Single ended amplier

[Diabolical Artificer]

Got a bit fed up with fixing solid state amplifiers so decided to build a simple 3w ish single ended class A amplifier for the craic. A single ended amp is never going to be great as they're prone to high distortion and they're inefficient.Obviously I can't better the Mullard 3-3 but I don't intend to build this, it's just for fun.

I started with the power supply which needed to have very low ripple as single ended amplifiers can't reject hum and noise like a P-P amp. I used a 200v power tfmr I had and two 100u/350v caps with a big 1k5 50w resistor so that these formed a high pass filter. the ripple is very low, less than 5%. A choke would have been far better but I didn't have one at the time that I knew it's ratings, so hence the big resistor. It's rectified by a bridge ( not shown in schematic)

I had in mind something that would amplify high line voltage devices like CD/MP4 players so used 1v AC RMS as my input "parameter". I used a 3u cap and a 50k pot on the input which should form a high pass filter, with a 1hz approx roll off, to block DC. However I'm not sure about this as the "sums" are correct if 50k is put in the formula, but the pot doesn't sit at 50k, it varies, so I couldn't find out how to calculate this. Would putting another fixed grid resistor on the grid of the ECC82 have been better?

I tried to bias the triode on the straight part of the curve so it's OP was as linear as poss, so little THD and used a 12k anode resistor so I didn't have much gain, IE, 25. The measured OP is 25v and is low distortion. This is a bit high, as the EL84 only needs about 4v but I wasn't sure how much it would drop when connected to the EL84 grid. It's biased at -5v.

All good so far, but things fall apart at the OP stage I think. The EL84 OP stage is pretty bog standard, biased at -7.3v. I started with a 180 ohm cathode R, but changed it to 220 as there is less anode current than anticipated. Cathode cap is a pretty standard 100u. I'm using a HT of around 250v as the OP tfmr I'm using is rated for 40mA, this is a bit restrictive as the tfmr drops about 20v. I started with a 220 ohm grid stopper only but a had to put a 2k2 in series to drop g2 as it was higher than the anode, this drops it about 7v. I also decoupled g2 as there was a bit of instability.

The two valves are coupled using a 0.1 cap, with a 1k grid stopper and 330k grid resistor. The frequency response is pretty good being pretty flat from 50hz to 20khz, although there is a bit of bass attenuation, nothing to worry about though.

The OP is about 2w tops, but the OP is very distorted above about 1/2w. NFB improves matters applied to the junction of the ECC82's cathode but reduces power OP. I put a 20 pre set pot as the NFB R, but if I reduce the NFB, it sounds aweful. A 1k sq wave input results in an unrecognisable OP, so I couldn't really try to use a NFB cap. I would have tried to calculate both of these but tried the suck it and see approach first.

Looking at the Mullard 3-3 design, I see they used a self biased EF86 really throttled back so they could couple it direct to the EL84. Two components I'm not sure of - C3 which is in parallel with the EF86 and R9 which connects the EF86's g2 to the cathode of the EL84. One other thing puzzles me about the 3-3 is that they use a very high HT which will take the EL84 over it's maximum peak anode voltage of 550v when on the low current half the cycle.

Conclusions, to make a simple half decent SE Class amp I would need to use a EL34 or similar with a lot better OPT. I did toy with the idea of using maybe four or more ECC82's in parallel, but the effort would be wasted without a decent OPT. All I found in my OPT stash were odd ones that would need higher ohmage speakers. I did find a "choke" OPT that looks interesting, that would solve the PSU issue somewhat and some big "matching" tfmr's, not sure if these would be suitable for valve OP stages though.

Any thoughts most welcome, Andy.

[Mark Hennessy]

I had in mind something that would amplify high line voltage devices like CD/MP4 players so used 1v AC RMS as my input "parameter". I used a 3u cap and a 50k pot on the input which should form a high pass filter, with a 1hz approx roll off, to block DC. However I'm not sure about this as the "sums" are correct if 50k is put in the formula, but the pot doesn't sit at 50k, it varies, so I couldn't find out how to calculate this.

Spot on here...

It would be different if something else followed on from the pot, but as a valve has a very high input impedance, there's no problem with the calculation.

