Audio Transformer Inductance

[MatthisD]

I jotted down the model of the handheld meter somewhere, I wouldn't know myself but would it not be more likely fixed at 1K?

 

[lassoharp]

That's a good question.  Many of the less $$ meters won't say but 1K seems likely. 

Some of the nicer hand helds give two, sometimes three freq choices.  Low freq selection is usually 100HZ or 120Hz.  I think I may have seen one meter that had single fixed freq stated as 100Hz but I can't recall for sure.

 

[ricardo]

Here's the response of a transformer from a microphone.  It's fed by an emitter follower running 1.37mA (Rsource = 19R), 4u7 and feeds a 2k mike preamp.

At low level, the "Primary referred" inductance and 4u7 form a HP filter which is underdamped with a Q of nearly 2 (red curve).  At higher levels, the inductance rises which reduces both the frequency of the peak and Q (black curve).

At HF, the leakage inductance and stray, cable and preamp capacitance form another resonant filter, this time a LP.  Leakage inductance doesn't change with level.  The upturn in the red curve is a measuring artifact cos the level is low and our bandwidth is high.

The change in response/inductance with level is fairly typical.  Better core materials change less but they all change enough to see easily on an electrical response curve.

So Inductance changes with frequency and level too.  The best we can do is note the frequency (as low as possible) and level at which they are measured.  As this is may not be available, perhaps note the Measuring Instrument and what it is set to.

My "3% thd level at specified frequency" is good cos it can be scaled accurately with frequency.  see my earlier post

Primary Inductance, DC resistance and 3% thd level are interrelated and together, decide the core size and material, which is the biggest cost in a transformer.  They determine LF performance

The winding and sectioning decide the Leakage Inductance and HF performance.  It is a good indication of the expertise of the maker and is probably the most Black Magic component.  This is even less often quoted than Primary Inductance.

 

[CJ]

note that DCR is at a 90 degree angle to XL,

so if DCR is 100 ohms pri, and XL is 100 ohms at 20 hz,

then total Z will be 141 ohms, not 200,



rt triangle with 45 degree corners ,

A is 1  B is 1 C is root of (A^2 + B^2) = root 2 = 1.414

typical transformer should have DCR at 10 percent of lowest XL,

so XL is 100 ohms, DCR will be 10 ohms,

imagine a triangle with 100 as side A, 10 as side B,

side C will be 100.2 or some low number, so the effects of the DCR are minimized by the lagging current,

the effects of DCR will be miniscule due to the 90 degree angle and 10:1 factor,

 

[ricardo]

> the effects of DCR will be miniscule due to the 90 degree angle and 10:1 factor,

The issue of DCR is not to do with Inductance but is a design consideration.

For a given core, you can have more Inductance with more turns.  This also gives less flux so you get more level before 3% thd sets in.  But more turns on the same bobbin needs thinner wire so more DCR; ie Cu & resistive losses.

These losses are important even if you are feeding a valve or FET directly as they raise the noise level.

Transformer design (indeed all good design) is a balancing act.

.. and I bet some of you didn't know transformers could be noise sources ...  ;D

 

[PRR]

> fed by an emitter follower running 1.37mA (Rsource = 19R) and feeds a 2k mike preamp.

The 2db@10Hz - 4db@15Hz bump on your plot suggests a non-infinite capacitor, probably blocking emitter-follower DC level. Knowing that cap's value adds clues.

There's also a small cap somewhere, probably on secondary, possibly just secondary self-capacitance (though if load is really 2K, it suggests a lot of stray C).

 

[ricardo]

The 2db@10Hz - 4db@15Hz bump on your plot suggests a non-infinite capacitor, probably blocking emitter-follower DC level. Knowing that cap's value adds clues.

4u7 in this case.

There's also a small cap somewhere, probably on secondary, possibly just secondary self-capacitance (though if load is really 2K, it suggests a lot of stray C).

I think there was a lot of cable to the preamp.  Then there's the RFI bits on the mike preamp.  The upturn in the Red curve above 100kHz is a measurement artifact cos the signal level is low and bandwidth high.

You can flatten both LF & HF with various Zobel & matching load tricks but only at one level.  Did a lot of this type of thing in my Calrec days to build supersonic & subsonic filtering into mikes.

But I posted the curve to point out that transformer Inductance varies not only with frequency but also with level .. depending on core material.

There are other factors which might give "similar" behaviour.  eg some types of Ceramic caps.  But in this case, it's the core material.

 

[mitsos]

Ian,
On the carnhill ungapped output description you write that it's not suitable for DC on the primary, how do we know how much DC an ungapped trafo can take, is there any rule of thumb to get a starting point? My understanding is that there is a grey area, the trafo could handle a bit of DC before inductance spirals down.

