Home Brew EL84 Output Tx Test Results

[retailer]

If moderators feel this belongs in the audio section then feel free to move it.

This is a follow on from these two posts.
https://www.vintage-radio.net/forum/...d.php?t=205313
https://www.vintage-radio.net/forum/...d.php?t=205132

I've done a few tests on the EL84 PP output transformer I've just wound, as the Tx was already stacked with ordinary power transformer laminations I left it like this for the first test, the idea of the test is to compare the 'ordinary' laminations against M6 GO laminations using the same windings, same circuit and same component values, I don't doubt parts of the circuit could be tweaked for better results with the ordinary iron but that is not the aim of the test/comparison, I did the test at what I consider ordinary listening - around 4.5 - 5 watts. The test amp is mostly built with salvaged hardware and components - EL84's are Russian Reflector and the 6U8A is a used Minwatt - max power out is just on12.5watts, the circuit I used is from Dynaco and calls for a B+ of around 380V, my B+ is a bit down on this at 320V.

Sinewave frequency response was actually quite good considering the size of the transformer and the type of iron used, the response was quite flat from around 65Hz to just over 23Khz - the -1dB point was just under 24Khz and the -3dB point was just under 30kHz, below 65Hz the amplitude of the output waveform started to rise progressively as the frequency was lowered - going from 6.2V at 65Hz to 7.5V at 31Hz and then dropping back to 6V at 17Hz, it was still 7.2V at 21Hz.

Square wave test was done at the same output level around 4.5 - 5 watts - 3 tests were done 100Hz 1Khz and 10Khz - not ever having done a square wave test before I'm not quite sure what to make of the scope traces.

Over the next day or so I'll redo the square wave test and expand the trace to get a better look at the ringing, and also hookup my home made distortion meter and get an idea of the distortion at 100Hz 1Khz and also 10Khz

I put an ordinary cig lighter next to the transformer in one of the pics so forums members can gauge it's size.

[Radio Wrangler]

OK the first squarewave picture.

100Hz There are three features. The slope on the top and bottom show there is low frequency roll off The amplifier is AC coupled and has a transformer at the output, so there are several timeconstants all rolling off at once. The slope on the tops and bottoms isn't just the exponential decay of a single order roll-off, it is the composite of several.

There is a fast raise and fall time on the edges, The digitising scope won't show it well until we're going a bit faster.

There is a little overshoot on the leading edges, both the negative going transition and the positive going one. Again, the digitising scope won't show it well until we're going a bit faster.


The middle picture,1kHz squarewave:

By now the LF roll-off has all but disappeared. If it had been a simple single timeconstant, it wouldn't have reduced so quickly. So there are multiple zeroes in the response and the eye would have spotted them on a circuit diagram. This all pertains to the low frequency stability margin which isn't much understood by most of the amplifier folk.

We can now see the finite risetime a little better and ditto the overshoot. We begin to see that the overshoot is simply a single one and not an underdamped diminishing oscillation train. We'll see better going faster but this picture told us the overshoot damping was good.

Last picture, 10kHz squarewave:

Now we're getting fast enough sampling from the scope to measure rise and fall times. These are a guide to the high frequency roll-off. From prior knowledge, I'd say that the HF roll off and its attendant phase shift are a bit too low down for you to deploy a lt of feedback to linearise etc the amplifier.... IF this is an open loop response. IF it's a closed loop response, it isn't doing badly at all, but you don't wamt to try increasing the amunt of feedback you'd run into problems faster than you ran into advantages.


We can now see the overshoot and it fits in with the comments on roll-off. As it stands, it would be a very good example if it's closed loop.


Hope this helps a bit. Most undergrads run away from control theory maths and stability analysis so these things aren't general knowledge.

David

[Gabe001]

I'd be pretty proud of myself if that were my transformer. Good FR, no ringing. Looks great, I wouldn't mess around with any component values.

Bargain basement OPTs will show a greater tendancy towards ringing when feedback is applied, have greater transformer power losses and a narrower frequency response

[retailer]

I probably should have posted the circuit for those that aren't familiar with the first 2 threads that started this journey, so here it is.

