Home Brew EL84 Output Tx Test Results

[trobbins]

Insight into the performance of the laminations/core may be more easily gleaned by low frequency response, and how that changes with feedback level, and power output.

Feedback, as a simple dB number is almost universally based on the drop of mid-band gain compared to when no feedback is applied. So simply use a mid-band sinewave of circa 1kHz, and with R37 disconnected (ie. no global feedback) set the input level to keep the output level suitably below max output - then reconnect R37 and measure how much the output level drops (perhaps to 10% for 20dB nominal feedback).

Knowing the feedback level can allow comparison to what the original amp provided, to confirm your transformer is actually acting as it should at mid-band. It can also put in context why you may see any minor peaking of the bass frequency response.

Your testing at 5W could be redone at the maximum output prior to clipping at midband (1kHz), and then take a few output power measurements as frequency is stepped down. That may then indicate when the core is losing its ability to pass power. If you can get close to 17.5W mid-band then you have a direct comparison with the original as it gave 17.5W flat down to 20Hz. That original power performance down to 20Hz was also with similarly low THD, so that is another comparison related to your core.

[kalee20]

I agree with trobbins - and as per my Post #19, the NFB is cancelling out any distortion anyway, which makes it harder.

I'd suggest:

1) Disconnect the NFB completely.

2) Set the signal generator level to give 50% of maximum power at 1kHz, and do a frequency response sweep.

3) Connect your distortion analyser, adjust the signal level for 5% distortion, measure output voltage and calculate power. Repeat over the entire frequency range, to give a graph of Pout against f, at 5% distortion.

Swap laminations and repeat 2 and 3.

My prediction is that there will be negligible difference at the HF end; that the GOSS will give a better frequency response at the lower end (thanks to its greater permeability giving more inductance, hence less bass loss); and that power output will be greater at the LF end with GOSS (thanks to its higher saturatioin limit).

As stated earlier, it there is a significant LF roll-off at a higher frequency than the transformer,'s roll-off, it will dominate the situation and you may never hit the transformer's limitations - whatever core material you use! So raise the coupling and decoupleing capacitor values to expose the transformer.

[Valvepower]

Hello,

I’ve just remembered I have an ST-35 I came by quite some years ago before prices went gaga… Being a ST-35, it has the 7247/12DW7 driver instead of the 6U8/ECF82 as used the STA-35, which up until now, I didn’t know.

If you need some comparative measurements, I could retrieve it from the outside workshop and get it on the bench and power her up, which I can see any problems with though, as it had been serviced with new capacitors etc., however I’m not entirely sure how much use these measurements would be with the 7247/12DW7 driver as we wouldn’t be comparing apples with apples.

Terry

[retailer]

Finally found some time to try the test amp with the negative feedback disconnected, I should mention the original Dynaco circuit runs the EL84 anodes at 365V whereas mine is closer to 315V, the chassis I used already had a PT wired up to a 5Y3 for the power supply which gave 300V - I removed the 5Y3 and wired in a pair of silicon diodes to boost the HT as much as I could, the end result is a lower output power at around 12.5Watts rather than the Dynaco's 17.5Watts.

Output voltage was measured at just under clipping (10Volts) into an 8ohm dummy load with a 1Khz sinewave, procedure was to run the input signal up until I detected clipping and then back off until no clipping detected, unplug the input cable and plug it into my audio millivoltmeter to get the input signal level.

No feedback - 10mv for full power 10 Volts output
45K feedback resisitor 115mV for 10 Volts output
27K feedback resistor 200mV for 10 Volts output

After a bit of googling I came up 22dB of feedback with the 45K resistor and 26dB of feedback with the 27K resistor - this seems like a lot, I'm unfamiliar with negative feedback calculations, most of my work with amps has been with repairs and transformer rewinds, so I may be wrong - please correct me if I am.

With the 27K resistor in place I did a quick frequency response test - no real change to the bottom end response, the top end however improved, the -1dB point was now around 27kHz.
The three scope pics are square wave at 100Hz, 1KHz and 10KHz with 27K feedback resistor, gave it a short listening test, sounded fine, the next stage is the try the M6 laminations.

Valvepower:- Thanks for the offer to get some ST35 measurements, I do have a SCA35 if I need to compare, and I may do that once to M6 laminations are in place.

