HP3458A restoration

[Ice-Tea]

Well, got one of these, but not in the best of shapes. The front bezel is chipped, the back bezel is broken (but can perhaps be glued to an acceptable state), pushbuttons on the front are missing, as well as the rods. The banana post front assembly is wobbly. Not too worried about most of that, because amazingly you can simply order parts from Keysight. I was first hoping 3D files would be available and this would finally be a good excuse to get a decent 3D printer but it seems I won't have to..

Unfortunately, it also comes with a number of errors:

RAM 1 TEST LOW: I assume depleted NVRAM batteries
205, "HARDWARE FAILURE - TEST VALUE OUT OF RANGE : 108"
According to the ALRM, should be on the A1 assembly somewhere.
110, "Calibration Required"
Yeah, figured as much
114, "SYSTEM ERROR - Non Volatile RAM checksum"
Yeah, figured as much

Once you get rid of the errors and warnings, unit actually seems to work: measured a coin cell voltage and current and both were OK. So, starting with the NVRAMs:

Seems like one of the earlier boards. And the NVRAMs are all original (datacode 89 I think) which, in this case, is not good. I think it's safe to say the cal data is long gone, no point in trying to retrieve it and all of them need to go.  I already figured ou the DS1235 can be replaced with the 1230. What I still wonder: these were also offered with more NVRAM as an option (see empty sockets):
- is it worth expanding the memory?
- is it as simple as plopping in 4 more chips? Will the SW recognize it?

Let's see how this goes

[Ice-Tea]

Grpmff... Replaced all NVRAMs. Now comes up with "RAM TEST 1 LOW". Can't reset beyond that. Which one is "1", by the way? Left or right?

[Dr. Frank]

Hello,
the 1220 should be replaced with DS1220AD-120+, and both 1235s by DS1230Y-150+ 
This RAM TEST error you still see usually refers to the 1230Y - 32kx8 nvRAMs, U121 and U122, because both form a 16 bit word, whereas the calram 1220 is 8 bit only.

The 4 empty sockets U122 - U125 are intended for 62256 SRAMs (volatile), to increase the data buffer to 148kB. nvRAMs should also work, but are a waste of money.

You might also update the firmware in the 6 EPROMs to latest version 8 or 9.

Frank

PS: RAM TEST 1 LOW should refer to lower byte, that is U121.
U121 and U122 also contain the working RAM space for the processor, therefore latter can't operate at all, if there's an error.

[Ice-Tea]

Well, I've put in 1230ABs. Should be as good as Y's as the power supply voltage is well within range (5.138) of the AB versions. And the 1220 is a 150+ version but so was the original one.

I should have socketted it. 

[bsdphk]

You should update the firmware as the first thing, version 4.6 has "issues" which I would not want to deal with, while trying to debug the hardware at the same time.

[Ice-Tea]

I was really hoping I wouldn't have to 

I aslo thought I didn't have an adapter for my programmer, but come to think of it, the base of the programmer probably *is* the adapter in this case 

[Ice-Tea]

Also: what kind of issues?

[Dr. Frank]

I just named the most convenient and cheapest types, but your choice seem to work also.

So maybe there's a problem with the new nvRAMs, or you have a soldering problem, or something else on the PCB is fishy.
Maybe you should really desolder all three nvRAMs again, and put them on precision sockets, and then you could also swap the high and low byte - RAMs, to see if this error moves to the other byte.
For testing, you might also replace them by 32kx8 volatile RAMs, you only would get some initialization errors on each boot.

Ah btw., have you saved the content of the old calram, maybe?
I think that its data were not yet corrupted and could be copied to the new 1220, if you own an appropriate programmer.

Frank

[Ice-Tea]

Maybe you should really desolder all three nvRAMs again, and put them on precision sockets, and then you could also swap the high and low byte - RAMs, to see if this error moves to the other byte.

Should have put it on sockets right away. Stupid. But swapping is next on the list.

For testing, you might also replace them by 32kx8 volatile RAMs, you only would get some initialization errors on each boot.

Hadn't considered that. Good idea.

Ah btw., have you saved the content of the old calram, maybe?

No. I didn't think chances of it containing anything at all after 30 years were nihil. In addition, I didn't think they'd be of much use after 30 years of component ageing. You thihk it's worth trying to read it?

[Dr. Frank]

The blank calram will be initialized with the default values like 7.2000V and 10.0000kOhm, so the new readings will be far off.

The old RAM might still contain the old values which for sure have drifted a lot, but that will at least give you a much better reading, in first instance.

Also, the RF calibration values might be preserved, which is an advantage, as this calibration otherwise requires special equipment, whereas zero, 10V and 10kOhm might be easily available from other volt-nuts.
So, it's better than nothing 
Frank

[Kleinstein]

The SRAMs can sometimes hold there contend even with a voltage considerably lower than given in the DS. So it is no longer guaranteed that data are still correct, but there still is a good chance the data are still there.
The is a message for SRAM checksum error, but no such message for the cal RAM. 

[Ice-Tea]

Message received. Will try to get the call data tomorrow

[bsdphk]

I'm not going to give you a very clear and detailed answer, because this is code which runs in a live product, and therefore by definition not a legitimate "dataarcheological" target.

To be honest, I am not entirely sure what is going either.

The fact that this version is named "4.6" where all other released versions are single-digit is ... strange.

All the other versions have the ROM checksum code in the bottom of the low eprom pair, where also the reset vector lives, so that you can meaningfully detect problems in the other four eproms.  In 4.6 everything is ordered differently, and the ROM checksum code lives near the end of the top eprom pair, meaning that 4 eproms must (mostly) work before you can even run the built in diagnostics.

Looking at v2, v6 and v4.6 (anybody have v3 and/or v5 ?) the differences indicate that 4.6 was more of a side-branch than a point on the mainline.

Almost as if somebody somewhere else were told "Please patch these critical bugs NOW! we can't wait for the next release..." or something.

Service note 12-B has the public release notices for the firmware versions, and gives plenty of reasons on its own to upgrade to v8 or v9

[Dr. Frank]

phk,
that sounds a little bit between the lines, as if this FW 4.6 might contain errors which prevent a proper initialization of blank nvRAMs.
Anyhow, if a UV eraser and programmer for these old EPROMs is available, I would directly upgrade to these FW 8 or 9, there is practically no difference between them.
Frank

[Ice-Tea]

Ah, crud. UV eraser  Hadn't considered I was going back that far in time  Programmer should be OK, I'll have to find something for the UV.

[Kean]

Ah, crud. UV eraser  Hadn't considered I was going back that far in time  Programmer should be OK, I'll have to find something for the UV.

If you don't want to buy/build a UV box, then the sun can work pretty well - it can just take a day or two of direct sunlight.

[bsdphk]

The only differences between 8 and 9 are the version number and two changed boundary values, which are probably for a component change to a larger tolerance.

[Ice-Tea]

Ah, crud. UV eraser  Hadn't considered I was going back that far in time  Programmer should be OK, I'll have to find something for the UV.

If you don't want to buy/build a UV box, then the sun can work pretty well - it can just take a day or two of direct sunlight.

Direct sunlight? In Belgium? 

Seriously, though. Ordered some cheap OTP's. So I can keep the originals around and give v8 or 9 a spin