Eddy currents are proportional to the rate of change of flux. So higher frequency will produce higher loop voltages, hence higher loop currents, and resultant higher bucking flux.
Domain switching may be fine for magnetic material, but not copper, aluminum, brass.. The proximity of conductive materials to the time varying field of an inductive storage device tends to exclude the magnetic energy. The result is a lowering of the system inductance. Inductance is defined via an equation which relates the total amount of energy stored as a result of a current. Lenz effects reduce the total energy stored, therefore lowers inductance.. That will appear to be a change in the permeability of the flux path.
Cheers, John
Unfortunately, transformers are known to have problems similar to tape heads, and this is WHY they are LAMINATED in the first place, and Ferrite cores are used at virtually everything above the audio bandwidth. Too much eddy current loss with mumetal or steel.
The method of how laminations are produced is interesting, but not necessary to know implicitly, in order to select a transformer or a tape head, but the difficulty of making THIN laminations implies WHY some transformer manufacturers would go for thicker laminations, just to save time and trouble. Of course, at line level, the difference between thicker and thinner lams may be almost irrelevant, but at MC levels, I consider it important.
Now I, at this time, have elected to forgo using a transformer in my all-out design. The transformers that I have been exposed to, though recommendation, catalog, or direct measurement, all concern me with various problems. However, this does not say that a transformer is not useful for another design, especially one that NEEDS it.
You see, without a transformer, I have a BALANCED INPUT and a self noise of 10-20 ohms. Most here would start with an UNBALANCED INPUT and a self noise of 60 ohms or more, usually much more. Here, an input transformer can give you a balanced input, and lower the equivalent noise to 5 ohms or so. Given eddy current losses, DC winding losses, etc, I would not expect much better than this, EXCEPT for special transformers made for ribbon microphones, for example, that have their own 'tradeoffs'.
Some of the other features that transformers offer such a bandwidth limiting and galvanic isolation is rather pointless in my design, but it could be helpful in others.
The original reason for me to concern myself with low noise design, starting 43 years ago, when Ortofon showed me the door, after I mentioned that I wanted to build a solid state replacement for the transformer (they thought I was some sort of a nut), and how Ortofon adapted my patented design (without paying me of course) 10 years later, to REPLACE their input transformer, shows that transformers are difficult and expensive to do right, and should be avoided if possible, with low noise solid state electronics.
The method of how laminations are produced is interesting, but not necessary to know implicitly, in order to select a transformer or a tape head, but the difficulty of making THIN laminations implies WHY some transformer manufacturers would go for thicker laminations, just to save time and trouble. Of course, at line level, the difference between thicker and thinner lams may be almost irrelevant, but at MC levels, I consider it important.
Now I, at this time, have elected to forgo using a transformer in my all-out design. The transformers that I have been exposed to, though recommendation, catalog, or direct measurement, all concern me with various problems. However, this does not say that a transformer is not useful for another design, especially one that NEEDS it.
You see, without a transformer, I have a BALANCED INPUT and a self noise of 10-20 ohms. Most here would start with an UNBALANCED INPUT and a self noise of 60 ohms or more, usually much more. Here, an input transformer can give you a balanced input, and lower the equivalent noise to 5 ohms or so. Given eddy current losses, DC winding losses, etc, I would not expect much better than this, EXCEPT for special transformers made for ribbon microphones, for example, that have their own 'tradeoffs'.
Some of the other features that transformers offer such a bandwidth limiting and galvanic isolation is rather pointless in my design, but it could be helpful in others.
The original reason for me to concern myself with low noise design, starting 43 years ago, when Ortofon showed me the door, after I mentioned that I wanted to build a solid state replacement for the transformer (they thought I was some sort of a nut), and how Ortofon adapted my patented design (without paying me of course) 10 years later, to REPLACE their input transformer, shows that transformers are difficult and expensive to do right, and should be avoided if possible, with low noise solid state electronics.
My goodness..what a production..
1. Buy the material in sheets a foot square.
2. Apply the bonding film to both sides of the sheets (or buy the sheets coated).
3. Stack the sheets to the desired thickness using a fixture which applies a force normal to the sheets. Either use a weight and gravity, or a flat plate with lots of bolts and bellvilles. Preload to the proper force.
4. Cure the stack in a furnace with the proper time/temp profile.
5. Send the finished stacks to a wire EDM facility with the core profile.
6. Debur the edges where desired.
Wire EDM is easily capable of 50 micron accuracy. 25 is doable, but it may cost a bit more.
With half mm lams, this will get you 98-99% packing factor. 1 mil lams, maybe 60 to 80% depending on film thickness.
See...problem solved...simple..
(I should be on the "Daily Show"..)Cheers, John
Core noise temperature is almost certainly the same as the core physical temperature.
Calculating noise for a particular frequency is no problem, provided you have resistance or loss measurements for that frequency. A frequency dependent resistance is still a resistance, so at a particular frequency it produces thermal noise which can be calculated from its resistance and temperature.
