Keysight DMM - Yay or nay?

[Traceless]

@blackdog thanks for reporting back. Awesome that you have all those meters and can provide such a nice comparison, altough this makes my life a *lot* harder now. I really do like the Keithley and its features, but I'm really struggling with myself if I can/want to accept the touch screen. To make things even harder I also stumbled across the Keithley DAQ6510 which now also entered the ring of contenders, unfortunately it is also touch-controlled.

[tautech]

The GDM906x meters seem to have rear terminals, that I don't see with the sigilent.
At least the GDM9060 seems to use a comparable SD ADC chip to the Sigilent.

There is also a Teledyn (T3DMM6-5), that seems to be essetially the same as the Sigilent - possibly an OEM build.

Correct, no rear inputs on SDM3065X unless you get the 16ch scanner card but these have limitations for voltage and current.

[Traceless]

@Kleinstein, @tautech Yes you're right the Siglent doesn't seem to have rear terminals, good point. Also the T3DMM6-5 really looks like a rebranded Siglent, it seems to be much more expensive though (according to a quick google shopping search).

[blackdog]

Hi, Traceless,

You're welcome, I also have two Tektronics DMM4050 which are similar to the Fluke versions.

Of these I can say that at the 10V DC range they are also very stable.
They were last calibrated about seven years ago and the deviation is now about 4PPM at the 10V range at normal room temperature.

I don't think I would buy this model (Fluke/TEK) now. don't get me wrong, these are very good measuring instruments, but the price... silly Fluke.
One negative point is the input sockets on the front, on one of my two meters I have already had to repair it.
The machanic design of the connection bushings is a big Marketing mistake.
I am very careful with my measuring instruments but it still breaks down.
Soon, on the second meter, I will also redo the soldering of the input bushings.

Software for logging is also not widely available from the Fluke/TEK meters, that is also a consideration when weighing.

The two TEK meters are my workhorses, so don't think now that I don't value them. :-)

Kind regards,
Bram

[Traceless]

@mawyatt to rule out that blackdog just had very bad luck with his KS meters, did you observe similar auto-cal issues with yours compared to the DMM6500.

Also since both of you @mawyatt and @blackdock have access to a DMM6500, can you by chance tell if a) I can ignore the touch screen and access all functions with the buttons next to the display and b) if the touch screen can be operated with gloves like this?

[blackdog]

Hi Traceless,

The Keysight meters are not bad! I use them almost daily.
I have received over the years very good support from KeySight Netherlands!

But the Autocal is just not well done and not worth spending money on if that is the only function why you buy a 34465a instead of a 34461a.
The calibration/reports from Keysight very also very good, If you buy a 34461 or a 34465a you get a nice calibration report.

With the Keyley DMM6500 you only get one a4 with on it that it is calibrated.
I don't know if this difference is important to you.

Kind regards,
Bram

[mawyatt]

@mawyatt to rule out that blackdog just had very bad luck with his KS meters, did you observe similar auto-cal issues with yours compared to the DMM6500.

Also since both of you @mawyatt and @blackdock have access to a DMM6500, can you by chance tell if a) I can ignore the touch screen and access all functions with the buttons next to the display and b) if the touch screen can be operated with gloves like this?

Haven't experienced any Autocal issues with either KS34465A, and the DMM6500 & two KS34465A agree well with our home brew 10V references based upon LM399 (2) & LTZ1000. Like mentioned we also wanted to "see" how good the DMM6500 was compared to the KS34465A and not having any experience over our career with Keithley somewhat concerned about how the DMM6500 would preform. Well it performed admirably and we have no concerns recommending this superb instrument. As mentioned the KS and DMM are good just about everywhere and have specific areas where they excel. Honestly you can't go wrong with either.

No, you can't access all the menus on the DMM6500 with buttons, you'll need to employ the touch screen. Don't have any gloves like shown, but did just use Nitrile Exam Gloves, and no issues with DMM menus.

