Your turns ratio is wrong because you, in addition to use wrong "real world math", you use wrong "real world physics".
I bet that you haven't done any transformer in the real world, if so, you used the wrong way, because real transformers do have losses.
Here I go again, and I'm pretty fed up...
The OP claims
Zp=3500 Ω Zs=8 Ω ip=120 mA fo=18 Hz W=30 W
From attachment of post#101
N = (Uac x 10^8) / (√2 π fo S Bac) (1)
Bdc = (L x 10^8 idc) / (S N) (2)
Where
L = Zp / (2 π fo) ≈ 31 H (3)
My choice was
Bmax = Bdc + Bac = 1.3 T (4)
(1), (2) and (4) determine a linear system of equations, solving it
Bdc = 0.6223 T (*)
Bac = 0.6777 T (**)
From generalized Ohm's law
U = i Z (5)
Can write
W = U^2/Z (6)
For a real world transformer, as a consequence of energy conservation principle
Wp = Ws + Wloss (7)
Then
Ws = η Wp, 0 < η < 1 (8)
Combining (1), (6) and (8)
(Np/Ns)^2 = (Zp/ηZs) (9)
That I wrote
My last OPT, made with the same EI 155 (EI 114 for eu), with a core area of 38x38, double insulation 0.25 mm primary wire (about 0.3 mm) has more than 3800 turns, bobin-epoxy-wire-epoxy-polyester-epoxy-wire...wire-epoxy-NOMEX winding scheme for each winding, 10 primary windings and 8 secondary windings oversized with 0.65 mm wire all paralleled and even leftover me window enough to many more primary layers, so no much problem for about 4300 turns.
Even I could make it without changing secondary wire, nor sixteen 0.4 mm NOMEX insulation.
Sorry, it's 0.45 mm, my bad, I'm getting old, that was for another OPT.
For imaginary transformer of post#101, with a core effective area of 38x38 (EI 114 38x43 with 0.9 stack factor)
Np ≈ 4316
For reasonable η=0.9, using (9)
Ns ≈ 196
Three of possible winding schemes for each half bobbin
3P-2S-6P-2S-8P-2S-6P-2S-3P
3P-2S-7P-2S-6P-2S-7P-2S-3P
4P-2S-6P-2S-6P-2S-6P-2S-4P
P=83 turns S=49 turns Фp=0.25 mm Фs=0.45 mm
For both three schemes
26 x 0.3 mm + 8 x 0.5 mm + 9 x 0.4 mm = 15.4 mm
Sorry, was needed another insulator to finishing the last winding, so 18 in total, instead of 16 as I was wrote, however that last insulator doesn't need to be NOMEX because of the distance to the core, capacitance won't be a problem and would be used 0.2 mm Mylar, saving a little bit more space.
Just in case you want use your crystal ball again, mind you that mine double insulation polyesterimide resin coated copper wire had been measured carefully with a micrometer, a long time ago, and is consistent with manufacturer data.
Current densities become
Jp≈2.45 A/mm^2 Js≈2.89 A/mm^2
For an optimized imaginary transformer, using (*) and (**)
Np ≈ 4140
For reasonable η=0.9, using (9)
Ns ≈ 188
One of possible winding scheme for each half bobbin
3P-2S-6P-2S-7P-2S-6P-2S-3P
P=83/82 alternated turns S=47 turns Фp=0.25 mm Фs=0.45 mm
For this scheme
25 x 0.3 mm + 8 x 0.5 mm + 9 x 0.4 mm = 15.1 mm
For an optimized imaginary transformer, using (*) and (**) and non split bobbin
Np ≈ 4140
For reasonable η=0.9, using (9)
Ns ≈ 188
One of possible winding scheme
3P-1S-6P-1S-6P-1S-6P-1S-3P
P=173/172 alternated turns S=47 turns Фp=0.25 mm Фs=1.00 mm
For this scheme
24 x 0.3 mm + 4 x 1.05 mm + 9 x 0.4 mm = 15 mm
Current densities become
Jp≈2.45 A/mm^2 Js≈2.34 A/mm^2
I bet that you will criticise fiercely those imaginary transformers, just in case, mind you that
I consider my work done, even with very small cores and wires for the job.
QED.
Popillin your "alternative" combinations are really bad, especially regarding power loss and/or frequency response. Forget the formulas and make one and you will know why! Or learn to calculate the leakage inductance and power loss.....