JC is not making things up. The paper he referred to shows this. It is unfortunate that poor nomenclature and, at times, weak grasp of physics has made this debate much longer and harder than it need have been. Frayed tempers and SHOUTING have not helped either.
Calculating noise for a particular frequency is no problem, provided you have resistance or loss measurements for that frequency. A frequency dependent resistance is still a resistance, so at a particular frequency it produces thermal noise which can be calculated from its resistance and temperature.
JC is not making things up. The paper he referred to shows this. It is unfortunate that poor nomenclature and, at times, weak grasp of physics has made this debate much longer and harder than it need have been. Frayed tempers and SHOUTING have not helped either.
Agreed for the most part.
It's the noise proportional to frequency part that is the interesting part... One can use a temperature term which is proportional to frequency (which I don't like), or transfer the losses to the coil resistance and have a frequency dependent resistance term..which I like better..
I believe using a freq dependent resistance would be mathematically easier, and I believe the actual core physical temperature can be used to proportionally increase the resistive term.
However, as I said (in a roundabout way), I am far too challenged to work the math required for an analytical approach to calculating the actual noise of a part exhibiting frequency dependent eddy losses.
I've not said he is making things up. I've claimed that he made a statement/correlation between lam thickness and resultant noise with no physics justification....so was ignored for the most part. I understood the mechanism, so was NOT going to allow JC's assertion to die away. He was correct.
Frayed tempers and shouting?? You've been nothing but pleasant, I've not shouted nor been angry...did I miss something?
Cheers, John
This means that the metal strip has to be treated for electrical isolation application after stamping and abrasive bathing. I was under the impression that laminates were purchased with the insulation already applied.
Is this acid pickling the electrical isolation application I wrote above?
If yes, do you know what acid is used for silicon iron laminates?
Yes, I have looked with a X10 loop.
I thought so too.
When it comes to very small thicknesses, how easy it is to see these defects?
Do the small signal transformer manufacturers test for such electrical shorts between laminations? (Large transformers can be checked by monitoring the excitation current not exceeding a certain upper limit)
OK. But it comes handy as such (isolator) sometimes, No?
Regards
George
Makes sense to pickle before stamping, easier to handle. But even with my sheet metal toys I have never made a transformer.
Acid pickling is the insulation techniques as mention in my college days, never tried it myself or seen it done, so I have to say "I don't know." (These words may cause some on this thread to pass out!)
One varnishing technique I am familiar with involves putting a load of transformers in a vacuum chamber, removing most of the air and then letting in the varnish. That gets it into every nook and crany but does not guarantee it will hit every possible short. I do not know if anyone sprays the material before stacking.
Jacco,
Maximum punch life I am told is when the shear zone is 40-50% of the thickness,
John,
A turret press is used for quantities of 1000 or less. Wire EDM probably 50 or less. A special die in a commodity press is used for greater numbers. That is based on costs. Of course you have a bigger budget!
If you look at an E-I core you punch two "E"s and two "I"s at the same time and get almost no waste! That was worked out long before CNC (Computer Numerical Control) machines existed.
Last in general the normal clearance you require is that the die hole should be larger than the punch by 10-15% of the thickness of the material being punched. In a turret press the punch and die rotate in a giant massive C frame. The accuracy is better than .002 inches for alignment. (Bearings, machining tolerances even thermal expansion play a role in this.)So that is why .032" thick sheet metal is the commodity item .014" is doable. But any thinner just gets smeared! Of course if you wanted to punch thinner stuffer a smaller lighter and more accurate press could be made.
ES
Last edited:
Ya think????
A typical lamination is 1mm thick and a meter by a meter. Salient features have to be in the 25 micron accuracy range. Stacking a two meter long final "device" prior to cure and keeping a 15 micron lam to lam alignment is VERY difficult.
One way is to use a two step stamping, with the required features locked at pass 2. Another is to cure lams together after one stamp, then wire EDM the final features.
I specified wire EDM because it's easy enough to send out to a vendor, it relieves us the issue of actually trying to machine to those tolerances, and the machining cost is easily determined (by quote).
If I were in the business, I'd have in stock various overall thickness cured laminate sheets with half or 1 mil thick starting foil, and I'd send these cured sheets out based on custom work orders.
It'd be cost effective for small quantity/fast turnaround..
Cheers, John
Last edited:
In theory, the opposing current of the eddy currents mainly prevents the flux from penetrating the core material completely resulting in uneven flux distribution. This effect is almost completely ameliorated by using thin laminations in combination, to a lesser effect, with using high resistivity laminations. I'm sure you are aware of all of this, but my point is that in high quality audio cores hysteresis loss swamps eddy current losses and does more in contributing to a tenfold decrease in permeability when increasing frequency from 400Hz to 20kHz.
John
Yes. And I'm confident that everybody contributing to this discussion is also. The thread is just too long to review now...