[Radio Wrangler]

OK, that's the full amplifier loop with feedback operating, C21 providing the overall HF roll-off control as well as providing HF phase compensation of the loop.

So, it's a well-behaved amplifier and the transformer suits it.

Good result!

David

[trobbins]

Were you able to determine the feedback level in dB ?

Your bass response may be showing some feedback induced minor peak (although you only give a voltage level at 65Hz so not sure if that is the same as mid-band). That peak is at 20-30Hz, so likely due to the transformer roll-off, and well above the identifiable RC corners (R33-C18, R35-C19).

Your bass response would change with power level, so may be extended down at 1W, and may retract at your full power level of 12W as that would be dependent on the laminations used.

[retailer]

Not sure on the negative feedback amount, I checked the SCA35 assembly manual, negative feedback is mentioned but no actual figures are given - the Dynaco ST35 assembly manual states it has 20dB - as this uses the same transformer, it may well be that the amount of feedback will be the same. I'm not too fussed about trying to tweak the circuit values to get better performance, the transformer as it stands only has ordinary 0.5mm power transformer iron that has been salvaged from small mantle radios, I was just interested to see how this type of transformer iron compared to the 0.35mm M6 GO iron that I will eventually use.

[Gabe001]

Are you happy with how it sounds, noise floor and all?

[Radio Wrangler]

Negative feedback amount is a graph, anyway, not a single figure. The forwards amplifier gain falls with frequency across some of the audio band. The negative feedback reduces this gain to a more uniform value, and so the amount of gain 'thrown away' by the feedback arrangement falls across some of the band, so the distortion and output impedance worsen towards the high frequency end. Spot values get bandied around, often without even mentioning where the spot is!


David

[kalee20]

Originally Posted by retailer

I've done a few tests on the EL84 PP output transformer I've just wound, as the Tx was already stacked with ordinary power transformer laminations I left it like this for the first test, the idea of the test is to compare the 'ordinary' laminations against M6 GO laminations using the same windings, same circuit and same component values

This is a thread which I'll watch with interest!

Comparing core materials... I think it's an unfair assessment. In order to make a valid comparison, you really need to compare like with like - in other words, the best that can be achieved with utility transformer iron against the best that can be achieved with grain-oriented silicon steel. Which may mean slightly different windings.

After all, nobody would want to compare germanium vs silicon as transistor material by unsoldering an OC71 and replacing it with a BC177 and leaving everything else untouched! Otherwise you'd conclude that the silicon distorts at a much lower output level. You'd want to adjust the bias circuit values as a minimum, and possibly even reconfigure the circuit.

Comparing the two core materials... in the circuit shown, by just swapping lams, I'd say there would be so little difference, that you'd conclude that the extra cost of GOSS is just not worth it.

But if you take into account the higher saturation flux capability of GOSS, you could probably reduce the number of turns by 10% throughout, for the same bass response (first change). And then you could beef up the wire gauge, to then occupy the same window space with the fewer turns (second change).

That will give a reduction in winding resistance and so an improvement in efficiency, so slightly greater power output. And the fewer turns will give lower leakage inductance, and lower winding capacitance - which will then give an improvement in HF response. You'd see squarer edges to your square waves.

Next, look at the circuit. It's clearly a decent amplifier as an entire assembly, but as a means to assess transformers, it leaves quite a bit to be desired:

Originally Posted by Radio Wrangler

OK the first squarewave picture.

100Hz There are three features. The slope on the top and bottom show there is low frequency roll off The amplifier is AC coupled and has a transformer at the output, so there are several timeconstants all rolling off at once. The slope on the tops and bottoms isn't just the exponential decay of a single order roll-off, it is the composite of several.

Exactly - the transformer limitations are diluted by the rest of the circuit.

What's needed is something which exposes as much as possible the Device Under Test - so you'd want the amplifier's internal RC roll-offs to be 10 times lower in frequency than the transformer roll-off - you can then be sure that the square-wave 'droop' is almost entirely down to the transformer. The same of course goes at the HF end of the frequency spectrum.