[retailer]

Small update on the transformer - swapped out the laminations today and as the bell ends I had on order arrived I decided to treat the transformer to a 'cosmetic' upgrade. I powered up and connected a dummy load and scope to the output, first thing I noticed is more output, I now had just on 12 volts into 8 ohms at clipping, not 100% sure where the extra output came from but I'm assuming the M6 laminations have lower core losses, it was a pleasant surprise. I did a quick distortion test at 1KHz, first at 5 Watts and then at 10 Watts - just under 0.1% in both cases. I should have carried on with more testing but I got a bit carried away and started listening to some music, I may have just been fooled by the extra output and imagined it but I think it sounded different - better. Over the next few days I'll endeavor to do frequency response, distortion and square wave tests and post the results.

[retailer]

In the previous post I mentioned I had a bit of extra output - in the workshop today I noticed that my dummy load was set to 16 ohms, is this where my 'extra' power came from, I should have realised - to confirm I powered up and checked, just over 13.5volts into 16 ohms is approx 11.5 Watts - switch back to 8 ohms and the output dropped to 10 volts into 8 ohms which is 12.5 watts.

[Valvepower]

I’ve been there…

I feel this is a function of the output resistance; this is obviously ignoring the effects of any mismatching to the output valves etc.

This was one of the problems in the recent ‘EL84 monoblocks’ thread and the oddball Armstrong with the ‘unusual’ OPT125 output transformer and determining the actual output load impedances.

Terry.

[trobbins]

Some output stage datasheets show the variation in output power for a changing load resistance - eg. the RCA 6L6GC 1960 (https://frank.pocnet.net/sheets/049/6/6L6GC.pdf)

[Ed_Dinning]

Possibly of more interest in this application might be a plot of % distortion with changing load resistance.

Ed

[Valvepower]

Hello,

I took the 6L6 data and extend the graph load (x2) and distortion axis to see what the distortion could be with a higher load.

I’m making umpteen assumptions here as its just my ramblings, however it gave me a rough idea of the rise in distortion with the increased load.

This concurs with Edwards thoughts of measuring the distortion.

Attached my scribble.

Terry

[retailer]

For those that are interested I've just done a few tests on the transformer with the M6 laminations installed - I found an online source of log graph paper so sketched up a quick frequency response graph which is good enough to get a good idea of the response, not sure what causes the +2.4dB rise that starts around 35Hz and peaks at 20Hz, and I'm not concerned about it I seriously doubt I'm going to hear it. Also did a few distortion tests with my home made distortion meter:-

1W 1KHz - 0.15%
10W 1Khz - 0.1%

1W 100Hz - 0.2%
10W 100Hz - 0.28%

10W 10Khz 0.6%

It's been a learning experience for me which I think shows a half reasonable output transformer can be made using ordinary laminations, however there is a worthwhile improvement with the M6 laminations, for now I've finish with the testing, I'm happy enough with the transformer as it is and will use the test amp as a music source in the workshop until I find time to make a second transformer for my ST35 amp project.

[Gabe001]

Is it just me, or do the square waves in the former tests (with the ordinary laminations) look better the latter ones with the m6 laminations?

THD is broadly the same, so what's the benefit? Am I missing something?

The low frequency peak is due to the feedback for low end frequencies being limited by the output transformer frequency bandwidth (and some other circuit design choices which act cumulative ly, but let's omit these for now). It is usually cured (if you want to) by placing an RC filter with a -3db point frequency higher than that of the output transformer outside the feedback loop, generally at the input.

[kalee20]

I really wouldn't expect there to be much benefit between M6 GOSS and utility transformer iron, with the same winding, except right at the bottom end. That's where the magnetic flux in the core will be at its greatest, that's where the core losses will be greatest, and that's where the poorer saturation characteristics will give the more distortion.

At 100Hz, flux will be half that at 50Hz, so it's not surprising that distortion, even with the bog-standard iron, isn't an issue (there'll be more from the valve characteristic curvature). You really need to measure distortion at progressively lower frequencies, at the same power output, for the two core materials.

I'd certainly agree you can make a good O/P transformer with ordinary laminations. You can make a better one with grain-oriented silicon-steel laminations, and an even better one with strip-wound GOSS (such as a C core) but the improvement will be of the order of a few tens of percents, not orders of magnitude.

