It was time to take the 814 SE Class A2 amplifier measurements. The challenge though, is that the amp is so heavy that I will never take it up to the workshop. Therefore, I decided to take my workshop PC down from the loft this time to see how the 814 really responded.
First test was to do a THD analysis as a function of the total output power. As you don’t want to do this with your speakers, and also the classic wire wound resistors (Alu-clad) are inductive, you want to use non-inductive resistors like this test jig:
I bolted on to a large heat-sink an array of resistors to form 8Ω in value by using a pair of 10Ω in parallel and three 1Ω in series. I added a set of binding posts and connectors for the speaker cables. this way you can easily wire your speakers and connect your audio test set to take the measurements.
The first measurement taken was the THD response versus Power. A tricky test as I didn’t want to blow away the crowbar protection when asking the amp to do more than it should. I nearly did, though as wanted to see what the response looked like when 15W or more were squeezed out of this amp! It would be an expensive test, so I desisted from doing any further adventurous sweeps:
As you can see the amplifier measures as good as it sounds. Distortion is really low despite it operates mostly in class A2. The A2 breakpoint is around 3.5W and you will see in a harmonic plot that higher harmonics kick in at this power.
The way I biased this amp was restricted by my HT supply which was a legacy of my previous 4-65a amplifier. However, when playing with the operating point, I found out that the 814 not necessarily will perform better when increasing the anode voltage (and reducing the anode current). The 814 is extremely linear at low grid voltages (providing it has an optimal driver for A2 grid current) whereas pushing it harder at higher voltages will increase the distortion as grid curves get further crunched at the bottom righthand corner of the diagram below.
Nearly at 12W is when the curve crunching becomes more evident and distortion grows exponentially. In my case above 1% whereas in the ideal simulation it was predicted above 2.5%. There is a nice harmonic cancellation taking place between the 46 and the 814 which reduced significantly the THD of this amplifier.
Now there is a dip in the H2 level that occurs at around 11W where in the simulation is predicted to occur at around 14W. It may be due to harmonic cancellation the difference shown. What is interesting to see is the exponential growth of higher harmonics, as predicted it’s reflected in my measurements:
Looking now to the frequency response, we can see that the 814SE behaves quite well from very low frequency at 5.5Hz to 17kHz. A bit low on the HF side but I’m resisting to add anode-to-anode feedback to improve BW. There is an extra peak at 30kHz due to lack of Zobel however I can’t notice this during listening to this amp.
When looking at where the HF drop cause was, I checked the 46 DHT driver and found that the response is very nice from 7Hz to 48kHz with no problems. Therefore my conclusion is that the output stage could be improved if OT was optimised
Finally, I’m pleased to measure what I hear. The 814SE performs astonishingly well for a class A2 amplifier. It’s rare to see this low level of distortion in class A2 for a single-ended amplifier.
Ale,
Great work and excellent THD results.
How would you compare the sound of 814 to the sound of 4P1L ( final stage)?
Is it worth the effort with power supplies and high voltage, and all the other features that adds to the complexity of this amp? I like that the 814 is reasonable priced. Do you know what is the lifetime of the 814?
I am surprised of the HF response considering the OT you used.
Best,
Radu
Thanks Radu.
Hard to answer your question and can’t do side by side tests at this moment so it won’t be a fair comparison. It is also very subjective view, so I rather don’t make such statements.
The cost of the valve is peanuts compared to the iron of this monster.
I’ve been playing for many hours this amp since November last year and so far so good. I will expect the 814 to be reliable, albeit I can’t recall coming across the reliability data on the datasheets I have. Will check.
Ale
Hey Ale, nice results. That extra peak is unavoidable, it is a bit low in frequency though and not damped (I assume it is a 5K OT with 1.6K source impedance). The combination of leakage inductance, capacitance and source impedance is not optimized. What is it? With a different winding geometry, while keeping the same turns and ratio, even with the same resonant frequency, the response could smoothly roll off above the resonance. Zobel networks are not good at output stage, IMHO. I don’t like them at all but here is the worst place, really.
Cheers
Hi Paolo,
Thanks. The OT is wired on 6K5. I thought the same when I remember our discussion of the 814 in PPP and the OT requirements. For now, I will leave it as is. It does sound great.
I’m not keen on Zobel networks anyway so haven’t added any of them. I actually removed the one from the IT as wasn’t properly measured. I used ideal values and had an HF loss as the peak was over-damped. I was planning to add a variable capacitor and a pot and measure the response when I just tried the sound of it after removing it and did notice the return of the treble lost there. The graphs above confirm there is no peak produced by the IT and the 46 driver.
Cheers
Ale
Oh yes the driver is absolutely fine. The input impedance of the 814 does the job
Wow, beautiful presentation. I am happy with the 4P1L but then that is with 104db horns. If you would like a better OPTs I might be able to put some together for you.
Thank You. It all depends on specs and price, of course. Let me know