EEVblog 1382 - Keysight EDU34450A 5.5 digit Bench Multimeter TEARDOWN

[tszaboo]

For the NPLC stuff, the higher end DMMs have specific circuitry to detect and deal with that, this one doesn't for cost reasons.

I specifically opened the 34401A schematic and looked up what circuitry this could be.
My guess would have been a optocoupler and maybe a TL431. But no, it is simpler.

This was the BOM of the NPLC detection:
One resistor, 42.2KOhm
One capacitor, 100nF
One zener diode, 6.2V. BZX84C6V2
You might need a lower voltage zener, but my estimated BOM cost of this is 2.1 cents. Surely it would've fit to the BOM. Just own up, that this is intentionally crippled.

Isn't the DCA range a bit weird too?

You mean the "DCI" range? 

They've been calling it DCI and ACI for the past 30 years. Just move on.

[temperance]

-This "unconnected" comb like structure, might it be some test coupon which is specified in IPC2221B?
-TH component solder joints. What do you people think about the quality of the solder joints.


But maybe it's a good idea to learn students about relevant digits and basic accuracy. Till now I've met two engineers not able to correctly interpret the numbers shown on a DMM.

Case 1: this PFC circuit outputs 425.1V and the second one 426.2. Replaced many things and damaged the board while doing so...Charged the customer €800 for a new board in the end.
Case 2: this reference measures 5.12562V and the second one 5.142354 and the third one,..... But my circuit is also unstable because the last two digits are fluctuating.

I will need one of those when I'm 102 year old. I will be able to read its display without powerful binoculars at that age.

[3roomlab]

(4/U/L/R/G/S/P = 4wire/USB/LAN/RS232/GPIB/software/powerlinecycles)
prices are rounded up to $100 (*uses fast/medium/slow topology). XX = not std rack sized
+AAA = units with generally better overall specs than EDU34450 and $1k or less

no   model         Vrange/1yr spec         input-imp      features            price

0a    34401a          -10V 0.0035 +0.0005    @10G 4/R/G/P +LXI+labv 0.5k-data    (USD ~0.7k used)    +R    
0b    KS 34460A       -10V 0.0075 +0.0005      @10G         (USD ~1.3k)          +R
0c    KS 34461A       -10V 0.0035 +0.0005      @10G         (USD ~1.5k)          +R [replaces 34401a]
00    EDU34450A       -10V 0.025  +0.005       @10M 4/U/L/S/* 5k-data          (USD ~1k)          XX
01    KS 34450A       -10V 0.015  +0.005       @10M 4/U/G/S/* 50k-data       (USD ~1k)          +R
02   K2110          -10V 0.015  +0.002       @10M 4/U/S/P +labv 2k-data       (USD ~1.1k)         +R?   107mm
03    K2000          -10V 0.003  +0.0005    @10G 4/R/G/P 1k-data          (USD 0.6k used)       +R    
04    MP730028       -10V 0.015  +0.004       0.1G 4/U/L/R 1M-dat         (USD 0.7k)          +R? 110mm

05    8808A/SU       -20V 0.015  +0.005      @10M 4/U/R/S/P data?          (USD ~1k)          +R          
06    8845A          -10V 0.0035 +0.0005    @10G 4/R/L/P 10k-data          (USD ~1.3k)          +10vD       +R
07    DM3068          -20V 0.004  +0.0005    @10G 4/U/L/R/G/S/P +LXI 512k-data    (USD ~0.9k)       +R?   107mm    +AAA
08    DM3058          -20V 0.015  +0.04       @10M 4/U/L/R/G/S/* 2k-data       (USD ~0.7k)          +R? 107mm
09    SDM3065X       -20V 0.002  +0.0004    @10G 4/U/L/S/P 599k-data       (USD ~0.8k)          +R?   107mm    +AAA

10    SDM3055          -20V 0.015  +0.004       @10M 4/U/S/* 10k-data          (USD ~0.5k)          +R?   107mm   
11    BK 5493CG       -12V 0.0035 +0.0005    @10G?4/U/R/G/S/P +labv 10k-data   (USD ~1.2k)          +R
11    BK 5493C       -12V 0.0035 +0.0005    @10G?4/U/R/S/P +labv 10k-data   (USD ~0.9k)          +R          +AAA
12    BK 5492CG       -12V 0.01   +0.004       @10G 4/U/R/G/S/P 0.5k-data      (USD ~0.9k)          +10vD +R    +AAA
12    BK 5492C       -12V 0.01   +0.004       @10G 4/U/R/S/P 0.5k-data      (USD ~0.7k)          +10vD +R    +AAA
13    GDM9061       -10V 0.005  +0.0006    @10G 4/U/R/L/S/P 100k-data       (USD ~0.8k)          +R          +AAA
14    HMC8012       -40V 0.02   +0.002       @10M 4/U/L/S/* +LXI +labv 50k-data    (USD ~1.1k)       +R

