Single ended amplier

[pwdrive]

Mark, for one of the best selections of valve data on the web try here:

http://www.shinjo.info/frank/sheets0.html

Here's one for Ia plotted against Va for different Vg:

http://www.shinjo.info/frank/sheets/020/e/EL84.pdf

Lawrence.

[Mike Watterson]

What Mark says. Test each stage on its own.

[Murphyv310]

Hi.
On an amp of this calibre thermal noise won't be a problem and I belive it's confusing the issue.
The op has asked why there is a problem and what I have explained is to replace the incorrect components with preferred values, then it is timely to then explain why.

[Mark Hennessy]

Hi.
On an amp of this calibre thermal noise won't be a problem...

I agree - that's exactly what I said earlier.

...and I belive it's confusing the issue.

Sorry, but I don't think this is the case. There is no confusion here - he's chosen 50k because he plans to use it with modern sources, and has selected 3uF to work with that. Job done

The op has asked why there is a problem and what I have explained is to replace the incorrect  components with preferred values, then it is timely to then explain why.

The values you've suggested seem eminently sensible to me, as I said earlier. Speaking totally personally, I'd like to learn more about how one chooses those values - beyond "that's what was typically used" - so I'm guilty of leading the thread astray in that regard. Give me transistors and I can design an amplifier and justify the presence and value of every single component, but I'm not there with valves. But keen to learn.

Notwithstanding that, it's important to establish if the input stage really is causing a problem. Re-reading the original post suggests that actually, all might be well with that:

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.

So perhaps focusing on the input stage is what's confusing the issue?

Turning to the output transformer, I did find some data on it here: http://www.vintage-radio.net/forum/showt...hp?t=25657 - I've taken the liberty of stealing the datasheet from that thread:

   

This confirms what Andy said in the OP - the primary is rated for 40mA. It has a DCR of ~1k, and he says it's dropping around 20V DC in use, suggesting that he has biased it fairly well (though 7.3V across 220R is 33mA). In which case, I wonder if it has a shorted turn or similar? Or perhaps a construction fault that isn't apparent from the picture?

Either way, we need to simplify and focus on each stage of the amplifier in isolation. How many times have we said that now?

For anyone learning electronics, here is a good bit of advice: https://www.youtube.com/watch?v=_pGk6GYDnes (jump to 1:40 and watch until 3:17). In short, you learn most when it doesn't work! Ain't that the truth

[Diabolical Artificer]

I've made a bit of a boo boo when I said the OP of the ECC82 is 25v, it is 25v but that's P-P, so that would be 8.8v rms. The datasheet for the EL84 says a 4v input is needed so I used a 82 after looking at mutual characteristics graphs for the 81/2 and 3. I was looking to get a low linear OP, IE as little THD as poss. Using a ECC83 is overkill, too much gain as the input is 1v. The 82 has high straight slopes so with a 12k anode R and it biased at -4.5v this gives a linear OP, IE 12.5v anode V for a change of .5v input g1.
The OP from the triode voltage amp is clean at the EL84 g1 and IMO has sufficiant gain for the OP stage. As I only wanted a 4v OP, this is why I used a 12k anode resistor. I'll post some load lines later when I get the time and all will become clear hopefully.

Yesterday I tried another OPT. The OPT's I have, have on the whole too low an impedance ratio. In theory for center biased class you find the optimum Zpri by HT^2/Pa which is 250^2/10 = 6250k or a ball park 5k. The EL84 datasheet gives a Ra of 5.25 or 4.5k in similar operating conditions to that which I'm using. So I tried another OPT, I found the Zpri by sticking a sinewave on the primary,( this I established by measuring the DC res, IE high DC R + primary) measuring the OP and using Vin/Vout^2 and times the result by load. I couldn't find one that was suitable, so used what I had, which I think had too low a Z ratio. This was born out after I used a 32R resistor instead of 8R, and got some improvement. I the end I found that with a sec load of 200 ohms I got a 2w clean undistorted OP. this OPT came off a radiogram i think, and had three speakers wired to the secs, one bass and two "tweeters" in series on different windings. As I have no schematic replicating this load is impossible.

I take your point Michael that it would be better to build a P-P amplifier, but I'm not going to build this amp and use it. I'm only doing it for a bit of fun and to try and go deeper into some aspects of amp building that I havn't done before and become familiar with things like open loop gain, NFB etc.

Mark is right about the input stage, for a modern AC source 10k ish is about right. I could have popped a 1meg pot on it but read somewhere that this is too high a value. I used this formula to find C in - 1/(2 x pi x f x R) ... 1 / (6.28 x 1* x 50000) = 3.184713376 -6, therefore a 3u cap is what I used. * Why 1? We want to block DC and pass all frequencies over DC.

I'll re read your posts in more detail later and try some of your suggestions and get back to you. Andy.

[Mike Watterson]

The transformer is certainly a big issue for SE over 500mW. Up to about 500mW you can often use a suitable ratio 3VA to 6VA mains transformer.

Well, sounds like you found the problem. You can try a 20W mains transformer (EI not toroid) of Square Root (6250 /8) = 28 times, so about 8V AC from 240V (a 6V to 12V transformer will do to test).

[ppppenguin]

Mark is right about the input stage, for a modern AC source 10k ish is about right. I could have popped a 1meg pot on it but read somewhere that this is too high a value. I used this formula to find C in - 1/(2 x pi x f x R) ... 1 / (6.28 x 1* x 50000) = 3.184713376 -6, therefore a 3u cap is what I used. * Why 1? We want to block DC and pass all frequencies over DC.