[retailer]

While it doesn't really look as if the GOSS laminations are much better, there is in fact a difference that makes it worthwhile, - less distortion - most of the distortion measurements on the standard laminations were done at around 4.5-5watts as this is where I figured most listening is done, looking at the notes I made at that time I did do 100Hz and 10KHz at 10 watts and the distortion was up to just on 1% at 100Hz and over 2% at 10KHz 10watts with the standard laminations - the GOSS laminations have 0.28% distortion 100Hz 10Watts and 0.6% 10Khz 10watts.The low frequency peak started at a higher frequency - around 65Hz - and had a greater amplitude at approx +3.8dB, the GOSS laminations have the peak start at a lower frequency and with a lower amplitude.

While there is every possibility I'm being fooled I do believe the GOSS laminations sound better, based on listening tests with a single Realistic Minimus 7, while it certainly isn't a HIFI speaker it also is not total rubbish, this morning I carried the amp inside and played it through one of my VAF DC7 column speakers for a short while, I'm now looking forward to building a complete stereo amp.

I wind output transformers for a local amp maker - when I started winding for him he was making PP 6BM8 and EL84 amps - over the space of a few years he now almost exclusively does SE triode amps, I know the early amps used ordinary iron as he used to give me the lamination stack so I could make a winding tube the right size, lately he's been giving me just the hand made bobbins, but I'm sure he is still using ordinary iron. I've listened to his PX25, and 2A3 amps and they do sound very nice through his 2 way speaker systems - 12inch Electrovoice bass coupled with Celestion horns. The winding layout is nothing fancy 1/4P - S - 1/2P - S - 1/4P, the secondaries are both wound in the same direction and connected in parallel - he doesn't use negative feedback.

[Radio Wrangler]

Grain orientation improves magnetic properties for flux in one orientation at the expense of worse performance for flux in the orthogonal direction.

GOSS comes into its own for magnetic devices where the flux-path is one-dimensional. In tape-wound cores like toroids, R-cores and C-cores it is at its best.

In flat laminations like E and I lams ot U and T lams, the flux path always has a fair proportion at right angles to the optimised direction. If the worsening transverse to the optimum direction is less than the improvement in the optimum direction, there can be a net advantage even with flat lams, but it seems a shame to not be playing GOSS to its best advantage.

Even with GOSS, thinner strip will impede circulating eddy currents in planes perpendicular to the flux flow, hence lower loss, especially at higher frequencies.

David

[Gabe001]

Thanks for taking the time to explain. As you say, the distortion figures have improved and probably so has the low frequency extension. Can I ask broadly what the price difference is between Goss and standard ordinary iron laminations?

[kalee20]

Originally Posted by retailer

The low frequency peak started at a higher frequency - around 65Hz - and had a greater amplitude at approx +3.8dB, the GOSS laminations have the peak start at a lower frequency and with a lower amplitude.

That's largely as expected. The reason for the peak, is you've got several roll-offs happening at similar frequencies - intervalve coupling CR networks, and the output transformer inductance and load resistance giving an RL roll-off, and they're all within the negative feedback loop.

More than a single roll-off gives a phase shift greater than 90° so the feedback starts to become positive - hence overall gain rises.

The higher permeability of the GOSS gives you more inductance in your O/P transformer, so the RL roll-off happens at a lower frequency - as you found. And, because the roll-offs are now separated in frequency a bit more, the peak isn't as high.

You could have reduced the peak rise by playing with the coupling capacitors anyway, that's simple to do. But your extra inductance, giving a response extending to lower frequencies, is the benefit of GOSS. I did predict that the benefit would be seen at low frequencies - I kept my fingers crossed that I wouldn't be shown wrong - and at least your measurements bear this out!

I'm now thinking about distortion, however, having NFB wrapped around the whole thing is muddying the water as it's nullifying the non-linearities anyway.

[retailer]

I should put my Dynaco SCA35 on the bench and see if it has a similar lump in the frequency response curve, but Xmas is looming and my wife is already starting to panic - oil the deck, paint the screens, wash the windows etc. so possibly something for the new year.

Originally Posted by Gabe001

Thanks for taking the time to explain. As you say, the distortion figures have improved and probably so has the low frequency extension. Can I ask broadly what the price difference is between Goss and standard ordinary iron laminations?

The EI-76 GOSS laminations were $15.50/kilo + shipping or around $22 for 1 transformer, almost 1/4 of the cost was for shipping - the ordinary laminations I have are recycled from mains transformers so a bit hard to compare, I wouldn't call the GOSS laminations expensive, total cost of all materials for one transformer would be around $40.