The Mad Science Challange -DIY MIC IPT

[solkatten]

Hi this will be trial to hand wound a usable MIC IPT on toroid core. Without the aid of expensive tooling and in the true spirit of DIY
What I would like is a good performing 1:8 MIC IPT to my 4in1 mic preamp.



Nanocrystalline toroidal cores. Vitroperm 500F

https://www.mouser.se/datasheet/2/599/V102-238079.pdf
This is a pretty large core for a mic trafo, but it will keep the primary inductance high and THD low and the leakage inductance is low anyway on toroids...

The discussion started in this thread:


https://groupdiy.com/threads/valve-mic-preamp-design-incoherent-rambling.77747/page-10
Pros: Toroidal core has natural ability to reject noise and has low leakage inductance. Cons: The challenge is to minimice the distributed capacitance in the secondarys that tend to degrade the performance...

The primary winding consist of a singel layer 0,25mm. 340T=10H. Which give good primary inductance and perfect symmetri for good cmrr.

The secondarys is to be layed in a progressive pattern. Im planning on using 0,15mm wire. I not sure if I manage to lay down perfect progressive pattern if that not practical I will resort to small and evenly spread section, where you basically lay the turns on top of each other....

Pictures from other trafo projects...

 

[solkatten]

The plan is lay down the sec. winding this way. With .15 mm wire it will be a ca. 1:8 trafo with .2 mm it will be ca. 1:5

 

[shabtek]

I recall (MIA CJ) describing 'pi-wound' to reduce capacitance (iirc) --something like winding 1 quadrant, then the quadrant across, then back 1 and then the last--like 1324 is that similar tho what you mean be by progressive pattern?

nice project!

 

[solkatten]

I recall (MIA CJ) describing 'pi-wound' to reduce capacitance (iirc) --something like winding 1 quadrant, then the quadrant across, then back 1 and then the last--like 1324 is that similar tho what you mean be by progressive pattern?

nice project!

Yep I think its simular. Lay the associated turns on top of each other to reduce capacitance... The most common would be 5 clock wise, 5 counter clock wise, 10 clock wise... until you circle the toroid...

The plan here is to use teflon tape between pri and sec, it makes it easier to lay down the wire if the surface is some what slippery so you can push the turn together by hand...

 

[shabtek]

random edited CJ+Abbey quotations:

Holy Cow how did you get that many turns?
You must have spooled up some wire on something you could pass thru the hole.
I bet you have a splice or two.
You might be able to pull the pri into the sec turns, use heavy wire, you want low dcr anyway, plus, you do not need many turns.
Imbed the pri into the sec. Good coupling.
You do not need interleaving with a mic xfmr, becuase a little roll off can be a good thing.

I had bad luck with capacitance stealing all my high end on a torroid I wound.
So I did some research and found that if you wind in "pies", you can reduce capacitance.

The theory behind this is that you are putting the turns with the most voltage between them on opposite sides of the core, thus reducing stray C.

The prefered method is North West East South.
You want the pies opposite each other as you wind. This is, of course, impossible to do for all windings, but this is better than nothing.

Here is an exagerated sample of what I mean.
I kind of just squeezed the turns into 4 quadrants.
Your actual coil should not have those big divisions between the sections, you should gradually expand the pie as you add layers to creep out to the boundry of the adjoining pie.

...Capacitance is the main obstacle here. Especially with the big core>longer wire length.
So we will use every trick in the book in order to deal with it.

One trick is to wind two pri's, side by side, ala UTC A-10 (sec) .
What this does is lower the "plate" area of the winding by two.
If you connect the two "pies" in series, you lower the C by 1/4 th, which is why UTC and others used this on ratios in the uper integers.

We will wind a bi fi primary to provide balanced DCR and capacitance.
Since this bi-filar winding is going to be used as a continous primary, and not a pri-sec bi fi, like the API, we do not need to worry about pri-sec capacitance.

Actually, the API uses this capacitance to it's bandwidthal favor.
A xfmr can be viewed as a transmission line. Calculate the L's and C's in this line and you can get an idea of what the performance will be like.
Some transformers can be modeled as a two pole filter, (like a 600:600, as all the L's and C's are the same on both sides, pri and sec)

Others can be modeled as a three pole filter network.
Pick the ripple you want in db along the pass band, and you can predict the rolloff in the stop band.
This is why the graphs I did on unterminated transformers resemble rolloff curves off some of the better known Bessel and Tolyschchenov type filters. (sp?)

Inter layering, while decreasing leakage, increases capacitance, so we do not want too many layers. Plus, the bi fi pri will greatly reduce leakage, but we have to watch voltage gradients.

