New Amplifier Project

[mole42uk]

Just look what the nice men at VVT made for my new 20W amplifier project (there's a £1 coin on top to give you an idea of size):

[mark pirate]

Nice
I take it that you will be using a push-pull circuit?
What valves are you going for?

Mark

[GrimJosef]

VVTs website says the VTP12496 is rated at 30W with 3k5 anode-anode load. They suggest it would suit two pairs of EL84s wired in parallel push-pull.

Cheers,

GJ

[Aub]

As the OP says it's a 20W project, I would guess it's the Mullard 5-20, so EL34's

Aub

[mole42uk]

I'm using four EL84 valves per channel with distributed loading. I have a pretty good idea that this will give me a cooler, more stable amplifier than EL34s. The only problem so far is that there's a lot of chassis real estate involved! Eight EL84s, a couple of rectifiers, phase splitters and driver stage.....

[yestertech]

Wow ! That's going to be quite a boat anchor
Cant wait to see some pics ...

For my own amp project ( severely stalled ) I opted for the 2 x monoblock approach.
At least then you can carry one in each hand (allegedly )
The downside is I need two matching mains transformers !

Andy

[Aub]

Well that threw me! Presumably the 4 x EL84s will be running towards class A to give only 20watts.

Good luck with it anyway. It should look and sound great.

Aub

p.s. don't normally post at 3.15 am - got a nasty cough keeping me awake.

[mole42uk]

I have thought about the monoblock approach, but this chassis shouldn't be too much bigger than, say, a Leak Stereo 20. The wound components are quite heavy, but our diminutive reception lady delivered the parcel to my office door without too much physical injury! Hopefully the complete amplifier should be manageable without a fork-lift.

If it gets a bit out of hand, I might consider the Decca approach and use a power supply chassis and two separate power amplifiers. Alternatively, I could build the power amplifier and the pre-amp into the same box and just use a separate power supply.

I did some tests yesterday, putting 240v AC onto the primary of the HT transformer, and hooking the 'scope up to the primary of an output transformer. I can get the transformers pretty close to each other (3/4" or so) before there's any measurable 50Hz on the output transformer. Obviously, I'll have to do a more comprehensive test before cutting metal for the chassis, but I'm pleased with the initial results.

Photos will be posted as the project progresses. I'm currently waiting for RS to ship the 10 x EL84 valves which appear to have been delayed into stock.

[valveaudio]

You might have a look at this design it would work with your transformers.

http://www.geocities.ws/tjacodesign/.../valveamp.html

Trevor

[mole42uk]

Interesting, but it's not Distributed Loading and I'm not keen on using MosFETs in a valve amplifier!

I do have a pretty much developed circuit which I will reveal when I have time to get some parts put together and some testing done.

[mole42uk]

Originally Posted by Aub

Well that threw me! Presumably the 4 x EL84s will be running towards class A to give only 20watts.

I'm inclined towards class AB because the distortion is easier to sort out. Although the power in AB seems too high, I'm hoping to run the valves at 300v with enough feedback across the whole amplifier to control the distortion and the frequency response.

Because I'm able to build something that might need tweaking to keep it stable under all operating conditions, I'm happy to accept some development work along the way. I started with the power supply, which is why there is a HUGE mains transformer and two 10H chokes, then selected the output transformers to give me the maximum amount of variability to get the whole thing right.

I haven't yet decided on the phase splitter, although Mr. Williamson and Mr. Baxandall pretty much sorted that out years ago, and the input valve might be the ubiquitous EF86. Sorting out the output stage seems a bit more important than the driver right now.

[Radio Wrangler]

I'd say that the phase splitter was really sorted out by Dr Bailey for Mr Radford.

David

[mole42uk]

As a Bradford-born engineer who grew up in the shadow of the Wharfedale works I hang my head in shame.

[Radio Wrangler]

It's a neat bit of design. The basis is a long-tailed pair but the subtlety is that with a single-ended input one half of a long tailed pair looks like a direct triode amplifier whilethe other half looks like a folded-cascode amplifier. So one side suffers from Miller effect more than the other.

The solution is to cascode the 'direct' side to balance up the thing. Rather than having a separate bottle to do the cascoding, the current can be kept as an electron flow in vacuum by using a pentode and we wind up with the counter intuitive result that a pentode versus triode long-tail pair is better balanced than a triode-triode pair.

