Improving the low end ACV of the HP3457A

[Tryer]

I bought my device end of last year from eBay.
It is built in 89. with a main controller 03457-66511 Rev.B  and an AC board labeled  09-8950.
 
Playing around with it, i noticed that in ACV, on all ranges,
whith shorted input, i had a quite "big" reading.
In the ACV 300V range this was something like 0.4V
The AC measurement in the 3457 is done by an RMS to DC converter the AD637 (U305).
I measured a small DC voltage at pin 13, which was present even with shorted inputs.
If only AC were needed, a coupling capacitor would solve the problem.
But in 3457 there is also the AC+DC mode and this is propably the reason for the DC coupling.
Searching around i found that i could bring it to zero by changing slightly the input offset
of the U308 opamp.
This improved the situation significantly.
This opamp is part of a DC canceling feedback for the input of the AD637.
For ACV it is configured as an integrator  with a time constant
τ1=200k*C323 or τ2=200k*(C320 C323) depending on Q300 (HFAC).
(This is propably controlled by ACBAND command).
The "fast" AC mode gives 4.4ms, the slow AC mode 48.4ms.
Again this is not an "AC-path".
To make it short, the next thing was to put a socket in U308 and replace it.
I had a chop OPAMP in my drawer the MAX420. The two 100nf caps, which are necessary for the MAX,
were placed and wired on the tot of it.
This improved the situation significantly.
An absolute 0 reading, even with 0V DC, cannot be reached, as the input is relative wide (>1MHz) and
all the noise will be treated as an AC signal. 
With shorted inputs, i have the following readings after the change:
ACV 30mV  : ~6uV
ACV 300mV: ~25uV
ACV 3V       : ~0.3mV
ACV 30V     : ~2.4mV
ACV300V    : ~25mV
All above values were ~15 times higher.
The high-end values and AC response are not influenced by that change.
This "tweak" makes only sense, if the low end readings are unacceptable high