Project P3459, building high-end DMM by using scrap 3458A parts

Intro and motivation

This is work in progress project, updates will be posted frequently.

It’s been a long while since we looked at famous 8½-digit DMM HP 3458A last time. Good eight (yes, 8!) years passed since acquired and repaired first broken HP 3458A. All that was documented and shared in first online 3458A repair article back in 2015. That 3458A is still going strong with non-stop 24/7 operation every day, delivering DCV measurements and handling metrology experiments daily at Since 2015 xDevs helped and repaired more than 10 different HP and Agilent era 3458As and have three healthy units in constant use all year long with periodic performance verification and stability checks.

Links below are useful to refresh reader’s memory to some of the past HP3458A units publications:

Article here is aimed to show us HP3458A glory some more. This time is not about some easy epay-special HP 3458A repair with board and Elantec comparator swaps. This time we will raise the bar and try to rebuild one fully working 3458A out of destroyed scrap donor and reject boards and parts from past 3458A repairs. This is also truly a community project, as number of people donated their busted components and bits to make it happen.

It will be a very interesting build and will give us a chance to dig into HP3458A operation like never before. Key motivator for this new project, codenamed P3459 is NOT just repair or rebuilding yet another boring 8½-digit HP3458A DMM but actually test some of the insane ideas on 3458A hardware. Few of these ideas are listed below for a teaser:

  • Replacement of the LTZ1000A reference on A9 Voltage Reference to lower noise ADR1000 zener chip
  • Testing different lower-noise modern opamps on A3 A/D converter board instead of original LT1001
  • Removing input protection on DCV ranges to test noise limitations of input front-end
  • Testing different lower noise preamps for DCV path on A1 DC board
  • Trying battery powered operation of 3458A, eliminating transformer and front panel display
  • Testing temperature coefficients of various circuits and functions
  • Testing better voltage divider for 100 and 1000 VAC/VDC ranges
  • Test operation without fan, air draft covers and other voltnut-approved paraphernalia
  • Testing Ohm current source noise and stabililty, trying to improve it for better resistance transfers.
  • Testing electronic front/rear inputs switching with relays to automate comparisons, like in Datron 1281/Fluke 8508A/etc.

All of these silly ideas require quite intrusive modifications of original circuitry and lot of high-risk experiments. It would be heart-breaking to waste a good fully-working HP3458A DMM to answer these questions. However with P3459 project rebuilt from scrap parts such experiments would be justified. In worst case we just made bad rejected parts … well, just still bad. No loss of good 3458A would be involved.

But before we may begin, main job is to bring all this random assembly of stuff to a working condition so it could function as normal HP3458A DMM.

As always on, all photos in this article are clickable for high-definition version.

People are still wondering with question “Is it worth to buy broken dead 3458A for XXXX$?. Well, simple answer is – No, it’s a big risk. If you asking yourself about this, then you are not ready for it. Buying broken or used 3458A for “good price” and fixing it might not be an easy way of getting good 8½ DMM, when multiple hidden costs considered. If unit stated as “untested/unknown/don’t know how to test” and not to have shown self-test result, it could have hardware failures. Paid cost for such a unit WITH proper repair can easily exceed cost of buying a fully tested unit. If unit’s photo shows it as powered up but you can see “ERR” on display, it also means there is fault detected. Keysight have standard 3458A repair + recalibration cost at [$2660, 2015 price] now $3.4k USD. Calibration alone of fully functional unit is likely to be over $1000 USD. Also old units manufactured earlier than 2005, would have Dallas NVRAM batteries already near end of life, with risk of losing calibration data any day. So even if you buy meter for “cheap” 1000$, you can be easily be held back by 2000-3000$ more to fix it and get calibrated.

Damaged parts (and current status):

  • A2 board missing TBD
  • A1 board damaged, untested TBD
  • A3 board not working, U180 damaged TBD
  • A4 inguard PSU board Good
  • A5 Processor board TBD
  • A6 board is broken No power to CPU A5 board Fixed
  • Missing fuse holder on A6 donor board TBD
  • Missing fiber optic cables. TBD
  • Damaged plastic front panel TBD
  • Front panel has parasite pixels and weak VFD TBD
  • Mains power transformer. Good
  • No legs Fixed
  • Old electrolytic capacitors on A6,A4 PCBs TBD
  • Prone to explosions Schaffner FN 323-3/05 EMI/mains filter TBD


Redistribution and use of this article or any images or files referenced in it, in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

  • Redistribution of article must retain the above copyright notice, this list of conditions, link to this page (/fix/p3459/) and the following disclaimer.
  • Redistribution of files in binary form must reproduce the above copyright notice, this list of conditions, link to this page (/fix/p3459/), and the following disclaimer in the documentation and/or other materials provided with the distribution, for example Readme file.

