Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
HP 3456A Problem Solved
|
A few weeks ago, I reported a problem with my 3456A where it would fail at startup with "--------" on the readout and fail Test 3. I reseated one of the microprocessors and the problem seemed to go away. Since Service Note 3456A-13A states to discard the chip sockets for all three microprocessors and solder them to the board, I figured the problem was in those red chip sockets and decided instead to replace them with good quality machined sockets as I believe others have done as well.
Unfortunately, after replacing all three sockets, the problem persisted so I started down the troubleshooting steps in the manual and one of the very first things it says to check is +5V on U13 (one of the microprocessors). I was surprised to find that to be at 4.7V or, possibly, a few hundredths under that. Since I had replaced all the electrolytics in the power supply, an off voltage like that was something I did not expect at all. Checking that rail at the power supply board, though, showed it to be around 5.1V. Knowing that Service Note 3456A-14A addresses an issue with faulty grounds (via some riveted ground lugs), I decided to ensure the In Guard board was properly grounded so I jumpered one of the GND test points on that board to one of the GND pins on the power supply board and the meter immediately started working. If I remove the jumper, the meter stops in its tracks and restarting it produces that same "--------" and failed Test 3. I can detect about 0.5 ohms from the pin that connects to the power supply and the pin that connects GND to the In Guard board and apparently that's just enough resistance to drop the voltage low enough to cause the processor to stop so better grounding is in order. While I'll probably go with the Service Note's recommendation and replace the rivets with #2-56 screws, nuts, and lock washers, I'm wondering why it might not be a better plan to simply route ground wiring from pin to pin as needed. I see where that was done from the factory from one or two other pins and it would seem that a separate, soldered wire might be better off since tiny screws might tend to need cinching up from time to time to ensure they're still tight. Has anyone here performed the rivet replacement? I suppose I'll stick with that since it's recommended (and will include a drop of DeoxIT under each lug as well) but was just curious as to whether a separate wire would be a better plan. Thanks, Barry - N4BUQ |
|
Bob Albert
Larry, I had a similar symptom with my 3456A.? It was intermittent when I got it, but since it worked a good part of the time I decided to keep it and live with it. After a while the intermittent stopped and it's been perfect ever since.? I decided it must be that it hadn't been used enough and some oxide built up in critical places. After reading your post, I know what to look for in case it acts up again. Bob
On Saturday, March 7, 2020, 08:13:44 PM PST, n4buq <n4buq@...> wrote:
A few weeks ago, I reported a problem with my 3456A where it would fail at startup with "--------" on the readout and fail Test 3.? I reseated one of the microprocessors and the problem seemed to go away.? Since Service Note 3456A-13A states to discard the chip sockets for all three microprocessors and solder them to the board, I figured the problem was in those red chip sockets and decided instead to replace them with good quality machined sockets as I believe others have done as well. Unfortunately, after replacing all three sockets, the problem persisted so I started down the troubleshooting steps in the manual and one of the very first things it says to check is +5V on U13 (one of the microprocessors).? I was surprised to find that to be at 4.7V or, possibly, a few hundredths under that.? Since I had replaced all the electrolytics in the power supply, an off voltage like that was something I did not expect at all.? Checking that rail at the power supply board, though, showed it to be around 5.1V. Knowing that Service Note 3456A-14A addresses an issue with faulty grounds (via some riveted ground lugs), I decided to ensure the In Guard board was properly grounded so I jumpered one of the GND test points on that board to one of the GND pins on the power supply board and the meter immediately started working.? If I remove the jumper, the meter stops in its tracks and restarting it produces that same "--------" and failed Test 3. I can detect about 0.5 ohms from the pin that connects to the power supply and the pin that connects GND to the In Guard board and apparently that's just enough resistance to drop the voltage low enough to cause the processor to stop so better grounding is in order. While I'll probably go with the Service Note's recommendation and replace the rivets with #2-56 screws, nuts, and lock washers, I'm wondering why it might not be a better plan to simply route ground wiring from pin to pin as needed.? I see where that was done from the factory from one or two other pins and it would seem that a separate, soldered wire might be better off since tiny screws might tend to need cinching up from time to time to ensure they're still tight. Has anyone here performed the rivet replacement?? I suppose I'll stick with that since it's recommended (and will include a drop of DeoxIT under each lug as well) but was just curious as to whether a separate wire would be a better plan. Thanks, Barry - N4BUQ |
|
Which versions of the 3456A does this apply to?? Or all three(?) -Dave
On Saturday, March 7, 2020, 08:13:45 PM PST, n4buq <n4buq@...> wrote:
