I stupidly said (confusing the issue)...

No, the AVC is definitely is NOT negative with reference to the chassis,
but the EMER773 states reference must be to the slider of RV3, where is
should be -1.2v. That is the reading that used to be OK and now is not.
And I've checked and rechecked aagain all the changes I made.

....and Richard said...

I found somewhere on the web a manual for the CR-88. I no longer
remember where but I suspect you can find it. If not I can send via
e-mail, its about 3Mb. It has a different voltage chart. All voltages are
measured to ground. For instance, using a VTVM the voltage at pin 4 of the
first RF tube is -18 volts with the RF gain control at minimum and -2
volts wtih the control at maximum. I can't understand measuring to the
slider of the RF gain pot because the voltage there will depend on the
position of the pot. I suspect this is an error of some sort.
Note that the whole RF gain/AVC sytem gets a negative voltage from a
devider string in the center tap return of the power transformer. That is,
the RF gain control is at a negative potential with regard to ground on
both sides.
Actually, there are two strings from the center tap, one consists of
R-46 (RV-3) and a 6800 ohm resistor called R-55 on the CR-88 schematic.
The other string consists of R-43, 44, 45 and the taps supply a fixed
negative bias to the audio driver tube and audio output tube. If you are
not getting a negative voltage on the grids of these tubes its a good
indication that something is pulling this point positive. Almost the only
thing there is the first filter cap which returns to the the top of the
negative voltage deviders rather than to ground. If this is the case you
will also see a positive voltage at both sides (and the slider) of the RF
gain control. The lines from all of the negative supply points to their
respective grids have rather high value resistors in them so I think its
unlikely that something on that side could pull the entire chain positive.
If you replaced the first filter cap check its wiring and check to see
that it isn't presenting a low resistance path from positive to the
devider chain.

OK, time to bring a few things together, I feel, especially as I introduced
a massive red herring in the first comment that I've reproduced in this
thread...

I meant of course that the AVC was not *positive* with respect to chassis,
as indeed it shouldn't be. Measurements taken from chassis are OK for the
negative grid bias that must exist. I've just measured and if the RF gain is
at minimum I get minus 18.5 volts on the grid of V1 (pin4) and if the RF
gain is at maximum I get minus 3 volts (using a DVM, so impedance similar to
VTVM). All pretty much in line with Richard's figures for the CR88. BTW I
also get minus 0.8 on the grid of V10 (1st AF) and minus 17 on the grid of
V11 (output). The grid voltages are all the result of the chain R43, R44,
R45, as has been said.

The problem is rather more subtle. You really need to download the document
EMER 773 from the VMARS website to understand it, but I'll quote the
relevant bits. There is a list of test resistances and voltages round each
valve on page 4 and 5. The relevant part is the grid voltage for valves 1,
2, 5, 6 (AVC line) (and incidentally the anode and cathode of v8, with the
AVC off (ie V8b shorted), *against the slider of RV3*. Now, I quote, 'RF
gain set to full gain'. That is why the measurement to the slider does not
depend on its setting -- or rather the setting is specifed.... 'full gain'.
The desired voltage is specified as minus 1.2 volts and is what I originally
got and do not get now. Looking at the circuit, it can be seen that this
measurement must be a result of the voltage drop across R42 (390K) and that
can only be the result of a current flow (about 3 microamps). I first
queried why there should be *any* flow as there should not be grid current,
but someone kindly pointed out that there is a path to earth via R47 (2.2
Meg), RV1 (66K) and R49 (33K). I still can't make sense of this using Ohms
Law (requires more volts than would exist at this point?), but it appears
that the test to the Anode and Cathode of V8b is to verify all those
resistors as well as the main AVC chain, plus the operation of the shorting
part of the AVC switch. Testing minus 1.2 volts on the other grids also
tests for continuity of the AVC chain and incidentally, if any valve was
gassy there would be grid current and different readings just for that valve
as the current would cause a drop on eg R2 (2.2 Meg). Incidentally, the
filter caps (4 mfd) C96, 97, 98 don't come into this. I have not touched
them. The caps I have replaced are the tubs.

Summarizing: Normal grid bias apparently OK. Specified reading to slider (as
fixed posotion) WAS OK (so not a mistake in the manual*) but now not. A
final bizarre possibility exists. Suppose the manual WAS in error and the
reason why I got the origina readings was a compensating error in that the
tabbed capacitors were leaky? Then I replaced them with good ones and now
the error in the manual showed up! That would explain why the application
of Ohms Law to the relevant part of the circut does not produce the quoted
desired results either!

I think I'll have another coffee now...