Just for background, let us pretend that the following stage was something else - doesn't matter what - that had an input impedance of 10k. In this case, the input impedance - as seen by the source, and as seen by the input capacitor - would vary from 50k (pot at min) to 50k||10k (pot at max). 50k in parallel with 10k being 8.33k. The capacitor needed to deal with that would be more like 20uF.

I'm (patently!) no expert with valves, so others will hopefully chime in, but my initial thoughts might be aimed at the output transformer. SE amps obviously put a DC current through the primary, which puts flux in the core. As you try to swing either side of that quiescent position, the core will saturate at some point. So if there is distortion on the peaks of a sine wave corresponding to maximum transformer current (lowest voltage at the anode), that might be a clue. Of course, the opposite happens if you don't have enough standing current. But all that applies irrespective of the technology, and is probably blindingly obvious (sorry!)

[Mike Watterson]

The SG resistor seems high, but I'd need to check.
Really the only reason for SE was reduced cost. Push Pull, (even if class A!) cancels out the first even harmonic and means any suitable dual winding 120 / 240 (110 /220) mains transformer can be used as Audio out as the DC bias currents cancel. Voltage ratio of one winding to speaker equals Square Root of Impedance ratio. PP also reduces ripple out output. Add extra Resistor and 32uF cap for preamp and output screen grid (with then no resistor) HT supply
The other half of the ECC82 can be the "gain = -1" phase inverter for the second EL84.

NFB from speaker to driver amp input is best, though you can simply leave of cathode resistor decoupler (SE or PP), and use one resistor for both EL84 if PP.

A SE transformer if not an Audio type (a gap) may even saturate at bias current.

[Hartley118]

Well, Andy this is puzzling. You can get 2W out of the EL84, but it's horribly distorted above 0.5W. The triode input stage looks OK, with plenty of voltage drive available for the EL84 so, as you deduce, there's something very wrong in that EL84 output stage.

NFB, by the way, never cures 'horrible' distortion because the 'horror' generally implies that the open loop gain is practically disappearing at some point on the waveform. That's why a NFB amp generally runs into clipping more suddenly than one without NFB - the open loop gain suddenly disappears at the clipping point.

My first thought was to suspect the EL84 valve itself, but you report that it seems to be drawing a sensible anode current, and a few volts either way on the screen will be neither here or there. I wonder, could your test rig be going unstable as you increase the gain to raise the output above 0.5W. Does the horrible distortion include any squeals?

I also wonder whether Mark may be on to something in looking in the direction of the output transformer. If the core were to be saturating, that would kill the primary inductance so that it would distort badly at low frequencies. Even if that were the case, I'd still expect it to perform reasonably at, say, above 1kHz. However, you say it's rated at 40mA, so should be OK - it should have gapped core and they don't suddenly saturate above rated current. I wonder whether it has a shorted turn somewhere. Might be worth testing it on an oscillator or just by feeding the secondary with, say 6V AC and see what comes out of the primary.

Martin

[Murphyv310]

Hi.
Personally I think the circuit is flawed.
The Triode stage should be a Voltage amplifier but the values you are using is hardly conducive to that.
The 12k anode load should be 10 or more times higher than that between 100k and 330k. The Volume control should be 470k to 1m and the cathode between 1.5k-3.3k, the ECC82 should really be an ECC83.
I suspect you have insufficient drive with the values you have and the drive is distorted and low.

The output stage looks OK although there is no need for the 100 ohm on the G2, the 2.2k is quite sufficient. Some NFB from the output transformer secondary to the cathode of the triode via a suitable RC network would be beneficial.

The Mullard 3-3 can use a higher HT as the cathode resistor is high so the V a-k is in respect not 300v. The big advantage in the 3-3 is it's directly coupled so eliminating phase errors.

[Mark Hennessy]

Yes, now you mention it, it does seem rather low. But when I saw "25V p-p AC" on the diagram, I assumed that the triode stage was working OK. But was that value the intention rather than the actual value observed?