All this is really very uncritical. With a high level low impedance source almost anything will work well and not degrade performance significantly.

It's also time for a note about excessive precision. You've specified the values in the formula to a maximum of 3 significant figures yet you've given the result to umpteen, simply because the cacluator says so. This is spurious precision. you have quite rightly roundedthe result to 3. But even this has some implications. The nearest common value caps are 2.2uF and 4.7uF. These electrolytics will have huge tolerances, often as much as -10% to + 50%. And it doesn't matter here!

It really doesn't matter if the frequency response is down to 1Hz or 10Hz as the output transformer won't work that low. Neither will the speaker. Using 3 for pi is perfectly adequate too. You could use a 220n cap, roll off at about 12Hz and it would be fine. I doubt you even notice using a 100n cap. if this is a learning exercise why not try it and see. THough be careful not to delude yourself into hearing a difference.

[Kalee20]

Comments:

1) 50kΩ for the volume control... "why is 1MΩ too high?" Well, the wiper of the control is feeding thye ECC82 grid, which is really hgigh impedance, and virtually capacitive. If the wiper of the control is at mid-position, then the grid would effectively be fed from a resistance of 500kΩ in parallel with 500kΩ (the two halves of the control's resistance), which would give some sort of treble loss when associated with the ECC82's effective input capacitance. But as this is going to be of the order of about 15pF, you can work out using f = 1 / (2πRC) that the frequency roll-off would be 32kHz, well above human hearing. So, while a 50kΩcontrol would extend the upper limit even further, it won't be audible! And as for loading a modern signal source, yes it would drive 50kΩ, but loading with this value rather than 1MΩ, again won't make any noticeable difference.

2) 3µF because "I want to block all frequencies above DC." Why 'all frequencies?' It's actually better to restrict input frequencies - the output transformer is unlikely to work well with much below 50Hz, so putting anything lower won't reach the loudspeaker anyway. But, if your input source has mega-bass boost or whatever, and starts shoving 10-20Hz frequencies (at significant amplitude) into your amplifier, you might find that the input stage is overloaded and distorts, causing clipping of the signal waveform, even though the 10-20Hz then gets stopped right at the output. So, better to use 0.22µF (with 50kΩ volume control), or 0.01µF (with 1MΩ).

I made the mistake, many years ago (teens!) of making a 3-stage amplifier using 0.22µF / 1MΩ coupling between valves, and I couldn't understand why the sound kept disappearing and reappearing every couple of seconds. It wasn't until I started measuring with a voltmeter that I found that the whole amplifier was oscillating at around 0.5Hz and the output stagewas gently going right into and out of cut-off!

3) 12kΩ load for the ECC82. If you have used load lines on characteristic curves, then you've done right, although you are operating the ECC82 at more current than it needs. From knowledge of the '82, I know that the curves are practically as non-curved at lower anode voltages and lower anode currents... 100kΩ would be as good. But that's the only comment. You've got something that works!

4) Load on the EL84. For an 'ideal' pentode, you need to load it with a load equal to: Anode voltage divided by standing anode current. That way, as you wind up the input signal, on one half cycle the thing will just start to distort as the anode voltage reaches zero at the peak of the signal sinewave (EL84 can't turn on any harder) and on the other half-cycle as the current just reaches zero (EL84 can't turn off any harder) at the other peak of the signal sinewave. For a real pentode, the anode-cathode voltage doesn't reach zero before the valve starts getting highly non-linear, so the actual optimum load is more like 0.9 times the 'ideal.' And you also need to allow for cathode bioas, and optput transformer DC winding voltage drop. But hopefully you will get the idea.

Using the above idea, if you want to give your EL84 an easier life by reducing the power, you can! And you can calculkate the optimum load for your reduced current. Obviously the output power will be less, but with the optimum load you can maximise the power available at your chosen standing current.

Have lots of fun - which it seems like you are!

[Murphyv310]

Comments:

{snip}

Well said that man............. My feelings entirely but put down far better than I could.

[Mark Hennessy]

1. An issue with a 1M pot that hasn't been mentioned yet is hum pickup. Depending on the physical layout, a high value pot might mandate a screen cable that would not be needed with 50k, say. Ultimately though, there is nothing wrong with Andy's choice of a 50k pot given his design objectives.

2. LF rolloff is a complex issue, and one I was avoiding until the other problems have been resolved. But in short, it's usual to define the required -3dB at one part of the chain. All other "poles" would be an octave or more lower in frequency so that they are spaced away from this point, meaning that the LF rolloff is a nicely defined "Gaussian" function formed by that one pole. For what it's worth, I did write some words about this subject some 15 years back - http://www.markhennessy.co.uk/preamp/ana...search.htm - where I show that setting the main LF rolloff with the input capacitor is quite feasible given an anticipated wide range of source impedances. At the same time, I also show that setting the HF rolloff point is not feasible at this point - any HF filtering at the input should be regarded as ultrasonic/RFI filtering, and the HF rolloff proper needs to be defined elsewhere.

So here, while I'm happy to reassure Andy that his understanding of how to calculate Cin was good, I would probably change it to 100n (~32Hz) or 47n (~68Hz), depending on what the output transformer dictates. The other time constants in the circuit are in the sub-10Hz region, so probably OK, but increasing those caps a little wouldn't hurt. As an aside, it's well-known that using electrolytics for filtering is bad for distortion, so it's normal to "oversize" them (totally academic here, naturally - but the idea of positively defining the LF -3dB point with a nice film capacitor at a known part of the circuit and spacing all the other poles well away is common practice and a good principle to learn).

As I said at least once earlier, some rail decoupling between stages would be a good plan.