(3840:188)^2 into 8R is 3.34K. Wrong??? You could adjust it a bit of course but not changing the geometry. Have you considered the DC resistance? Do you think will make a difference in the real world???? Make some amps too. All the rest is just hogwash...
Use craft (or latheroid) paper or nomex for insulation.
(3840:188)^2 into 8R is 3.34K. Wrong??? You could adjust it a bit of course but not changing the geometry. Have you considered the DC resistance? Do you think will make a difference in the real world???? Make some amps too. All the rest is just hogwash...
Use craft (or latheroid) paper or nomex for insulation.
Last edited:
As I told I need to go back to Italy. I don't have a lab here in UK and don't think will have one as I don't need it. There are more important things for me to experiment with my system other than valve amps which I have already optimized sometime ago.
The amp is the easy part of the stereo system.....
No measurement can show the final results otherwise all the best power amps would be solid state only....
Maybe someone else is interested in some calculations to make his own transformers. In the meantime your thread is kept alive waiting for someone that has your own hobby as I don't think so many people spend time in measuring transformers but are rather interested in the amplifier. A transformer alone is useless....
Curiously you don't mention interwinding capacitance, the biggest challenge, and the Achilles heel of proposed designs, especially regarding frequency response.
That show your poor understanding about transformers, if you are a physicist, you hide it very well.
Just in case your crystal ball don't work properly
Neglecting edge effects, we can obtain a reasonable approximation to the value of capacitance C.
For two parallel plane plates of area A separated by a distance d, filled with a dielectric medium of dielectric constant ε, from Maxwell's equations
∇.D = 4πρ
D = ε E
Integrating in a volume V
∫∇.D dV = ∮D.ds = D A = ε E A = ∫4πρ dV = 4πQ
Then
E = (4πQ) / (ε A)
Also
E = - ∇φ = dφ/dx
Integrating
Δφ = ∫dφ = E ∫dx = E d
By definition
C = Q/Δφ
Then
C = (ε A) / (4πd)
Where A is the average winding area, and d is the insulation thickness.
About power losses, you are right, proposed designs have high power losses, but that was not the point, did you really read/understand my posts?
My Tarzan-English don't help much, but you don't need a cryptologist.
And as I said before, I prefer copper losses against core losses, mind you that hysteresis loop is a curve, why do you think that core losses are expressed in WKg?
Maybe you realized that your enormous cores aren't always the best choice.
BTW, if you have equations to calculate core losses, please post them.
I can't forget the formulas, that is my way of design, how do you do your designs?
Sorry, I forgot your crystal ball...
Power losses are easy to calculate, especially copper losses, but leakage inductance not so much
Ls = (0.417 Np^2 Tl [2 n c + a)] / (10^9 n^2 b)
Where
Np = number of primary turns
Tl = average turn length around bobbin in mm
n = number of dielectrics between windings
c = thickness of dielectric between windings in mm
a = winding height in mm
b = winding traverse in mm
Put numbers here and you will see that proposed designs have very low leakage inductance.
Yes, this is wrong by far, seems that you used Eq.(9) of post#143 with efficiency factor η on the numerator.
Looks like you have don't noticed that, when increase power, primary impedance decreases due to power losses, thereby increasing distortion as more loads the output valve, maybe you understand this with Eq.(9) written as
Zp = η Zs (Np/Ns)^2
Curiously you still insist on power losses when you really ignore them.
Yes, of course I do, beginning with resistivity
ρ = R (a/l)
But this is also trivial, and please stop talking about real world, you don't understand it.
I make amps for happy customers, transformers are for that amps.
Paper is a poor insulator, besides its low dielectric constant is only when dry.
Curiously you mention NOMEX right now after thirty posts when I did.
Conclusion: along the whole thread you only was capable to criticize the work or ideas from others.
No contribution, but a few links.
I answered all your questions/critics, one by one, with solid fundamentals and you are still criticising, but without any fundament.
Finally I don't want to keep boring people with equations, so I quit from this thread, that's time you criticize to another, don't count on me anymore.
