Hewlett-Packard 3314A function generator

[tggzzz]

I've been reading about 1N4148 failures and one of them is very similar to what I've been experiencing.

That is curious, i am curious.

Any URLs/summary?

[synx508]

It was this page https://gideonlabs.com/posts/glass-diode-failures

I think that was a problem with a particular batch, from the way I'm reading it, so it might not be relevant, but it seems like a way that the diode can be intermittent that I'd not considered before.

[tggzzz]

In my ignorance seals between different materials has always felt like a weak point, especially as metals corrode over time.

[Zenith]

I never think of 1N4148 as things that fail. Obviously like anything else they can fail, it's just that for most applications they are very over-specified for the voltages and currents they have to cope with. You'd have to be determined to damage them by soldering, unlike germanium parts. I've never heard that they had ageing problems either - unlike carbon resistors, paper capacitors, Rifas, germanium transistors growing tin whiskers, electrolytic caps and a few others.

There and again, I've come across a few things which I thought never failed, but have. Small signal silicon transistors operating well within their specs, which suddenly gave up the ghost. 78 and 79 series regulators which died or behave strangely. Dead logic chips.

Hameg scopes used 1N4149 everywhere, which is similar, but specified for a higher current. I'd guess they may have had a need for 1N4149 in some positions. With the numbers they were ordering, there'd be no difference in price and standardising on one type kept stock control and manufacturing simpler and cheaper.

[Zenith]

In my ignorance seals between different materials has always felt like a weak point, especially as metals corrode over time.

Intuitively that's a reasonable thing to assume as a weak point, but a lot of effort was put into eliminating it. For instance with valves. With B9A and B7G valves (the smallish all glass valves used from the late 40s in TVs and made by the million) a lot of R&D went into choosing an alloy for the pins which matched the expansion of the glass and caused no other problems. Valves went from room temperature to too hot to touch in less than a minute. Loss of vacuum because of the pin to glass seal breaking down was nearly unheard of.

[synx508]

I've switched it off for the night, there were no restarts since the diode replacements. That's the longest it has run non-stop since I've owned it. Maybe a good sign, maybe not. I'll repeat the test tomorrow.

If I've fixed that problem I will try to source a couple of replacement 1k x 4bit µPD444 static RAMs, for almost as much as I paid for the function generator, or I could just leave the 62256 in there, hanging on its rats nest of wires. It needs a new battery, too, as it's below 3V.

[Zenith]

You've got us on tenterhooks over this. Who'd have thought 1N4148s could play up like that? Have you retained them for testing to see if the fault can be reproduced outside the 3314A?

As for the 1k x 4bit µPD444 static RAMs, "for almost as much as I paid for the function generator", I don't think you should be too quick off the mark. You seem to have a talent for spotting components at swapmeets, and as sure as eggs is eggs, if you splash out, you'll find a tube of them for a quid at the next rally.

[synx508]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them. I decided it'd be quite a lot of work to build a test jig to get them to the right temperature to replicate the conditions inside the 3314A. I might fish them out if I decide that would actually be fun.

My component-seeking abilities worked in reverse with these SRAMs, I'm 90% sure that I threw several of them out last year because why would I want those ancient things?

[synx508]

Also, I'm clearly too tired to be typing stuff into the internet, I used a 6264, not a 62256. Still 16 times more capacity than each of the original chips.

I *think* the 6264s came from a pile of Tandata Prestel terminals that I bought, must've been around 1990, from a very strange junk shop that was run from what was an abandoned church in Brentford (St. Lawrence's, it looks like it has been spruced up recently, on streetview). A friend and I had cycled to Brentford from Windsor along the canal towpath together them after I'd spotted them for sale on a previous trip, we returned home with these things tied all over both bicycles, I think there were about 15 altogether. They contained a socketed TCM3105 chip, which was used in amateur packet radio modems. It's a neat chip, with excellent performance for the time, IIRC it uses switched-capacitor filtering and can dig modem tones out from noise rather well. Those chips were some of the things I sold at the first radio rally that I attended as a seller. I can't actually remember which rally it was, but probably DDRC at Stockwood Park, where I borrowed a bit of a G4 friend's stall to flog the chips. Pretty sure that was also the rally where I met the people from RadioFax, so it was a few years after buying the chips.

[tggzzz]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them. I decided it'd be quite a lot of work to build a test jig to get them to the right temperature to replicate the conditions inside the 3314A. I might fish them out if I decide that would actually be fun.

Is the temperature that critical? A quick and dirty test would use a desoldering hot air gun (mine goes down to 100C), or even a hair dryer.

