Audio Transformer Inductance

[trobbins]

ricardo, the Williamson design dramatically escalated audio amp design efforts during the 40's and 50's (http://dalmura.com.au/projects/Williamson.php), and certainly focussed effort on OT design and influence, and spurred on advances in amp design in general. 

The example of Bert Der Klerk's effort (http://ideas.home.xs4all.nl/amps/index.html) shows that many are still interested in recreating and improving - and even flagwaving their efforts in the black art of OT design, especially with respect to lowering leakage inductance whilst keeping primary inductance sufficiently high.

 

[ricardo]

IM very HO, anyone building a Williamson instead of a Mullard 5-20 needs an OT with Unobtainium core wound with the tresses of and by virgins.  I have this from our own transformer guru, Marik.  Swedish, Suffolk and Californian virgins (Lundahl, Sowter, Jensen) are now in short supply but if you send $500 in used bank notes to Cooktown Recording and Ambisonic Productions, I'll put you on my waiting list.

To get back to Ian's original article and question ..

It would be good to put the stuff around pages 8 & 15 from the original Williamson articles

http://www.sowter.co.uk/pdf/Williamson%20Amplifier.pdf

in Ian's.  This would allow us to predict -3dB response from Primary and Leakage Inductance as well as derive these figures from the -3dB specs of reputable makers.

It would also point out the dangers of assuming a single value of L.

But we need "some" value when choosing transformers.  As this is level & frequency dependent the best we can do is to note the Instrument, Level and Frequency if available).

The level will probably be OK for large transformers but too high for small mike transformers.

The DCR tells us what LF compromise has been chosen for the core and the Cu losses.

The last spec should be a 3% thd Level spec at some low frequency.

Not all of these need to be filled in.  Only the last, 3% thd level spec, involves more than using a meter or copying from maker's spec.

Maker & Part No.
Description
Ratio
        Design Source & Load Impedance This often bears no relation to the ratio.  eg a valve mike transformer may transform 10K to 200R but is expected to work into a 2k preamp load.  see the next spec.
        Is it a Bridging Trannie? These will be 1:1 but designed for 600R feed and 10k load.  Maybe leave this out as its answered by the previous spec.
Primary DCR
        L
        Leakage
Secondary DCR
        Leakage Do we need this or is covered by the Primary leakage?
Instrument
        Measuring frequency
        Level
-3dB response spec.
?% thd level preferably 3% thd level
        @ frequency

I've also put further explanations on my post with the frequency response of a small mike transformer.

 

[ruffrecords]

I have updated the document to include results provided by groupdiy members. When I get time I 'll start measuring DCR and leakage inductance.

Cheers

Ian

 

[trobbins]

So Ian do we still have the concern that any stated primary inductance measurement/spec at low frequency is nebulous unless the excitation voltage is stated along with the frequency?  Or is this just an issue for amplifier OT's in general?

 

[CJ]

i have measurements for some stuff, like A 10, HA 100X,
HS 56, K-241-D, A 24, HS 52,

all the usual suspects,

let me know if you want me to round it up,

numbers will be unique to the meters i used, but no biggy,

 

[ruffrecords]

So Ian do we still have the concern that any stated primary inductance measurement/spec at low frequency is nebulous unless the excitation voltage is stated along with the frequency?  Or is this just an issue for amplifier OT's in general?

Yes and no. We are sating the frequency wherever possible. Excitation remains an issue but the primary inductance is still the key parameter.

Cheers

Ian

 

[trobbins]

The Vertexion advert in the WW compendium for Williamson amp specs primary inductance of 95H at 5V, and 142H at 20V.

The nominal silicon steel curve of permeability vs flux is quite different from that of GOS - two very common cores for OTs.

It seems more sensible to just say the inductance is 'high' compared to offering a value of L without giving source voltage and freq for an OT, as giving a value of L offers little in the way of design per se, and possibly leading more to misinterpretation.  The Vertexion example may have given say a 50H reading at 1V, and would have been denigrated as not acceptable on just a 50H measurement from some modern LCR meter.  :-\

 

[abbey road d enfer]

It seems more sensible to just say the inductance is 'high' compared to offering a value of L without giving source voltage and freq for an OT,

Simply "high" has no meaning. Just because the conditions of measurement must be known does not mean that the inductance measurement is inadequate or useless. If it was the case, frequency response measurements would have been rejected, because FR without indications of +/-dB variation is not significant.
Indeed, this is a concern mainly for OT's; input xfmrs generally have much less variation with level.

as giving a value of L offers little in the way of design per se,

It depends what one calls "design". Most contemporary vacuum-tube gear "design" is to real design what painting by numbers is to Michelangelo. The Williamson article is examplary in that respect, since almost at the beginning, right after the topology and choice og output tubes, he assesses the performance in terms of primary and leakage inductance.

 

[ruffrecords]

i have measurements for some stuff, like A 10, HA 100X,
HS 56, K-241-D, A 24, HS 52,

all the usual suspects,

let me know if you want me to round it up,

Yes, please if you would.

