EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: SharpEars on October 08, 2014, 02:15:13 am
-
I just bought an HP 3456A multimeter that was classified as non-operational. The symptom of its non-operation is that the unit itself turns on, begins to sample as indicated by the sample rate annunciator LED flashing, but the rest of the display stays off.
So, first thing first, I checked all of the voltages on the power boards and they all seem to check out. There don't appear to be any bad electrolytics either. Then, I went for the A4 board (Keyboard scan logic, display and annunciator). Voltages on that board seem fine. I see level changes on U29 and U30. I see a clock feeding U28 (an octal latch). The 5V/Gnd test points on the board measure 5 V. Data seems to be arriving at the D0-D7 pins. Nothing looks wrong or suspicious. I decided to disconnect the ribbon connecting from J4 to the A2 display/button board. I examined the ribbon and it is in excellent condition by the way with no damage. I pulled out the A4 board itself, cleaned that J4 connector on the board to which the display/keyboard ribbon connects. Made sure when I plugged the board in that everything is snug - both the ribbon that connects to J4 as well as the two rows of pins that hold the board.
However, the problem persists - nothing on the display lights up, not the 7-segment LEDs and not the LEDs next to buttons. The funny thing is that the keys seem to work and perform their proper functions, although I only have the sample rate annunciator LED to tell me that since it's the only darn thing that lights up. None of the 7-segment LEDs nor the LEDs near the buttons light when I press the buttons near them. The buttons themselves seem to work, because they have an effect on the flashing of the sample rate annunciator LED - an effect that seems reasonable for their intended purpose.
The funny part about all of this is that the button LEDs, the annunciator 7-segment display and the other 7-segment displays are all being driven by different seemingly functioning driver chips (I checked input voltages on them for +5/Gnd, by the way) on the A4 board. Yet, none of the LEDs on A2 (Keyboard, Displays and Annunciators) light up other than that one sample rate annunciator LED which flashes in utter defiance of the rest of its neighbors!
I also tried pressing the TEST button which is supposed to light up all of the LEDs, at least on the 7-segment panel, but nothing lights up even though the test seems to function and cycles like it should, which I hear from relays closing/opening, indicating (probably) successful tests cycling.
I am at wit's end. The A2 board with the actual LEDs and seven segments is very hard to get at without a lot of diassembly so its difficult for me to get probes in there, but I still managed to test some of the 7-segment pins and see voltages reaching them (both, several volts and a few hundred millivorts north of ground).
So, what could be causing all of the LEDs and LED segments but one which are driven by multiple driver chips not to work? It seems statistically impossible that all but one of the LEDs just decided to die.
I know many of you guys have HP 3456A troubleshooting experience, so thanks for any and all advice, I really appreciate it!
-
Is there a clock signal on either or both pin 3 and 11 of U24 ?. Whats on pin 1 and 13?.
U29 and 30 have +5v ?, in particular pin 10 ! which is the shutdown pin !! ie no output.
And that +5v rail is clean with no ripple!!!!.
-
Is there a clock signal on either or both pin 3 and 11 of U24 ?. Whats on pin 1 and 13?.
U29 and 30 have +5v ?, in particular pin 10 ! which is the shutdown pin !! ie no output.
And that +5v rail is clean with no ripple!!!!.
OK, here's what I measured. Note that I don't have a scope at home, but do have a mutlimeter with frequency counter:
U24 has a 52.51 Hz clock present on both pins 3 and 11. Both pins 1 and 13 read a voltage of 4.52 V. U29 and U30 are properly supplied with +5 V and Gnd at their respective Vcc/Gnd (pins 19 and 28 respectively) and both show a voltage of +5 V on pin 10 which is good (i.e., SHUTDOWN on low is not enabled). Pin 8 (WRITE STROBE-L) on both U29 and U30 indicates a 26.21 Hz clock present.
Sadly, I can't check for ripple on the 5 V rail, because I have no scope.
-
Do you know someone with an oscilloscope that you can borrow?.
You really need to be sure of the power supply, as quite a lot of troubles with these DVM's stem from bad electrolytics, even if they look okay!.
-
Do you know someone with an oscilloscope that you can borrow?.