My component-seeking abilities worked in reverse with these SRAMs, I'm 90% sure that I threw several of them out last year because why would I want those ancient things?

That's traditional

Would UM6264-12 be of any use?

Personally if the diodes are faulty, I wouldn't preemptively replace the RAMs

[tautech]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them. I decided it'd be quite a lot of work to build a test jig to get them to the right temperature to replicate the conditions inside the 3314A. I might fish them out if I decide that would actually be fun.

A good friend and mentor, a Dr of EE Prof shared he had given his students the task to find the current limits of IN4148......20A for 1us !!!!
Pretty darn good for a 300mA device.....

The dumb stuff one remembers.....

[synx508]

Would UM6264-12 be of any use?

Personally if the diodes are faulty, I wouldn't preemptively replace the RAMs

I've got a 6264-12 in there at the moment, it's in a socket, soldered to some stripboard, with wires soldered between the stripboard and the top side of two more µPD444C sized sockets, which are then in sockets that I put in place of the 444s, which I snipped out as the leads were folded over. I could probably salvage the old chips but it'd be a lot of work.

Well yes, the diodes…, but then I didn't know diodes could fail in that way and the usual voltage and ripple checks on the PSUs were fine. I hadn't captured the voltage ramp during the glitch when I pulled out the NVRAM, it was showing up corrupted after the glitches and at the time I didn't know it also stored the working copy of all the settings so I thought it must be causing the glitches rather than being battered by them. The cable fault made it harder to establish the real fault and even when I homed in on the +!5V regulator section as the problem I got the wrong target, the op-amp, on the basis that I have seen op-amps intermittently fail. And working backwards, the +5.1V line could fire resets into the CPU for many reasons and the +15V could go low for multiple reasons, too. Then the first reason for CPU resets looked very much like the noisy, misbehaving CMOS inverter that I started the thread with, which had indeterminate logic levels on its output pins. A 15V rail dropping to 11V for 25ms once every few hours is definitely less visible than the computer resetting, but that was just a symptom. At least I haven't changed any of the posh tantalums, apart from the one I changed then changed back…
I was thinking of all the 1N4148s in all the keyboard matrices in 45 year old gear and how you'd never notice if they went open for 25ms, but when they're holding up the power supply it's a lot more obvious.

[Zenith]

I thought the problem was the computer detecting a power out and panicking. In most cases where a 1N4148 is used, you probably wouldn't notice if it glitched for 25ms every few hours. You might see a fleeting flicker on an analogue scope screen or something like that. Intermittent problems with 1N4148 seem very uncommon anyway.

[synx508]

The computer has a time constant assisted inverter to run its reset line, that's connected to a comparator that checks various things about the 5.1V supply. The 5.1V supply watches the 15V supply to check it's >13V, the 15V supply was dropping to 11V because of the 1N4148.

And now I've found that problem the flaky interconnect reared its head again and may well have taken out a number of buffer ICs or worse. Utterly unbelievable stuff, I was about the screw the lid back on, of course. My fault for not changing it.

No output, even on the sync socket. A quick continuity test of the interconnect shows two opens, that's on the wire itself, not even taking the dodgy connectors into consideration. Unfortunately one of the opens was GND, so it could've damaged the A2 mode board, however the A1 generator board is also not generating any signals and that hasn't happened before.

[synx508]

Looks like the A1 and A2 interconnects are the problem, not just the conductors but also the connector continuity. I've managed to get sync to come back but only when I'm pushing the connector to maintain the connection. The main output isn't working, though. I am so tempted to just hard wire it all.

[EC8010]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them.

I'd be very interested to try measuring if it's just one or two. I can improvise a computer-controlled curve tracer and put the DUT in a temperature cabinet. Curve tracing is usually quite good for spotting intermittents and popcorn noise.

[Zenith]

I've just made a mental note not to buy an HP3314A function generator if I spot one.

[synx508]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them.

I'd be very interested to try measuring if it's just one or two. I can improvise a computer-controlled curve tracer and out the DUT in a temperature cabinet. Curve tracing is usually quite good for spotting intermittents and popcorn noise.

I found two of them by sorting through the scrap bin, but I'm not sure if I've recovered the one with the interesting fault.

[synx508]

I've just made a mental note not to buy an HP3314A function generator if I spot one.

This applies to the 3325B, too, in terms of interconnects. I've put my 3325B into storage because the interconnects were driving me round the twist. I don't have the same problem with the 3325A, which uses slightly different parts.