Best,

[Traceless]

Hi Bram,

well the main reason I chose the 34465A over the 34461A is that it has a much higher reading rate 50k rdgs/s vs the 34461A's 1k rdgs/s, I just want to avoid the grief I have with my current meter, which has very slow auto ranging and sluggish AC readings. Aditionally the 34465A's 1µA range is nice to have. Other than that any of these meters probably even the Siglent is more than good enough for my purposes and a quantum leap forward from what I have currently. If I get a new meter this is probably going to be a life-time purchase (unless the device breaks down and can't be reparied) so I do my best to make the right call this time, even if that means stretching my budget a bit.

[Traceless]

@mawyatt: Thank you very much for reporting back so quick and also for doing the glove test, you are awesome!

[blackdog]

Hi mawyatt, :-)

Regarding the Autocal, Keysight investigated that problem and came to the same conclusion as the measurement report I sent them.
They sent me a measurement report showing the same effect measured in their factory as what happened to me in my lab.

If you use the Autocal function at e.g. 25C lab temperature and say the Keysight DMM shows +2PPM difference from its Cal temperature.
The meter usually corrects that with Autocal, but usually after about 20 minutes your meter will read the same value as before the Autocal function.
This also happens when the internal temperature of the DMM remains the same.

Kind regards,
Bram

[blackdog]

The AC ranges of the 3446x series meters are really good, very broadband and you have the possibility of three measurement speeds for the low frequencies.
Also, the dB function of the 3446x meters is very useful with a very high resolution.

As for Autorange, that feature on my multimeter is using less and less.
I usually choose a range 1 step above the range needed for maximum resolution, the meter is then very fast.

Also choose the number of NPLC well, here the basic setting is 10 NPLC, but depending on the measurement function it can also be 0.1 NPLC.

The digitizing function I actually do not use, but for you it may be important.
The Keithley is a lot faster in this respect.

I think you will definitely be happy with a 34465a DMM even though I have problems with the autocal measurement function of this model! 

Kind regards,
Bram

[mawyatt]

@mawyatt: Thank you very much for reporting back so quick and also for doing the glove test, you are awesome!

Thanks. If this is your lifetime purchase then this tilts the scale in favor of the KS since it "may" contain some heritage from the timeless HP34401A

Best,

[mawyatt]

Hi mawyatt, :-)

Regarding the Autocal, Keysight investigated that problem and came to the same conclusion as the measurement report I sent them.
They sent me a measurement report showing the same effect measured in their factory as what happened to me in my lab.

If you use the Autocal function at e.g. 25C lab temperature and say the Keysight DMM shows +2PPM difference from its Cal temperature.
The meter usually corrects that with Autocal, but usually after about 20 minutes your meter will read the same value as before the Autocal function.
This also happens when the internal temperature of the DMM remains the same.

Kind regards,
Bram

Hi Bram,

We hadn't noticed that drift, and will look for it next time we get the pair of 34465A & DMM6500 "looking" at the references for a few days. Will report what we find.

Thanks for pointing this out.

Best,

[Cerebus]

Thanks. If this is your lifetime purchase then this tilts the scale in favor of the KS since it "may" contain some heritage from the timeless HP34401A

Technologically the TrueVolt series are very different from the 34401A, in fact the technological basis of the range was first tried out in the 34410A. The 34410A is the last that HPAK published schematics for - a comparison of those with the 34401A's schematics shows a whole raft of differences, in particular the core ADC is totally different. The TrueVolt series, in particular the 34461A, was explicitly designed to be compatible with the 34401A both from an operation and automation point of view, and in the case of the 34461A, identical measurement characteristics so that they could be a drop-in replacement for the 34401A.