You really need to be sure of the power supply, as quite a lot of troubles with these DVM's stem from bad electrolytics, even if they look okay!.
Sadly I don't know of anyone with a scope I can burrow. I was hoping that if someone ran into a similar situation where all the LEDs (except for the sampling one) and 7 segments were off, they could pinpoint me toward a particular place on the board to look. I would imagine that a power supply ripple would interrupt it enough not to function, but it does seem to function and test, except it doesn't display any numbers. I mean it starts seems to go through the startup test and begins sampling. The buttons effect the sampling rate in proper ways as far as I can tell. There's just no display of measurement.
-
Just got a digital copy of the HP 3456a service manual that is actually readable, including the scope traces and gray areas on circuit board diagrams (i.e., a proper high-resolution scan). I am going to go through it again and see if I can figure out what to test next.
-
The board with the LED's and buttons is very simple to remove. You will see 4 or 5 screws on the top and bottom of the frame of the meter. The ones on top might be covered by the plastic strip if its still there. There is also one larger machine screw on the top inside of the case. Remove the screws, the ribbon cable, and push the whole front panel out from the inside. Watch the red plastic LED cover as it can slide out from the side. Once the whole assembly is out, you can remove the board from the front face, there are about 8 screws holding it on. Then you can inspect the LED's or other potential damage to the board. If nothing there the display driver IC, usually an Intersil branded one, on the processor board is probably shot.
-Mark-
-
The board with the LED's and buttons is very simple to remove. You will see 4 or 5 screws on the top and bottom of the frame of the meter. The ones on top might be covered by the plastic strip if its still there. There is also one larger machine screw on the top inside of the case. Remove the screws, the ribbon cable, and push the whole front panel out from the inside. Watch the red plastic LED cover as it can slide out from the side. Once the whole assembly is out, you can remove the board from the front face, there are about 8 screws holding it on. Then you can inspect the LED's or other potential damage to the board. If nothing there the display driver IC, usually an Intersil branded one, on the processor board is probably shot.
-Mark-
Thanks Mark, I'll follow your advice and see if I can get that display board out. Half the trouble I am having testing is accessing that board to test the signal levels at the various displays.
-
U28 is common to the display and button LEDs. I would start there since you suspect the unit seems to be otherwise working. It's possible, but I doubt U29 and U30 are both dead.
Definitely do inspect for physical damage, but noted the "RATE" LED is blinking indicating the ribbon cable is probably intact.
You said you already measured a signal on U28 CLK (pin 11). Are the outputs (pins 2, 5, 6, 9, 12, 15, 16, 19) changing?
-
I just found a copy of the schematic, and MarkL above is right, the U28 latch IC looks to be a common denominator with both the annunciator LED's and the segments, and probably where the trouble is at. If so, it's an easy one to find on fleabay, SN74LS377N.
-Mark-
-
I will take a deep look at U28. I managed to get the display board out and inspect it and it looks clean as a whistle. Also, as I was power cycling several times, I noticed that during powerup sometimes some of the key leds and 7-segments very briefly flash, before they are turned off properly as part of initialization. This tells me two things: 1) Odds are the displays are working and 2) The ribbon cable is good. Now back to A4 and U28. I'll report my findings...
-
U28 update: OK, it looks like the clock frequency is in kHz and not Hz and my meter can barely keep up in its frequency counter mode.
Anyways, pin 11 is getting a clock signal. But, it seems like the Ds are not getting to the Qs on the chip. I also made sure that U28 is properly getting power at Vcc/Gnd and it is.
Now the weirdness: Pin 1 on the schematic in the service manual should be connected to ground, but in my unit pin 1 shows a steady 5.0 V. Since pin 1 is an inverted enable for the entire chip, why the heck is it showing 5 V and not ground. I don't know where the 5 V signal is coming from, because I can only see the top part of the (plugged in) board and on this side, nothing is connected to pin 1. This may be the source of the problem. I am going to see if I can figure out why in the heck pin 1 of U28 is high. Is it a cold solder joint and its high by default (but if that is the case it should show a voltage that is slightly less than 5 V), or is something actually feeding it +5 V and disabling the octal D flip-flop? The schematic in the service manual shows it just connected directly to ground - weird.
