Merlinb,
I'm speaking about "Core loss".
Mayhaps "amount of iron" is a poor choice of words, however the bigger the transformer the larger the amount of metal in it.
Most all the transformers discussed so far appear to use EI loss-less laminations in their core.
The more power a transformer is designed to handle, the greater the cross section area of the core has to be.
This also dictates a larger winding window to fit the windings and insulation layers.
This also means a longer magnetic path.
All of this contributes to a larger core and greater mass of metal.
One concern I have with respect to core size is:
It takes a finite amount of energy to reverse the magnetic domains. Part of the energy is lost and can be considered part of core loss and does not contribute to the output driving a speaker.
The larger the core, the larger this loss is.
At what point does this loss make a noticeable loss of detail?
I'm speaking about "Core loss".
Mayhaps "amount of iron" is a poor choice of words, however the bigger the transformer the larger the amount of metal in it.
Most all the transformers discussed so far appear to use EI loss-less laminations in their core.
The more power a transformer is designed to handle, the greater the cross section area of the core has to be.
This also dictates a larger winding window to fit the windings and insulation layers.
This also means a longer magnetic path.
All of this contributes to a larger core and greater mass of metal.
One concern I have with respect to core size is:
It takes a finite amount of energy to reverse the magnetic domains. Part of the energy is lost and can be considered part of core loss and does not contribute to the output driving a speaker.
The larger the core, the larger this loss is.
At what point does this loss make a noticeable loss of detail?
Last edited:
Core losses vary roughly as the square of the flux density. So operating a large OT at a lower level can get you to lower losses than the equivalent power small OT (which is running near max flux density).
The high frequency end could be maintained, if the large OT windings were kept to a thinner profile (allowed by lower power operation, smaller wire), but that is not what the usual economics dictate.
The high frequency end could be maintained, if the large OT windings were kept to a thinner profile (allowed by lower power operation, smaller wire), but that is not what the usual economics dictate.
...I'd expect losses due to magnetic hysteresis to depend on the material and amount and express itself as high-frequency loss. But I'm not a qualified engineer.
So are there any calculations to find the sweatspot for transformer size for a given tube and topology?
It sounds like HF loss and detail loss is due to higher capacitance and core losses in larger output transformers. This capacitance is harder to drive for lower power tubes then higher power tubes. The tube power to transformer capacitance ratio is lower for an EL84 pushing a CXSE25-8-5K then a 300B pushing a CXSE25-8-5K. Am I understanding this correctly? Personally I would think that the capacitance would act like a high pass filter no matter what tube was pushing it.
How big are the capacitive differences between a CXSE25-8-6.5K and a GXSE10-4-5K, are we talking pF, uF?
Does this tradeoff change when we look at PP transformers?
It sounds like HF loss and detail loss is due to higher capacitance and core losses in larger output transformers. This capacitance is harder to drive for lower power tubes then higher power tubes. The tube power to transformer capacitance ratio is lower for an EL84 pushing a CXSE25-8-5K then a 300B pushing a CXSE25-8-5K. Am I understanding this correctly? Personally I would think that the capacitance would act like a high pass filter no matter what tube was pushing it.
How big are the capacitive differences between a CXSE25-8-6.5K and a GXSE10-4-5K, are we talking pF, uF?
Does this tradeoff change when we look at PP transformers?
Last edited:
Core losses increase with frequency, but for audio, the flux density decreases with frequency too. With a square factor for the flux density related loss, net result is insignificant core loss at high audio frequency.
Capacitance in a large OT can be less, since it takes fewer turns. But typically the wire will be bigger for more power, so comes out a wash. You should look at the frequency spec for the OT.
Capacitance in a large OT can be less, since it takes fewer turns. But typically the wire will be bigger for more power, so comes out a wash. You should look at the frequency spec for the OT.
Last edited:
I'd think of that magnetic hysteresis as very roughly how much the core material itself does or does not become magnetized one way and subsequently re-magnetized the other way, as no iron core material is perfect. So does it depend directly on the amount of iron or does the bigger iron have less flux density (more iron magnetized less)? Is this an initial loss associated with reversal, or is it linear loss with current/flux?
Core loss certainly does increase with volume (proportionally, all else being the same). But lowering the max flux will do wonders with a squared term in control.
There is a distribution curve of energy to flip magnetic domains, with the lowest energy ones also being the highest permeability ones. So lower flux density operates on the easy ones first (in the linear region). Higher flux operates successively more on harder ones. (into the non-linear region) Once the flux gets high enough though (like say 1/4 max flux) they start avalanching (non linear hysteresis region, effective permeability peaks here too due to this), causing distortion effects. A low plate impedance (plus the load) will mostly prevent this distortion problem however (by constraint, any flux change produces voltage output, loading constrains it).
There is a distribution curve of energy to flip magnetic domains, with the lowest energy ones also being the highest permeability ones. So lower flux density operates on the easy ones first (in the linear region). Higher flux operates successively more on harder ones. (into the non-linear region) Once the flux gets high enough though (like say 1/4 max flux) they start avalanching (non linear hysteresis region, effective permeability peaks here too due to this), causing distortion effects. A low plate impedance (plus the load) will mostly prevent this distortion problem however (by constraint, any flux change produces voltage output, loading constrains it).
Last edited:
I want to point out something I just noticed. Take a look at this link
http://www.edcorusa.com/c/60/singleended
The GXSE series is rated for 40hz-18khz while the larger CXSE is rated for 20hz-20Khz.
http://www.edcorusa.com/c/60/singleended
The GXSE series is rated for 40hz-18khz while the larger CXSE is rated for 20hz-20Khz.