[razvan784]

If I may offer a personal opinion:
* if you are after precision, accuracy and stability, and you absolutely want a new meter, I would skip anything besides Keithley and Keysight. I used a 34461A and liked its interface. I think i'd like the 6500 interface better, but haven't had the opportunity to try it.
* if you are after repairability, I would skip anything new and go for a used 34401A in good condition (or even better, a 34410A if you can find one for a reasonable price). These can be hooked up to a PC via RS-232 (or even to something like a Raspberry Pi) where you can easily do logging, fancy graphs and whatnot. An old Keithley 2000 is also an option (I also like its interface better) but these are not as easily repairable - the circuitry is more complicated and official schematics have not been published.
For a general-use DMM, sure, capacitance is nice to have, and stuff like continuity and diode mode are important. For a precision bench DMM on the other hand, I don't think those are relevant selection criteria. I also use a DE-5000 for capacitors and I'm quite happy with it.

[blackdog]

Hi,

First I want to make this clear, I can still recommend Traceless the 34465a from Keysight it is not a bad DMM!

The things I show below like this concern small deviations at the PPM level when using the Autrtocal function.
My assumption when I bought the 34465a was that if my lab temperature was stable at 26C and I used the Autocal function then the few PPM that the Autocal function corrects,
would remain stable if the temperature in the measurement environment did not change.

The corrected value of the autocal function drifts downward at two instruments I tested.
And a 34465a reference meter at Keysight's factory to which my meter was compared had the same drift after using the Autocal function.

And again, this does not make the 34465a a bad DMM, just that the Markereers at Keysight wanted an extra feature on their 34465a version to be able to sell the 34465a for a higher price.
So I paid the higher price, but the functiue doesn't work well, and I find that a bit sad.

I can still take accurate measurements as needed as I also have an Agilent 3458a standing by. but for good stability this DMM will need to be on for at least 12 hours,
I choose to keep 24 hours on and that also means the sound of an F16 next to you from the fan. 

OK some pictures,
Here you can see the drift on my 34465a after using the autocal function.
This meter was paralleled to two 34461a and my 3458a DMM for verification so that I had some assurance that there were no measurement errors involved.


.
This is one of the measurements the Keysight factory made of my Unit and their reference 34465a DMM.
My 34465a is the orange line and the blue line is the reference DMM.
In the green frame that I placed in the picture is a special test from Keysight, as far as I understand it, there the Autocal function that is always present in this DMM series is disabled.

I was told that this function is not available to the users, as it is only for the fabrication or calibration process.
But the drift in both instruments is readily noticeable.
And again, the orange line is again my DMM.


.
I have no idea what the behavior of the Autocal function is at larger temperature deviations, say e.g. as a 34465a in a 19-Inch rack mounted with equipment and whether the Autocal will also drift more percentage-wise.
It may well be that the few PPM I have experienced is the "base" drift and if the ambient temperature is e.g. 25 degrees higher in a 19-Inch rack that this base drift I am measuring will cancel out against e.g. 15PPM drift that the Autocal function corrects.
I have no way to check this, I do not own a climate box to do good measurements in this regard.

Some tips
If you are going to make precision measurements keep in mind these things, among others:
Provide a measurement room with a stable temperature.
No varying Sun, no drafts.

Do not place your DMM in a place where it is heated by other measuring instruments.

Clean the inputs of your DMM regularly with alcohol and Q-Tips.

Preferably do not use banana connectors, they are slow to warm up and if you do want to use them, make sure they are gold plated and also cleaned with alcohol.
For most precision measurements use a pair from a CAT5e cable that is full copper and a pair of Q-Tips to secure this pair in the inputs of the DMM and clean them with alcohol.

Keep the wiring as short as possible, wiring always twisted!

Keep telephones and other transmitting equipment away from your test rig, and yes, also the phone in your pocket or that of your colleagues.

Also think about SMPS power supplies, these are often lighthouses as far as EMC is concerned, always think about that when you have strange outliers in your measurements, it's really not always the 1/F noise of your voltage reference or your D.U.T.
I'm sure a good number of people here on the forum can add to this list with tips.

