Designing an Interstage Transformer

[TheGimp]

1. If cost were the only consideration I would buy transformers. Considering I will need a winding jig (Make or buy is still a question), and in time more laminates wire, etc this is not cost effective per se.

I am doing this because I wish to learn. I am going back and studying things I forgot years ago.

2. Thanks for the input on winding technique, however for now I would prefer to stick to the electrical design and once finished move on to issues like interleave, segmentation, multifilar winding, etc.

Back to design.

To address Simper-Fe's questions:

5 X Rp was chosen as I understood that for inductor loading, one would normally like to have load impedance 5X Plate resistance. If this is not necessary I can relax this spec.

Why 10Hz, because the design example I was reading used that value. With this core I am sure I will have to make compromises, but for a first pass calculation. So, I'll recalculate with 20Hz which is more typical of F1 for audio amps any way.

Wire size will be addressed as part of the design but in a later phase.

So back to the design.
there was a miss-calculation in the first post. Using 5XRp and 10Hz should give 436H instead of 218H. 218H is correct for F1=20Hz

Calculating N we re-arrange the equation:
L=(3.2 * N * N * u * A )/ (10^8 * l)

solving for N one has :

N = SQ-RT (L * 10^8 * l )/(3.2 * u * A)

The cores dimensionally look real close to EI-87 (26guage), and may be.,so I'll use that data. The fact that they were used for line transformers gives a high probability they are GOSS but low silicon content.

u of 4% GOSS is around 5000
u of steel is around 300
so a conservative guess would be u=1000 for these cores.

l (MLP) = 15.35cm
A = 0.88sq in ~= 5.68sq-cm

this yields: 13568 Turns!!! Yikes, this is not going to happen.

So, where to start making compromizes?

1) What is a reasonable value for XL if not 5X Rp?

Is XL = Rp reasonable? It meets maximum power transfer theorum.

2) Is u=1000 reasonable or should it be higher?

[iko]

Originally Posted by SemperFi

But this is diy isn't it? If any one here does diy to save money I think they picked the wrong hobby.

Not generally true. While difficult, one can build an 833A amp for less than one can buy a wavac.

[iko]

What about the DC component? Is that taken into account in your calculations?

[GoatGuy]

[1] there is no such thing as a "lossless" E-I core.

[2] 10 Hz spec ... is just silly. SemperFI is right on this. Shoot for 17 Hz or 20 Hz

[3] Bifilar winding is a must have aspect of this proposed transformer. The enhancement of both top-end and mid-range coupling is substantial, and desired.

[4] Your biggest "problem" will be core saturation, regardless of whatever henry-value you choose for the windings. Its real, and it sucks.

[5] Because of [4], you'll find in turn that you'll likely have to compromise the low end rolloff.

[6] And... while you're using transformers as interstage coupling devices ... you will want to look at using tri- or tetra- filar windings ... so that you might step UP the output of stages for that grand old coupled-inductor voltage-amplification effect. Its useful, it works, and that's what the attraction is to using transformers.

[7] Or, though difficult, consider using separate insulated overlapped layering. Since the transformer is a one-off (i.e. you're not in production!) the extra labor of winding one layer of primary, then 1 layer of secondary (with a different wire gauge and # of windings), then a paper gap, and another primary, and another secondary, and another paper gap, and so on ... until the desired primary/secondary is built up ... isn't all that bad. Better control of interwinding capacitance.

Well there you are. I'm happy indeed to no longer be winding my own transformers. There is just so much MORE one can do to become happy with electronics than this particularly odious task.

GoatGuy [PS: Jensen ... are good people. They make great transformers.]

[popilin]

Originally Posted by TheGimp

1. If cost were the only consideration I would buy transformers. Considering I will need a winding jig (Make or buy is still a question), and in time more laminates wire, etc this is not cost effective per se.

I am doing this because I wish to learn. I am going back and studying things I forgot years ago.

2. Thanks for the input on winding technique, however for now I would prefer to stick to the electrical design and once finished move on to issues like interleave, segmentation, multifilar winding, etc.

Back to design.

