HP3548 / HP3457 OLED display

[Xyphro]

It looks exactly as my hp3457a allthough that was not the question you asked. It has the same press fit 2 rows by 8 pins connector. Also the mechanical mounting points look identical.

[Xyphro]

And you can just pull the connector out. It is not soldered, but has a spring mechanism.

[Le_Bassiste]

 I have to disappoint in the feasibility to reuse this exact same display design as it stands for the HP66xx supplies. Mechanical it is very easy to make it fit (just a notch has to be inserted in the PCB outline). BUT (and a big one) the protocol seems different.

If I'd have all time in the world and this power supply would be extremly interesting for me to get it repaired I'd immediately jump on it to get it done, but I would here prioritize other things first. Maybe.... maybe I'd take this up in future. Decoding it should not be too tough, but time is an issue here.

have a look in https://www.eevblog.com/forum/repair/replace-lcd-on-hpagilent-664xa-psus/msg2137255/#msg2137255, might help figuring out the hp66xx display communication.

[Xyphro]

Thanks for sharing. Actually the protocol is the same! Just the connector has a different pinout.

[Kean]

And you can just pull the connector out. It is not soldered, but has a spring mechanism.

I can confirm.  I have my HP 3468A open on the bench, and the DIP connector has split pins which press fit into the PCB.  Just be gentle removing it.

I took a quick peek into one of my HP 3478A's, and it has the same display PCB labelled "5061-5212".  A google on that brings up lots of hits.

[Kean]

Hmmm... newer ones seem to use a "5181-2813" display PCB.  I wonder what the difference is.
If it wasn't already past midnight here I'd check the service manual. 

[Xyphro]

However this turns out with those multimeter, just want to say you get my full support in adjusting the AVR Software whereever/ifever required :-)
This would especially work well in case you can make logic analyzer dumps. I have a method to play back logic analyzer traces physically.

[RichardM]

Ok

Does the whole thing come off the board ? Or is half the dip socket remain after removal ? The spring is easy tonremove and suggests half the socket remains ?

Richard

[Xyphro]

The whole thing comes off the Board, so the cable including all the plastic parts in one piece. Just apply some force between the lcd pcb and the plastic Part where the cable ends up in.
Don't be afraid, it is not a lot of force required.

Best regards,

Kai

[Xyphro]

Or to explain it differently: after you have the dip press fit removed, the display lcd had no plastic parts, but you just see holes in it. The cable side has all plastic parts and metal spring pins that were pushed before removal into the PCB.

[RichardM]

Great thanks

I am willing to give this a try on my 3468a.

Regards

Richard

P.S If anyone else in Australia wants to make the same modifications I would be happy to share the PCB costs, I certainly don't need 5 boards

[Kean]

P.S If anyone else in Australia wants to make the same modifications I would be happy to share the PCB costs, I certainly don't need 5 boards

Richard, Yes!  I am in Sydney and have 1 x 3468A and 3 x 3478A.  I'd like to try this on at least one of them.

[Xyphro]

Lucky guys, you have summer, while we have it freezing cold here :-)

[Hexley]

The datasheet for the OLED display has a section on "Image Sticking" that refers to a ghost image that can remain if a fixed pattern is displayed for long periods.

From the datasheet:
"If you remain a fixed image on OLED Display for a long period of time, you may experience a phenomenon called Image Sticking. Image Sticking - sometimes also called “image retention” or “ghosting”- is a phenomenon where a faint outline of a previously displayed image remains visible on the screen when the image is changed. It can occur at variable levels of intensity depending on the specific image makeup, as well as the amount of time the core image elements are allowed to remain unchanged on the screen."

They go on to say that it is not true burn-in, and can be reversed by displaying a black screen for an extended time. They recommend not showing a fixed display for more than 2 hours.

A DMM display will be a mix of dynamic elements (the numbers) and static elements (the mode letters and the annunciators). The latter may remain unchanged for long periods, and thus invite ghosting.

This suggests that it might be good to add a screen saver mode to the Kai Display. Perhaps something that follows this logic:

1. Create a watchdog timer with a period of, say, 30 minutes. This will determine the timeout period.
2. Start the timer at power on.
3. Restart the timer if there are any signs of activity, such as:
     * Any change in the operating mode, as determined by a change in the right-most 4 characters of the display. E.g., if those go from "MADC" to " VAC", that shows the user has interacted with the instrument.
     * Any change in the annunciators. E.g., if the "Shift" annunciator comes on, the user has interacted with the instrument.
4.  If the timer times out, display a screen saver image, such as:
     *Black screen.
     *Animated image, e.g. the "flying geese" of the HP41 calculator.
5. When in blanked mode, unblank (restart the timer and restore the normal display) if any user activity is detected. See step 3 above.