My TEK DMM4050 meters and also my Keithley 6500 do not have an autocal function for the user, so I cannot compare my 34465a with that,
And again the 34465a is NOT a bad DMM!

I hoop this helps and if there are question, i like to hear them. 

Kind regards,
Bram

[Traceless]

* if you are after repairability, I would skip anything new and go for a used 34401A in good condition (or even better, a 34410A if you can find one for a reasonable price). These can be hooked up to a PC via RS-232 (or even to something like a Raspberry Pi) where you can easily do logging, fancy graphs and whatnot. An old Keithley 2000 is also an option (I also like its interface better) but these are not as easily repairable - the circuitry is more complicated and official schematics have not been published.

I actually do like old DMMs and also thought about that but there are a few catches. For some reason the good brands are unobtainum in Germany. You do find stuff on ebay but:
- prices are stellar
- they're often in pretty rough shape (cosmetically) and sometimes technically have tired VFDs etc.
- you don't know if the meter has any quirks/defects
- I don't know when they were calibrated last and if calibration is still good

Since I don't have any equipment to test the accuracy and recalibrate such a meter I'll likely end up having to send it to Keysight/Keithley for calibration who will then tell me that there are a few issues with the meter that need to be fixed along with a 1000€ quote for repairs and calibration. At that point just purchasing a new one will be cheaper plus I have warranty.

Once I actually own a proper multimeter I can use as reference I can totally see me buying a damaged used one for restoration - just as a fun project/learning experience.

An old Keithley 2000 is also an option (I also like its interface better) but these are not as easily repairable - the circuitry is more complicated and official schematics have not been published.

I think they are still sold new too bad though, that there is not much documentation around, but I assume that is true for most modern DMMs. If I really want good documentation I'll have to look for the HP branded stuff (which is really hard to get here).

For a general-use DMM, sure, capacitance is nice to have, and stuff like continuity and diode mode are important. For a precision bench DMM on the other hand, I don't think those are relevant selection criteria. I also use a DE-5000 for capacitors and I'm quite happy with it.

Since this DMM is going to be my only (bench) DMM I'd rather not like to miss out on diode mode, my secondary meter currently is an Aneng 8009. I could live without capcitance though if I absolutely have to.

[tautech]

Since I don't have any equipment to test the accuracy and recalibrate such a meter I'll likely end up having to send it to Keysight/Keithley for calibration who will then tell me that there are a few issues with the meter that need to be fixed along with a 1000€ quote for repairs and calibration. At that point just purchasing a new one will be cheaper plus I have warranty.

100% !
Still for basic use the purchase of say of a DMMCheck provides a sanity check that accuracy is appropriate for your needs otherwise the Voltnut rabbit hole awaits.

You may not be aware Defpom has a website for adjusting Cal constraints for the SDM range:
http://www.thedefpom.com/siglent_sdm_calibration.php

[Kleinstein]

The Keithley 2000 are no longer sold new, but there are rather similar meters as Keithley 2100 or Picotest M3500.

In Europe the situation with used higher grade DMMs is not as good as in the US.

[Traceless]

Hi Bram, thank you very much for the detailed explanation and the effort to post your measurements and compiling all the high precision measurement tipps here I really appreciate that. Also I find that autocal-issue you found quite interesting and would be curious if this is a hardware design flaw or a firmware bug. Probably only Keysight can answer that question. I think you should post yor discovery in the Metrology Forum I'd love to see where the discussion about this goes.

As you say the 34465A is still a good meter despite the autocal issue, the number one question I'm asking myself now is: "Does the 34465As autocal issue make the 34461A actually the better meter just because it is not constantly drifting and recalibrating?"

Ultimately I probably shouldn't worry too much about the autocal problem because the environment conditions the meter will operate under are far from ideal. May lab is right under the roof and like in most german houses there is no air conditioning. That means that especially in the summer time I'll have temperature fluctuations from around 21-26°C in the morning up to 38°C in the afternoon, and I'll have to open the window because otherwise I'll be cooked alive . (Mental note to self, don't do precision measurements in summer but rather go out for an ice cream instead ) At that point the auto-cal issue is probably going to be negligible.