15   R6552         -30V 0.015  +0.006?      @10M 4/R/G/* 1k-data         (USD ~0.3k used)       +R
16   ADCMT 7351      -20V 0.015  +0.002?      @10M 4/U/P/* ??-data         (USD ~0.8k)          +R
17   PXI4060         -25V 0.024  +0.001?      @1G  4/? ? ?               (??USD)             XX
18   Hioki 3239      -20V 0.016  +0.005?      @10M 4/R/G/* ??-data         (USD ~1.2k)          XX
19    DMM6500       -10V 0.001  +0.0004    @10G 4/U/L/S/P +LXI+labv 7M-dat   (USD 1.7k)             +10vD       +R
20   Hioki7276-02   -10V 0.0009 +0.00012   @10G 4/U/L/G/P 5k-data         (USD ~2.7k)          +R

21   3458A         -10V 0.0008 +0.00005   @10G                     (USD ~11.6k)
22   34470A         -10V 0.0016 +0.0002   @10G                     (USD ~3.5k)
23   K2002         -20V 0.0010 +0.00015   @10G                     (USD ~11.5k)
24   ADCMT 6581      -10V 0.0005 +0.0002      @10G
25   DMM7510         -10V 0.0014 +0.00012   @10G                     (USD ~6.4k)
26   8588A         -10V 0.0004 +0.00006   @10G

[maxwell3e10]

Regarding NPLC detection, I am trying to figure out all possibilities:
1. Detect actual line frequency (every cycle or every few cycles) and use that as sampling interval
2. Trigger on power line but use a fixed sampling interval (automatically detect 50 or 60 Hz).
3. Trigger on power line and use a user-set line frequency (50 or 60 Hz) to determine sampling interval
4. Not trigger on power line and use a user-set line frequency to determine a sampling interval
5. Not trigger on power line and use a minimum sampling interval of 0.1 sec as common for 50 and 60 Hz.

I am not sure where different meters fall on this set of options.

[free_electron]

i am thinking the production line has ongoing problems with surface resistivity of the PCB.

standard structure to test surface contamination. aka greasy fingers on precision circuit flagger.

[free_electron]

this thing is cool but disqualified for one and one reason only : you can't stack em or rack mount em. game over.

Their target market doesn't stack'em and rack'em though.

i disagree. for education (or any other purpose) it is nicer to be able to stack a number of 1/2 rackwidth machines. a e34xx series power supply and a 34401 for example.
they need to kepe to 1/2 rackwidth machinery with all the same depth. it makes it easy on the bench to stack em. test equipment should be 19 inch or half width and should all be the same depth. so it can easily be combined into a 'stack'

I hate those non-standard width/depth devices.

[nctnico]

this thing is cool but disqualified for one and one reason only : you can't stack em or rack mount em. game over.

Their target market doesn't stack'em and rack'em though.

i disagree. for education (or any other purpose) it is nicer to be able to stack a number of 1/2 rackwidth machines. a e34xx series power supply and a 34401 for example.
they need to kepe to 1/2 rackwidth machinery with all the same depth. it makes it easy on the bench to stack em. test equipment should be 19 inch or half width and should all be the same depth. so it can easily be combined into a 'stack'.

I hate those non-standard width/depth devices.

I agree to some extend. The fact is that you can fit a much larger (touchscreen) display in an 'oscilloscope' formfactor device. But it comes at the cost of stackability. Last year I bought a function generator with an oscilloscope form factor. It isn't stackable but the big screen makes it a whole lot easier to use.

[bdunham7]

Handily, I've seen it.

OK, I'll take your word for it.  I probably haven't properly considered the cost of the screen and such.  Still, the price/performance competition in entry-level test equipment is pretty brutal, although for some reason bench DMMs don't seem quite as competitive an area as oscilloscopes and handheld DMMs.  The big screen and form factor may carry the day for some people, but the performance and feature set is more in the 4.5 digit handheld category.  It may still be a good fit for some actual educational settings, but it's going to be tough to put many of those on working technicians' and engineers' benches.

