Transformer Measurements 1

[CJ]

A nickel core should follow the same slope. If you want to see a flatter line try a gapped core design.

 

[Wavelength]

A nickel core should follow the same slope. If you want to see a flatter line try a gapped core design.

Yes and No.... While Nickel is great it can be problematic with DC current. Grades of IRON down to M3 is a better option or permadore/cobalt. I have outputs with Nickel, Cobalt, M3 and higher in stock. The nice thing about nickel/coblat for some reason is it shields most noise. Last buy on cobalt really hurt though something like $32K MOQ. In Push Pull outputs we usually slip in like 6 pieces of nickel in the core center so that low level information is not lost in the wash.

 

[The Electrician]

Here are some measurements made on 3 different 600:600 ohm transformers, an Edcor, a Jensen, and a Cinemag. The Jensen and Cinemag have nickel containing laminated cores, and the Edcor has silicon steel laminations, probably M6.

I've set the Hioki to sweep from 4 Hz ( the lowest frequency it has) to 20 kHz showing curves of the self inductance of 3 transformers superimposed. There are 3 images each with a different stimulus voltage per image, but the same stimulus for each of the 3 superimposed curves. At the B marker (1 kHz), the Cinemag curve is the top curve, the Jensen in the middle and the Edcor is the bottom curve. The three curves are in that order in all the images I'll show. The first image shows the measurements with a stimulus of .1 volt:


Here's the .5 volt stimulus result. Notice that the Cinemag curve has an upward slope for a short distance at the far left (4 Hz). At the very low frequency of 4 Hz, .5 volts applied to the winding is enough to push the core slightly into saturation:


Here is the .01 volt stimulus result. These 3 curves are very interesting. Notice that the Edcor curve becomes nearly constant at about 2 henries starting at about 300 Hz. Does anyone know why this might be? That flat spot makes me think of a post where CJ speculated that maybe the magnetic domains may need to be shaken awake. Why do the nickel cored transformers show a significant change in slope?

 

[The Electrician]

Here are two more images showing the effects of saturation. This image shows the result of applying a 1 volt stimulus. The Cinemag is well into saturation at 4 Hz, and the Jensen shows just a little saturation:


Finally, with a stimulus of 3 vots, Even the larger Edcor is beginning to show some saturation:

 

[ruffrecords]

Interesting. In both cases the ratio between 20Hz and 1KHz inductance is close to 7:1 and ratio of the frequencies is 1:50. Is this an inverse square law?

Cheers

Ian

 

[Wavelength]

I would bet the CineMag is using High Nickel which is usually like ~80% and Jensen is using standard Nickel which is ~49%.
Nickel always seems to saturate early depending on the quality, were as iron does not. I will try and grab some sims that Dave Slagle did for me when we designed a few loading chokes and grid chokes for some high end audio stuff I build.
Thanks for this, I might get Dave to chime in.
Gordon

 

[The Electrician]

Interesting. In both cases the ration between 20Hz and 1KHz inductance is close to 7:1 and ratio of the frequencies is 1:50. Is this an inverse square law?

Cheers

Ian

I suspect it has to do with the mathematics of magnetic skin effect. The core is laminated, but even though the laminations are relatively thin, at high enough frequencies the flux can't penetrate all the way into the lamination. This means that all the core material isn't used, and the effect is as if the core is smaller. PRR mentioned this: https://groupdiy.com/threads/audio-transformer-inductance.47515/post-597897

All this stuff I'm showing with the fancy LCR meters is well known. I'm just showing actual measurements of known effects because it can aid understanding--it makes a believer out of a person.

These curves really show why mic transformers use nickel cores. With nickel, the permeability stays high even at very small excitation compared to M6.

 

[CJ]

Try sweeping a coil without the core,

 

[The Electrician]

I'm unwilling to destroy any of the audio transformers I have, but as it happens:
Years ago I bought a Radio Shack filament transformer, 120VAC in and 12.6 volts @1.2 amps out. For some reason this transformer didn't have the lams varnished, and I have used it on various occasions to make measurements with the core in and the core out.

Here the sweeps of the primary and secondary self inductances with the core in. I'm not sure if the core is even as good as M6.


Here are the sweeps with the core out:


Here are sweeps with core in and core out, superimposed:


Here are sweeps with core out all the way to 1 MHz:


And just for grins, here is a sweep of the primary with the core out showing the impedance magnitude and Rs (resistive part of the complex impedance) out to 5 MHz. The value of Rs shown in the upper right is 45.57 ohms. That's the value at the end of the sweep. At the low frequency end the value of Rs is 33 ohms, which is the DC resistance of that winding.

 

[CJ]

Interesting to see a sweep of cores laced 1x, 3x. 10x with the same coil and some DC offset,

 

[The Electrician]

Interesting to see a sweep of cores laced 1x, 3x. 10x with the same coil and some DC offset,

I don't know what the red part means.

Edit: Does this mean where the primary and secondary are segmented and then interleaved?

Interleaving a secondary which is opencircuited while the primary is swept should have no effect at low frequencies, but at high frequencies the differences in parasitic capacitances would probably lead to differences among the different amounts of interleaving. Is this what you are looking for?

 

[CJ]

Lamination stacking, 1 lam this way, 1 lam that way, 10 lams this way, etc.
Sorry for the confusion.

It used to be listed as Lap 1 or Lap 3 on our transformer blueprints so that is what I am use to but apparently we had the only engineer who used that terminology.

 

[The Electrician]

OK, here are the sweeps with differing stacking of the lams. I'm also showing Rs as measure of core loss. I did these sweeps within 15 minutes of each other so there shouldn't be much difference due to environmental factors such as temperature. I made sure the butt joints in each case were tight.

The inductance and the losses are slightly higher with 1X than with 4X. I wasn't sure what to expect.

The first image is with the lams stacked 1X:


This sweep is with the lams stacked 4X:


Here are the two previous sweeps superimposed.

 

[Wavelength]

Lamination stacking, 1 lam this way, 1 lam that way, 10 lams this way, etc.
Sorry for the confusion.

It used to be listed as Lap 1 or Lap 3 on our transformer blueprints so that is what I am use to but apparently we had the only engineer who used that terminology.

That is called BUTT stacking it's a way of introducing a small air gap, typically used for low level DC current usage.
For instance we make a 4000H grid choke that is 49% nickel and this helps on output tubes as the DCR is like 2K, the ACZ is super high so the amp delivers more output because when the grid starts drawing current the offset in bias is not effected. Ok off subject...
We also BUTT stack this with M4 or M6 and use it with a 6GM8/ECC86 in a parallel feed circuit at 1ma into a 15K:600 output transformer in a USB DAC I make called the Cosecant. Parallel feed is when you put the choke/inductor to B+ and cap couple the output transformer. Since both the choke and the output transformer have very high impedance the design of the primary in the output does not have to be so difficult to design. The higher the primary impedance and the DC current and required inductance can be very hard to do. Remember it's all diminishing returns on designs like this.