All information posted here is hosted just for education purposes and provided AS IS. In no event shall the author, site, or any other 3rd party, including HP/Agilent/Keysight be liable for any special, direct, indirect, or consequential damages or any damages whatsoever resulting from loss of use, data or profits, whether in an action of contract, negligence or other tortuous action, arising out of or in connection with the use or performance of information published here.

If you willing to contribute or add your experience regarding HP/Agilent/Keysight instruments repairs or provide extra information, you can do so following these simple instructions

Manuals references

We would also highly recommend book Sampling with 3458A, ISBN 978-961-94476-0-4 written by recognized metrology expert Rado Lapuh. Book showcases performance of 3458A functions and practical programming and application examples for high precision AC and DC sampling with HP3458A DMM.

Service notes

There is also number of service notes/engineering changes were published during years of 3458A’s lifecycle. If your unit is old, worth to check if any of them required to do.

Rebuild workflow

Before starting any repair, good idea to read thru repair manuals, to get idea of board functions and purpose. HP has both assembly and component level repair manuals, which include full schematics and board assembly drawings. AFAIK it’s the only 8½ DMM which have complete schematics. I still remember repair efforts and countless hours without schematics, deciphering circuitry in Keithley 2001 DMM.

HP3458A Prototype #74 unit history, in memory of George Hnatiuk

Before we go deep into hardware and measurements I’d like to dedicate this project in memory of George Hnatiuk whom we lost on September 12, 2022. George was one of the key Hewlett-Packard engineers on 3458A project back in 1980s responsible of A1 DC board design. Thanks to his and team expertise at HP we today have still unrivaled DC performance of 8½-digit 3458A DMM, even 40 years later today.

I’ve had privilege to discuss metrology, DMM projects and engineering in general with George in 2021 and 2022. He was very supportive to test equipment community, helped countless number of people online and offline with their HP3456A, HP3457A and other HP-era equipment. So I hope a little insight into his prototype HP3458A engineering unit #74 can be helpful to inspire today’s young generation to pursue career, sharing and transferring knowledge and passion in test equipment instrumentation and measurements industry.

George kindly shared some photos and videos of his early engineering prototype meter. It was unit that George used for development and testing various ideas back in 1983-1985 year during the development. This unit had engineering prototype number #74 and is very different to production 3458A hardware we are all familiar with.

His prototype unit was based around HP 3457A chassis with customized 3457A-style front panel. Overall design is somewhat close to modern 3458A units, but the circuit boards are far from being same. This meter was early enough not to have ACAL functionality just yet, but otherwise it was still working and functional at the time in 2021.

VFD glass is almost same as current HP3458A but extra status annunciations are also marked on the display lens. Buttons are following old 3457A style with plastic caps instead of modern rounded larger rubber pads. At power on EP74 3458A does not do self-test but gives a message “READY FOR ACTION” briefly.

Interesting to note that in resistance mode instrument had ability to show full 8½-digit resolution readout. Modern 3458A firmware truncate value to smaller resolution on display, even though have full 9½-digit measurement performed internally and provided over GPIB bus.

Main processor A5 board is somewhat close to initial production 3458A DIP-chip boards, but have main processor in ZIF-socket, two oscillators and number of dead-bugged logic chips. Interconnect to A6 board has special blade connector, unlike standard 2.54mm pitch pin header we have on current PCBAs. Just like with production processor boards this prototype used same pair of Dallas DS1235 and DS1220 NVRAMs for RAM and calibration ROM data storage.

A4 inguard circuit board use large TO-3 packaged regulator mounted on heatsink and had some additional capacitors. This board is simple enough with just few LDO regulators, so not much other interesting stuff happening here.

Even relays used for current range switching on A1 are marked with “Sample” stickers, hinting for customized or special order relay parts in this prototype. Relays used in production DMMs are DS2E-SL2 type, but perhaps they were just starting to manufacture them back in 1985, as ones on prototype boards don’t even have any labels or part-number marking.