The value of the volume control is irrelevant - it merely sets in the input impedance (and influences the value of the input capacitor). Andy specifically mentioned the source equipment he plans to use with this, and it's modern gear that works into low impedances like 10k or 50k with no issues. In audio design, always go low if you can

While I can see that the value of the cathode resistance on the triode depends on the "Vgs" of the valve and the desired "collector current", I lack the experience to say if he's chosen well there or not. With 5V there, that's 7.5mA. Across the 12k resistor, that results in a drop of 90V, which sounds pretty plausible on the face of it, but what would be the optimum "collector current" for that valve? If nothing else, that 12k will dissipate more heat that it needs to if a lower current gives equal or better results. Certainly, the circuits I've seen do indeed use the values you suggest...

Forgive the use of solid-state terms in quotes, but I find the analogies help me.

Taken in the round, insufficient voltage gain from the triode would result in insufficient voltage gain overall. But that won't give distortion, surely? As there's no overall feedback yet, it would be easy to check the triode in isolation. And indeed, it would be easy to check the second stage, totally in isolation, providing Andy has an oscillator with a highish output voltage. I'm still suspicious of that transformer. But now I look again, some supply decoupling between the two stages wouldn't be a bad idea

[Murphyv310]

Hi Mark.
What we want here is a decent drive for the EL84. On the face of it 25V P-P will be fine but with the values used to get that amount of drive there will be loads of distortion. With the "correct" values you could be getting 100V P-P or more with the same distortion but your EL84 will be driven hard so reducing the input or volume will reduce the distortion and you'll still have enough drive to get 3-3.5 watts of undistorted output.
Personally I don't like low value volume controls for the reason that it can cause bass roll off due to the time constants of the input "R" of the VC vs coupling cap or the coupling in the source.
I'm very much a firm believer that traditional values work better than some of the so called "new" designs in valve circuitry.

[pwdrive]

I agree with Trevor's earlier post, use an ECC83, the max voltage amplification with the ECC82 won't exceed the valves amplification factor which is around 18, draw a load line for that 12k anode load you've used for your ECC82 and you will soon see what's what.

Lawrence.

[Mike Watterson]

Test by driving EL84 direct.
12K anode load does sound low. Suitable bias with 120K load will give more gain.

[Mark Hennessy]

Hi Mark.
What we want here is a decent drive for the EL84. On the face of it 25V P-P will be fine but with the values used to get that amount of drive there will be loads of distortion. With the "correct" values you could be getting 100V P-P or more with the same distortion but your EL84 will be driven hard so reducing the input or volume will reduce the distortion and you'll still have enough drive to get 3-3.5 watts of undistorted output.

I assume you're talking about distortion from the first stage? I'd like to learn more about this. I'm studying the datasheet for that valve, but I'm not seeing a transconductance curve. I expect to see something that shows the exponential relationship between Vg and Ia, and from that curve, I'd be able to choose a decent quiescent current for the valve to work at. From the curves I can see, I don't really see a problem with 7.5mA, beyond a lack of voltage gain because of standard "gm times Rl" stuff - the same basic principles you apply to solid state amplifiers.

There are some tables in the datasheet that report distortion for various values of anode current and anode resistor, and they don't vary much (4-6%, which seems pretty high to me). The lowest resistor I see is 47k, and that's with a 1k2 cathode resistor.

Either way, I did suggest that each stage can be tested in isolation - this is a really valuable technique and I don't understand why it's not already been done as part of the troubleshooting, given the luxury of no overall NFB. Just a photo of a 'scope trace developing 25V pk-pk would be interesting. The results from a distortion analyser is even better if Andy has one. But, break it down, and work out the problems of each stage in isolation. Isn't that what everyone does?

Personally I don't like low value volume controls for the reason that it can cause bass roll off due to the time constants of the input "R" of the VC vs coupling cap or the coupling in the source.

One drawback of a high value volume control is thermal noise. It depends on how much gain follows, and how noise-free that gain is, but with good solid-state design, the noise of just a 50k pot could easily be the dominant source of noise in the entire system. Perhaps not in this context, sure, but that's the philosophy behind avoiding high values, and it's basic, well-established stuff. Some state of the art devices use 1k potentiometers! The high impedances seen in valve gear is more about cost saving that anything else.

As I've said, LF roll-off in this context is academic with modern gear. Different if you're building an amp to work with valve sources (that are often too mean to include a cathode follower and a decently sized output cap), but again, Andy is building this for modern gear.