I've been interested in winding a 7:1 output 30K:600 that is not cap coupled. (like the UA PA5946, original output for the vintage UA610 and one of the comps, 175 or 176 don't remember)

Inductance spec per the rev eng of the 610 schematic (from john hinson) is minimum 55HY with 7mA DC.  Some people have said it should be gapped, but some say just make it bigger.

I'm of the "make it a big steel monster" camp, can you guys shed some light on the effect of DC on primary inductance? 

thanks!

 

[abbey road d enfer]

how do we know how much DC an ungapped trafo can take, is there any rule of thumb to get a starting point?

The starting point is that the B/H curve inflects as soon as DC shows its nasty nose, less with iron than Ni, so basically it's a "linear" process by which THD increases as soon as DC comes in.
Let's say you have a xfmr where the max level is quoted at a given power.
You will deduct the maximum instant current.
P=RI² => I=sqrt.P/R Imax=1.414.I
Operating in class A, maximum DC current should be less than half Imax and usable max instant current will be half Imax, resulting in less than 1/4th power. This will be obtained at the cost of quite a high 2nd harmonic THD.

 

[mitsos]

So, am I reading this correctly... we need a trafo with 4 times the (0V DC) power to give us the necessary inductance in class A? Simplistic, sure, but this is all theoretical so far.

This will be obtained at the cost of quite a high 2nd harmonic THD.

Some might say that's why the old ones sounded better!  8)

 

[ruffrecords]

Ian,
On the carnhill ungapped output description you write that it's not suitable for DC on the primary, how do we know how much DC an ungapped trafo can take, is there any rule of thumb to get a starting point? My understanding is that there is a grey area, the trafo could handle a bit of DC before inductance spirals down.

Difficult to say but I suspect it is rather small. Once, when designing a headphones amplifier I wanted to take some NFB from the primary of a cap couple ungapped Sowter transformer with a mumetal and M6 core via a largish resistor (about 47K I think) to the cathode of a tube. The dc current would have been in the region of 50 micro amps. I asked Brian Sowter if this would be OK and here is his reply:

Hello Ian

I think you might get away with this.

The current will tend to saturate the Mumetal part of the core but certainly
won't touch the M6 part.

I think however that it would only partially saturate the Mumetal.  Any
result will be loss of inductance which would cause an increase in the LF
cut off frequency.


Best regards

Brian Sowter

I guess you have no mumetal in your intended design so maybe you might get away with 7mA of dc.

Cheers

Ian

 

[mitsos]

Thanks Ian,

I don't think I'll use nickel at all on this, but if my deduction from abbey's comment is correct, I'm gonna need a huge trafo.. we'll see what happens.  I'll be visiting my winder friend in the next couple of weeks, I'll try to get this designed and will post back with any results.

 

[CJ]

Maximum Flux Density for an inductor with DC current is

Bmax=B-AC + B-DC

B-DC  = 0.4Pi*N*I-DC*10^-4/(Lm/delta u)

see the Lm part?  that is magnetic length.

the longer the magnetic length, the more DC the thing will handle.

You can increase the magnetic length of the transformer or inductor by adding an air gap.

inductance goes down quite a bit so you need to increase size of core or add more turns, or both.

inductance for a gapped transformer or inductor is

L = 0.4pi *N^2*Area(cm)*10^-8/lg + (L-m/u-r)

lg is the size of the gap, so you can see as the gap increases, inductance goes down,

let me know if you want more formulas, yuk,

 

[david-p]

I agree that the primary inductance of a transformer is a principal determiner of frequency response for a specific source resistance, and therefore an important parameter to have available when deciding which transformer to use for a given application.  I would like to see measurements done on the tramsformers that are used for valve (tube) condenser mikes.  There are about half a dozen that are regularly mentioned here.  E.g. various varieties of BV8, etc.

I am not clear, however, about what Ian has actually measured and what he calls primary inductance.  Does not primary inductance also include the reflected (transformed) inductance?  Can they be separated?

Why not make the determination by loading the secondary with the intended resistance (usually 2k in the case of microphone inputs) and doing a frequency response measurement to determine the LF and HF 3dB limits?  This would also show up any resonant effects and might be more useful to designers.

David

 

[abbey road d enfer]

Does not primary inductance also include the reflected (transformed) inductance?

Primary inductance exists by itself. The secondary, being unloaded, reflects nothing on the primary.

Can they be separated?

Yes, primary and secondary exist on their own. That's why you can use just one winding of a transformer as an inductor.

Why not make the determination by loading the secondary with the intended resistance (usually 2k in the case of microphone inputs)

Many mic xfmrs are intended to be loaded by a much higher load, check tube mic pres - the secondary goes directly to the grid without a grid leak resistor.

and doing a frequency response measurement to determine the LF and HF 3dB limits?

