HP 8647A Signal Generator - empty RAM backup battery?

[Greybeard]

What data is stored in the battery-backed RAM of the HP8647A?
Will the calibration data be lost if the battery is empty?
Is there a procedure to save and write back the calibration data before the battery is empty?

[edpalmer42]

I have an 8647A with a dead battery circuit.  Some unknown fault will drain the battery within weeks - even if I use external AA batteries!  Since schematics aren't available, I just gave up on the battery.

When I power the unit on it complains that the RAM is corrupt and then reloads the cal data from eprom.  The only thing that's actually lost is the previous state of the machine so things like frequency and level return to default values.  You also can't store any settings for recall later.

[Greybeard]

I have an 8647A with a dead battery circuit.

Do you know if your calibration is still valid or can you check it?

I found a calibration ROM in the 8648 block schematic in several assemblies (A5 Sig Gen Synth, A6 Output, A7 Attenuator). But that's true for the 8648 not necessarily for the 8647...
I couldn't find any schematic or block schematic for 8647.

The 8647 has some error codes (503-512) of type "failure in comparing calibration data
between RAM and the calibration ROM, a data restore was done" type, as listed in the Operation and Service Guide [08647-90006].
That could mean the cal data is stored basically in EEPROM in several assemblies and is copied to normal (volatile) RAM at startup, so the cal data would not be corrupted by potentially empty RAM backup battery.
But I'm not certain about that...

I still don't have any error message at start up, but I want to exchange the old battery.
If just the device settings were stored in battery-backed RAM it would not risk to completely brick the device.

[edpalmer42]

I can't remember if it's even possible to calibrate this unit.  It might be factory calibration only.

Anyway, when set for 10 MHz, after a few minutes to warm up it's putting out 9.999 999 924 MHz.  It's slowly drifting down because it needs some time to warm up.  Mine has option 1E5 - High Stability Timebase.  The frequency will probably settle to something like 9.999 999 9 MHz.  A 0 dBm setting gives ~0.62 Vpp as compared to 0.63 Vpp according to calculation.  I think these values are still within spec.  This was an economy series rather than a lab series.  It was intended for manufacturing and testing uses.

I normally have this connected to an external frequency reference.  I had to repair mine because a coupling capacitor was missing.  Using the external reference puts mine dead on frequency.

FYI, I figured out how to step the frequency by 1 Hz rather than the 10 Hz minimum on the display.  You can directly enter 10.000 001 MHz on the keypad, but you can't use the dial to step the frequency up and down by 1 Hz.  But if you press 'INCR SET' and enter 0.001 KHz,  you can then use the up and down arrow to change the frequency by one increment.  After changing by 5 increments, the display will change to show the new rounded off frequency.

Ed