Keysight's new 34465A (6.5 digit) and 34470A (7.5 digit) bench multimeters

[economist]

Did anyone get the Digitize (DIG) option? I would love to hear what maximum sustained sample rate is possible when simultaneously transferring data to a PC with USB and LAN connections.

I'm interested in the targeted use case for the digitizing functions of these DMMs (both the two Keysights and the Keithley 7510). I know the usual response is "if you can't think of a need for it you don't need it". But being a hobbyist I've stumbled onto needs I had no idea I would have at the time of purchase. A good example is buying a scope and then later coming up with a project where I really needed serial decoding -- a concept that was entirely foreign to me a year prior when making the scope purchase decision. So, what is the best use of these digitizing options? Don't get me wrong -- I can certainly imagine having fun with that functionality and using it. I just wonder what a real professional would do with it.

[AndyC_772]

A few years ago, I measured the output of a prototype DC power supply with a cheap Tenma DMM (a tool which, I should add, I used on a daily basis for many years and which was generally both accurate and trustworthy).

The reading was a few volts higher than it should have been, and I spent a while probing around the circuit to try and figure out why.

It turned out to be noise on the output, which was upsetting the DMM and causing it to misread. Give or take a capacitor or two, the PSU itself was fine.

The noise was visible with a scope, but that's less convenient, and not isolated. A 34465A with digitiser could, potentially, have shown not only the correct DC voltage, but also detail of the noise.

If I had any other ideas, I'd have the DIG option for my 34465A.

[Lizerd]

I use the 34411a at work all the time in the 50KHz sample rate mode, I built a C#/WPF application that handles the trigger and graph function with data analyses. Really awesome.
Both for fast signals and slow.
Thats why I will buy the 34465A soon. I was delighted when I saw that the 65A was coming.
but I still will use the K2000 because it's just awesome (mostly UI update speed wise).

[6thimage]

A few years ago, I measured the output of a prototype DC power supply with a cheap Tenma DMM (a tool which, I should add, I used on a daily basis for many years and which was generally both accurate and trustworthy).

The reading was a few volts higher than it should have been, and I spent a while probing around the circuit to try and figure out why.

It turned out to be noise on the output, which was upsetting the DMM and causing it to misread. Give or take a capacitor or two, the PSU itself was fine.

The noise was visible with a scope, but that's less convenient, and not isolated. A 34465A with digitiser could, potentially, have shown not only the correct DC voltage, but also detail of the noise.

If I had any other ideas, I'd have the DIG option for my 34465A.

The digitising mode would more than likely have shown the noise, but also the secondary peak measurements would have (and they are not an optional extra).

[6thimage]

Did anyone get the Digitize (DIG) option? I would love to hear what maximum sustained sample rate is possible when simultaneously transferring data to a PC with USB and LAN connections.

The datasheet says that the 34465/70A can do up to 40k readings/s in ASCII to the bus (LAN or USB connected computer). I have also been reliably informed that both LAN and USB have been tested at the maximum 50k readings/s but the data has to be in binary format (configured by the FORMat:DATA SCPI command - page 264 of the manual).

[Sparky]

Did anyone get the Digitize (DIG) option? I would love to hear what maximum sustained sample rate is possible when simultaneously transferring data to a PC with USB and LAN connections.

The datasheet says that the 34465/70A can do up to 40k readings/s in ASCII to the bus (LAN or USB connected computer). I have also been reliably informed that both LAN and USB have been tested at the maximum 50k readings/s but the data has to be in binary format (configured by the FORMat:DATA SCPI command - page 264 of the manual).

It's unfortunate, and I don't understand why, the instrument configuration for max readings/sec is not clearly stated.  One should not need reliable informants on-the-inside for the configuration that yields the publicly disclosed specs...

I hope it's just an oversight and Keysight publish it soon enough...

[6thimage]

Did anyone get the Digitize (DIG) option? I would love to hear what maximum sustained sample rate is possible when simultaneously transferring data to a PC with USB and LAN connections.

The datasheet says that the 34465/70A can do up to 40k readings/s in ASCII to the bus (LAN or USB connected computer). I have also been reliably informed that both LAN and USB have been tested at the maximum 50k readings/s but the data has to be in binary format (configured by the FORMat:DATA SCPI command - page 264 of the manual).

It's unfortunate, and I don't understand why, the instrument configuration for max readings/sec is not clearly stated.  One should not need reliable informants on-the-inside for the configuration that yields the publicly disclosed specs...

I hope it's just an oversight and Keysight publish it soon enough...

I'm also hoping that it is an oversight - the datasheet should have both the ASCII and binary data rates (the binary mode is new for the 34465/70A, I assume purely so they can get the required data rate). I'm guessing the people putting the datasheet together had a check list of things to include, with one being the bus reading rate - I don't think it occurred to them that there might be two different rates.