To address Simper-Fe's questions:

5 X Rp was chosen as I understood that for inductor loading, one would normally like to have load impedance 5X Plate resistance. If this is not necessary I can relax this spec.

Why 10Hz, because the design example I was reading used that value. With this core I am sure I will have to make compromises, but for a first pass calculation. So, I'll recalculate with 20Hz which is more typical of F1 for audio amps any way.

Wire size will be addressed as part of the design but in a later phase.

So back to the design.
there was a miss-calculation in the first post. Using 5XRp and 10Hz should give 436H instead of 218H. 218H is correct for F1=20Hz

Calculating N we re-arrange the equation:
L=(3.2 * N * N * u * A )/ (10^8 * l)

solving for N one has :

N = SQ-RT (L * 10^8 * l )/(3.2 * u * A)

The cores dimensionally look real close to EI-87 (26guage), and may be.,so I'll use that data. The fact that they were used for line transformers gives a high probability they are GOSS but low silicon content.

u of 4% GOSS is around 5000
u of steel is around 300
so a conservative guess would be u=1000 for these cores.

l (MLP) = 15.35cm
A = 0.88sq in ~= 5.68sq-cm

this yields: 13568 Turns!!! Yikes, this is not going to happen.

So, where to start making compromizes?

1) What is a reasonable value for XL if not 5X Rp?

Is XL = Rp reasonable? It meets maximum power transfer theorum.

2) Is u=1000 reasonable or should it be higher?

Watch out !
In SE have DC circulation, you should consider an air gap !

I hope some of this would be useful

Designing Transformers with J.C.Maxwell

[GoatGuy]

Dunno how you did your calcs there...

But I came up with 3800 turns. While that clearly is a high count ... it isn't outrageous. Just will take some time.

(I used u=2500, and H=432)

GoatGuy

[316a]

Originally Posted by sjs

I have to say you are starting with probably the most difficult audio trans to wind imho.

For good performance from interstage driver transformers, bi-filar windings are by far and away the best method to get good bandwidth, but you will need high quality double insulated wire, and v good winding technique to reduce risk of primary to secondary breakdown.

I would suggest your best bet is to buy commercial from Hammond, Audio Note or Monolith Magnetics.
Best of luck

Hammond 126C and 126B , to me sound superb for a £30 a piece lump of iron . These bifilar 1:1 IT are a bargain , probably the biggest bang for buck out there

316A

[TheGimp]

I've been using the formulas from here:

Transformer Math

In his example he used plate load of the transformer as 5 X Rp.

I've been typing the equations into a spreadsheet.

I'm now using u=4000 and lg= 0.0076cm (.003")

I get a ue of 1340.

Here are a screenshot and a zip of the spreadsheet

[nigelwright7557]

Originally Posted by GoatGuy

Well there you are. I'm happy indeed to no longer be winding my own transformers. There is just so much MORE one can do to become happy with electronics than this particularly odious task.

I found you have to try and keep things tidy. Keep the tension the same. Lay the windings next to each other neatly.
For dual secondaries wind both at the same time to keep inductance the same.

I have done loads of class d output coils.
I have also done quite a few SMPS transformers.

[TheGimp]

OK, Since I don not know the permeability of the core I figure I should measure it so we can all agree on a value.

I used a variac to drive the 16R secondary to common taps. I measured the current using a 1R sense resistor. I will strip the windings and count them once I have sufficient data. From the number of turns and the current I can get the Ampere-Turns at saturation

Readings were taken at 5V intervals.

1) Is this sufficient data to get an accurate measure of permeability?
2) Where on the curve does one define saturation since it is not a step function?

Current doubled from sample 6 (30V) to sample 7 (35V). Is 30V then a good point to chose? Should I take more readings between 25 and 35V?

I also tried measuring it with a 100W light bulb between the variac and transformer. When I got to 20V the lamp came on and the voltage across the transformer didn't change but 4V as I increased the variac to 125V out. From that it looks like saturation occurs at 20V.

If permeability is the slope of the BH curve, it seems I should get more readings below 25V and calculate around 15 V.