Sorry, the above is incorrect, I mistook U25 for U28. The correct update:
U28 pin 1 is low as it should be. Power is correctly going into the chip. The clock is present at about 46.6 Hz as stated before. The weirdness is that the D inputs are being fed signals at crazy frequencies 160-320 kHz depending on the particular D pin of the flip-flop and nothing seems to be transferring to the Qs.
All of the Qs on the U28 D flip-flop are at a steady 4.45 V DC.
-
First, good luck and keep us posted. I think you are on track. The manual is fairly detailed on dead front panel troubleshooting, I think its labeled Service Group A.
I'd like to know which vintage 3456a this is, what is the first 5 digits of its serial number? If via eBay, do you still have a link to it if you don't mind sharing? Just curious about the shape this unit is in.
-
If I were you, I would wait with this project, until I'd have an oscilloscope and maybe a small logic analyzer. A second 3456A is nice too, since you can swap boards and thereby narrow a fault to a particular board.
A frequent fault are the caps, swap em all with bigger ones, I used 63V and 100V caps with same capacity. Then comes the Mostek RAM and ROM
-
The signals going in to the latch IC come from the U15 main controller uP, so the buck would pretty much stop there. As you can hear the meter going through the self-test routine, it’s safe to presume the processor is probably OK, at least we can hope so. You can get like 20 of those latch IC’s for $15.00, much cheaper than buying any signature logic or other test equipment, and just swap one in and see if it fixes it rather than go in to troubleshooting outland. But I’m a gambling man that way where others might not be. ;)
-Mark-
-
OK, my current plan of action is to yank U28 out and test it on a breadboard. If it fails, I'll order up some 74ls377 chips and replace it. If it passes, I'm not sure what to try next. I've never desoldered from a double-sided circuit board before, so that will be a first for me. I do have a bulb suction based desoldering iron - hopefully it will be able to handle the job.
Any desoldering advice for taking a 20-pin DIP off of a two sided circuit board that you guys can give would be appreciated!
For whoever asked about the unit specifics, it's a 2201A1xxxx that apparently went dead before the electrolytics gave out (since the voltages are all great - ripple aside, since I have no way to test that).
-
These can be a major PITA to remove from a double sided or multi-layer board without a really good vacuum de-soldering system, though those with more patience than me have done it successfully. For me it's usually sacrificial- with a new IC in hand I clip the old IC out and unsolder each lead end individually, then suck the old solder from the holes with a vacuum bulb. Adding some clean solder before sucking it out makes it easier and cleaner. This way I know I won't risk damaging the board traces.
-Mark-
-
Well hello! I pull A4 out and start forming my desoldering plan of attack and what do I see - rosin around the pins of U17. I take a closer look and it looks like someone got their intro to soldering lesson replacing that puppy. Cold solder joints on half the pins, especially on the pads on the front of the board. I think I am going to give that chip some soldering TLC first and see if the unit magically springs back into action. Actually, scratch that - I'll just test connectivity on both sides and see if those solder joints are actually cold or not.
-
You'll probably destroy more than you repair without first diagnosing the actual problem. And without some basic test equipment, that's impossible. My guess is that these chips are completely OK, except maybe the cold solder joints.
BTW, a heat gun is a perfect tool to remove large chips from multilayer boards, gently push two small screwdrivers under the chip to remove and then heat the chip while gently wiggling the screwdrivers under the chip. Another nice trick is to run some wire under the chip to pull it out, using the same method as above, if there's no room for screwdrivers or pliers.
-
OK, time for some "pics or it didn't happen" action:
A4/U17 front and back:
Front:
(http://s24.postimg.org/dk4lox1hd/u17front.jpg)
Full screen image link (http://postimg.org/image/dk4lox1hd/)
Back:
(http://s27.postimg.org/p6n2ch3lr/u17rear.jpg)
Full screen image link (http://postimg.org/image/p6n2ch3lr/)
And as I was looking around the board some more, :wtf: check this botch out! Did somebody slip with the soldering iron while doing the piss poor job of soldering U17, because this required some friggin' force? If this appeared in one of Dave's EEVblog videos, he'd probably spend five minutes making fun of it :palm::!
