3458A - black edition

[Kleinstein]

There are a few points that could be improve at the 3458. The relatively easy part would be changes in the digital part, like adding standard SCPI support (if not already done) and a few more statistics features.
Another point would be reducing the power needed, so the meter can run cooler - this alone could improve some leakage specs.

It could be in deed very difficult to beat the 3458 as a direct 100% compatible replacement, because of the rather special ADC that is accurate and fast in one unit. A more modern design would likely go for 2 separate ADCs: one for the slow modes (e.g. 1 ms and up) and one (e.g. good SAR chip) for the fast modes. The fast part is relatively easy as modern chips already beat the 3458 ADC at the higher speeds.  Not needing the high speed modes can help to get better low speed performance.

With modern CMOS switches one may not even need such a special analog ASIC with JFET switches to build a good ADC.  It could still help to get a custom CMOS switch chip - but this is no real magic and nothing that special. One may even get away with of the shelf chips.  Much of the custom chips are more like custom resistor arrays of very good quality. The reference part is still relatively easy - the LTZ1000 reference is still available and the A9 board (ref) can still be the old design.

However a completely new design from scratch would not be 100% compatible. So even if better there would still be demand for a direct replacement. However even if inferior in some aspects a good one could cover much of the market, especially those who only something like a fast 7 digit DMM. Chances are the 3458 is not as linear anymore when using 1 PLC or faster.

[notfaded1]

I get that no matter what you do you can't change physics... but let's use my field as an example.  Many times we thought we had reached the limits of Moore's law.  Time and time again we've destroyed the limits shrank and improved on processor speed at closer to room temperatures.  Perfect example are the materials... many scientists thought we had reached the limits of silicon... well IBM strained the silicon and put other materials next to it and low and behold our architectures are now orders of magnitude smaller and faster and work without super cooling.  Just look at the how the transistor has changed.  We no longer have 2D silicon it's now 3D.  My point is... Everytime scientists said it can't be done... we did it with new novel technology.  I work with many RF engineers and what they're done with radios today is just short of amazing.  Our Holy Grail right now being sought is quantum encryption.  Unbreakable encryption where any observer can be detected.  I will say this the RF engineers I know do stuff today that's black magic to me.   Sometimes they do seem like a dying breed but there are new RF engineers coming out of college.  The older generation is always saying the younger engineers are missing something and low and behold they come up with something that changes the entire game.  Just like you were saying... Analog ASICs... for all we know analog devices could have a new product in development we don't know about.  I wouldn't ever count out the engineers I see as interns today.

Bill

[Dr. Frank]

I'm just  about all of your speculations, w/o any substantial background.

Please take for granted, at least from what the KS sales man has told us @ Metrology Meeting, that the black 3458A has been redesigned by an external engineering group in Germany, and that many parameters have been improved. I already mentioned 1/2 DCV noise, and that's simply because now, 30 years later, much better and less noisy transistors are available, but that's not the only reason.

branadic has already seen the interior of the new 3458A, so you may guess where and why.

Henrik_V obviously also has seen the black 3458A live at the PTB , when they demonstrated the swapping of A/D boards. Guess why @ PTB.. probably they were also involved in the improvement of the A/D converter, where else should that happen, alternatively in Loveland facility only. Here again, you may guess that of today, better OpAmps, better PCB material and more knowledge about layout techniques are available, than 30 years ago.

Frank

[notfaded1]

That'll be great then Doc... if you need a new A/D board for your 3458 you can still buy a black edition one and put it in your old 3458.  Sounds like a win for existing labs to me    It also means many of us probably won't be getting one anytime soon.   

Bill

[SilverSolder]

That'll be great then Doc... if you need a new A/D board for your 3458 you can still buy a black edition one and put it in your old 3458.  Sounds like a win for existing labs to me    It also means many of us probably won't be getting one anytime soon.   

Bill

The good news is that we are very close (if not there already) with viable hobby level A/D designs that can approach 8.5 digits (see other threads in this forum).

It seems to me that kind of project is the next step up from building your own LTZ1000 etc.  reference...  and could be very entertaining to try to achieve!

[FrankBuss]

The good news is that we are very close (if not there already) with viable hobby level A/D designs that can approach 8.5 digits (see other threads in this forum).

It seems to me that kind of project is the next step up from building your own LTZ1000 etc.  reference...  and could be very entertaining to try to achieve!

I don't think this will be possible anytime soon. It is a big step from a voltage reference to a fully working DMM, especially with this accuracy.

But maybe something less accurate, like the Fluke 8842A, would be possible to build yourself. I couldn't find a circuit diagram for the 3458A, but there is one for the 8842A. Someone would just need to replace the obsolete parts with modern parts, then reverse-engineer the ADC algorithm, and then it might be possible to build it cheaper with a modern microcontroller.

[SilverSolder]

^Of course a hobby level effort is never going to be an industrial DMM like Keysight's flagship, but people on this forum have achieved spectacular results within the limitations that you only have DC volts and one range.  These efforts will only get better in future, it seems to me.

[Kleinstein]

...
I don't think this will be possible anytime soon. It is a big step from a voltage reference to a fully working DMM, especially with this accuracy.

But maybe something less accurate, like the Fluke 8842A, would be possible to build yourself. I couldn't find a circuit diagram for the 3458A, but there is one for the 8842A. Someone would just need to replace the obsolete parts with modern parts, then reverse-engineer the ADC algorithm, and then it might be possible to build it cheaper with a modern microcontroller.