[free_electron]

I agree to some extend. The fact is that you can fit a much larger (touchscreen) display in an 'oscilloscope' formfactor device. But it comes at the cost of stackability. Last year I bought a function generator with an oscilloscope form factor. It isn't stackable but the big screen makes it a whole lot easier to use.

Why does a multimeter needs such a large (color) display. If it's a 51/2 digit machine all it needs is 6 7-zegment or starburst displays or a vfd. Let alone a touchscreen...
those things like chart plotting and recorder are gimmicks. if you really need recording you hook em up to a computer. That tiny display is useless. and nobody in his right mind will sit there zooming in and out on the chart. you need the data in excel or some other file anyway to do analysis.
color touch UI's for multimeters don't make sense. especially not when the machine still has 27 classic buttons next to the display as well... you either go all touchy ui without buttons or not at all.
another issue with touchy user interfaces is that machinery on the bench is typically stacked just at the border of reachability. you want to keep the bench open for the item under test. Touchy user interfaces lead to very tired arms very fast...
not very ergonomic.

[nctnico]

I agree to some extend. The fact is that you can fit a much larger (touchscreen) display in an 'oscilloscope' formfactor device. But it comes at the cost of stackability. Last year I bought a function generator with an oscilloscope form factor. It isn't stackable but the big screen makes it a whole lot easier to use.

Why does a multimeter needs such a large (color) display. If it's a 51/2 digit machine all it needs is 6 7-zegment or starburst displays or a vfd. Let alone a touchscreen...
those things like chart plotting and recorder are gimmicks. if you really need recording you hook em up to a computer. That tiny display is useless. and nobody in his right mind will sit there zooming in and out on the chart. you need the data in excel or some other file anyway to do analysis.

That depends on what you need and what the device can do inside the box. If the device can show a chart and calculate min/max, average and standard deviation then you basically have 99% of the analysis requirements right there in the box. The graph gives you a visual view of the stability during the measurement and the numbers show you the statistical information. Again, it depends on what a device can do by itself. Personally I prefer to have all the analysis being done by the device itself since hooking it up to a computer usually isn't that straightforward and takes hours to mess with crappy software, drivers and incompatible versions. Been there, done that. Hard pass on anything that needs NI Visa.

color touch UI's for multimeters don't make sense. especially not when the machine still has 27 classic buttons next to the display as well... you either go all touchy ui without buttons or not at all.

IMHO that is mostly for people who still think they need to have buttons. On most touchscreen enabled test equipment I hardly use the buttons but some people insist on having buttons. MicSig for example offers versions of their tablet oscilloscopes which have buttons nowadays after starting out with button-less designs; it seems there is a large enough market for it.

[Kleinstein]

Regarding NPLC detection, I am trying to figure out all possibilities:
1. Detect actual line frequency (every cycle or every few cycles) and use that as sampling interval
2. Trigger on power line but use a fixed sampling interval (automatically detect 50 or 60 Hz).
3. Trigger on power line and use a user-set line frequency (50 or 60 Hz) to determine sampling interval
4. Not trigger on power line and use a user-set line frequency to determine a sampling interval
5. Not trigger on power line and use a minimum sampling interval of 0.1 sec as common for 50 and 60 Hz.

I am not sure where different meters fall on this set of options.

The first option is not really an option, if at all this would requite a PLL to generate a line synchrounous clock. It's quite some effort, but some old meters (Prema, Solatron, some Datron106x?) with long integration did that. It is less desirabel for shorter integration.

A user set option is found in few meters, especially those with battery option and still a higher speed than 10 reading per second. The multiples of 100 ms is very popular with battery powered meters as it is OK with a pure display orientation.

Auto detect between 50/60 Hz is quite common in mains supplied meter. It sometimes comes with the option to also trigger line synchronous. Depending on the meter internals it may not be needed for a SD ADC, as there are no separate signal and zero measurements as with more classical MS DMMs with higher resolution. As usual for a SD ADC chip or DMM chip set, there is no extra choice of a auto zero or non auto zero mode. In theory a SD ADC could have more options for the digital filter and could have simultaneous 50/60 Hz suppression even with less than 100 ms, at the cost of non equal weights.

The EDU34450 is kind of a odd meter, with a large screen and arguably nice UI, but the actual meter perfromance is more like at the lower end of the 5.5 digit meters. Not sure why they made the effort to save a few buck from the old 34450, which is allready not that great performance. A new 5.5 digit meter design, based on a modern of the shelf SD ADC chip should not be that complicated.

