Calling all tube heads - recommend me an alternative output transformer.

[GK]

Hello all.

Here is an amplifier I am working on:

http://www.users.on.net/~glenk/pp20/pp20.htm

However, as explained, (see last paragraph especially) the performance is a flop due to the crappy high frequency performance of the Edcor
output transformers with a high 10k primary impedance.

I need to find a decent alternative transformer that will work well with the EL34 tube pair, that doesn't cost a retarded amount of money!

[poptones]

Can't you find a Hammond with the proper 4-5K primaries?

[GK]

I haven't seem much good said about the Hammond transformers. However I'd consider a pair if decent frequency response plots were avaliable.


I just pulled one of the Edcor CXPP50-MS-10K transformers out of the amplifier chassis and conducted some frequency response measurements.

Edcor specify the frequency response of their CXPP transformers as being within -1dB from 20Hz to 20 kHz. However they conveniently don’t specify a test condition as far as I can see.

I’m willing to bet that they conduct their measurements with the primary winding of the transformer driven directly from the low impedance (either 50 ohms or 600 ohms) output of a signal generator.

I’ve attached a chart of my frequency response measurements for this transformer, for various driving/source impedances of 150 ohms, 1k, 3k and 10k.

This transformer can only meet its frequency response specification at 20 kHz if it is driven by a source impedance of less than 150 ohms!

Unfortunately there isn’t a general purpose Hi-fi audio power tube in existence with a plate impedance that low. For pentodes like the EL34, several k-ohms is the norm.

With a 10k source/drive impedance the frequency response is a whopping 4dB down at 20 kHz. The low frequency response of this transformer however is fantastic (obviously due to the high inductance of the high impedance primary winding). Would be good for a subwoofer amp.


The high frequency (20 kHz) double-digit distortion above 10W output caused by this "50W" transformer, driven by the plate impedance of the EL34's, has to been seen on the scope to be believed. The thing is, with its high primary impedance of 10k, this transformer can only be properly utilised with high voltage tubes, which, as a general rule, have high plate impedances.
If you actually wanted to get the full 50W out of it, you'd need a drive voltage of 1000V peak, which means a plate supply voltage for the push-pull configuration comfortably above 500V - in the vicinity of 600 to 700V.  However the only way you could possiby get moderately acceptable 20 kHz distortion performance and bandwidth out of it is by driving it with tubes having a rather low plate impedance; either triodes or pentodes/tetrodes connected as triodes ("triode strapped), and possibly multiple pairs, connected in parallel push-pull. However, that limits you (with common audio power tubes) to a plate voltages well under 600-700V.

This transformer is an arse-over-head heap of shit! And what pisses me off even more is that I bought, for another, much bigger amplifier project, an even more expensive pair with the same 10k primary impedance, but with a power rating of 100W. They can only perform worse than the 50W ones.

[GK]

OK, I've found a use for these crappy Edcor transformers. They're now requisitioned for my 19" modulator rack. Instead of HiFi in my living room the 811A’s triodes and associated driver/power supply circuitry will be instead assembled into a rack case and put into class E MOSFET AM transmitter drain (aka “plate”) modulator duties in the shack. The Edcor’s are perfectly adequate for 300Hz – 2400Hz speech.

However I still need to find something for my EL34 amp. Something like the ultra-linear Partridge transformers as used in the famous Mullard 5-20 amplifier would be ideal...........

[poptones]

I had no idea. I was under the impression Edcors were pretty nice. Then I stumbled upon this and it confirmed pretty much exactly what I suspected when I first saw your frequency response plots. Apparently this transformer maker needs to go back and read Williamson.

http://www.users.on.net/~glenk/pp20/pp20.htm

[GK]

Err, yeah, that is my website...........   

[poptones]

ROTFL.

[GK]

I’ve found a manufacturer of tube audio transformers down here in Oz:

http://www.avat.com.au/Mambo/index.php?option=com_content&task=category&sectionid=5&id=20&Itemid=48

Their site is a bit sparse in information and product pictures, but it is at least devoid of audiophile BS, which is heartening. They offer output transformers for the Mullard range, including the 5-20, which are apparently reverse-engineered, re-makes of the originals. That’s great, but I’m not keen on forking out $620 for a pair of 20W transformers.

However I have another option. I can try to re-wind the crappy Edcors myself. How hard can it be? The Edcors have horrendous leakage inductance. I’m guessing the interleaving between the primary and secondary winding is minimal to simplify construction.

Given the exemplary low-end frequency response, these transformers are probably designed for too high a primary inductance. They don’t need to be so flat down to 10 Hz, and making them that way just compromises the performance at the high frequency end.

But anyway, first things first. What do I use to liquefy the lacquer and get the laminations apart? I’ve got acetone and turps.

[poptones]

You don't need to liquefy anything. Just get the cores apart and toss the windings. Get a copy of Williamson's treatise and go to town. Would be an interesting read.

