus making L4 to be .411H
For some reason the Transformer Models Rev4 spreadsheet that El Sid is using calculates primary inductance as Z/(4.pi.f) instead of Z/(2.pi.f).
Z/(2.pi.f) is correct when referring to the entire primary winding....
8000 / (6.28 * 6.3Hz) = 200H for the entire primary winding Inductance...
However, in the model the OT is Center-Tapped, thus each half of the winding is 1/2 the turns, therefore 1/4 the impedance... we now have 50H per side of Center-Tap...
In true Class A output stages...the Center-Tap essentially disappears and is of no use in AC signal analysis, since the AC currents are 180 out of phase, and therefore there is a net 0 current flowing through the Center-Tap...
Yes, it works out the total primary inductance then divides by four for each half. It's the formula for total inductance (Lp in the spreadsheet) that seems wrong, see attached. Otherwise I find it very useful.
I found a different earlier version of the transformer model spreadsheet - it seems that the 4.pi.f/2.pi.f is accounted for in the calcs because they both ultimately produce very similar values for Lp and Ls. These values are also similar to the generic 8K models.
Part of the problem I have is that there is no data whatsoever available on these OPTs - all I have is the resistance measurements. The low and high -3dB points are just what is used in that spreadsheet for the generic transformers. Could I measure those with a signal generator?
Alternatively i will just build the bloody thing and see where I end up....
Part of the problem I have is that there is no data whatsoever available on these OPTs - all I have is the resistance measurements. The low and high -3dB points are just what is used in that spreadsheet for the generic transformers. Could I measure those with a signal generator?
Alternatively i will just build the bloody thing and see where I end up....
In real life there was no sustained oscillation without zobels but definitely some strong ringing on square waves. I ended up with zobels on the primary (6k+470p for each half from memory) + one in parallel with the first stage anode load (values would be circuit dependent). The transformer upper f3 was about 60 kHz (measured), and I entered 5Hz into the spreadsheet as the lower f3. Inductance of each primary half was then 33H (but probably should have been 66H)
.SUBCKT Partridge8k P1 B P2 O16 O8 O4 Com
* Push Pull transformer
* 8000 to 15 ohms, 3db 5 to 60000 hz
* speaker taps at 3.75 and 8 ohms
* Partridge8k
*
* model generated by TransformerModels.xls 2019 Jun 26
*
LP1 1 B 33.3322249136949
LP2 2 P2 33.3322249136949
LA1 5 O8 0.0182899874315516
LA2 6 O4 0.0133357246271817
LA3 7 Com 0.0628609636710886
KALL LP1 LP2 LA1 LA2 LA3 0.999619370999275
RP1 P1 1 189
RP2 B 2 189
RS1 O16 5 0.2
RS2 O8 6 0.2
RS3 O4 7 0.4
.ENDS Partridge8k
- Status
- Not open for further replies.