The Valhalla power amplifier project

[Bandersnatch]

It was clearly *MY* misunderstanding. Sorry about that.

The 'Magic' that SE, DH Triode users claim as their own is the sort of thing possible with conservatively run, lightly loaded, class A push-pull. I was never able to achieve it with a loop NFB path.
cheers,
Douglas

[demogorgon]

a very neat looking chassi you have there jane.

hope the amp matches the casing
i'l be watching your project onward for shure.

-Marius
deeply offended by the bashing comittet toward us norwegians.
the simpsons rule btw..

[jane]

A little progress report on my Valhalla project:

I have completed most of the hardware assemblies for the high tension side of the amp. That side is now ready for final wiring.



On the signal side I have (almost) finished one channel. I have tried to stick to the original wiring plan that I made a few mounts ago, but I had to improvise a little bit, and you always see better solutions when thinks become practical.



On the picture you will recognise the driver stage at the left side and the output tubes on the right hand side. I think the result became quite tidy and well arranged. No cables are fastening with strips yet; so it still looks a little bit like "Polish free-jazz".
I did not use a particular kind of colour scheme for the wiring, but from the driver board, I used green leads to the 6SN7s grids and red for the plates. Ground is black, B+s are red and C- is orange. Bias is green for tube #1 and yellow for tube #2 (sounds like Ringo Starr singing Yellow Submarine: Sky is blue and sea is green… )

I did not want to mix the high tension (560V) with rest of the wiring so I elevated the B+ cables approx 3 cm from the chassis bottom (or top when I turn it around). I made a sort of pillar attached to one of the terminal strips to elevate them. I will also use this route for the screen leads from the OPT.



The plan is to finish the other side next weak, wire the HT-side, put in the Sowter OPTs and do some preliminary tests and measurements--I hope.

[jane]

Final measurements, Valhalla 2x75W power amplifier:

Frequency response 1 watt L/R ch:
20Hz – 20kHz (+/- 0.1dB)
20Hz - 40kHz (-3 dB)

Power into 6.6 ohms load L/R ch:
2 x 24V rms, 2 x 87 Watt rms

Harmonic distortion:
1W rms, 1kHz: -60dB, 0.1% 2nd harmonic (dominant)
1W rms, 10kHz: -56dB, 0,2% 2nd harmonic (dominant)
80W rms, 1kHz: -50dB, 0.32% 3rd harmoniic (dominant)

Global negative feedback (NFB):
17dB (7x)

Signal/Noise ratio:
Less than 2mV rms output, better than -80 dB (ref 75 watt)

Output impedance:
0.5 Ohm, Damping factor greater than 15x (1kHz)

All measurements done with KT88s (50mA bias) except for output impedance measured with KT90s (60mA bias)

Frequency response:


10kHz square wave:


Distortion pattern 400Hz sine, 80W rms output:


A picture of the amp:

[Family_Dog]

Congratulations, it looks great! And I bet it sounds as good as it looks too...!!

-Eric

[Tubes4e4]

Congratulation Jane, looks excellent

One question though:

With a gNFB loop around five stages, although only 17dB and two DC-couplings between stages, is this circuit really (unconditionally) stable into complex loads like speakers?

Tom

[jane]

As far as I have tested, yes. But I have just tested with two different sets of speakers: high efficient (99dB/W/m) semi horn speakers and low efficient (ca 88dB/W/m) three ways book shelf speakers.
The open loop gain is not that high, approx 50dB (6SJ7 triode wired) and there are just two pairs of coupling caps (plus one decoupling cap in the LTP) in the feedback loop.
I had to reduce the value of one of the coupling cap pairs to establish a well-defined dominant pole in the low frequency area.
In the high frequency domain I’ve got ca 60 deg phase margin at 3dB roll off (40kHz). If I increase the forward roll-off in front of the LTP I can have more margins, if needed.

My biggest problem was the OPTs that had some serious resonances at ca 100kHz due to leakage inductance and winding capacitance. The open loop frequency response was not looking promising at all until I put a RC network plate to plate across the KT88s. (you can still see the peaks at 100kHz)