[GK]

I intend to chuck the windings. The laminations are bound together by the lacquer.
I have the original Williamson article from WW. It is not a treatise on transformer design, just an overview. So far this is more informative:


http://www.tubetvr.com/transformers.pdf

[poptones]

Ah. My experience was to use naptha.

[GK]

I've now done enough reading to be confident that I have what needs to done to rewind this transformer with adequate performance down pat.

Firstly, I've measured the primary winding inductance. As suggested by the frequency response measurements, it is way to high. It measured 313H. It really only needs to be one third of that. Winding a primary with excessive inductance just compromises the HF performance of the transformer by increasing leakage inductance and winding capacitance.

The attached scope shot show in part how I measured the primary inductance. The secondary winding was left disconnected during this test. A 10V peak squarewave of 1kHz was applied to the primary winding via a 100 ohm series resistor in the ground leg as a current sense. The blue scope trace shows the voltage developed across this resistor, and thus the coil current. The voltage ramped 16mV, which corresponds 160uA, in 5mS.

L = (Vin*t)/I

= (10V * 5mS ) / 160uA = 313H.

I currently have the transformer soaking in acetone. It seems to be doing a good job at removing the lacquer without attacking the plastic that the bobbin is made of.

I won't know for sure how the winding are arranged until I can remove the bobbin from the laminations and unwind them, but thus far it appears, as far as I can make out, that there isn't any interleaving between the primary and secondary windings. The bobbin is slightly translucent and by holding a strong light against it I can vaguely make out the windings. It appears that the secondary is wound on first, with the primary on top.

[xygor]

These guys have some nice toroidal transformers.
http://www.plitron.com/standard-toroidal-transformers/tube-audio-transformers/tube-output-transformers/

Also check out the "tech notes" tab.

[GK]

The Piltrons do look nice, but IMO the price is a joke. I’d also be concerned with the sensitivity of the toroidal core to DC current due to any plate current imbalance between the tubes.

I have finished dissecting my crappy Edcor CXPP50-MS-10K now. Attached is a simplified schematic that I managed to draw of the winding scheme. The acetone soak over night did a good job at liquefying the lacquer that the whole thing was dipped in, but it was still a bit of a tedious pain to get the sticky E and I laminations apart. I had to gently prize each individual lamination from the one underneath with a razor blade.

As it turns out, the primary and secondary windings are actually interleaved in sections, but only in three sections for each winding. From the Hans Beijner transformer design article I linked to in a previous post:

“Many output transformers are winded using 5 primary sections and 4
secondary sections; this seems to be the best compromise in many cases.”

I think the crap HF performance of this transformer can be but down to two main issues:

1)   An inadequate number of interleaved layers and
2)   A far too high primary winding inductance.

As for #2, you just can’t have an awesomely flat frequency response down to 10Hz (as I’ve measured) and exemplary HF performance at the same time. There is a compromise between the two and this particular transformer design is just far too lilted towards the LF end.

An interesting thing revealed by the dissection is that the secondary winding (unlike the primary winding) is not electrically divided into three separate winding layers connected in series, but three separate layers connected in parallel.
   
 

[notsob]

The 'Australian" brand equivalent for naptha solvent is 'Shellite'

[poptones]

Well there's your capacitance. Why on earth would they interleave them in parallel?

[GK]

While I'm at it, an irate e-mail from an audiophile has already arrived. I'm apparently guilty of denigrating the great products of Edcor!

Firstly, I'm only "denigrating" two of their products; The CXPP100-MS-10K and CXPP50-MS-10K transformers. I'm my opinion these transformers are not suitable of HiFi. Edcor's design philosophy, as revealed by the dissection of my transformer as shown here may very well be quite fine for making audio output transformers with adequate HF performance if the primary impedance is only a few k-ohms, but it is, IMO, totally inadequate for a transformer with a primary impedance of 10k. I'll stick with conclusions drawn from what my measurements reveal!
 
To make plain just how bad this 10k transformers high frequency performance actually is, I've attached some scope shots. The top two photos show the output signal of my amplifier power output stage open loop (no global negative feedback) delivering a 1kHz sinewave into an 8 ohm load; the photo to the left below clipping, the photo to the right well into clipping. The vertical scale is 10V per division.

While that 1kHz performance looks perfectly fine, the lower two scope shots show how it all falls into a heap at 20kHz. This level on HF non linearity just isn't acceptable!

The output tubes and operating conditions are two EL34's pentode-connected with 470V on the plates, 360V on the screens with a bias current of 30mA each tube, accurately balanced. The grids are driven from a relatively low source impedance of 27k.

However, the poor HF linearity is only half the problem. The transformer has a very distinct, under-damped resonance at only ~55 kHz. The phase shifts associated with this, along with the limited bandwidth (-3dB at 20kHz driven by the EL34's as described) makes it next to impossible to wrap a decent amount of global negative feedback around the amplifier towards the upper octaves of the audio frequency spectrum whilst maintaining stability.