Pri Sec Pri is a popular choice...

it might have been "pie" wound.

this is a technique you can find on the web.
it is used to minimize C so that your high end has a nice bell curve.

you wind a bunch of turns, then you rotate the core 180 and wind another pie.

then you only get to roate 90, but then you get to rotate 180 again, but now this section is right next to the start, and on and on.

you could wind the low end all as one pie, then spread out the high end windings.

i do not think any special high end audio winding was going on, as this was a cheap telephone eq.

you could also try 120 120 120.
this would have equi distant voltage gradients, which might require a bench experiment to verify.


you can machine wind pies.
you just set a small pitch, wind a pie, then hit stop, rotate the core with the L or R button, and start winding again.

i do not think that those toroids were hand wound.
possibly the antiquated toroid winding machines made it look that way.

today the toroid cores are a lot better,

so you can use less turns to get the same inductance,

this means you can skip the fancy winding techniques, like pie winding, back and forth winding,

leakage C in a toroid has to due with the pitch at which the wire is wound,

so you speed up the rotation of the core when machine winding.

EI lams will work, keep the turns down so you get less hum,

toroid self shielding properties are nice but if you shield an EI inductor, you should be ok.

Leakage capacitance is the sum of many unitary capacitances, which have different impact according to the placement. Capacitance between adjacent turns is not very significant because they end up in series, but capacitance between layers may be worse if the voltage between them is high. Reverse winding may be good or bad. Using RW and ending with the end close to the start is bad. On a toroid, it may be good to reverse wind in order to make the winding short and separated from the other winding (that's how safety transformers are made). You have to look at the geometry of the finished xfmr to evaluate all the possibilities.
So well, RW influences leakage capacitance, but does not necessarily reduces it. Better a large-ish but balanced cap than a small unbalanced one.
As to balancing "the primary-secondary capacitance", consider the voltages at the primaries. If one primary has the "hot" point on top and the other has "hot" on bottom, the secondaries will receive the same capacitive coupling on both sides (see attached). RW is a necessity in humbuck configuration, so external fields cancel and internal fields add up. It's equally valid for both cases, segmented or separate. You answered your own question. I've never seen a case of a solid-state audio circuit that needed segmented construction for this reason.

 

[shabtek]

curious how that core behaves well below 10kHz

 

[abbey road d enfer]

I recall (MIA CJ) describing 'pi-wound' to reduce capacitance (iirc) --something like winding 1 quadrant, then the quadrant across, then back 1 and then the last--like 1324 is that similar tho what you mean be by progressive pattern?

Pie-wound or beehive is mandatory for RF, not so much for audio, mainly because it results in poor filling ratio. Basically a beehive coil will have about 25-30% less turns than a standard layered winding.

 

[solkatten]

curious how that core behaves well below 10kHz

Most nano/amorph cores have a flat inductance you dont see the rise in H at low freq.

 

[solkatten]

random edited CJ+Abbey quotations:

Words of wisdom

 

[abbey road d enfer]

Yep I think its simular.

I don't think so. See my answer to shabtek.

Lay the associated turns on top of each other to reduce capacitance...

There are ways to minimize the voltage between turns, by being creative with geometry, but the ultimate is chambering.

The most common would be 5 clock wise, 5 counter clock wise, 10 clock wise... until you circle the toroid...

I'm not sure I understand what you mean... (just a way of saying I don't understand)
Keep on, we're following.

 

[solkatten]

Yep they are 2 diffrent things pie wound and progressively wound

The clock-wise-thingy:

 

[gyraf]

..note that the thinner the wire, the less local surface area, and thus less resulting inter-winding capacitance. And you probably want to minimize this to get high frequencies decently behaving at your 15K to 50K secondary..

/Jakob E.

 

[abbey road d enfer]

The clock-wise-thingy:

Ah! OK. It's chambering without the chambers.

 

[solkatten]

I´ll think I will do the trafo in 2 diffrent versions.

1. The easy to do version. Pri: singel layer 0,25 mm. Sec like the picture toridal secondary.jpg. Many small sections trough out toroid were you lay the turns on top of each other but not in perfect order (a.k.a. semi progressive)

2. The hard version. Primary: same as above. Secondary: 10+9+8+7+6+5+4 on top of each other (beheaded-pyramid-style)

As my earlier experiments have indicated there is no advantage in interleaving pri/sec/pri/sec it made the capacitance higher...
The singel layer primary is the best solution, its easy to wound and it got the best cmrr... what can be improved upon is the winding of the secondarys.

A 10 H primary should give -3db roll off at ca. 3Hz with 200ohm source.

My earlier toroid 1:6,5 had a -3db at 45Khz with a 200ohm source, indirectly measured with 1k3 serial resistor. The high roll off goes hand in hand with the reflected impedance. The higher impendance ratio 1:64, in my1:8 step up trafo, will make the parasitic capactance more capable of rollin off treble...

 

[solkatten]

The new core…

 

[abbey road d enfer]

The new core…

If you want to make a dick contest against LL, you're gonna win easy!

Seriously, is it molded in some kind of plastic or what?
I think I"ve never seen one...

 

[solkatten]

If you want to make a dick contest against LL, you're gonna win easy!

Seriously, is it molded in some kind of plastic or what?
I think I"ve never seen one...

Yep the nanocrystalline core sits in a molded plastic container

 

[abbey road d enfer]

Yep the nanocrystalline core sits in a molded plastic container

Thanks!

 

[solkatten]

Primary winding…

 

[solkatten]

Pri = 360T
Sec = 2700T/75 sections = 36 T per section