A few weeks ago, I reported a problem with my 3456A where it would fail at startup with "--------" on the readout and fail Test 3.? I reseated one of the microprocessors and the problem seemed to go away.? Since Service Note 3456A-13A states to discard the chip sockets for all three microprocessors and solder them to the board, I figured the problem was in those red chip sockets and decided instead to replace them with good quality machined sockets as I believe others have done as well. Unfortunately, after replacing all three sockets, the problem persisted so I started down the troubleshooting steps in the manual and one of the very first things it says to check is +5V on U13 (one of the microprocessors).? I was surprised to find that to be at 4.7V or, possibly, a few hundredths under that.? Since I had replaced all the electrolytics in the power supply, an off voltage like that was something I did not expect at all.? Checking that rail at the power supply board, though, showed it to be around 5.1V. Knowing that Service Note 3456A-14A addresses an issue with faulty grounds (via some riveted ground lugs), I decided to ensure the In Guard board was properly grounded so I jumpered one of the GND test points on that board to one of the GND pins on the power supply board and the meter immediately started working.? If I remove the jumper, the meter stops in its tracks and restarting it produces that same "--------" and failed Test 3. I can detect about 0.5 ohms from the pin that connects to the power supply and the pin that connects GND to the In Guard board and apparently that's just enough resistance to drop the voltage low enough to cause the processor to stop so better grounding is in order. While I'll probably go with the Service Note's recommendation and replace the rivets with #2-56 screws, nuts, and lock washers, I'm wondering why it might not be a better plan to simply route ground wiring from pin to pin as needed.? I see where that was done from the factory from one or two other pins and it would seem that a separate, soldered wire might be better off since tiny screws might tend to need cinching up from time to time to ensure they're still tight. Has anyone here performed the rivet replacement?? I suppose I'll stick with that since it's recommended (and will include a drop of DeoxIT under each lug as well) but was just curious as to whether a separate wire would be a better plan. Thanks, Barry - N4BUQ |
|
Barry
toggle quoted message
Show quoted text
The argument for restoring the original ground points has often to do with "ground loops" inside the unit. While not obvious most measurement levels, there are possible unexpected anomalies or even errors you are unaware of which can occur specifically when making very low readings, etc Dave On 3/7/2020 11:13 PM, n4buq wrote:
A few weeks ago, I reported a problem with my 3456A where it would fail at startup with "--------" on the readout and fail Test 3. I reseated one of the microprocessors and the problem seemed to go away. Since Service Note 3456A-13A states to discard the chip sockets for all three microprocessors and solder them to the board, I figured the problem was in those red chip sockets and decided instead to replace them with good quality machined sockets as I believe others have done as well.-- Dave Manuals@... www.ArtekManuals.com --
This email has been checked for viruses by Avast antivirus software. |
|
The Service Note states Serial Number Prior to 2201A06299. I haven't checked the serial number on mine but would be surprised that they changed the construction of those grounds. Maybe so.
toggle quoted message
Show quoted text
Thanks, Barry - N4BUQ ----- Original Message -----
From: "Dave Seiter" <d.seiter@...> |
|
I would tend to liberally solder ground wires for distribution. It may be more obvious that only several points are needed. Drilling on a HP circuit board seems like a really bad idea. Especially given the age. Unless maybe they are safely at the edge. But thanks to the thread when the day comes on my 3 X 3456a's I know what to watch for. Thank you Paul |
|
The mod doesn't involve drilling a circuit board. There are header pins that are mounted with a ground lug to the appropriate pin. All the boards are pretty easily removable (the front panel board is a but more effort and most likely doesn't need to be removed for this) so the drilling is done with the boards removed.
toggle quoted message
Show quoted text
I'm still a bit curious as to why that mod mentions a serial number and below. It makes me think they did something different at that point to address this problem. Thanks, Barry - N4BUQ ----- Original Message -----
From: "paul swedberg" <paulswedb@...> |
|
That is precisely why they mention a serial number and below.
toggle quoted message
Show quoted text
HP, and others, were on a process of continuous improvement. -Chuck Harris n4buq wrote: The mod doesn't involve drilling a circuit board. There are header pins that are mounted with a ground lug to the appropriate pin. All the boards are pretty easily removable (the front panel board is a but more effort and most likely doesn't need to be removed for this) so the drilling is done with the boards removed. |
|
I think they may be other places, but a web search shows them from this site:
toggle quoted message
Show quoted text
Some links on that page point to the service note files: As an update, with the boards removed, I took a pair of needle-nose pliers and very gently rotated the ground lugs, every so slightly (just enough to see movement) around each rivet. The resistance appeared to drop to a consistent value for all ground lugs (primarily the lead resistances). I reassembled the boards and when I powered it up, it worked. I measured the +5V at the CPU I measured yesterday and it was back up to 5V as well as the other 5V test points on the other boards. I'm not convinced that's a very reliable "fix" and still plan to replace the rivets but did find it interesting. Thanks, Barry - N4BUQ ----- Original Message -----
From: "garp6" <hrgerson@...> |
|
Recently i found the same problem with two HP 3437A system voltmeters that we have. After one of them showed occasional digital errors (e.g. failure of display multiplexer) i looked inside. While the power supply generated 4.9 V at 1.5 A, some corners of the digital board received only 4.73 V or so, with a large part of the loss along the ground network. I just added two wires (Gnd and +5V) from one corner to the opposite one as redundant connection and the losses are less than 20 or 30 mV now. The faults are gone.
The digital boards of that time had very few buffer caps. In the 3437A i measured up to 1 Vpp noise on the +5 V net. When looking at the 3456A digital boards (especially those with CMOS logic) i am missing buffer caps that are so common nowadays, especially near the isolation interface digital receivers. Regards, Dieter |
to navigate to use esc to dismiss