Instead of custom four COTO 0490-1658 reed relays found here we can find more traditional armature high power rated S2EB-L2-12V latching type. However large high-voltage relay COTO 5000-0124 and three low-thermal COTO 3500-0043 are still here. In modern A1 boards same HV reed relay has custom COTO part-number 0490-1657 designated for HPAK. Low thermal relays are marked COTO 0490-1748 on modern A1 PCBA. RoHS 03458-26511 Rev005 PCBA also have same COTO 0490-1748 K1, K3 and COTO 3500-0044 for K4.

A9 reference module is quite different to modern revision and used cheap 1% film resistors instead of Vishay S102 resistors. Otherwise it’s still original LTZ1000A closely repeating typical datasheet circut design from Linear Technologies.

On the bottom side of the instrument we can see most of the difference to modern 3458A hardware. Transformer is different with multiple wires going to all electronic assemblies. A6 outguard power supply board with Revision A5 datecoded with 1986 week 4 and has some parts hanging on the wires. DC +5V regulator have just metal braid instead of modern metal heatsink/bracket mount to chassis. Large bulk capacitor with smaller 6800 µF x 35VDC rating is radial type around the board center. Location of IEC mains input power filter and 110/230VAC switches is different too. Front panel electronics is placed on front panel board, and not on A6 in this prototype unit. DC/AC module to supply VFD with filament and high voltage is same as production units.

Famous A3 A/D board from 1986 week 44 have very different U180 hybrid chip in ceramic CDIP48 package, installed in socket. It is actually even has different part designator U111. This was due to use of much smaller less stable resistance elements together with active FET-switch die. Later this hybrid chip design was improved and changed into U180 we know today with larger separate chip resistor network placed in much larger ceramic square PGA package.

Analog A/D reference voltages +12, -12 and +5V on this old A3 board were generated same way with external LT1001 opamps U108, U109 and U107. There are some bodges with resistors and JFETs soldered onto the parts here and there, showing it was work in progress. Key integrating capacitor C102 is much larger film type in blue epoxy package. In modern boards integrating capacitor is just small ceramic MLCC type. Also instead of Elantec EL2016 comparator here we can find LT1016 with designator U110 instead in the board corner.

Gate array chip on prototype board by japanese manufacturer Fujitsu has P/N MB651305 and placed on a blue socket. It’s cool to see ES mark on the gate array PGA package, meaning it is still “Engineering Sample” grade part. Clock source for A3 operation is 20 MHz RASCO 1 oscillator manufactured by Motorola in week 14, year 1986. In-guard microcontroller manufactured by Intel P/N L8360206 also in pretty CERDIP40 package with transparent glass window to allow UV-erase for firmware updates. There is also dead-bugged DIP14 chip with some wires next to MCU, perhaps just some glue logic to handle digital control signals.

AC board also have number of test jump-wires and circuit patches, unmarked DS2E-type relays. Guard switch is mounted to chassis instead of PCB. Voltage regulators for this board was generated on elsewhere instead of local like its done now in production 3458A. That perhaps explains additional LDO on A4.

Notable capacitor build from piece of PTFE-insulated coax wire is already in place on prototype board here. Overall this old A2 board looks less busy than modern production, perhaps with smaller AC voltage functionality set in EP74 3458A unit.

Sourcing missing components


Bad PCBAs repairs


A6 PCBA Out-guard power supply

With A6 installed and connected to transformer I got no power to CPU board or display. After sniff tests and poking around the culprit was quickly discovered. L2 filter inductor next to large blue axial capacitor had broken pin and no contact to actual PCB.

And since this inductor was at the input of main +5VDC regulator, nothing else on A6 board was getting power.

Fixing was simple removing some of the plastic to get access to pin remains, soldering extension pin and populating this contraption on the PCB. In future I’ll replace the whole inductor with a new one, but for now this will work.

And since we are at it, let’s replace old electrolytic caps on A6 PCBA as well.

A quick look and on DC-AC module to generate power to vacuum fluorescent display on front panel. There are two 47 µF 63VDC rated caps and two 220 and 330 µF; capacitors at the input.