[maxwell3e10]

Regarding NPLC detection, I am trying to figure out all possibilities:
1. Detect actual line frequency (every cycle or every few cycles) and use that as sampling interval
2. Trigger on power line but use a fixed sampling interval (automatically detect 50 or 60 Hz).
3. Trigger on power line and use a user-set line frequency (50 or 60 Hz) to determine sampling interval
4. Not trigger on power line and use a user-set line frequency to determine a sampling interval
5. Not trigger on power line and use a minimum sampling interval of 0.1 sec as common for 50 and 60 Hz.

I am not sure where different meters fall on this set of options.

The first option is not really an option, if at all this would requite a PLL to generate a line synchrounous clock. It's quite some effort, but some old meters (Prema, Solatron, some Datron106x?) with long integration did that. It is less desirabel for shorter integration.

I think 3458 almost does it, it can measure and set an uneven line frequency, but I guess it doesn't do it continuously and automatically.

The EDU34450 is kind of a odd meter, with a large screen and arguably nice UI, but the actual meter perfromance is more like at the lower end of the 5.5 digit meters. Not sure why they made the effort to save a few buck from the old 34450, which is allready not that great performance. A new 5.5 digit meter design, based on a modern of the shelf SD ADC chip should not be that complicated.

I am not sure how EDU34450 can decide between 50 and 60 Hz if it doesn't have a trigger circuit and as far as I can tell from the manual doesn't have a menu option either. Perhaps it is a 50/60 Hz notch filter in the ADC.

[Kleinstein]

There are no separate specs for 50 Hz and 60 Hz operation.  Even for the medium mode, there may be some additional averaging before display, as one can not really read the number at 50 updates per second - this would barely be the screen update rate. 10 SPS update rate is still relatively fast for the screen.
The old 34450 spec also show only one set of speeds - so maybe it is not using exactly 1 PLC for the medium mode either.

Chances are the slow mode is a multiple of 100 ms and thus natural simultaneous 50/60 Hz suppresion.

With a SD ADC adjacent readings may not be fully independent. So the reading rate may not directly reflect the integration time and the integration may not have a sharp start or end. Chances are the medium mode is a kind of compromise mode. There are kind of compromise filters, that can give some 60 dB NMRR for 50 and 60 Hz even with a reasonable fast rate.
https://www.analog.com/media/en/technical-documentation/application-notes/AN-611.pdf
It takes a little more settling, but still faster than 100 ms.

[free_electron]

IMHO that is mostly for people who still think they need to have buttons. On most touchscreen enabled test equipment I hardly use the buttons but some people insist on having buttons. MicSig for example offers versions of their tablet oscilloscopes which have buttons nowadays after starting out with button-less designs; it seems there is a large enough market for it.

mixsig has gotten it right. an oscilloscope leans itself nicely for touch . pinch to zoom etc. especially the newer version that has a couple of rotary knobs is really nice.

For a mulitmeter .. i don;t know. you can show some stats on the display if you want. but chart recording ? whenever i needed recording it was to plot behavior of a device over long time ( like battery charge / discharge ) , or pulsed behavior ( like a cellphone going from deep sleep to sleep to contacting the tower , to sending a message and going back to deep sleep. those are complex charts. not handy on a tiny screen. you want that data in a computer.
Same thing if you make a setup where you have an input multiplexer in front of your dmm. like when characterizing a D/A convertor.step ,measure, step , measure. you do that on mulitple devices, then run statistics between devices. none of that can be done on the machine itself. you need a data dump in excel or matlab.

ni-visa is the work of the devil. just install the base ieee-488 driver nd talk directly from excel. i have a small vba library that allows me to send scpi directly from excel and i made 'virtual instruments'

Code: [Select]


dim mydmm as new hp34401(17)   ' new 34401 at gpib 17
dim vcc as new TTI340(12,1) ' new thurlby tandar power supply gpib address 12 , channel 1
dim vee as new TTI340(12,2) ' new thurlby tandar power supply gpib address 12 , channel 2
mydmm.range = voltsdc20  ' set to fixed 20 volts range DC
mydmm.digits = 6 '
mydmm.nplc=10
vcc.voltage = 5
vcc.currlimit  =0.1
vee.voltage = 3.3
vee.currlimit = 0.1
tti340(12).alloutputson
tti340(12).attachexceptionhandler shutdown(event)

for x = 1 to 10
    vcc.voltage = 5v + (x/10)
    activecell.text = mydmm.read
    nextcol
    activecell.text = vcc.voltage.set
    nextcol
    activecell.text = vcc.voltage.read
    nextcol
    activecell.text = vcc.current.read
    newrow
next
tti340(12).alloutputsoff


sub shutdown(event)
     select case event
     case overcurrent
           tti340(12).alloutputsoff
           msgbox "overcurrent detected. system safed and stopped"
          end
     case else
         ' there ain't any others
     endcase
end sub

this would give me a table with dmm reading , set vcc , actual vcc (measured through the psu) , current draw on vcc
if i need ot do that on a device that has 4 output : instantiate a 34970 relay multiplexer and put another for-next loop that switches the relay.