In autumn, winter and early spring the temperature will be much more stable though also not perfect (likely between 21 and 24°C). Also I usually only warm up my meters for 5 minutes or so, they don't run continously.

[nctnico]

If you want to do really precise measurements then you'll need a room with climate conditioning (temperature + humidity) but by that time you enter calibration lab territoy. I'm not saying a 6.5 digit DMM is useless, but you'll have limitations in normal room / office conditions.

[Traceless]

@nctnico: Well I could get an 3458A IIRC those have an oven to compensate for room temperature - no just kidding. The 34465A is probably already overkill for most/all of my applications. I'm just a hobbyist and don't do high precision circuit design or anything like that.

[Kleinstein]

The autocal function is not so much about a changing temperature, but more about using the meter for many years without calibration. The ACAL function does internal tests to remeasure resistor ratios and this way can reduce the drift. This can also help against temperature dirft, but the main point is the long term drift. It is rare that one needs a really accurate reading with a temperature way off.

AFAIK one does not have to use ACAL on the 34465 and the 34461 does no offer this option / point.
It is just inconvenient that the function does not seem to work all the way as intended. Chances are this is a SW problem, maybe related to some HW oddity. For the 34470 the ACAL seem to be even more problematic and for some reason KS seem to still have not really fixed it. They seem to have a problem if ACAL actually makes a difference to the normal cal factors. The problem with the 34465 may well be related though there are differences.
I still have some hope that they finally will fix it - though one never knows.  Some SW problems persist until the product is obsolete (e.g. extra LF noise in many Keithley meters, that is likely due to a SW quirk).

The ADC in the 3446x series is not the absolute best INL and linearity is an important factor in how good ACAL works.
The 34465 still has more current ranges, higher speed and tighter specs, though much of the internals seem to be very similar.

[Traceless]

The autocal function is not so much about a changing temperature, but more about using the meter for many years without calibration. The ACAL function does internal tests to remeasure resistor ratios and this way can reduce the drift. This can also help against temperature dirft, but the main point is the long term drift. It is rare that one needs a really accurate reading with a temperature way off.

So autocal is probably worthwhile despite the drift then, since I plan to use the device for a long time and probably will never send it off to a lab for calibration.

AFAIK one does not have to use ACAL on the 34465 and the 34461 does no offer this option / point.

According to Bram you can't turn off autocal on the 34465A, so if I really didn't want autocal I'd have to go for the 34461A.

It is just inconvenient that the function does not seem to work all the way as intended. Chances are this is a SW problem, maybe related to some HW oddity. For the 34470 the ACAL seem to be even more problematic and for some reason KS seem to still have not really fixed it. They seem to have a problem if ACAL actually makes a difference to the normal cal factors. The problem with the 34465 may well be related though there are differences. I still have some hope that they finally will fix it - though one never knows.  Some SW problems persist until the product is obsolete (e.g. extra LF noise in many Keithley meters, that is likely due to a SW quirk).

Who knows maybe if Bram starts a thread about his discovery in the metrology section and enough people start investigations, it might get Dave's attention. If he makes a video about that there is a good chance that Keysight will address the problem, unless it really is a HW issue that is not fixable via firmware.

The ADC in the 3446x series is not the absolute best INL and linearity is an important factor in how good ACAL works.
The 34465 still has more current ranges, higher speed and tighter specs, though much of the internals seem to be very similar.

The high speed is exactly what im interested in. Also the additional ranges are nice.

[nctnico]

As an owner of the 34461A I do have to say that the resistive range is lacking at the lower end of the range. Despite that the 34461A has 4 wire measurement, it is not  suitable to measure milli-Ohm resistors. It would need an extra (lower) range for that. The DMM6500 lowest resistance range is 1 Ohm where the lowest range on the 34461A / 34465A is 100 Ohms.