Two larger 680 µF 50V capacitors C1 and C11 already started their dirty business with leaking electrolyte. PCB solder mask started to flake near one of the pins. It’s good idea to replace old caps and clean up damaged PCB on old HP instruments, even when externally everything still looks fine and instrument working properly. Don’t delay, replace your capacitors today. :)

Negative pin on C11 position suffered more noticeable damage, perhaps because it was closer to warm regulator nearby.

I didn’t have good replacement in stock at the moment, so for now installed 1100 µF 50V and 820 µF 50V capacitors in C1 and C11 place. Later we can install better sized capacitors here.

Board is now ready for tests and should be working fine.

First power-on with A4, A5, A6, front panel and bad A3 board installed in chassis.

After quick check 3458A powered up and CPU board greeted us with first messages on display. Obviously we’ll get ISOLATOR FAILURE without any inguard electronics. So A4 and broken (drifty U180 hybrid) A3 were installed to poke around some more.

A4 inguard power supply board was tested and confirmed working, with good +5 VDC, -20V, +18V and -18VDC outputs. I don’t have original HP fiber, so here everything connected with aftermarket much longer fiber cable between A3 A/D converter board and A6 outguard power supply/guard crossing interface board.

Don’t have A2 board yet, so it is still part to source later on. 3458A would NOT operate even in DCV-only function if the A2 AC board is missing. There are some communication and signals going between A1 DC board, A2 AC board and A3 A/D converter, so all of these need to be working to allow even simplest DCV measurement from the instrument. Otherwise firmware is upset and will not give you any readings.

It is possible to build DIY-version A2 PCBA, populating whole A2 assembly with parts on aftermarket copy of original PCB layout. But I’d expect many of the original spec components be obsoleted and expensive to acquire today, so perhaps we keep that idea for Plan B, when it comes to that.

Without A1 board installed meter boots up and after quick RAM/ROM TEST goes to major fault 202, HARDWARE FAILURE: DC BOARD. So next step is getting A1 into the instrument.

A1 DC analog board repairs and tests

How that we are done with easy stuff and power supplies are working, time to get to real difficulties in the repairs. I have here one poorly handled A1 DC PCB, donated by long-time supported of project Todd. It had blown chip right in the middle, which I’ve cut out years ago with dremel. Also board is missing some of special components like shunts, high voltage resistor divider network and other minor parts used in other projects. We’ll need to put missing parts back in to make it fully function. And fix everything wrong with this board otherwise, so it might be a deep rabbit-hole here with this one.

This A1 PCBA had few MLCC decoupling caps catch fire, so perhaps suffered major overvoltage on power rails. Hole in board can be repaired with specially designed patch PCB that would restore missing connections and parts back in place.

Estimated dead parts are bootstrap amplifier/driver U12, related Q7 dual JFET transistor array and bunch of passives around the affected area. I’d think JFETs connected to sensitive node at Q7 input likely destroyed too, despite visually undamaged condition.

U3 and U4 are also were in the blast zone, so these will need to be repaired as well.

Perhaps before I connect this board to actual meter power supplies it would be better idea to power it from laboratory power supply and see if anything heats/cooks abnormally. If there are any parts that shorting power rails, we’d see it right away from excessive power draw and heat. This way there is no risk to kill good power supplies and transformer in 3458A we repairing.

To test this I’ll use quad-output HP 6629A power supply together with Keithley 2400 and Keithley 2425 Keithley SourceMeasure units. A1 board needs +5 VDC, +18 VDC, -18 VDC and -21 VDC power rails to operate.

First power up with lab power supply to A1 PCBA

Laboratory power supply connected to A1’s P3 input with provided +18, -18 and +5 VDC rails. This will allow to check if anything horribly wrong with board, before installing it into actual 3458 frame.

Connections are temporary, so just hacked the temporary adapter wiring from Agilent 6629A to 2.54mm pitch pin header. Red wire cable comes from Keithley 2400 to supply -21 VDC rail.

First attempt caused reed relay to click and shows power draw as outlined in table below.

Power rail Voltage from PSU Draw
+5 VDC +5.20 VDC 0.29 A
+18 VDC +18.20 VDC 0.09 A
-18 VDC +18.20 VDC 0.05 A
-21 VDC +21.00 VDC 0.01 A

Current compliance limit was set at 0.5 A on each rail, except -21V which was limited to 0.1 A. Nothing heated out of normal and no smokes were observed. Now I’ve installed random A9 DC Voltage reference board to see if we have good zener voltage output for A3.