Addio
Last edited:
With all this animosity one ventures out rather carefully ......
but I also missed that. Somewhere I read about 8 secondaries or so. Leakage reactance is dependent on several factors so perhaps a sweeping statement, but as far as secondaries are concerned, I have found over different transformers that going beyond four secondaries invariably makes my h.f. response more dependent on interwinding C than on Ls. In one particular transformer Ls gave a -3dB point at about 160kHz, whereas C started playing a role round about 80 kHz. This naturally depends on many other factors as well, but just to illustrate.
The interwinding capacitance increases proportionally to the number of primary-to-secondary interfaces whereas the leakage inductance is inversely proportional to the square of interwinding interfaces. From this one might be tempted to make as many windings as possible but in reality it is not a good idea firstly because that means more and more space is wasted for insulators. Then the Q of the resonance also is important and so the ratio of leakage L and stray C.
The capacitance can be redistributed. One case where one does it "automatically" is the PP OT with intermediate flange.
One way to verify this is measuring the resonant frequency with the secondaries shorted in two cases for a single cave SE OT: grounded and floating (with the core and the primary always having reference to ground). You only need a generator in series with the primary and close the loop with a small 10R precision resistor. In parallel to that resistor you put a precision milli-voltmeter. You will get a minimum current at the (first) resonance (parallel type). When you switch from shorted grounded secondaries to shorted floating secondaries the resonant frequency will increase because you are redistributing the voltages and in the latter case it is as if the capacitances were all in series.
If then you do the same with PP OT and balanced generator you won't get that increase! Because of the geometry the distribution is already as if they were all in series.
Last edited:
Measurements can show faulds in the design, not many people understand this.
Personal preferance can make a differance, 10Hz or 30Hz -3dB can make a differance for the low frequencies, the same for high frequencies.
Squerewave for resonance. Best is have both good but even then personal preferance..... Etc etc
Personal preferance can make a differance, 10Hz or 30Hz -3dB can make a differance for the low frequencies, the same for high frequencies.
Squerewave for resonance. Best is have both good but even then personal preferance..... Etc etc
Yes I know that some basic measurement can show flaws but I am not the other people....
45: "The interwinding capacitance increases proportionally to the number of primary-to-secondary interfaces whereas the leakage inductance is inversely proportional to the square of interwinding interfaces"
True. popilin, your calculations are valid for ONE primary to secondary interface. The total
capacitance depends on the the interlaving pattern, and the voltage differencies involved between layers.
esltransformer:
"In real live the best transformer companies use c-core for there "high end" series."
Yes they do. Main reason, the usual marketing b*******.
45 has explained it thoroughly: We don't care so much about core losses in audio, because we DO NOT operate the iron near saturation. It's NOT a power transformer.
What remains as an advandage of C core, is the higher permeability, because in all parts of the transorformer core the grain orientation is in the same direction with the magnetic filed lines.
No big deal either, since we have an air gap. Enen in PP output, we do need an airgap: Not only to accomodate for DC imbalance, but also to linearise the hysterysis loop of the iron.
popilin: "has more than 3800 turns...0,25 mm"
Wrong design. Very high cooper losses, even for a 10 Kohm / 8 ohm trans. We DO care about cooper loss, since it is present at every level of excitation. Use a -much- larger core.
True. popilin, your calculations are valid for ONE primary to secondary interface. The total
capacitance depends on the the interlaving pattern, and the voltage differencies involved between layers.
esltransformer:
"In real live the best transformer companies use c-core for there "high end" series."
Yes they do. Main reason, the usual marketing b*******.
45 has explained it thoroughly: We don't care so much about core losses in audio, because we DO NOT operate the iron near saturation. It's NOT a power transformer.
What remains as an advandage of C core, is the higher permeability, because in all parts of the transorformer core the grain orientation is in the same direction with the magnetic filed lines.
No big deal either, since we have an air gap. Enen in PP output, we do need an airgap: Not only to accomodate for DC imbalance, but also to linearise the hysterysis loop of the iron.
popilin: "has more than 3800 turns...0,25 mm"
Wrong design. Very high cooper losses, even for a 10 Kohm / 8 ohm trans. We DO care about cooper loss, since it is present at every level of excitation. Use a -much- larger core.
True, I didn't say the otherwise.