 

[GK]

Well there's your capacitance. Why on earth would they interleave them in parallel?

I don't think there is a capacitance issue with that aspect of the construction. The major capacitance issue is primary winding one, associated with the massive number of turns required to achieve a 313H inductance.

Thinking about it, I think the reason the secondary is interleaved in parallel sections is because otherwise (if done so in series) there wouldn't be enough turns per winding section to fill a complete layer. That would, I guess, make for a poor coupling factor and a high leakage inductance. A 10k - 16R transformer has a rather high turns ratio after all.

If my reasoning is correct it seems to me a clever technique. Maybe a transformer expert out there can chime in?

[SeanB]

3 in parallel will enable use of a thinner winding wire, possibly one that was in stock in large quantity but which otherwise is not commonly used, or it was otherwise too thick a wire for the winding machinery to handle happily.

[GK]

I'm sure resistive losses play a role, but I'm also pretty sure, considering the numbers, that the real reason is to have a sufficient number of secondary turns so that there are complete secondary layers to interleave with the primary winding; thus maximising mutual coupling.

Now I have a plan. I've decided that I'm going to rewind the transformers for a lower primary impedance, of 8k instead of 10k. This will yield a slightly higher power output from the amplifier, but still within the capacity of the power supply. As the primary impedance is raised, it becomes exponentially more difficult to keep a limit on things such as leakage inductance and winding capacitance and thus achieving and adequate high frequency bandwidth.

8k versus 10k may not sound like much, but it makes things quite a bit easier overall. Secondly, I will not aim for a stupidly high primary winding inductance - for the reasons already adequately detailed. For the interleaving, I will use the supposedly established standard of 5 primary sections and 4 secondary sections. I will use the parallel winding method for the secondary to maximise mutual coupling with an adequate number of turns. I'm also going to retain the pentode configuration of my power output stage, with the fixed DC potential on the screens, so I will not be be incorporating Ultra-Linear taps into the primary winding, as they thefore have no purpose. 
 
I will initially wind a primary inductance of 100H. The only real, vaguely questimated variable at the moment is how well this plan will juggle the ratio of leakage inductance to the primary winding capacitance. It is these two elements that determine the HF roll-off of the transformer. The leakage inductance forms a 2-pole (12 dB per octave) low pass filter with the winding capacitance. For optimal squarewave/ step response, you don't want this filter response to have an excessively high Q and thus overshoot/ringing. You don't want it to be overly damped either. A Q of 0.71 (Butterworth) would be ideal (as Mr Beijner mentions in his paper), but IMO anything in the range of 0.5 to 0.71 would be perfectly OK.

I may very well end up winding and scrapping a few iterations before I wind up with a transformer that I am happy with. I'm going to resort to a little carpentry and metal turning to knock up a hand cranked winding jig that supports both the bobbin and the wire spool. I have a plan in my head that will be pretty simple to make. So far I have raided my junk box and come up with a pair of these (lovely vintage    ) "Silver King" mechanical counters (see attached), that advance one count for each full turn of the input shaft - perfect for counting wound turns! I also have a nice stash of high temp yellow transformer tape that I use for my SMPS ETD's.

Also before I can start calculating required turns and winding my transformer I'm going to have to workout the Al value of the core as I really have no idea who makes it and don't have the original manufacturers data. I think the easiest way to do this will be to wind a full layer of (counted) turns onto the bobbin, reassemble with the E-I's and accurately measure the inductance. Knowing both the number of turns and the inductance value I'll then be able to compute the Al value.

However all this will have to wait until next weekend before I can get started. Maybe some out there find all this interesting, I dunnno.


 

[hazard]

Hello, I know this thread was long time inactive, I want to know if this project still active. if I can help you with the design of the audio output transformer, you can consult me, i've design & construct my actual in use stereo amplifier, two years ago, i design and construct the output transformer too, and i document the project for a spanish electronic forum, if you wish, you can take a look to that paper.



sorry for my really bad english, i'm from argentina south america.

[GK]

Hi,

This is still a project sitting on the shelf, waiting until I get around to it again. I very much lost interest in audio a while back and moved onto other things, though I still do intend to rewind those Edcor transformers, eventually. I do appreciate the offer, but I'm much to busy with other things right now to correspond about audio transformer design in a meaningful way. I still have future plans to build a winding machine, but right now my workshop is in a state of moderate disarray as right now I'm pulling down this little secondhand EFI alloy chevy for a >500hp carburetor induction rebuild to make my 260Z coup track car go a little better. Also have to finish dummy fitting it and ancillaries into the car, cutting bonnet for the towering Mr Gasket air intake scoop, etc. No room right now for homebrew transformer winding machines 

   

[calexanian]

If you can stomach the price I have had good luck with these transformers. No HF problems here!! I typically will wind my own but have used these as replacements in old sun amplifiers which used Dynaco transformers and have been impressed.


http://triodeelectronics.com/dytr.html

[GK]

Compared to some other suppliers their prices actually look reasonable.