814 SE Amplifier: Custom Output Transformers

 Improving the 814 SE Amplifier

After more than a year listening to this fantastic amplifier, it was time to do the first significant improvement to it despite I resisted to modify it after so much work and effort put into the design and build. The evident upgrade was the output transformer. When frequency response was measured, it was evident to see that the HF response was lower than expected. This is the result of the transformer and its configuration in this circuit. The LL9202 is a better OT for higher impedances and in this circuit, it is used in the 6KΩ /8Ω mode.

The higher leakage capacitance and inductance of the LL9202 made this amp to roll off early below 20kHz. It sounded very nice to me, however I knew it could sound much better if this were to be improved.

My friend Paolo – who I consider an erudite in valve amplifiers – suggested he could help out by designing a custom OT for this amp. The price wasn’t cheap, however it was a bargain if you consider value for money. I would have paid much more for a branded transformer meeting these design specs though!

Paolo summarise some of the above points clearly here:

[…] Lundahl, like many others, don’t specify frequency response because they don’t use the standard Japanese method (which consists in a source impedance equal to the primary impedance of the transformer) because this makes it look worse at low frequency, I guess, and don’t built most of their transformer for specific devices. They only give leakage inductance which in not enough to determine high frequency response. One needs to know the total capacitance as well. This can be estimated (as I have done) or figured out precisely from resonance frequency measurement. Once you know the leakage inductance and the resonant frequency you know the capacitance.[…]

After some discussions with Paolo about this, he came up with two options. As we know is all about balancing size, power loss, frequency response and cost. It was promising enough to risk and go for this adventurous upgrade.

The transformer design

The design of the transformer was carefully undertaken by Paolo. I was aiming for a FR up to 25-30kHz at least whilst balancing the low frequency response and power loss:

Paolo wrote:

” […] The main point about the OT was simply getting a more balanced result in comparison to the Lundahl. I think the Lundahl are more balanced designs when used for the higher and intermediate impedances (e.g. 23k and 11.5k in you case).

OT design, as always, is about trade-offs so I traded some size and some inductance for lower power loss and extended high FR. I believe your LL9202 has some 45-50H vs. 40H of mine, although impedance is a bit higher. This case also shows that a C-core and toroidal core don’t make so much difference regarding high frequency behaviour (leakage inductance in particular) without an optimized winding scheme which is always the most important thing.

Your transformer has the following main specs:

1) 5.9K primary impedance for 8R

2) 40H inductance (33H minimum for small signals, i.e. usually 3-5V at the primary)

3) Design FR 8-25 KHz (-1 dB)

 4) 0.3 dB power loss (using your measured values for DC resistance)

[…]

Measuring the transformer

Before fitting the new heavy-weight transformers, I took some time to do some basic tests to derive the transformer key parameters. I measured all parameters (i.e. resistance, primary and secondary inductance and step down ratio) before proceeding to measure the leakage capacitance and inductances with the secondary grounded:

 

The challenge was to model the complex impedance of this transformer. I followed Morgan Jones’ recommendations as before which gave me reasonably accurate results as I only wanted to characterise this OT. Best way of testing the FR is with a similar circuit as what is used in reality as you want to emulate the anode resistance, current and signal drive as well.

You can see a couple of resonances at above 30kHz and 60kHz.

The first resonance is somehow approximated with 2,500pF/10mH leakage values which should give at least -3dB at 50kHz with the 814.

If instead we try to fit to the second resonance of this complex impedance:

The model fits somehow close up to 100kHz and above response is due to the testing setup as I didn’t worry to use short/twisted cables when measuring this response. Either way, the 6mH of leakage inductance and 1,000pF of shunt capacitance is quite promising as would indicate a good response up to 80-100kHz at least.

Here is Paolo’s analysis of this situation:

[…]Regarding HF frequency behavior you have gone high enough to look at whole story which usually you don’t see from manufacturers measurements. The series of peaks and dips is normal! It is normal because the total shunt capacitance is mainly relevant to the first resonance. At higher frequencies, as the transformer is sectioned ans stratified, you have their relative (lower) capacitances redistributed and so you get other resonances at higher frequency. I think in a normal frequency response graph you usually don’t see them they don’t go up to 1MHz and because of the 8R resistive load. […]

[..] Even with secondary grounded and resonance at 40KHz I think it will be substantially better than the LL9202. I am confident HF response will be no worse than -1 dB down around 22-25 KHz and -3 dB down around 35-40 KHz with a good damping of resonance itself.

It is not really straightforward to predict exact HF response with a specific valve because, in addition to parasitic elements in the transformer and even the circuit, I think at those high frequencies the plate behaves like a true complex impedance rather than a simple resistance. I have seen many times the HF response improving just switching from triode to UL configuration or viceversa, with same valve! No preferred solution for triode or UL of course. It depends on the specific case. Actually even with the EL84 in pentode mode I have been able to get 100 KHz without any fb while switching to triode it went down to 60-70 KHz. Of course the slope of the roll-off itself will be a bit different as well. […]

 

So, despite not achieving my modelling objectives, I settled for the second option in my simulations, albeit I’m aware that it will be a bit optimistic in the HF response:

 

I did a quick FR test on this transformer like this. I used my audio tester interface and a CCS to provide 100mA of quiescent current. Not precisely a real life test but good enough. The challenge is that the CCS shunt capacitance is unknown and in the region of 500-1000pF. The flat response up to 35kHz is due to this total unknow shunt capacitance which I also modelled below:

So let’s take the worse case and assume the CCS has a shunt capacitance of 500pF. Then the OT total shunt capacitance should be around 2,500pF which matches the first resonance model derived earlier. In this case then the response of the transformer should be up to 50kHz. Much better than the Lundahl even if we add any extra parasitic capacitances!

Upgrading the amplifier and listening tests

It didn’t take long to replace the OTs despite some mechanical adjustments to the fitting. This is te advantage of my breadboarding construction! No chassis drilling, etc.

The adjustment of the circuit was a bit of a pain. The lower primary resistance of the OT forced me to change the bias points of this DC-coupled design. I ended up reducing the HT level as I was running out of headroom in my SSHV2.

I fit a new pair of NOS 814s and a matched pair of Sylvania 46s and calibrated the entire circuit again. I haven’t touched it for 12 months, so it was a good opportunity to do this.

I was astonished by the new treble. I was wrong before suspecting that the improvement wasn’t going to be considerable. Particularly when playing  vinyl as my MC stage can perform well above the 20kHz range. You can notice this. Cymbals, drums and brass upper register is noticeable. The amp is better balance from bass up to high treble. The detail is there, the clarity is wonderful.

I’m very happy with this result and am looking forward to bring down from the attic the testing gear at some point to do a new frequency response of the amp with this configuration. But for now, I will enjoy the sound of it…

Ale

Custom transformer supplier

If you are interested in getting a custom transformer like this yourself, you can contact directly Mr. Vincenzo using the form below or via his website (in italian only).