A4/U20:
(http://s29.postimg.org/qq2o5qjzn/u20front.jpg)
Full screen image link (http://postimg.org/image/qq2o5qjzn/)
And, yes, pins 2 and 3 are shorted even though it is not certain from the pic.
Update: The joke's on me, pins 2 and 3 are connected with a trace on the bottom of the board, so they should be shorted. The question is, are the pins dislodged from the DIP package being as contorted as they are. Time for more testing...
Looks like I've got bigger fish to fry than replacing U28...
-
That's the wonderful world of surprises from technicians past when you open something up......
Hankie says: I did the best I could, I promise, but that big hot soldering pencil scares me :-[
(https://fbcdn-sphotos-a-a.akamaihd.net/hphotos-ak-xfa1/v/t1.0-9/10690035_848855938467641_6623930284238107581_n.png?oh=f3eda02e1cdd35adb985a1fd28d2a586&oe=54C851D6&__gda__=1421448416_b149b77494eaf76ce577bbe3feb36b12)
-Mark-
-
After much testing and perhaps sadly, both A17 and A20 seem to be functioning normally. Oh well, I got my hopes up and now I'm back to U28, which looks perfect, is soldered properly on both sides, but seems not to work right.
-
You've got quite a mess there.
U20 pins 2 and 3 are supposed to be shorted (altogether with 10 and 11 too). Look at the schematic A4-10 (fig 8-71 on page 8-86).
Cutting and then (neatly!) resoldering pins on DIPs is sometimes used to isolate the pin(s) when debugging a problem. Looks like someone used a hatchet on U20.
As ugly as U17 looks, from the pics it looks like all the pins are connected.
U20 pin 4 output is called "DISPLAY READY-H". Its operation is described in sec 8-271 and could be affecting the CPU writing to U28. I would take a look at pin 4. It's also interesting that someone has been poking around in the display circuit. Maybe repair of the problem has already been attempted. Something to keep in mind.
I'm still highly suspicious of U28, but I agree with the others that without a scope or LA to verify if it's being written, we're operating blind. If you don't have any options to get more equipment on the job, I'd be tempted to try replacing it.
On removing DIPs, if you don't care to save the chip, cut all the pins and remove them from the PCB one at a time. Then suck the remaining solder out of each hole It's the best method to minimize the risk of damaging pads. Edsyn SoldaPullt is a good solder sucker without having to invest in a heated desoldering station.
-
Ah, this is making more sense. My big guess is it was sold as whomever owned it before couldn't get it repaired, which you are now giving a go.
The problem I find is that its unclear what the original fault was that triggered the repair and given its appearance, it may not have been a thorough troubleshooting job. Thus, the device's true fault can still remain elsewhere and the attempted repair may have added another problem.
I second the others who suggest you'll need a scope to make best use of your time but I'd also send lowimpedance a pm since he's repaired a number of these in the past.
https://www.eevblog.com/forum/testgear/hp3456a-service-notes-for-owners-info (https://www.eevblog.com/forum/testgear/hp3456a-service-notes-for-owners-info)!/
As for the serial number, sequence before 2201A0**** had factory service notes for faults that are similar but its not where the original owner did the repairs. I've seen one or two old 3456a act this way, even above that serial number and without any repairs, a reset or power down once will 9/10 get the machine back up and running, but then the problem recurs later. Reseting it > 2x, with the above fault but without repairing it, got the machine running 100% of the time so often the service note repairs were not done.
-
I was typing this as Saturation posted, but more or less saying the same thing..... You have to hope the hack in there previously wasn’t troubleshooting some other problem and ended up with the blank display as the fruit of their labor before giving up on it.
-Mark-
-
Yes this is turning into a high risk repair job. I got the unit for $80 though including shipping, so if I'm willing to spend maybe another $20 before calling it a day and selling off the parts (or the unit as a whole to someone with a scope)...
-
In which continent do you live? In case it's Europe, I could repair it for you for some cash or would buy it before it ends up as scrap
-
U28 removal update. I have successfully removed the chip. Thanks to everyone who suggested to clip the pins prior to suction desoldering them.