This is indeed what many people do; that's called indirect measurement. It's correct for someone who designs mic pres but does not want/can't design a mic xfmr. Someone who designs xfmrs must assess the primary parameters, i.e. inductances and DCR, some other parameters are measured indirectly, such as coupling factor/leakage inductance, but it is always preferrable, on a scientific point of view, to measure primary parameters directly, not indirectly, particularly in the case of xfmrs, which have a high level of interaction with the associated circuitry.

This would also show up any resonant effects and might be more useful to designers.

Indeed, and that's a necessary complement to the primary measurement.

 

[mitsos]

let me know if you want more formulas, yuk,

I'd rather have the winding info for the PA-5946!!  ;D

hmm, found an interesting CJ winding diagram for an RCA trafo. I'll start another thread now, I've derailed this one enough.  Appreciate the insight from all, if anyone wants to help design this thing, I'll be in the other room! thanks!

 

[david-p]

As the initial post referred to measurements of already-designed and on the market transformers, I did not take it to refer to new designs.  In view of the complexity of a transformer, that the elements are not "lump" but "distributed", it seems to me that it would be more valuable to measure performance in specific test situations rather than individual parameters that can only be used for predicting performance when "everything else is equal" -- as it rarely is.

For instance, as I suggested, I would be interested in seeing such results tabulated for the various microphone o/p transformers that are currently offered for sale.  As already noted, the manufacturers offer little data beyond turns ratio and price.

David

 

[lassoharp]

performance in specific test situations

Not all mic amp circuits DIYers build will be the same.  How would they cover them all?

I would be interested in seeing such results tabulated for the various microphone o/p transformers that are currently offered for sale.  As already noted, the manufacturers offer little data beyond turns ratio and price.

Sowter offers circuit specific response charts for some of their transformers.  Output will play a part in the whole picture.  For something like a BA6A clone you'd want to see a shot of the whole amp.  Then that's specific for all Sowter iron and others may use a mix of different iron.

Jensen and Cinemag give you suggested zobel networks for their inputs.  You may still have to re-tweak.  How many different mics the input will see.  So many factors . . . how can you standardize to suit every application?

I do agree that L specs would be helpful.  As ricardo & others have pointed out,  freq and level will be relevant.  I've never seen L specs in any of the old catalogs either

And then there's the subjective end of things.  What may be not so good from a designer's perspective may be very adequate to an end user for a given application.  There a lots of older circuits that "cheat the iron" using various techniques.  If the iron is weak on the bottom or top and we "cheat it up"  - how does the iron handle it?  And how does a manufacturer give meaningful specs for this?    This is where the DIY community comes in.  There seems to be very few who publish response charts & distortion figures for builds.

I'm just asking similar questions with you here.  I think it's really up to the end user to sort through the fine details and hopefully document them.  Sometimes it's hard to turn away from hard data (and the woes and worries it can bring) and just lock in on what our ears say.

 

[trobbins]

Some OT transformer manufacturers back in the 1950's,60's included primary and leakage inductance levels. 

The advent of the Williamson amp led OT manufacturers to publish primary inductance and leakage levels to support their products quality.  UTC advertised their LS-63 with 300Hy primary at 50V 60Hz.  Red Line advertised pri and leak inductances, but not measurement details.  Vortexion provided prim L at both 5V and 20V 50Hz, and leakage L at 1kHz with 10V applied.  Savage-Devizes provide pri L at 5V 50Hz, and leakage at 1V 800Hz.

A&R's 1965 quoted their primary inductance at 10V 50Hz for the hifi range.

Bert Van der Kerk prepared details on diy Williamsons, and came to the conclusion that 5V mains was the most practical measurement for prim L.

 

[ricardo]

First read http://www.sowter.co.uk/pdf/Williamson%20Amplifier.pdf  Much good stuff although there are better and simpler valve circuits like the Mullard 5-20.

Have a look at pages around 9 & 18.

The old gurus were certainly aware of all this stuff including the change in L with level.  Fig2 p9

That may be why some reputable transformer makers don't always quote L.  But the info is in their specs.  It's the -3dB points for their "frequency response".  See p9.  And its the "Incremental" value, point A on Fig2 if this is the low level LF response.  The L at max output is less likely to be quoted as its its harder to measure but would certainly have been considered in the design by a good maker.  It would be summed up in the max o/p at 20Hz etc figure.

The stuff here is most relevant for large power output transformers.  Small low level transformer cores with loadsa Nickel behave somewhat differently.

Bert Van der Kerk prepared details on diy Williamsons, and came to the conclusion that 5V mains was the most practical measurement for prim L.

Err.rrh!  I think Bert stole this from Mr. Williamson.  see p18  ;D