The screen saver function could be enabled/disabled by a jumper, perhaps. The parameters suggested above would need tweaking, no doubt -- exactly which annunciators should be monitored, for example.

There are probably some corner cases that would allow an undesired timeout, or perhaps prevent a desired timeout. But for the majority of uses, the algorithm above would probably work adequately, or so it seems to me.

I'm curious what others might think about this.

[Xyphro]

Agree, this would prevent issues. About 2008 when I used oleds the first time there were recommendations to limit burnin of pattern like shifting every few second the picture by a Pixel or so,... I hope those oleds are better, but I cannot tell that by experience.

The SW can certainly detect if the content did not change over the last x minutes and then darken the display, turn it off,... And on again when something in the content changed. Maybe the sample annuciator should be ignore then or a smarter time out method found.

I can do something in the Software if we agree on a scheme and reaction.


Possible actions I can imagine
- blank screen
- dim screen
- shift over time every few seconds/minutes the screen content up and down very few pixels

[RichardM]

P.S If anyone else in Australia wants to make the same modifications I would be happy to share the PCB costs, I certainly don't need 5 boards

Richard, Yes!  I am in Sydney and have 1 x 3468A and 3 x 3478A.  I'd like to try this on at least one of them.

Hi Kean, thats great. I am just up the freeway in Newcastle. I would be happy with 2 boards but to be honest we may as well order 5 each and combine postage.

Any ideas on a source for the OLED's ? Aliexpress is mentioned above ?

PM sent.

Richard

[Berni]

Yep burn in is a thing for OLED as the tiny LEDs inside it get worn out and become dimmer

However if you are just displaying static segments the burn-in might be very hard to notice. It's hard to tell if a segment is just slightly dimmer, so in order to see the burnin you typically need to fill up the display with solid lit pixels that way you can see the slight 'shadow' of the burned shape on the area of lit pixels (the difference between fresh new pixels and tired used ones right next to each other). So what the "shifting by 1 to 3 pixels" does is smear out that shadow images edge making it less noticeable in that case.

So id say for your use case OLED burn in is not an concern. It would only become an issue once the burning becomes really really severe to the point where you can notice some segments being dim. Same as what you can see on VFD display test equipment that has had a lot of hours put on it.

[mapleLC]

Beautiful work.

[srb1954]

The Hp66xx supplies have mechanical same displays, also with 12 digits. Of course supplies like E3632 and similar ones are very different.

Btw. My hp3547 shows when the input is left open in Volt Mode a very huge drifting voltage until I connect it to a source or I change the input impedance to the lower setting. Is this normal or an indication of something dying?

Perfectly normal. It is just the stray leakage currents generating a voltage across the very high (10G ) input impedance of the meter. Switching the input impedance to 10M is sufficient to drain the stray leakage currents.

[ArgyllGargoyle]

This looks like a fantastic solution for the stock display - I am probably going to give it a try.
I wonder if there is any EMI impact from the switch-mode DC-DC converters in this design? Has anybody looked at this aspect?

[sagias]

Tested 3478a works ok but i have smpl in place off srq,2w in place of azoff azoff in place of 4w readings are accurate
found 3478a hex icons and 3457a hex are the same compare in hex editor we need the right file

[wolfy007]

P.S If anyone else in Australia wants to make the same modifications I would be happy to share the PCB costs, I certainly don't need 5 boards

Richard, Yes!  I am in Sydney and have 1 x 3468A and 3 x 3478A.  I'd like to try this on at least one of them.

Hi Kean, thats great. I am just up the freeway in Newcastle. I would be happy with 2 boards but to be honest we may as well order 5 each and combine postage.

Any ideas on a source for the OLED's ? Aliexpress is mentioned above ?

PM sent.

Richard

Well if you fellas are going to give it a try, Ill have a go (3x 3478A, 1x3468A & 1x3468B) I have one 3478A going thru repair at the moment I could try it on. Im down Wollongong way.

[Kean]

Well if you fellas are going to give it a try, Ill have a go (3x 3478A, 1x3468A & 1x3468B) I have one 3478A going thru repair at the moment I could try it on. Im down Wollongong way.

Hmm, well I forgot all about this and I don't know if I ever ordered any parts.  I have acquired another 3478A which arrived with a slightly damaged display, so maybe I should look into this again.

[cncjerry]

I'm using a 3457 more lately and the display sucks in fluorescent light.  It looks like the parts used two years ago generally available?  I looked at the project and I'm not familiar with the OLED displays.  I see they come in 128x32 pixels?  I had thought they were just one character, but it looks like that size are all 128x32.  But those available are white, and I want blue!

thanks

Jerry

[sagias]

Fitted neen some trim on pcb and open the right hole a little to align,marks are messy due to wrong fw on github for 3478a,i dont know how to compile hex for 3478a from source, measurements are ok