Mullard 5-10 ECC83 balance

[mpegjohn]

Hi,
I have just built a Mullard 5-10 amplifier.
I was measuring voltages around the phase splitter, ECC83.
I noticed that the anodes were at different voltages.

So I changed the 100K anode resistors for 0.1% metal film types.

This has not cured the problem though.

The conclusion is that the two trodes are not identical.

This, I am guessing is normal. Do I need to worry too much.

Anode 1 = 223V Anode 2 = 238V
Cathodes = 76V, Grids = 74V

I thought about putting a balance pot between the 2 anode resistors, is that a solution?

Many thanks,
John.

[PJL]

The amp was warmed up for 10 minutes or more before you took the measurements?

The grid voltages should be the same. What sort of meter were you using?

The triode with the lower grid voltage has the higher anode voltage which would be correct so if its not the meter loading then C5 or the socket may be leaking.

It could be the valve is gassy in which case it should improve over time.

[GrimJosef]

Cathodes = 76V tells you that the two triodes are passing a total of 1.12mA. Since the average anode voltage is 230.5V and the paralleled anode load resistors are 50kohm your HT rail voltage would seem to be 286.5V or so. That seems rather high (I believe it's supposed to be 248V). Have you measured it ? Anyway, working backwards from there it would appear that one triode is passing 0.635mA and the other 0.485mA i.e the currents are mismatched by more than 25%. While that is possible for a used valve it does still represent quite a bad mismatch.

On the other hand this is an AC amplifier so strictly speaking you don't care about the DC conditions in the phase-splitter stage. It's much more important that the two triodes have a similar AC gain since it's that which controls the overall performance. Unfortunately AC gain and DC emission can be quite strongly correlated, so the large difference in emission which you're measuring might well indicate a significant difference in gain too.

Fixing the problem by altering the load resistors might be (sort-of) possible. But without knowing how well-correlated Ia and gm are it's always going to be a bit hit-and-miss. You'd be better off finding a valve in which the two triodes are better matched.

EDIT: In fact the nature of the circuit means that for perfect output signal matching the two anode loads in the phase splitter should have slightly different values (Mullard explain this in their description of the ECC83 stage in this amp). But in practice using 100kohm resistors in the anodes of both triodes results in only a small mismatch and the negative feedback pretty much takes care of it.

Cheers,

GJ

[barrymagrec]

Is it possible that there is a slight leakage in the coupling cap from one of the anodes?

The Leak TL12.1 used different values in each anode in a similar circuit.

[mpegjohn]

Very interesting all this.
I thought that both sides should be exactly the same, so the output stage halves were driven exactly the same but out of phase.

Differing anode loads will present differing amplitudes to the output stage won't they?

If so, why doe the Leak design have differing loads?

John

[Cobaltblue]

Quite a nice explanation of the circuit here .

http://www.r-type.org/articles/art-003e.htm

Cheers

Mike T

[barrymagrec]

I think its a gain +1 thing, as the write up says, the ECC83 has a gain of 100, or 50 as a cathode coupled phase splitter, the Leak uses an ECC33 which only has a gain of 35 to start with, therefore requiring different loads to even it up.

[Radio Wrangler]

Looking at the circuit and thinking of it with no AC signals present, the ECC83 has equal anode loads, but one grid is driven by an extra 1 meg resistor. If there is any grid current taken, this offsets the grid voltages a little and the voltage gain of the stage amplifies this. Similarly if C5 leaks a little. It only takes a little leakage to pull enough voltage across that 1Megohm resistor and multiplied by the stage gain to give you an offset.

If the 1 meg resistor has gone high, then the leakage current to do the deed is even less.

Measure the anode load resistors as a first check
Strap the two grids together with a piece of wire and measure the anode to anode voltage (floating a voltmeter between both points avoids the errors in two measurements and a subtraction). If there's still a difference, the valve gets the blame.

Look at the signal path to the left grid. Signal flows into this grid and out of the left anode. Any capacitance from grid to anode is multiplied by the Miller effect, and you get a roll-off with frequency.

Look at the signal flow to the left grid, then to the left cathode, then to the right cathode, then up to the right anode. This path sees a cathode follower driving a grounded grid amplifier. Cg-a is no longer multiplied by the Miller effect, though the Miller going on in the left valve influences the grid drive for both paths.

THe net result is that the anode output impedance FOR AC SIGNALS is rather different for the two sides of this phase splitter, and the high frequency roll-off is different. This doesn't affect DC conditions. An elegant solution to this AC imbalance was invented by Arthur Bailey.... the Bailey phase splitter, and was implemented in Arthur Radford's amplifiers. It gives less phase shift in the amp and let the two Arthurs use a bit more feedback.

David

[ukcol]

As GJ pointed out in post 3 your HT supply to the the ECC83 stage is about 20% high and although not critical it may be desirable to get it somewhere near to the correct 235v (Mullard manual).

As regards the imbalance, you may like to try this approach.

1] Remove the ECC83 and measure the voltage drop across R8. (this is the 1meg resistor between the 2 control grids of the ECC83. Use a meter with an input impedance of at least 10 meg. The voltage measured should be zero.
Any voltage drop across R8 indicates that there is a problem in the grid circuit external to the valve (e.g. leaky C7).

2] Still with the ECC83 removed measure each anode voltage, these should be pulled up to HT volts and be identical.
Any volts drop across either one of the anode load resistors will be due to a fault current; the coupling capacitors to the output stage being the first suspects in this case.

3] Refit the valve and measure the voltage drop across R8 again, this should still be zero.
If there is a volts drop across R8 in test [3] but not in test [1] then the ECC83 is drawing grid current.

If the above tests do not reveal a problem then the imbalance is likely to be down to differences in the characteristics of the two triodes in the ECC83.

[GrimJosef]

Originally Posted by mpegjohn

Very interesting all this.
I thought that both sides should be exactly the same, so the output stage halves were driven exactly the same but out of phase.

Differing anode loads will present differing amplitudes to the output stage won't they?

If so, why doe the Leak design have differing loads?

John

The point is that the inputs to the two phase-splitter triodes are not the same. The left hand one in the circuit has its grid driven directly by the signal from the EF86 stage. The right hand one has its cathode driven (its grid is held to AC ground by a nice big capacitor) and the drive signal comes from the cathode of the first triode. As always with a cathode-follower the signal at the cathode will be a bit less than the signal at the grid. Increasing the right hand anode load corrects for this 'bit less' drive signal to the right hand triode.

The size of the correction doesn't depend only on the gain of the valves. The ECC83 is indeed a high-gain triode. But Leak apply the same anode load correction (91k/100k) to the phase-splitter in their Stereo20 and that uses an ECC83.

Cheers,

GJ

[Aub]

I think this article encompasses much of what's been said, especially David Radio Wrangler's post.

http://www.r-type.org/articles/art-097.htm

Cheers

Aub

[G8HQP Dave]

In this particular case triode mismatch is the most likely cause, but grid current can be a problem in this circuit if the input stage anode voltage gets too high thus raising LTP current. However, the degree of mismatch you have probably doesn't matter.

[Cobaltblue]

I suppose the obvious step is to fit a new ECC83

I would be surprised if the imbalance shown would make an audible difference but you never know

And there may be a big difference between what I can hear and someone with a golden ear.

Cheers

Mike T

[PJL]

I am with David, don't write off the valve just yet. Try the suggestion of shorting the grids together and measuring the anode voltages.

[Cobaltblue]

Not suggesting writting off the ECC83 but a substitution is a pretty inexpensive 678-4114

The RS parts as far as I know are pretty slid

Cheers

Mike T