The HP3458A: King Of Multimeters For Three Decades

[Marco] looks at a lot of meters. However, he considers the HP3458A the best even though they were introduced more than 30 years earlier in 1989. Someone donated one to [Marco] but it presented some error messages on startup and exhibited erratic behavior, so he had some repairs to do.

The error codes hinted there were issues with the multislope analog to digital converter and that’s what sets the meter apart, according to [Marco]. The meter has 8.5 digits, so a normal conversion stage won’t cut it.

The good news about the problem is that it gives us an excuse to see inside the box. Each board inside looks about as complex as a modern PC motherboard. At this accuracy range, the board is covered in custom high-performance resistor networks.

A standard way of converting voltages to digital uses the time required to charge and discharge a capacitor and the required time indicates the voltage. This meter uses multiple possible slope resistors and [Marco] explains how the meter uses fast and less accurate slopes to get a rough reading and then uses slow and accurate slopes to refine the lower digits.

A custom chip has an IC and a custom resistor network and, if it fails, makes the meter virtually unrepairable without a trip to a factory service center for a new board at about $3,000. The custom chip appeared to work ok and replacing comparators that are known to fail didn’t help.

The next move? Buy all the parts you can find for the board (about $100) and simply replace everything. We liked how he removed excess component leads on the rebuild. At first, it seemed like this might work, but self-calibration failed. It appears the custom IC might have been bad, so he wound up replacing the entire conversion board.

That cleared the main error, but there were still problems with some measurements and that led to another board to repair. The circuits in question do RMS conversion for AC signals. The meter has several methods to measure RMS.

The video is a great detective story and you’ll learn a lot about what’s in a high-resolution meter. After it is all working, we get a tour of oddities like the cable that acts like a capacitor and the noisy fan.

[Marco] is no stranger to high-precision meters. We’ve looked at these meters before, too.

16 thoughts on “The HP3458A: King Of Multimeters For Three Decades

  1. I used to work with one of the engineers who designed the analog section of this. He said it was an enormous effort and took them a lot more work than they anticipated, and he thinks that’s part of the reason HP/Agilent/Keysight has started but never finished an upgraded version. Only Fluke has a comparable DMM, and arguably the 3458 is still the best. It’s too hard to mass produce one that’s better.

  2. I was told that the AVO8 was the best multimeter money could buy, and that it was carved in stone on the back of one of the tablets that Moses brought down the mountain on his Triumph. I was obviously misinformed.

    1. I craved an AVO 8 as a teenager but they were WAY more than I could afford. 40 years later and I have a Mk II on my bench. It amuses me to use a period-correct meter on the odd occasion I work on valve radios.

    1. All these quibbles about other multimeters being better stem from a misunderstanding of the HP3458A’s intended application. It is NOT intended for general fault-finding, but for semiconductor characterization where its accuracy in the uA & uV range really shines. The 4-wire measurement capability (see the 6 binding posts) and the HPIB control are additional proofs that it is primarily intended for characterizing semicon devices.

    2. I bought an old 5.5 digit Keithley and had it calibrated by a friend. It’s really been handy over the past year. From matching transistors to measuring the input impedance of an audio amplifier.

    3. The Fluke 77 may be a good general purpose meter, but is not the ‘best’ meter in any environment. Regardless of what your requirements are, Fluke sell something better: Automotive? 88V. Explosive Environments? 87V Ex. Harsh Environments? 28 II. General Purpose Industrial? 87V. Datalogging? 287 / 289. Industrial Process Control? 789.

      Any of those meters will handle any task that a Fluke 77 can, with better accuracy and a wider bandwidth, in addition to other tasks that the 77 simply cannot perform. Temperature? Conductance? PWM Duty Cycle/ Pulse Width? Frequency? Microamps? RPM? True RMS voltages? I wish you luck.

      It’s not even as if we could say that the Fluke 77 is a budget-minded option for a hobbyist when is sells for $300 on Amazon. Sure, it’s cheaper than the other meters listed, but that’s not saying much. (289 is currently on sale for $570 for those interested). The reality is that if you’re using your meter to make money, the correct Fluke will soon pay for itself. Maybe you only need the capabilities of the 77. Fine, buy a 77.

      Here’s the thing. Maybe a business user can tightly define their daily requirements. Maybe a business sends its technicians out with 77s, and the supervisor holds something more capable (eg 87 with thermocouple) for the rare occasion that a temperature measurement is needed. That might seem a smart thing to do, reducing up-front costs, risks due to theft or loss, etc., but every hour lost waiting on that meter could have paid for an upgrade to begin with.

      Hobbyists rarely have the luxury of tightly defined requirements, nor of depreciation schedules that can be used to amortize costs over several years. If we have to buy two meters it would often have been better to have bought the right one the first time.

      With patience, I eventually found my used Fluke 189 (the predecessor to 289) on craigslist at a great price. It looked like it had never left its case, completely unmarked. My advice to other hobbyists, buy the most capable used Fluke that you can afford. That may even be a 77.

  3. I’m never going to understand the inner workings of that class of gear. He does, apparently, and it’s very interesting to watch him fix what others, understandably, gave up upon.

  4. My every day carry meter is a Fluke 8060A which I bought back in 1983. It was a game changer meter back then when Simpson 260s ruled technician tool kits, and the 8060A still isn’t bad. I had to send my 8060A back to Fluke in about 1990 because the display driver chip freaked out, but after that repair I’ve been using the 8060A regularly. I have a recently calibrated Keysight 34461A 6.5 digit bench meter. The Fluke 8060 within its rated bandwidth agrees well within 1% of the 34461A when making casual voltage measurements. That’s not bad for a meter that’s been knocking around in a tool bag for 30 years since it’s previous calibration.

    1. I think that’s fair.

      Wasn’t there even a time when some companies specified the 260 when measuring equipment vo!tage? They gave sample voltage on the schematic, and specifying the meter ensured you got the same results. That could only happen if one meter was king.

      I went from a cheap generic VOM in 1972 to an RCA Voltohmyst, to a Radio Shack DMM in 1984. I do have an HP 410B.

  5. Have a 34420A which gets to 7.5 digits. Can’t imagine a need for it other than calibration of lesser meters. Also have three 6.5 digit Agilent meters and a Fluke as well. They are all very close as one would expect with these sorts of meters. All saved from a recycle bin simply because I would have cried seeing them get torn up.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.