Pics or it didn't happen:
Front:
(http://s17.postimg.org/pe0vbwk7v/u28removed_front.jpg)
Full size image (http://postimg.org/image/pe0vbwk7v/)
Rear:
(http://s27.postimg.org/dp5vtcepr/u28removed_rear.jpg)
Full size image (http://postimg.org/image/dp5vtcepr/)
I hope I did a reasonable job and didn't damage any attached pads too much.
Update: Replacements have been ordered: eBay item link (http://www.ebay.com/itm/121260532317)
-
Hey SharpEars, I've got an Xprotolab I can send you, if it would help. It's not much (2x200kHz), but it's something!
Sent from my Smartphone
-
Your desoldering looks very clean. Should be easy to drop and solder the new DIP in. Will keep fingers crossed for it to work :D
-Mark-
-
For whoever asked about the unit specifics, it's a 2201A1xxxx that apparently went dead before the electrolytics gave out (since the voltages are all great - ripple aside, since I have no way to test that).
I realize the debugging focus is elsewhere at the moment, but I wanted to point out that you can measure ripple with your DMM set on AC. It will measure just the AC component in the presence of DC. The limitation is that you won't see any readings if the ripple is out of its frequency range (like with a switcher), but it will at least show you any line frequency components from bad filter caps in a linear supply like this one.
-
I realize the debugging focus is elsewhere at the moment, but I wanted to point out that you can measure ripple with your DMM set on AC. It will measure just the AC component in the presence of DC. The limitation is that you won't see any readings if the ripple is out of its frequency range (like with a switcher), but it will at least show you any line frequency components from bad filter caps in a linear supply like this one.
Interesting, I'll have to try that. I've always thought that if you set the setting to AC for a DC circuit, you'll just get some bogus value that assumes your DC is a sine wave when it isn't. I didn't know that there would be a capacitor in series with the circuit (or something) to filter out the DC. Speaking of which, maybe a capacitor in series with the DC (and some alligator clips for connecting it) is a good way to measure for ripple - not sure if that would work though or what value of capacitance to use if it would.
-
Your desoldering looks very clean. Should be easy to drop and solder the new DIP in. Will keep fingers crossed for it to work :D
-Mark-
Mark, thanks for the kind words. I literally did this in about 15 minutes total. The idea to clip the leads of the original IC made the job so much easier. I used a special very fine nose diagonal clipper/cutter to get in between the pins. I think just the (careful) clipping took a quarter of the total time. Then it was all desoldering iron with bulb and technique to get me through the rest. Most of the time was spent on pins that were rotated/wedged in the through holes (i.e., difficult to suck out, requiring multiple attempts and unfortunately damage to unconnected pads).
I tried to apply heat to the side with unconnected pads, but sadly many of the pins could only be sucked out from the top due to being thicker and wider on the front as compared to the back. Sometimes I had to apply heat to both sides to get all of the solder out. Sadly, if you look at the bottom right pin in the (from the bottom) picture, I partially damaged the trace going to it. I will have to fix that up with solder when I put in the new chip and check for a proper low resistance connection that leverages the full width of that (power?) trace.
-
I realize the debugging focus is elsewhere at the moment, but I wanted to point out that you can measure ripple with your DMM set on AC. It will measure just the AC component in the presence of DC. The limitation is that you won't see any readings if the ripple is out of its frequency range (like with a switcher), but it will at least show you any line frequency components from bad filter caps in a linear supply like this one.
Interesting, I'll have to try that. I've always thought that if you set the setting to AC for a DC circuit, you'll just get some bogus value that assumes your DC is a sine wave when it isn't. I didn't know that there would be a capacitor in series with the circuit (or something) to filter out the DC. Speaking of which, maybe a capacitor in series with the DC (and some alligator clips for connecting it) is a good way to measure for ripple - not sure if that would work though or what value of capacitance to use if it would.
Well, I'll update my statement to "should". On old analog meters what you say is the true about displaying a bogus value. But on a modern DMM (at least all the ones I've worked with), there's a blocking capacitor or filtering circuit as you describe.
If you have a true RMS meter, you'll get a reading of the actual AC ripple. If the meter has no such claim, you'll probably get some reading computed from the peak with the assumption it's a sine wave. In either case, you'll get some kind of a reading and can determine if there's a possible problem with a filter cap.
If you're unsure of your meter try measuring AC voltage on a battery. You should get 0V.