Can be proven that if a capacitance C with a voltage U1 across it is to be referred to some other voltage U2, the effective value at reference voltage U2 is
Ce = (U1/U2)^2 C
For a transformer with grounded short circuited secondaries, (secondaries are almost 0V equipotential) referred to +B can be written as
Ce = (n P / Np)^2 Ci = (n P / N P)^2 Ci = (n/N)^2 Ci
Where
Ci = capacitance between two adjacent P and S windings
P = number of turns in the layer
n = number of layers in the winding
N = total number of layers
For a SE OPT, with this proposed winding scheme
3P-1S-6P-1S-6P-1S-6P-1S-3P
Assuming that referenced to layer midpoint, we have 8 interphases, and capacitance "seen" by anode of the valve is distributed as
Ce(total) = (2.5/24)^2C1+(3.5/24)^2C2+(8.5/24)^2C3+(9.5/24)^2C4+ +(14.5/24)^2C5+(15.5/24)^2C6+(20.5/24)^2C7+(21.5/24)^2C8
Capacitance is increased as we go to 0VAC (+B) to U VAC at anode of the valve.
In a rough aproximation we can take
C1 = C2 ... = C8 = C
Where C is the average capacitance between two adjacent P and S windings, then
Ce(total) ≈ 2.63 C
Last edited:
PP transformer designed properly with GOSS don't need any air gap, even toroid can handle up to 7mA imbalance (tested and verified in my own setup). And nothing needs to be linearized, voltage distortions caused by non-linearity of mu (below saturation) are very small as long as L << Req / (2 * pi * F).
In details this is described in Menno Van der Veen book - modern high-end valve amplifiers based on toroidal output transformers.
The point was another, and I must repeat it again and again...
OK, prove it, with equations please, we are talking about physics, not philosophy.
The purpose of my few scribbles is to illustrate the point that the OP numbers are totally possible, I can't to design a Hi-Fi OPT in just ten minutes, and if I could it, it would be a commercial secret.
OK, prove it, with equations please, we are talking about physics, not philosophy.
Nonsense! Depends....
Explain why? I know why but there is no free lunch, I am afraid.
Transformer distortions are low as long as one stays away from saturation but L will still depend on the signal anyway.
Wrong!!!!!!!!!!!!!!!!!!
Then look a the specs of his transformers and tell me why:
1) his SE OT for 300B isn't any better than mine with EI's...
2) his PP transformers don't have a specified allowed unbalance and he has created a servo bias circuit for the output stage. Those specs are true only for 0 mA DC unbalance at low frequency.
Last edited:
In post 31 you write that your transformer has 0,3dB loss and frequency response is 3Hz- 67kHz.
Comparing that with the Menno van der Veen VDV 3025 it is clear to me that your transformer is not as good as de the vdv3025. (Yours has 75% more copper loss, the 3025 has 35% more high frequency response but your transformer has better low frequency response, 17%)
And yours have still EI, more Barkhausen noise.
http://www.amplimo.nl/images/downloads/ds vdv/vdv3035.pdf
Yes it has 0.3 dB loss and so?
If I measure inductance on my transformer as he does (i.e. 240V at 50Hz) I get 44H! Now you could understand why at max specified RMS power I only stop at 0.75T @ 30Hz.....no significant barkhausen noise I am afraid! That transformer is NOWHERE near this value. At low frequency is worse, I am afraid. I could get the same inductance and get that power loss and even better HF response if I wanted. But then one has to see in practice which one is better.....make some amps!
84KHz in comparison to 67 Khz...maybe a bat can hear the difference. Do you want to run a blind listening test and see if you can spot which one is playing? Then we see if you can spot the difference in HF roll-off and power loss.....
And finally my transformer weights a bit less than 5.4Kg and has even a slighlty small footprint! It is only about 3 cm taller. Is is this also a problem? I don't thinks so. A 300B is taller even exclunding its pins.
esltransformer C-cores are a bit better in this case (and in other cases quite a bit better) certainly not toroids.....
Last edited:
All what you write is not clear, to much smoke, no graphs 240V 50hz 3500 Ohm, 16,5W? Is it 26H or 33H or 44H?
All i see is that the VDV 3025 has better performance so far.
You look like a salesman who try to sell a BMW to someone who ordered a Ferrari and say: go for a ride and you will not notice any differance.......
All i see is that the VDV 3025 has better performance so far.
You look like a salesman who try to sell a BMW to someone who ordered a Ferrari and say: go for a ride and you will not notice any differance.......
- Status
- Not open for further replies.