[radioFlash]

Some additional detail photos. The Resistor R324 is a Vishay precision resistor marked as VH102ZT 10K000 1%. Also, I did notice a small fan shield after looking at it again. It's just a small piece of bent sheet metal.

[G0HZU]

Don't get me wrong -- I can certainly imagine having fun with that functionality and using it. I just wonder what a real professional would do with it.

Obviously, there will be valid uses (for some users) for these features but at my place of work the logging bench DMM is a rare sight in the engineering labs. I doubt many people have used one for logging because they are usually way too slow. Even in this thread people are hitting very slow sample limits.

The usual solution is to use multiple ADC channels to log lots of stuff in parallel and this is faster, cheaper, takes up less space and can usually be designed 'in' to a product or its test jig. DMMs offer high resolution when logging but this isn't really needed in our case. If we tried to do all the above with DMMs we would need loads of them on every bench and they would still be too slow.

We do sometimes use logging DMMs in the ATE section because the ATE guys tend to use tools like Labview so they prefer to just plumb stuff together quickly.

However, for home use it's pretty easy (as in highly convenient) to log stuff with a DMM if you just want to look at noise and ripple on a dodgy power supply over time. But some people might just use a PC soundcard instead with some spectrum analysis SW and look at the spectrum up to maybe 90kHz using freebie analyser software. Some SW can capture the data and produce a wav file for repeated playback. It can be a very powerful tool even if not calibrated.

[LaurentR]

I am confused about the trend plotting in Data Log or Digitize modes. Is it available while logging or only when logging/digitizing is done? Does the graph get reset (and not redrawn) if you change the X axis? Can you also rebin the histogram without clearing it?

[Dr. Frank]

Some additional detail photos. The Resistor R324 is a Vishay precision resistor marked as VH102ZT 10K000 1%. Also, I did notice a small fan shield after looking at it again. It's just a small piece of bent sheet metal.

Thanks for these great photos!

Wow, that VH102T resistor, hermetically sealed, is really a very stable one!

It has a guaranteed T.C. of +/- 0.8ppm/K, 0.2ppm/K typical, and a long term stability (no load) of 2ppm in 6 years, although it's not oil filled.
The resistance specification of the '365 / '370 are  too wide (8ppm/yr. and 3ppm/K), especially as they really implemented some downsized form of the 3458A AUTOCAL function.
The 40k S012C resistor is not so stable, +/- 4.5ppm/K for this old C foil technology, but maybe used for the LM399H reference.

Frank

[KedasProbe]

The person that came up with the name "Digitize Mode" at Keysight should be fired.

[6thimage]

Some additional detail photos. The Resistor R324 is a Vishay precision resistor marked as VH102ZT 10K000 1%. Also, I did notice a small fan shield after looking at it again. It's just a small piece of bent sheet metal.

Thanks for those photos. That fan shield doesn't look like it would do that much - I was expecting something a lot more substantial. The back of the front panel has a few extra passives on it, but I can't see what I was looking for on it. To support thermocouples, a temperature sensor is needed. I have heard that there is only one on, or near, the front connectors - I'm guessing it is on the far side of the front panel board, as I couldn't spot any on the measurement board.

I'd also be interested if you could get the part numbers for the flash and ram (on either side of the spear processor), as I wanted to know if they have increased from 128 MB (1 Gb).

[6thimage]

I am confused about the trend plotting in Data Log or Digitize modes. Is it available while logging or only when logging/digitizing is done? Does the graph get reset (and not redrawn) if you change the X axis? Can you also rebin the histogram without clearing it?

Looking in the manual (pages 133-134), when data logging to a file, the trend plot will act in the same way it does when it is in the continuous mode. When data logging to memory, however, you can use the zoom, pan and cursors. As you can only digitise to memory, I assume you can always use the zoom, pan and cursors (provided you are below the memory limit). I am not sure how the re-binning works for the histogram - the manual doesn't mention it.

[SteveyG]

Don't get me wrong -- I can certainly imagine having fun with that functionality and using it. I just wonder what a real professional would do with it.

Obviously, there will be valid uses (for some users) for these features but at my place of work the logging bench DMM is a rare sight in the engineering labs. I doubt many people have used one for logging because they are usually way too slow. Even in this thread people are hitting very slow sample limits.

We regularly use them remotely for logging etc. I guess it depends what you're designing.

[TiN]

LM399 in socket? Beware wild voltnuts.
As a serious note, maybe that's how they swap reference to LTZ1000 for 34470A, if mainboard same? Just plug a LTZ-based ref board into socket.

[6thimage]

LM399 in socket? Beware wild voltnuts.
As a serious note, maybe that's how they swap reference to LTZ1000 for 34470A, if mainboard same? Just plug a LTZ-based ref board into socket.

The 34461A's LM399 is socketed as well. I doubt you could do a direct swap for an LTZ1000 though (even if it is on its own board), as I thought it required that the voltage output was separate to its input.