Yes, you can !
Consider vertical sectionizing
Yves.
No, if you make more sections you have more inner surfaces (i.e. the total surface between primaries and secondaries increases). The stray capacitances between primaries and secondaries are the dominant ones that determine the first resonance.
If m is the number of surfaces between primaries and secondaries then leakage inductance is function of 1/(m^2) while stray capacitances are function of m. So the product will be lower if m gets bigger and the resonance frequency that 1/[2*pi*SQRT(LC)] will be extended but capacitances can only increase if you increase m.
However you might also want to use the bigger space for more efficiency and not to waste too much space for insulation between sections. The number of sections can be a bit higher than a smaller transformer but no so much.
The main capacitances for the first resonance (i.e. what matters for frequency response) are those at the interfaces of primaries and secondaries. These are only dependent on the size (effective area), insulation dielectric properties and number of sections. They are independent of the number of turns.
Of course but there are no absolute rules. The overall specs are always the result of a how the rules combine.
You need to decide a minimum size first, to handle the power at low frequency at a max working flux. Already this spec can be variable. Low frequency could be 20, 30 or 40 Hz and max flux could be 8000 to 11000 Gauss. For example, my best compromise for SE up to 40W is the size that will handle max power at 30 Hz with less than 9000 Gauss (AC+DC). However you will find that most commercial transformers will be smallish for such requirement unless you are dealing with low power. Consider that the typical transformer specified for 25W and nothing else will typically handle about 9W at 30Hz because those 25W are referred to 50Hz. So a 5K impedance for such transformer with gap for 80-90 mA DC anode current could be enough for a 300B but highly insufficient for a 845.
Once you know the size you start to decide how many turns and sections you want (need) to make, insulation, further sub-sectioning etc. This is to point out that there is a balance that needs to achieved for best overall performance and if you favor one characteristic over the others it might not work better.
Last edited:
I thought it was clear there is no maximum size in theory. It's all practical: the sound quality depends on some many variables, the whole amplifier namely, that a discussion in this sense is rather pointless IMHO. The sound quality can only be judged by listening. There is no other way.
I am looking for a maximum, not only in theory, but also in practice. You might as well be saying you have to buy them all and listen to them to know if they perform well in a circuit. That seems rather pointless as well. Measurements and circuit design may only tell part of the story but it is a large part of the story.
There may be no maximum size in theory, but only if the model is insufficient.
There has to be a maximum size.
If made large enough, a low level signal will lose enough energy in core losses as to distorte the output significantly. Coupling is never 100%.
This may be 100 times the minimum calculated core size for whatever -f3 is chosen, but it will happen.
So the question is "when and under what conditions will this happen"?
Arguably it may be with such a large transformer as to be irrelivent. If so it would be nice if someone would explain the particulars instead of simply stating that in theory it won't happen.
Otherwise one is left with "chose the transforme you can afford and go with it".
Regretably, comparing transformers in a given circuit is difficult at best.
There has to be a maximum size.
If made large enough, a low level signal will lose enough energy in core losses as to distorte the output significantly. Coupling is never 100%.
This may be 100 times the minimum calculated core size for whatever -f3 is chosen, but it will happen.
So the question is "when and under what conditions will this happen"?
Arguably it may be with such a large transformer as to be irrelivent. If so it would be nice if someone would explain the particulars instead of simply stating that in theory it won't happen.
Otherwise one is left with "chose the transforme you can afford and go with it".
Regretably, comparing transformers in a given circuit is difficult at best.
Last edited:
If I read you right and you're feeling a little frustrated I'd say I understand your feeling but I'd also say that to get an answer you can be happy with you'll have to modify your view. You may be looking for the definitive answer honestly but that doesn't mean the absolute is there.
If you're not satisfied with answers here and you're seriously interested in the transformer question I'd recommend you go over to the Bottlehead site and ask Paul Joppa. He's a good guy (not that anyone here isn't) and just because he deals with Parafeed (uh.ohhhhh . . . now we're into it ) doesn't mean he don't know nuthin else. He might give an answer you'd like.This is an old one and partial but maybe it's the sort of detail you're looking for.
That was referred to E+I, this way transformer distortion will be very low. For HI-B core you can go higher.
Depends on the actual hysteresis loop (with or without gap). Linearity dictates how far you can go. Depends on your expectations as well. I prefer to minimize THD as much as possible. On top of this I always prefer to stay far away from saturation down to 30 Hz for SE and 20 Hz for PP at full rated power.
Last edited:
Reasonable size, weight and cost set a maximum before anything else for small-medium power. If you want a maximum at any rate then consider the OT's made for 211 and 845 or even the big OT posted by MAGZ but I don't think this is useful because you have to buy it and try it! Measurements can tell as much as you want but will NEVER give you the final answer.
I think this is your best "rule of thumb" answer. Also remember, it takes energy (current) to magnetize the core, so if I have a transformer that is built for 100ma optimal to go through it's primary but am using a 45 triode at 30ma, it's probably not going to sound so good because the 45 isn't pulling enough current to put the transformer in it's sweet spot.
I like permalloy cores because they magnetize much easier than standard silicon steel cores. Spec wise, the permalloy's may not measure up to the silicon steel cores, but to me, they sound better. And that's what diyaudio is all about; building something that is pleasing to my ears.
I had a bunch of Peerless 16431 outputs that I sold off because, personally, I didn't think they sounded that great with the tubes I like to use in amps. Sure they were awesome, spec wise, but they just didn't do anything special for my ears.
Go figure....
Daniel
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.