The Fluke 8842A with it's recirculating reminder ADC is a pretty odd starting point to build a DIY high resolution ADC. I would consider this more like variant of a pipelined successive approximation ADC. The adjustment alone is a nightmare.
The more classical multi-slope ADC (e.g. like the 3458, 3456 or Keithley 2000) is a much more promising starting point. Unless one needs the high speed of the 3458 there is no need to use a run-up so fast. Less frequent switching reduces demand on the switches so that normal CMOS can get you quite far. The real difficulty is very good linearity - this starts with the problem that verification of linearity better than 0.1 ppm pretty much needs a JJA. At this level it is also not just the circuit diagram -  a not so good layout can also ruin things. Not all the higher order effects that contribute to INL at the 0.1 ppm level may be known.  So a new board for the ADC can be a big deal.

[bsdphk]

"Unless one needs the high speed of the 3458 there is no need to use a run-up so fast."

That may be true at the margins, but it is certainly not true in general.

Fast-ish conversion is important for accuracy, as it keeps a lot of time-variant effects at bay.

Note though, that even the original HP3458 hardware could have run faster, but the speed was reduced to hit the sweet spot between different noise-sources.  (It's all in the original HPJ article, if you read carefully.)

[TiN]

Some interesting box showed up today with Keysight Technologies label on it. Wonder what that could be...

Ah, right....

[e61_phil]

Some interesting box showed up today with Keysight Technologies label on it. Wonder what that could be...

Ah, right....

That's easy. It's a power supply

[Zucca]

Ah, right....

it's the num 5, I bet the other four n 1-2-3-4 are on the other side of the room.

[iMo]

But we need pcb shots instead

[HighVoltage]

Why is it that I am not surprised that TiN is the first one with a black edition?
Wow!
Yes, now inside pictures please!

[chuckb]

You will of course have noticed that the "3458A Multimeter Assembly Level Repair Manual"  on the Keysight Website was updated for the Black edition. No update of the the schematics yet.

https://www.keysight.com/main/techSupport.jspx?nid=-31940.1242732&pid=2905513&cc=US&lc=eng&pageMode=PL

Included are some low resolution Pix of the PCB! Looks like DIPs replaced with SMD. Parts are in the same place.
A1 DC Input has new input relays, one Hermetic resistor in the current measurement section (no change), a rectangular black module near the voltage ref connector.

A9 LTZ1000 Voltage Ref looks the same.

A3 ADC finally has an updated, small, logic controller. Can't tell about U180, it's still under the shield.

Part Number for new A1 03458-66511
Part Number for new A3, 03458-66523
Part Number for new A9, 03458-66559 (normal,  03458-80009 (opt 2)

There are no photos for the new PCB in the replacement parts area yet.

[notfaded1]

That's pretty awesome TiN the Darth Vader 3458 and you got one already!  What's the device sitting on top of it called that seems to match it perfectly in color and looks like it's monitoring V and A from multiple inputs?  It looks like you've already hooked a ton of stuff up to it already as well!  Hopefully you can tell us soon everything different about it and it's brothers?  Those shield boxes you have over the inputs or outputs on some boxes looks interesting too.

Bill

[MiDi]

Number 5 is alive!
And it has a printed version of the HP Journal on top 

Who would put an inventory sticker on something borrowed?

[Dr. Frank]

Nice, and congratulations! Please do a Tear-Down and a characterization, please.
The black 3458A and the 5720A seem to agree within 0.01ppm.. even nicer.

Now please check DCV impedance, bias, noise, INL of DAC. If you can, noise spikes from the multiplexer, in comparison to the old instruments.
Check stability of DCI in terms of T.C. and time and consistency against 5720A artifact calibration.
FW should be 8 or 9, as it didn't change. U180  also did not change.

From pictures of the Repair Manual, take notice, that all the logic and the 68000 is included inside one gate array, apart from SRAM and Flash firmware.
On the A/D board, logic and µP obviously also was crunched into another GA.. how many layers does this PCB have, in comparison?

Btw.: An original paper version of hpj 4/1989 is still down in my basement lab... together with the complete original manuals 

Frank

[iMo]

U180 in plastic flatpack?

[Dr. Frank]

U180 in plastic flatpack?

[iMo]

Or is it covered by the shield (5) there in the picture?

[Dr. Frank]

Or is it covered by the shield (5) there in the picture?

You've got it!

[Macbeth]

OMG, the absolute madlad has only gone and done it! Looking forward to the calibration checks vs Fluke 5720A and the ultimate teardown 

[TiN]

The black 3458A and the 5720A seem to agree within 0.01ppm.. even nicer.

Anything below 0.2ppm is just nonsense, unless we are pouring helium. I've tricked output for a pretty photo.

Actual deviation between 5720A and 3458ABE from Day 0 is -1.2 ppm, which is not bad too.
I've calibrated primary 5720-5725 in August using 792X LTZ1000A FX reference, which was in turn calibrated to NIST PJVS 10V system at ITRI CMS NML.
My resistance capability is better, since I handcarried my SL935, calibrated to PTB QHR chip / National MINTL standards at ITRI CMS NML same time August as well.
3458ABE shows -0.19 ppm error on 10 kOhm resistance range vs 5720A as result, even with <24 hour warmup
My SR104 is at different friendly lab location, and I am not shipping it back until it's warm and nice outside to avoid any chance of stress. Don't want to loose it's <0.1 ppm/year capability.

Btw.: An original paper version of hpj 4/1989 is still down in my basement lab... together with the complete original manuals

Borrowed original HPJ issue from a friend of mine too, for pretty selfie.

I'm getting also some more 8.5-digit reinforcements from FedEx next week, and my lab will back to 8x8 DMMs again 

[bsdphk]

Hi TiN,

What does REV? report for the black edition ?

Does the XYZZY command still work ?