I would expect hysteresis losses more as an amplitude based loss mechanism, not so much a frequency based one. For the application JC initially spoke, I'd still believe eddies will dominate. Selecting a thin lam certainly will reduce eddy losses, so I'd expect hysteresis to dominate at lf. I'd have to sweep-measure Ls-Rs to see just how big the eddy loss is.
Cheers, John
Thin Lams
I have never designed the actual laminations for heads nor transformers. However I have had heads made to my specs by a couple of companies, both for baseband audio recording and high-speed duplication. In each of these cases they used photo-resist masking and chemical etching. In the case of the permalloy baseband audio head, they may have only done this because it was a small custom run, and of course the 1.28KHz>1.28MHz head with 8MHz bias had ferrite pole pieces. I am sure there is more than one way to skin this cat.
(meow)
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill
www.wxyc.org
1st on the Internet
I have never designed the actual laminations for heads nor transformers. However I have had heads made to my specs by a couple of companies, both for baseband audio recording and high-speed duplication. In each of these cases they used photo-resist masking and chemical etching. In the case of the permalloy baseband audio head, they may have only done this because it was a small custom run, and of course the 1.28KHz>1.28MHz head with 8MHz bias had ferrite pole pieces. I am sure there is more than one way to skin this cat.
(meow)
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill
www.wxyc.org
1st on the Internet
Last edited:
From Soft Magnetic Materials for Audio Transformer: History, Production and Applications:
For the higher frequencies in the audio range there is the
"skin effect," that is, the flux tends to concentrate on the
outer surface of the laminations, which accounts for the
fall in effective permeability, as shown in Fig. 16.
se
Because I neither design, nor will use an input transformer in ANY of my designs at this time, I will leave transformers now. If a real transformer designer wants to continue to seriously contribute here, I would appreciate it. Input and line transformers can be useful, and knowing them better as to the tradeoffs in design would be valuable.
Etching is great for thin stuff, the down side is that you get some side etching under the resist, so if you lose some width during etching. The loss is not uniform due to things such as non uniform etchant flow in corners.
Freq dependent resistance is the correct way. This is not a fiddle - you can actually measure the resistance. If you put current through it you get I^2R losses. If you assume that the whole transformer is at one temperature you don't even need to worry about the physical origin of each element of measured resistance. Just measure it, and calculate thermal noise.
No, but others have implied that JC was imagining things. I thought so at first, especially when he seemed reluctant or unable to offer a reasonable explanation. At times he resorted to shouting, or sneering, which made me even more suspicious.
It was only when I began to filter out all the noise and confusion, and think about the actual physics, that I realised that he was probably right but he seemed unable or unwilling to offer a coherent explanation. I don't know him, so I can't judge why this is so. Unfortunately, this seemed to cause his critics to assume he must be wrong and demand an explanation to support what they saw as probably invalid assertions. They seemed to continue banging on his door for an explanation even after I had offered one instead.
Sorry I wish had more time yesterday that was a hurried post. I don't think we need the esoteric references. Let me take a step back because I think I have an experiment that anyone can do.
Yes, except for heating and other non-ideal effects
The self noise of the coil is a signal and has to experience the same magnetic mechanisms as any other signal.
OK maybe you will buy this. Take the two coils of wire and resonate them with a lossless (compared to the coil) capacitor. The noise spectrum if measured will reflect the Q of the resonant circuit and if transformed from an parallel to series circuit will reflect the same measurement of Ls/Rs as your meter. This week is bad but I have some crap cassette recorder heads that should suffice for the experiment. A low noise FET buffer should do for the measurement. The 1A @ 20kHz might heat the system and raise the noise slightly and there might be excess 1/f noise but that would depend on non-idealities.
I have published the noise spectra of MM coils here, which shows the high frequency noise based on coil Q. Take John's tape head of 10k Ohms at 10kHz, if resonated at 10kHz would show 10k Ohms of resistance in the noise spectra at 10kHz. You could dispence with the mystery of the "black box"
Yes, except for heating and other non-ideal effects
The self noise of the coil is a signal and has to experience the same magnetic mechanisms as any other signal.
OK maybe you will buy this. Take the two coils of wire and resonate them with a lossless (compared to the coil) capacitor. The noise spectrum if measured will reflect the Q of the resonant circuit and if transformed from an parallel to series circuit will reflect the same measurement of Ls/Rs as your meter. This week is bad but I have some crap cassette recorder heads that should suffice for the experiment. A low noise FET buffer should do for the measurement. The 1A @ 20kHz might heat the system and raise the noise slightly and there might be excess 1/f noise but that would depend on non-idealities.
I have published the noise spectra of MM coils here, which shows the high frequency noise based on coil Q. Take John's tape head of 10k Ohms at 10kHz, if resonated at 10kHz would show 10k Ohms of resistance in the noise spectra at 10kHz. You could dispence with the mystery of the "black box"
- Status
- Not open for further replies.