And indeed we have good +7.1-ish VDC output at A9 output. Quick cheap with Flir E4 also shows reasonable temperatures around few chips on A1 board.

Next step would be to obtain some missing parts stolen from this A1 PCBA and install it into P3459 mainframe so we can have more stuff interconnected for further troubleshooting.

A4 In-guard power supply board

A4 is providing power to all isolated analog circuits and boards in 3458A. It is powered from separate windings of the power transformer and has linear regulators to provide +5, +18 and -18VDC. Also loosely regulated -21 VDC rail generated for use within auxiliary circuits on A1 PCBA. This rail protected by 65mA fuse F1.

Circuit board is designed with 2-layer FR4 PCB and generate some heat since it is 100% linear. This board is placed right in front of chassis fan to provide good thermal management in the instrument.

This module is actually only PCB that did not any repairs. This is because it was a spare board from one of the previous 3458A units.

Electrolytic capacitors in this PCBA were replaced with new modern components, just to make it slightly more interesting. I don’t expect any performance impact from this, but hopefully bit more reliability since new caps are +105 °C rated instead of +85 °C.

Capacitors that were replaced are located in position C07, C08, C09 and C17.

A5 CPU controller board repairs

A5 board designated to be used in this instrument has a “blinking cursor” issue and missing ROM/RAM chips in addition to stolen oscillators. This section is documented on video below.

A3 A/D converter repairs

No doubt A3 assembly is the heart of the 3458A performance. To try something new I’ve took one of bad drifting boards and baked to real well done condition. It was decided that in worst case I’d still have broken board and not loose anything except time. In best case we could obtain some interesting data if baking humidity out of the board change anything in behavior.

Obviously all plastic components like connectors, fiber transmitter/receiver need to be removed before whole board is heated up. Video showing reinstalling all parts and swapping damaged tantalum axial capacitors shown here:

Benchmarks and tests of restored DMM before any experiments

Lower noise A9 voltage reference PCBA with new ADR1000 6.62V zener chip

This is a combined modification of both hardware and firmware of the 3458A. Hardware part is easy enough, just take original HP A9 PCBA, replace LTZ1000A with Analog Devices ADR1000, replace resistors to match the new ADR1000 requirements and reinstall the modified module back onto A1 PCBA. Hardware part at this step is finished.

However to make the meter fully functional with much lower voltage zener reference chip we need to allow lowest limits for acceptable reference voltage from 7.1V down below 6.6V. Without this step meter will not be able to complete ACAL and unable to self-adjust all the ranges and functions.

Benchmarks and tests on final modified “P3459” DMM

Noise and resolution verification

Voltage coefficient verification

Temperature coefficient verification

Time stability

Conclusion and summary

Time log of efforts dedicated to this project

Activity Time log Dates
Source parts 20 hours 2020-2022
A6 outguard PSU repairs and tests 1 hours 31.DEC.2022
A4 inguard PSU tests 1 hours 31.DEC.2022
A3 reassembly for test, front terminals reinstall 5 hours 30.JAN.2023

Table X: Time summary on P3459 project

Component Post Shipping
03458-61603 – 2 pcs $20 USD $6 USD
Transformer $150 USD Included
Chassis $220 USD Included
03458-40224 × 2 – keycap, carbon black 17.74
03458-40225 × 1 – keycap, phantom gray 4.11
03458-43702 × 1 – pushrod guard 13.69
03458-43701 × 1 – pushrod front/rear 10.82
03458-40222 × 1 – front panel 210.00
03458-62121 × 1 – terminal block 189.00
03458-81922 × 1 – keypad 22.83

Table X: Cost summary on P3459 project

The author would like to express our appreciation to everyone contributed to this project, especially Tom, Martin, Todd and David. Discussion is very welcome thru comment section or at our own IRC chat server: (standard port 6667, channel: Web-interface for access mirrored on this page. If you have information and interesting ideas on 3458A hardware modifications not mentioned or listed in this article, feel free to provide them and xDevs will test and add them with next article update.

Author: Illya Tsemenko
Created: Dec. 29, 2022, 7:02 p.m.
Modified: Jan. 31, 2023, 2:02 a.m.