Cheers

Ian

 

[trobbins]

The concern is that giving a 'value' for primary inductance for an OT could be interpreted to range perhaps to 3x or 1/3 the 'value' presented, depending on what measurement voltage was applied. 

To me, that is as meaningful as high, medium and low - which is similarly not very meaningful.

It was interesting that the advert by Vertexion went out of its way to put a measurement in at 20V - no doubt the marketing arm said that giving only a value of 95H at 5V would be deemed as 'not hi-fi', as Williamson specified 100H minimum in his article  :-[

 

[Marik]

IM very HO, anyone building a Williamson instead of a Mullard 5-20 needs an OT with Unobtainium core wound with the tresses of and by virgins.  I have this from our own transformer guru, Marik.

Dear Ricardo,

This is not the first time I see it from you, so since you seem to be persistent, for the record that you know, first, I am not a transformer (or for that matter, any other) guru, second, I never had anything to do with unobtainium cores (whatever it is), let alone with tresses and virgins. If you have some kind of fetish with those, please keep it to yourself and do not put on me  8)

Best, Mark

 

[ricardo]

Marik modestly reminds us that Williamson OTs are not his field of expertise.  8)

But the various discussions & caveats brought up by this thread now allow the reader to compare, with better understanding, Marik's Samar Audio offerings against those of other renowned makers like Lundahl, Sowter & Jensen.

For ribbon transformers, it should be obvious that the Samar Audio transformers represent a significant advance in the state of the art.  They have larger Primary L, lower DCR and higher "undistorted" output in a smaller package.

http://www.thinksrs.com/products/SR554.htm  This isn't a good transformer but the datasheet shows clearly why you need to match source/load impedances to a transformer.  They affect both noise (with a given amplifier) & frequency response.

 

[lassoharp]

Due to general thread title this has been a rather broad discussion.  Williamson OTs and ribbon mic OTs are great subjects and hopefully will get their own thread if conversation becomes very focused.

One thing that hasn't been discussed in regards to input, specifically low level mic input transformers is secondary L.  I'm particularly interested in hearing more on the importance of this parameter for transformers that are primarily used feeding the open grid of a tube.  Over the years I've noticed that most of the higher quality input transformers for this purpose share similar values of sec L which tends towards the high side relative to others of same ratio.  What role does it play and how is the grid circuit broken down into simple elements to analyze and compute response?  It's a subject I have not come across in outside reading but have always assumed the higher L is there for a similar reason it is on the pri.

 

[ricardo]

One thing that hasn't been discussed in regards to input, specifically low level mic input transformers is secondary L.  I'm particularly interested in hearing more on the importance of this parameter for transformers that are primarily used feeding the open grid of a tube.

Secondary L is determined by the turns just like Primary L.  So the ratio of Primary : Secondary L is the ratio of Primary : Secondary design impedance = (Turns Ratio)^2

Thanks for bringing this up as I forgot the ramifications in my suggested list of transformer parameters in Reply #61.  There is no need to measure Secondary L if you've measured Primary.

The effect is best considered by transforming all impedances using the "ideal transformer" ratio to the primary.  You see examples of the type of frequency response obtained in my Reply #42 post and also at http://www.thinksrs.com/products/SR554.htm

The transformer can be designed to be loaded by it's characteristic impedance, source transformed by (turns ratio)^2, or to be unterminated.  All that means is the damping has to be ALL provided by the Primary circuit as the grid circuit will provide little or no damping.

At LF, it depends if there is a capacitor in series with a source.  This will give a resonant 12dB/8ve filter with the total L as in my little mike transformer example.  No capacitor means simple 6dB/8ve roll-off with no peak & no damping required.

At HF, there will be stray capacitance etc with the Leakage L, so 12dB/8ve which needs to be damped.  The SR554 curves are for unterminated transformer and you see how raising the source R damps the HF resonance.

You can play around with Zobels too which provide damping only at HF.

Pages around 9 & 18 of the Williamson link I posted from Sowters has more detail.  As has Chapter 5 of Radio Designer's Handbook.  Highly recommended for daily study.

 

[trobbins]

Lassoharp, I'm also interested in microphone step up for phono use.  As I understand it, the secondary leakage and SRF and secondary winding resistance are the dominant HF players when the load is effectively very low grid leakage capacitance (possibly millar impacted) and any damping resistance needed.

As such, can we assume L is fairly constant and hence can be meassured by just applying a low (eg. 1V sine) voltage and measuring current (eg. voltage across a relatively low series resistance).  SRF probably needs a sig gen that goes at least to a few hundred kHz, and a voltmeter with similar frequency response.

Ciao, Tim

 

[ricardo]

As I understand it, the secondary leakage and SRF and secondary winding resistance are the dominant HF players when the load is effectively very low grid leakage capacitance (possibly millar impacted) and any damping resistance needed.

The Leakage L on both sides need to be considered.  See the Williamson article (which is for a loaded secondary) and RDH chapter 5 for details and how to measure this.  There are several leakage inductances and it is difficult to separate them which is why it's easiest to transform them all to the Primary side.  I think the usual shorted Sec. measurement does this for us.