The same imbalance in single-ended drive into long-tailed pairs happens with transistors where there is a further imperfection. For them, the Miller feedback capacitance is that of a diode junction and so it behaves as a varactor... the capacitance varying non-linearly with changes in bias voltage. Cascoding here makes a one pole function into two separate poles and pushes up the frequency so you not only lose the problem of a high-frequency roll-off, you also lose the problem of said roll-off pumping in frequency with stronger signals.

David (Huddersfield-born)

[GrimJosef]

Originally Posted by Radio Wrangler

... The solution is to cascode the 'direct' side to balance up the thing. Rather than having a separate bottle to do the cascoding, the current can be kept as an electron flow in vacuum by using a pentode and we wind up with the counter intuitive result that a pentode versus triode long-tail pair is better balanced than a triode-triode pair ...

Interestingly it seems there were (at least) two reasons for Bailey's choice of an asymmetric arrangement for his phase-splitter.

The first is the one that David's described i.e. that the first element in the splitter would suffer from Miller capacitance if it were a triode. So it improves the performance to use a pentode instead.

But the second reason, which Bailey explicitly mentions (New Phase-Splitter, A.R.Bailey, Wireless World 68 (9) p411, Sept 1962) is one of cost/complexity. He points out in that article that the splitter might work even better if it were a symmetric design using two pentodes. But there was no double pentode available with the right characteristics and, presumably, the cost of using two separate single pentodes couldn't be justified. So a triode-pentode was chosen.

Curiously, in the construction write up for the Radford MA15 amplifier (A New High Fidelity Amplifier Part 2 - Design Considerations, A.R.Bailey and A.H.Radford, Hi-Fi News 7 (2) p90, July 1962) which as far as I know was the first product to use this phase-splitter, the possibility of a double pentode splitter is again mentioned. But there the authors claimed that there would be little advantage over a triode-pentode since the Miller effect in the second half is so very small.

To cap it all when the later very high power Radford professional amps were designed they did choose to use two pentodes in the phase-splitter, in part at least to get sufficient drive for the output stage. I'm currently working on an STA100 (a 100W unit from 1967 which spent many years in a recording studio) and there two EF184s are used. With a bit of care in the choice of the preceding valve, which sets the DC operating point of the phase-splitter, I've been able to get ~0.03% THD at 20W with no sign that things are going to get much worse as the power rises.

The usual argument against pentodes is that they are inherently noisier than triodes. Presumably this was not regarded as a problem a) because signal levels are already pretty high in these phase-splitters and b) because in the 1960s audio sources themselves couldn't deliver silences as quiet as the "inky black" digital sources we're all used to now.

Cheers,

GJ

[Radio Wrangler]

Yes, the pentode-pentode pair would be more elegant and would certainly avoid the surprise factor of the pentode-triode pair.

But with a single-ended input, the Gm of the first pentode is dominant, and the second half whether a triode or a pentode is operating as a unity current gain grounded-grid stage. I can understand Bailey/Radford saying that any difference between pentode-triode and pentode-penode is slight. They must have tried both variants once the idea had struck.

Me, I'd probably go for balanced push-pull right back to the input stage and follow the engineering elegance trail to the limit. We're doing these things for fun, not profit so there's no point in saving small amounts. Anyhow, I'm not going to design a valve amplifier again, transistors are somewhat more of a challenge which makes them more interesting for me.

David

[mole42uk]

There's an attractiveness about the pentode pair route, but given the reported slight improvement over the pentode/triode, and given that I already have two GZ34s and eight EL84s to accommodate I am very attracted to using a pentode/triode in a single envelope.

However, there is still the output stage to do before I get to working out how to drive it.

[Radio Wrangler]

I did wonder why four EL84 and not a pair of EL34 or KT88

David

[mole42uk]

1. I've been there with EL34s and have no desire to do it again. Mullard did it quite well anyway.
2. EL84s are not quite so phool-inflated so I can spend money more appropriately on the output transformers and a decent power supply.
3. A quad of EL84s rated for only 20W rms should give me lots of room to minimize distortion and optimise the frequency response.
4. I want the challenge.

[kalee20]

Looks like an interesting project! The transformers look workmanlike.

Philosophy behind the four EL84's is interesting, credible, and I'd agree. If you give each EL84 its own cathode resistor (decoupled of course), you should get good self-adjusting current sharing. You could even use six EL84's, each with a larger cathode resistor, to give each an easier life - as long as the total current drawn is the same, it won't affect the matching and total power output.

I can see a possibility of parasitic oscillations, so each valve ought to have its own grid stopper.

Look forward to how this develops!