(When you have your optimised transformer, of course, you do it the other way round so that the dominant roll-off is a low-level coupling stage. In that way the transformer never sees a signal to drive it to its limits. But that's amplifier good-practice, whereas this thread is all about the transformer).

[G6Tanuki]

I'm watching this with interest too; I need to make up a transformer to hook a couple of 6BW6 or EL84 to anode-modulate a 6146. Not really interested in audio bandwidth, it only needs to handle 300-3000Hz, my bigger concern is the effect of the static DC through the secondary and whether it will cause the core to saturate.

OTOH saturation does limit the peak positive drive, so might actually provide a degree of 'speech processing'!

[Bazz4CQJ]

Could someone just give us amateur amateurs a bit of guidance? Am I right that what's been looked at here is how good or bad the amp is with ordinary steel laminations from a power transformer, and that the big issue which might occur is poor high frequency response? An audio quality transformer would have more suitable lamination materials?

But the results reported seem to be better than might have been expected??

And moving on to the modest needs of a modulator, the results being discussed seem very encouraging?

B

[kalee20]

Originally Posted by Bazz4CQJ

Could someone just give us amateur amateurs a bit of guidance? Am I right that what's been looked at here is how good or bad the amp is with ordinary steel laminations from a power transformer, and that the big issue which might occur is poor high frequency response? An audio quality transformer would have more suitable lamination materials?

You can make a decent output transformer with low-cost transformer laminations. It's just easier with higher grade metal.

Really, the benefits of better material apply equally whether you're making a 50Hz power transformer or an audio output transformer.

A transformer is always going to have losses. If you want to reduce losses, you use bigger geometry, which allows thicker wire (lower winding losses) and you run the core at a lower level of magnetisation (which reduces core losses).

But you can also reduce losses by using a more exotic core material, which is lower loss at the same, or even greater, magnetisation. That's what the OP is aiming to do with grain-oriented silicon steel. The downside is, it's more expensive.

For extreme situations, where size and weight are at a premium (such as aircraft installations) you might use cobalt-iron alloy, which saturates at 2.1 tesla (compared with 1.5 tesla for commonplace iron). That allows the turns to be reduced by 30%, and wire gauge to be correspondingly increased. But it costs 8 times as much.

The number of turns for a given size is determined by the voltage and lowest frequency you want to use the transformer at. Once this is fixed, high-frequency performance depends on clever winding design, such as interleaving primary and secondary. It doesn't usually matter in a fixed-frequency power transformer, and won't matter in G6Tanuki's modulation transformer, which only covers a frequency range of 10:1 - whole primary followed by whole secondary (or vice versa) will be fine. But in a hi-fi transformer where you might want four decades of frequency coverage (three for 20 - 20,000Hz plus margin beyond the extremes so that negative feedback instabilities can be avoided), attention to winding geometry will certainly be necessary.

There's more to it than that - for instance, material permeability may enter the picture (if the number of turns to avoid saturation doesn't give sufficient inductance). But hopefully the above will give a guide.

As for single-ended situations, or modulation transformers where significant DC current flows in a winding, quite a different set of calculations needs to be applied, and an air gap added (in this case, material permeability doesn't usually matter because it's decimated by the air gap).

[Ed_Dinning]

To add to Kalee's comprehensive summary.

You may also find that a normal Stalloy has quite a sharp knee on the saturation curve. This is apparently liked by instrumentalists due to the distortion it adds.
On the other hand GOSS material has a smoother knee which reduces these effects (as will "C" cores), so they are preferred for Hi Fi applications

Ed

[Kentode]

I'm sitting in the back row, next to Bazz! Good Ol' Stainless Steel? Nope, Grain Oriented Silicon Steel!

[Radio Wrangler]

Grain oriented silicon steel has, unsurprisingly, directional properties. It doesn't work to its best advantage as E and I laminations or as U and T. In both these the flux flows East-West in some places and North-soutt in others. There is no optimum choice for the orientation.

GOSS comes into its own with tape-wound cores like toroids and C-cores. also R-types. Where the flux flows always in one orientation.

The lady is due to come round with the Butterkist and Kia-Ora any moment.