If you use your own blocking cap beware of DC leakage through the cap. It may be significant given the high impedance of modern meters. This can be minimized by shunting the meter with a lower resistance (say 10k or 100k), but then you've got a frequency-dependent RC circuit to take into account.
-
Raise your hand if you bet on U28 on board A4 :-+ :
(http://s8.postimg.org/y3w7ibuv9/3456a_repaired.jpg) (http://s8.postimg.org/y3w7ibuv9/3456a_repaired.jpg)
And it's calibrated (at least for low volt DC, 100 mV - 10 V)! Now, what do I do with the other nine 74ls377 chips?
Virtual beers to all...
Update: Just checked full DC calibration and it's friggin' perfect (I'm talking <5 ppm for all ranges save the 100 millivolts range that's off by around 10 uV and jumps quite a bit at 6 1/2 digits - it's a hard range to test). This was with the voltage standard I got last week from voltagestandard.com (within 1 week of calibration via a NIST certified Agilent 3458A). The 3456A matched all settings but the 100 millivolt one to within 5 ppm, the 5 and 10 V ones to within 2 ppm. The 5-digit Data Precision voltage standard in the pic is a tiny bit off. I can use the 3456A to calibrate it - the irony!
So, best $87 ($80 for the HP 3456A and $7 for the ten replacement 74ls377 chips of which I used one) I ever spent! OK, I guess the 8-10 hours I spent on the troubleshooting and repair have to be mentioned, but it was a good learning experience for me.
Update 2: Just measured a 480 ohm 1% metal film resistor to get a feel for the state of the resistance section and the reading came within 4 ppm of my Fluke 8506A. (476.404 ohms on one and 476.408 ohms on the other).
I am beginning to get the feeling that this unit was calibrated in the not too distant past and that that 74ls377 blew on the main board and it wasn't used since. I just wish it was in better cosmetic condition, but hey beggars can't be choosers :) .
Update 3: Found the "calibration on the unit's" Achilles Heel, the AC. It's showing about 1.5 V too high at 120 V US line voltage, compared to three other meters I have, which are all within a quarter volt of each other. I guess that can use some calibration in the future, but I don't measure AC all that much and have no way to really calibrate it either.
-
good job
-
:clap: Most excellent. Also a new reference for any such failure in the future. I haven't found a similar issue searching the net in the past many years, i.e., failure caused by that flip flop.
-
Congratulations on the repair and thanks for the details.
Sounds like a real bargain.
-
:clap: Most excellent. Also a new reference for any such failure in the future. I haven't found a similar issue searching the net in the past many years, i.e., failure caused by that flip flop.
I am very happy that in this thread a fairly thorough troubleshooting/repair guide has been created for others with display issues. Almost all of the other HP 3456A repair info on the net seems to deal with bad capacitors and PROMs and most of the info is rather vague (i.e., "I replaced the caps on the power supply board and now it works!"). Surely, that's not the only thing that can go wrong on these units after 25-30 years.
-
Your initial characterization of the problem was key to narrowing down the possibilities immediately. Getting a complete understanding of what works and what doesn't, and how it behaves with various stimuli is time well spent. Nicely done.
I never take anything off the "possibilities list", but your description put power supply related issues very low on the list. Absolutely check all power supply levels before proceeding with more in-depth debugging, but the knee-jerk reaction to replace all caps just because it's old equipment is not the right approach.
I really like these old 3456A meters. In their prime in the 80's, they were touted by HP as transfer standards and sold for US$4100 (1987 catalog). They're built like tanks and they don't have any NVRAM to go bad. You just have to have the bench space for them.
Enjoy!
-
Nice work :) It's always satisfying when you get something working again. It can and does happen occasionally, but it's very rare if electrolytics in 1980's HP equipment need to be replaced, and I would never do it just as a precautionary measure. I will remove and test them if they raise suspicion, and only if they fail for being out of tolerance or high ESR or current leakage do they get replaced. If they test OK they go right back in. For the bunches of 1980's HP equipment I have, only a few have ever needed replacement. In fact a piece of HP gear I'm working on now from 1962 has only needed about 50% of its electrolytics replaced, I think that from the piece of equipment sitting dormant for so long.
-Mark-