Lampy cd player

[Richard]

Ok, well that seems to be the case, got the formulas working and see he shows this circ as a CF augmented plate follower so I've borrowed his drawing and put our values in for 20x gain.

The 95V and 120V should have ?? by them, what will happen there Nick?

These take no account of losses in the cathode follower and in his formula terminology give,
R 71274
a 20.75
gain 20.38
Ratio 1219
rp 50000
mu 100
Zo (plate follower only) 55k46

Mike, I looked at the first stage again and think AG used the high plate load as he had a low HT and didn't need a lot of current. This gives a gain of 55, and distortion should be low guessing from the data sheet. It also keeps the coupled voltage lowish and will suit my heaters raised 40V.

So I'll put in on a board and see if it works. My maths are dodgy and Nick's idea for the IV is untested so it should be interesting!

[Richard]

Actually, no, I'm not sure about that virtual earth at the grid. It wouldn't allow current to flow so we wouldn't get IV conversion would we?

[Nick]

The 95V and 120V should have ?? by them, what will happen there Nick?

Well, the 95 is just from the plate curves, grid at 0v, cathode at voltage generated by current. The 120v will be something more like 96v, leaving 1v bias on the second triode grid.

Actually, no, I'm not sure about that virtual earth at the grid. It wouldn't allow current to flow so we wouldn't get IV conversion would we?

Why not, the action of the feedback is to maintain th egrid at 0v, as the grid is at 0v, there is a 82R in your case between the output of the dac and ground, so current will flow in just the same way as if the 82R was just taken to ground.

BTW, you dont need the 1M on the input. Even with the HT off, there is a path for the input to get to ground via 1100082 ohms of resistance, so close enpugh to 1M to not matter.

[Nick]

If you get that working, I would try replacing the second cathode with a 600v N channel MOSFET, even lower output impedeance.

It seems to work well for Croft preamps.

Maybe something low power like this

http://uk.rs-online.com/web/search/sear ... &x=24&y=20

Its has a lowish gate capacitance and should be fine in place of a ecc83 CF

[Richard]

The 95V and 120V should have ?? by them, what will happen there Nick?

Well, the 95 is just from the plate curves, grid at 0v, cathode at voltage generated by current. The 120v will be something more like 96v, leaving 1v bias on the second triode grid.

Ah, thanks, ok, just seen the cathode resistor is 150K so 25V bias will not be much current, just looked wrong at first.

Actually, no, I'm not sure about that virtual earth at the grid. It wouldn't allow current to flow so we wouldn't get IV conversion would we?

Why not, the action of the feedback is to maintain th egrid at 0v, as the grid is at 0v, there is a 82R in your case between the output of the dac and ground, so current will flow in just the same way as if the 82R was just taken to ground.

Thanks, you've just saved a lot of re-cacl'ing if it works

BTW, you dont need the 1M on the input. Even with the HT off, there is a path for the input to get to ground via 1100082 ohms of resistance, so close enpugh to 1M to not matter.

Yep, that looks good, the less the better, the blue figs are his, I guess the parallel 1M were to make sure the usual gridleak 1M wouldn't be exceded.

[Richard]

If you get that working, I would try replacing the second cathode with a 600v N channel MOSFET, even lower output impedeance.

It seems to work well for Croft preamps.

Maybe something low power like this

http://uk.rs-online.com/web/search/sear ... &x=24&y=20

Its has a lowish gate capacitance and should be fine in place of a ecc83 CF

Thanks, will look at it if this works. May be a few days now though as I'll get roped into doing the garden...

[Richard]

I should have known joining this site would end up with smart hifi looking like this

Only joking, some good progress.

Set up the board with reg heaters and variable 200-300V HT to let me test the circ last posted.

The IV doesn't work, not as shown anyway, but some more work has resulted in a very promising stage. I'll post figs and circ after some more checking.

[Neal]

Good one Richard, you're getting the hang of it. Check out the surplus yards for old barn doors and mount your stuff on that, seems to work wonders for the sound quality!

[Mike H]

I should have known joining this site would end up with smart hifi looking like this

Wouldn't look out of place on my sideboard. But then you haven't seen my sideboard ..


 

[Richard]

Thanks guys, I thought it had a certain olde englishe charm, mono too you might note...

So, some reworking. Basically with the previous circ the signal in from the gen was going straight through via the low values in the feedback loop and I was getting a decent output before even switching it on. When switched on the levels didn't match the calcs so I think the formulas perhaps don't apply at very low input resistance values. I can't say what would happen to IV from a dac.