[6thimage]

I am confused about the trend plotting in Data Log or Digitize modes. Is it available while logging or only when logging/digitizing is done? Does the graph get reset (and not redrawn) if you change the X axis? Can you also rebin the histogram without clearing it?

I have been given some answers on how the trend chart and histogram work.

When data logging to memory, the trend chart and the histogram both use the stored data - so you can re-bin the histogram without data loss. When logging to a file, the measurements are also stored into memory, allowing the use of cursors, zoom and re-binning. However, when the number of readings surpasses the memory limit, both the trend chart and histogram revert to continuous mode.

In continuous mode, the trend chart data is compressed into 'buckets' which have a constant time width. Each of these buckets keeps a high and a low value. For displaying of the 'recent' trend chart, the last 400 of these buckets are shown and scroll to the left hand side of the screen. When the 'all' trend chart is displayed, all the buckets are compressed to fit onto the screen - with the highs and lows being kept for each pixel column.

For the histogram, in continuous mode, the samples are binned continuously, with the only limitation being that each bin count is stored as an integer. The auto-binning can compress bins to cover a broader range, but cannot arbitrarily re-bin.

When digitising, the histogram is not actively updated, with its processing being postponed until the digitising is complete - this is done so that the measurement rate can be guaranteed. Additionally, the data for the histogram is only processed if the display is set to it (in continuous & data logging, the histogram is updated with every reading).

The person that came up with the name "Digitize Mode" at Keysight should be fired.

Why do you think that?

[EEVblog]

LM399 in socket? Beware wild voltnuts.
As a serious note, maybe that's how they swap reference to LTZ1000 for 34470A, if mainboard same? Just plug a LTZ-based ref board into socket.

You'll have to wait for my teardown, rendering now... 35 minutes of waffle about the reference and A/B comparison with the 34461A board.

[KedasProbe]

The person that came up with the name "Digitize Mode" at Keysight should be fired.

Why do you think that?

Because it doesn't say anything about what is different in that mode, it's digitizing in all modes, that's what a DMM is supposed to do. "Measuring Mode" would say equally as 'much'.
 "High Speed Mode" for example would already mean something.

[HighVoltage]

You'll have to wait for my teardown, rendering now... 35 minutes of waffle about the reference and A/B comparison with the 34461A board.

We wanted to order a new one today, but will wait for your teardown.
Thanks so much for all your teardowns and explanations.
It really helps in a buying decision.

[EEVblog]

We wanted to order a new one today, but will wait for your teardown.

Photos are up in the usual place for those who can't wait...
Teardown wise the 34470A is as good as you could expect it to be 

[Dr. Frank]

Thanks Dave!

LTZ1000A, 15k / 1k divider, 95°C. For 55°C Tamb., 8ppm/year stability only.
All SMD resistors, maybe also metal foil SMD?
And a lot of PCB slots!!
PCB fully thermally and air shielded, high temperature plastic, again.
 
They just copied the errors from 1988.. could have been done better..

Obviously, they just transferred the old 3458A reference PCB to modern design / components.
Even the alternative positions for R4 and R5 are available.. for pimping to 65°C / 1ppm/yr.


Frank

[EEVblog]

They just copied the errors from 1988.. could have been done better..
Obviously, they just transferred the old 3458A reference PCB to modern design / components.
Even the alternative positions for R4 and R5 are available.. for pimping to 65°C / 1ppm/yr.

What errors?
I'm not familiar with the 3458A reference, never opened one. But isn't it the ducks guts?
Do you doubt this can meet the claimed specs?
I was impressed with the plastic enclosure design and the separation of the driver transistor (for separate cooling?).

[Dr. Frank]

They just copied the errors from 1988.. could have been done better..
Obviously, they just transferred the old 3458A reference PCB to modern design / components.
Even the alternative positions for R4 and R5 are available.. for pimping to 65°C / 1ppm/yr.

What errors?
I'm not familiar with the 3458A reference, never opened one. But isn't it the ducks guts?
Do you doubt this can meet the claimed specs?
I was impressed with the plastic enclosure design and the separation of the driver transistor (for separate cooling?).

Hi Dave,
sure, this reference is something like the Holy Grale. We discussed that in the Ultra LTZ1000 blog intensively.

But anyhow, HP engineers originally designed this circuit to run on 95°C, also due to the required ambient temperature of 55°C.
That causes a higher drift than intended originally by LT: running on 65°C instead gives typically -1.. -2ppm/year, and other designs of 8 1/2 digit DMM, standards and calibrators use 45..55°C oven temperature for around 1..2 ppm/year guaranteed stability.


In this instrument, they could have used the LTZ1000, (non A), easily reducing the oven temperature by 10°C and increasing the stability by a factor of about 2, and maybe further down to 75°C by better thermal management, giving 3..4ppm/year for sure.

Obviously, the Keysight engineers did not want to create an improved circuit, which they would have to re-qualify elaborately.
That could also be true for the rest of the circuitry, where they may have copied the Multislope IV circuit and algorithm from the 34410/11.
Frank