I've managed to get the main output to do the right thing but until it passes the self-test it's very hard to get any usable output from it and I can't hold the interconnects in the right place long enough for that to happen. Maybe it'll be fine once I change the interconnects for something that conducts electricity reliably. I am not going to do that for a while, though. *moderate seethe*

[Zenith]

To be fair, interconnects and edge connectors are a likely trouble spot with any 30 plus year old kit. There are tarnishing problems and cables become stiff with time. I've had a couple of problems fixed by removing boards and re-seating them. Connectors will have a specified number of insertions, maybe 50, which no one takes any notice of, and in most cases is never approached. In this case it sounds as if previous owners did a lot of whipping out of boards to locate the problem. There's always a temptation to grab the cable and yank it out rather than pulling the connector body, especially when you are getting fed up.

[EC8010]

I found two of them by sorting through the scrap bin, but I'm not sure if I've recovered the one with the interesting fault.

That's a shame. But I'd need to be pretty confident that I had the guilty party. My temperature cabinet started life as a £20 tabletop oven from Aldi before being rewired and having its bimetal strip replaced by an industrial PID controller and a fan added within the chamber. But it's manual control from the front only so temperature tests have to be attended.

On the subject of interconnects, I think I may have fixed the booting problem on my XP machine. Bought this time last year for £30, it's a really nice machine in a huge (high quality) crate. I replaced its HDD with a SDD and gave it 4Gb of RAM, taking the cost to just over £100. It has a quad 2.4GHz processor and fairly rattles along. But it frequently didn't get to boot XP, getting stuck during autodetection of SATA drives. This morning, (with power still applied so as to not lose settings) I replaced the motherboard's coin cell. Gripping the cell with a neodymium magnet on the end of telescopic aerial, I was able to nudge the locking pin with a small screwdriver and withdraw old cell without it dropping onto powered bits. Fitted new cell. Old cell was 3V, new 3.3V. But in doing all this, I noticed a SATA lead that went nowhere, so I've removed it. I can easily imagine the BIOS being confused by reflections from an unterminated SATA lead. Machine has now booted cleanly three times in a row.

[AVGresponding]

I dropped the dodgy 1N4148s into my "scrap" bin but I could retrieve them.

I'd be very interested to try measuring if it's just one or two. I can improvise a computer-controlled curve tracer and out the DUT in a temperature cabinet. Curve tracing is usually quite good for spotting intermittents and popcorn noise.

I found two of them by sorting through the scrap bin, but I'm not sure if I've recovered the one with the interesting fault.

You'll find out which one it is when you use it to repair something...

[synx508]

To be fair, interconnects and edge connectors are a likely trouble spot with any 30 plus year old kit. There are tarnishing problems and cables become stiff with time. I've had a couple of problems fixed by removing boards and re-seating them. Connectors will have a specified number of insertions, maybe 50, which no one takes any notice of, and in most cases is never approached. In this case it sounds as if previous owners did a lot of whipping out of boards to locate the problem. There's always a temptation to grab the cable and yank it out rather than pulling the connector body, especially when you are getting fed up.

These interconnects have failed in many different ways. The original pins are pitted and have reacted with something over the years, they're almost slate coloured and crumbly looking. The insulation displacement arrangement looks as if it was never punched down evenly. The cable itself is single core, or at least tinned to the point that it may as well be and it's very brittle. The pins were never really long enough, they only just reach the contacts inside the housing. I added longer pins and the contacts visually perform much better but the springiness of these contacts could be better. I am hoping to find some ready assembled cables so I don't have to go fake-molex-crimp crazy making my own.

[synx508]

You'll find out which one it is when you use it to repair something...

Indeed, that's why they were straight into the scrap bin. I bought a big box of many, many thousands of Philips 1N4148, so I don't mind throwing out any that might be suspect.

[Zenith]

You'll find out which one it is when you use it to repair something...

Indeed, that's why they were straight into the scrap bin. I bought a big box of many, many thousands of Philips 1N4148, so I don't mind throwing out any that might be suspect.

Yes indeed. I bought 200 for a quid from the slightly dodgy, but now sadly defunct, N.R Bardwell about 30 years ago. They said they were 1N4148, I recall they are marked PH256. They've never given any problems where a 1N4148 was needed. I still have half of them left. I'm sure I have also bought the odd packet of 100 1N4148s at rallies for a quid or 50p. IMHO, you'd have to be very hard up to reuse a pulled 1N4148.

I'm not sure what these interconnects are like. Years ago I sometimes had to make up 50 way IDC cables for SCSI. There wasn't a proper press on site, and we couldn't justify buying one for the amount of use we had. I pressed them together using a bench vice in the machine shop. Then the narrower ribbon cables came out with thinner connectors. They pretty much had to be bought in.