Secondary resistance plays almost no part in damping cos it is swamped by the large input resistance of a valve.  But a Zobel will interact with it.  See the Jensen datasheets for examples.  You still need to design the secondary to have equal "Cu" losses to the primary cos it has noise implications even into a high Z valve input.

As such, can we assume L is fairly constant and hence can be meassured by just applying a low (eg. 1V sine) voltage and measuring current (eg. voltage across a relatively low series resistance).  SRF probably needs a sig gen that goes at least to a few hundred kHz, and a voltmeter with similar frequency response.

Better to feed a constant current, eg  via 10k resistor and measure the voltage across the winding.  To measure leakage, you need to short the other winding so you have to draw an impedance graph to get what you want.

What's SRF?

 

[trobbins]

Self resonant frequency - to give the effective distributed capacitance in the secondary (with primary shorted).

The equivalent secondary side circuit is the sec leakage L, sec winding R in series with sec leakage L, and a shunt C (effective distributed cap of sec).  The load on the sec (for a direct to valve grid circuit) would be some parasitic cap to ground, and any additional loading circuitry to tailor the response or the Q at HF.

Using say a 1V 50Hz AC source from a mains transformer winding would I think be a very simple test circuit, as just a relatively low value (eg. 10, 100, or 1k resistor depending on sensitivity of available DVM) series resistor is needed for current measurement and would only drop say <1% of source voltage.  Impedance of L would dominate any winding ESR or sense R.

 

[ricardo]

Self resonant frequency - to give the effective distributed capacitance in the secondary (with primary shorted).

Mea maxima culpa.  I neglected self stray capacitance.  Especially as both my examples show clear signs of this.  :-[

The equivalent secondary side circuit is the sec leakage L, sec winding R in series with sec leakage L, and a shunt C (effective distributed cap of sec).  The load on the sec (for a direct to valve grid circuit) would be some parasitic cap to ground, and any additional loading circuitry to tailor the response or the Q at HF.

For a properly designed transformer, I don't think it matters whether you transform everything to primary or  secondary.  Probably to whichever side gives you the most convenient measurements.

Using say a 1V 50Hz AC source from a mains transformer winding would I think be a very simple test circuit, as just a relatively low value (eg. 10, 100, or 1k resistor depending on sensitivity of available DVM) series resistor is needed for current measurement and would only drop say <1% of source voltage.  Impedance of L would dominate any winding ESR or sense R.

This is exactly the method Williamson describes on p18 of the Sowter reprint.

My caveat is that 1V is a small voltage for a power OT but a very large voltage for even a 10k mike secondary.  This method would give a good "incremental" small signal L for the OT but a very optimistic figure for a mike transformer.

For HF, if you are only interested in damping the resonance and only marginally in its frequency, just use a square wave input and scope.  The resonance is very unlikely to be above 1MHz so even the new cheap digital scopes will do a good job.  Adjusting a Zobel for no overshoot gives Q=0.5 and -6dB at the resonance.

In the valve mike preamp case, even with your secondary L and distributed shunt C interacting with secondary winding R, the majority of the damping will still be the source resistance on the primary at HF.  A Zobel attempts to provide some of the HF damping on the secondary.

Damping requires bigger source R and smaller load R

 

[trobbins]

Me myself I was mainly focussed on a phono setup for a moving coil cartridge, just for fun.  The mic transformers I have at the moment have 42k (Zephyr) and 125k (Woden MT101) output impedance nominal ratings - I haven't yet tried to deduce the likely max test voltage that I should attempt to apply for secondary L and SRF testing - hopefully they'll cope with up to 100mV.

I agree that the simplest test 'in circuit' to tune the HF response is to measure by oscilloscope the valve grid voltage on the secondary side of the step-up transformer, for a square wave source - noting the loading that the oscilloscope probe introduces, which should be negligible if 1M and a few pF.

 

[CJ]

"simplest test 'in circuit' to tune the HF response is to measure by oscilloscope the valve grid voltage on the secondary side of the step-up transformer, for a square wave source ..."

square wave is good for B-H loop testing and ringing,

plot a freq response curve with sine wave, 10 hz to 100 k hz,

your self res will be from 40 k hz to 100 k hz,

make sure you keep your signal constant, as you hit resonance, more energy is required to make that voltage peak, so your generator voltage may drop, depending on how "stiff" it is,

stiff would mean that the thing is running into a load that is bigger than the load you are putting on it, like the HP generator, 50 watts of power into a short circuit,

you drive a 20 mw tranny and it does not know it is there,

zobels, well, never seen a really big change in sound quality due to adding a zobel,

if anything, they cut high end,

most just squash one peak into two or three peaks of lesser amplitude, but the transformer still uses more energy at self res, even with a zobel, as higher freqs will pass more current thru the cap, because the cap is a short at some high freq,

zobels work at a fairly high freq, most people cut off at 15,750 hz fundamental freq,

effects of third, fifth harmonics might be shorted by the zobel, but not the fundamental that is within hearing range,