David

[retailer]

I managed to get some time to do a distortion test at an ordinary listening level of around 5watts - my spot frequency distortion meter is, with a little care it can read down to .01%. Distortion measured was:-

100Hz - 0.28%,
1KHz - 0.13%
10KHz 0.37%

Quite acceptable all things considered.

I did say I wouldn't play around with the amp component values, but I couldn't resist the temptation to play around with the obvious - the feedback resistor R37, I wired a 50K pot and a 27K in series in place of the 56K and setup distortion measurement again 5 watts @ 1KHz - it was quite easy to get this down to less than 0.1%, but I know there are trade off's so I left it at just on 0.1% and then had a look at the 1KHz and 10KHz square waves, they are different as expected but not by a massive amount, R37 was now 45K against it's original 56K. Played some music through the amp it sounded fine but a bit hard to tell with only a pair of Realistic Minimus 7's to listen to.

The test results have been a bit of a surprise for me and also has got me a thinking a bit, did I really need to buy the M6 laminations, I'm now hoping that they will show a measurable difference and so prove their cost was worth it.

[kalee20]

Originally Posted by Radio Wrangler

Grain oriented silicon steel has, unsurprisingly, directional properties. It doesn't work to its best advantage as E and I laminations or as U and T.

GOSS is still better than utility-grade iron, otherwise nobody would use it for laminations, but you're right - laminations don't make best use of the material's properties.

Originally Posted by retailer

Distortion measured was:-

100Hz - 0.28%,
1KHz - 0.13%
10KHz 0.37%

Quite acceptable all things considered.

The test results have been a bit of a surprise for me and also has got me a thinking a bit, did I really need to buy the M6 laminations, I'm now hoping that they will show a measurable difference and so prove their cost was worth it.

Your test results are commendably good , and TBH I wouldn't expect much difference with the M6 GOSS.

Reason? Firstly, unless you recalculate and rewind your transformer to make best use of the GOSS properties, you're not going to get their benefit.

Secondly, you have negative feedback around the amplifier, which will straighten-out the transformer's shortcomings, as it's inside the loop. It could be, for instance, that the closed-loop section is absolutely perfect, in which case tweaking the transformer will make no difference whatever - and the distortion is arising entirely within the first stage.

Unless you open the loop for your measurements, you're never going to know.

Good design does seek to make the response, distortion, etc as good as possible, leaving the NFB as the icing on the cake. (NFB can actually increase higher-order distortion, and particularly intermodulation distortion).

[Valvepower]

Hello,

Sorry, I’m late to this one, anyway the nice usherette has just showed me to my seat using one of those torches with the directional lens…

Some of the 'better' guitar amplifiers like the Hiwatt used Partridge transformers with GOSS laminations with a bit more of a sophisticated layering scheme.

I was recently looking at what I’d call 'work-a-day' valve output transformers and ended up back at the Mullard 5-10 design book and the Leakey & Gilson article in Wireless World both from the 1950s.

Terry.

[retailer]

I think a bit of background is needed for those that missed the first 2 threads that lead to this one.

I had a Dynaco ST35 amp copy build in mind and found the output transformers to be quite pricey so I decided to wind a pair using the same winding layout as the originals, a bit of research and I came across the winding layout for the Z565 OT as used in the Dynaco ST35, to the best of my knowledge the laminations used are M6 GOSS, the primary is actually 2 complete ultra-linear interleaved primaries connected in parallel, one however is wound in reverse, 6400 turns in total become 3200 turns once the connections are made, the secondary is done in a similar way. I ordered some laminations and went ahead with the wind and while waiting for the laminations to arrive I assembled the transformer with ordinary power transformer laminations - I wanted to do some voltage tests and also check the impedances etc. were all correct, the wind is not overly complicated, but with 16 primary wires to connect there is plenty of room for mistakes and if I had made an error it would have shown up then.

The voltage/impedance checks all came out good, but as the proof of the pudding is in the eating I really needed to try the transformer in an amp to see if was correctly wound, the laminations were still on their way so I went ahead building the test amp, a Dynaco circuit designed around the Z565 transformer, I then thought it might be interesting to do the first tests with the ordinary iron and compare once the M6 laminations arrived, so the idea of comparing was really an after thought.