Whatever, I know the 82R IV conversion works well on the front of the srpp stage so shunted it across this stage and set the amp up for 18x gain via the feedback resistor ratio 47k and 1M5. These nicely block the input from the output and the 82R does IV and gridleak.

Tested and it works very closely to the formula despite the slight unknown of the cathode follower, not much loss there apparent at all.

Formula says 18.37x gain and this is giving 18.1x as far as I can measure.

116mV in gives 2.1V out.

Distortion is 0.34% THD 1KHz which is not bad considering it is se and will be mostly second harmonic.

Interesting note here that winding the generator level up gave a reduction in distortion down to 0.2% then it increased again as you'd expect from an se triode stage. Not sure why, normally it just goes up with level (possibly signal to noise ratio, the clean signal was becoming a larger part of the sum being tested up to a point or is it the feedback loop?). I will build with metal films rather than these carbons and see if it improves but it may also be valve noise, any thoughts anyone?

The first stage is not bypassed so there is about 5.2dB local feedback then the global loop adds 7.7dB more and sets gain.

This tallies well with the Ecc83 data sheet with regard to expected distortion from a stage such as this. Feedback will have reduced the distortion 13dB, between 4 and 5 times, from the data sheet which would suggest around 2% for the basic stage.

I checked it into loads from 100k down to 5k3. Very flat frequency response rolling off slowly after 30KHz. It hardly changed in level or distortion between 100k to 18k loads then increased quite sharply at the 9k and 5k3 checks. So 20k to 100k loads = almost no change.

Set 50k load and added 220pF across the input but it made no change to distortion or frequency sweep.

Added 5n1 to provoke it and it made no change to distortion, just a smooth roll off of the highs to 1.2V out at 20KHz, so used into realistic cap on an input should be no problem. The cathode follower output and feedback loop will give a low output impedance, haven't worked it out yet.

I tried to improve it. Increased and decreased anode load and cathode values and tried increased voltage (seems very insensitive to voltage, no change in gain or distortion) but all were worse so this is a pretty sweet pairing from AG!

So will it sound any good? I will fit it tomorrow if I can find the chisel and plane

[Richard]

A long session yesterday saw the CF Augmented Plate Follower stage installed listened to and removed

It worked exactly to spec, a technical success, but sounded poor. Sat-on sound grey like solid state. Oh well.

Now having the test plank I thought I'd revisit Lucasz' stage.

What a surprise. The sheer simplicity and scribbled diagram hides a lot of work I think. It tests very well indeed. I tried different cathode resistors, 330R, 150R and 100R and all were worse so this has clearly not been arrived at by ear alone. I tried different IV resistors, 53R and 100R but found the 82R best.

That was enough for me and I rebuilt the Lampizator 6n2p. Wow, it does sound stunning. It's been a couple of weeks since taking it out but the old method of putting it back in certainly let's you hear the difference.

Thorsten's srpp stage was good and tidy sound. Clear with the Ecc88 and meatier with the 6n2p.

The optimised srpp version had even less distortion on test and sounded clearer and cleaner but was sterile and sharp.

The CF Follower stage was neutral but grey.

This stage though really has it all. Very powerful yet clear and untiring.

Merlin Blencowe describes it as a half mu stage on page 3 here,
http://valvewizard2.webs.com/SRPP_Blencowe.pdf

I think the difference must be that it is taking the signal straight off the anode of the input valve rather than working in srpp, or off the cathode of a CF.

Here then is how it's now set and a few of the figs into different loads.

[Nick]

Cool, going round in circles can be fun and informative sometimes

[Aardvark]

And how does it stack up against your reference player on rehearing?

[Mike H]

I note on page 6 it says:

"If we change the load we must change Rk too, [bias] or linearity will probably suffer (too many authors try to ignore the load when analyzing the circuit, which leads to quite misleading results).

"It is therefore a mistake to use an SRPP as the output stage or line-driver of a preamp, for example, because we can’t be sure what it will be plugged into and hence what Rload might be." . . . "The SRPP can only be designed properly if we know what the load impedance actually is, and treat the whole thing as a unified system (remember, an SRPP with no load is not an SRPP at all!)."

Further down this caught my eye also:

"What may not be immediately obvious is that absolute maximum power from the SRPP obtains when the load impedance is equal to ra so, in theory, it should be possible not only to optimise the circuit resistances but also to choose the optimum triode for the job! All we have to do is find one whose ra is equal to the impedance of the thing we wish to drive,"

 

[Mike H]

Further to the above, that's an aspect I hadn't considered (what is Rload) and I should have 'cause I'm sure I saw that PDF before. Need to experiment....