Code: [Select]

mymux= new hp34970(15,1) ' gpib 15 , relay card 1
for y = 1 to 4
mymux.channel = y
  '  blabla
next y

i have virtualisations for all the instruments i use. if tomorrow i dont have a 34401 but a keithley 2002 all i do is change the instance. the virtual machines use the same syntax. so i don't need to alter my code.
mydmm = new keithley2002(COM1)  ' COM1 is constant larger than gpib address 32. if the driver gets an 'impossible' gpib address it subtracts 32 and opens COMx instead of GPIB
mydmm.serialinterface (9600,n,8,1)

the rest of the code is unchanged.

[Kleinstein]

The chart recording is sometimes nice when checking if the reading is stable. Something to differentiate between settling , drift and noise. It can also help to see if a recoded set of data is good (especially including the part of interest), before sending it to a USB stick - in case the DMM is not directly connected to a PC.  For a trend plot the slightly larger screen can be helpful.

Quite some experiments would measure more than 1 votlage. So it may be nice to have the option to use one of the ohms sense input as a 2nd optional input (a bit like the ratio function found in some higher end meters). From the HW side this should work for DC voltage in the 1 V and 100 mV range. It is a bit limited, but would be a relatively simple software only part.

[maxwell3e10]

What one would want in a modern benchtop DMM at a minimum is a trend display and ability to save data to USB stick (also faster recording than once per sec).  Something like Owon XDM3041 and Siglent SDM3045X do that. Otherwise, one might as well get an older used HP or Keithley meter for less than $200.   

But if one is equipping a student lab with 10 DMMs maybe Keysight makes sense if its going to be more stable than Owon/Siglent and one is spending someone else's money.

[floobydust]

Any confirmation for what the USB port is good for?  F/W updates, stand-alone datalogging as a text file?
It can't display a histogram or bar chart or trend chart, unless you recall a log, I believe.

[SilverSolder]

why are the guard traces split?

It seems the guard emanates from the chip itself...  one on each side of the sensitive pin.  Maybe it just makes it easier to route the board?

Maybe the guard traces are driven?  Could be interesting to measure.

By not closing the guard traces - are there any advantages e.g. avoiding making loops that pick up magnetic fields etc. ?

[WackyGerman]

Could be that Siglent made this multimeter for Keysight . The Siglent rust is still included or is this an option which cost extra $ 

[tszaboo]

IMHO that is mostly for people who still think they need to have buttons. On most touchscreen enabled test equipment I hardly use the buttons but some people insist on having buttons. MicSig for example offers versions of their tablet oscilloscopes which have buttons nowadays after starting out with button-less designs; it seems there is a large enough market for it.

mixsig has gotten it right. an oscilloscope leans itself nicely for touch . pinch to zoom etc. especially the newer version that has a couple of rotary knobs is really nice.

For a mulitmeter .. i don;t know. you can show some stats on the display if you want. but chart recording ? whenever i needed recording it was to plot behavior of a device over long time ( like battery charge / discharge ) , or pulsed behavior ( like a cellphone going from deep sleep to sleep to contacting the tower , to sending a message and going back to deep sleep. those are complex charts. not handy on a tiny screen. you want that data in a computer.
Same thing if you make a setup where you have an input multiplexer in front of your dmm. like when characterizing a D/A convertor.step ,measure, step , measure. you do that on mulitple devices, then run statistics between devices. none of that can be done on the machine itself. you need a data dump in excel or matlab.

I have a standard, that calls for a temperature measurement, when the temperature settled, eg: It doesn't change more than 1K over 30 minutes. The output is the max value.
You can do average energy usage measurements for low power devices by just running it from the DMM. Onboard statistics gives you directly the end result. Only need to do a screen capture, and there is less possibilities for mistakes.
This is just two examples why it is useful.

[free_electron]

IMHO that is mostly for people who still think they need to have buttons. On most touchscreen enabled test equipment I hardly use the buttons but some people insist on having buttons. MicSig for example offers versions of their tablet oscilloscopes which have buttons nowadays after starting out with button-less designs; it seems there is a large enough market for it.

mixsig has gotten it right. an oscilloscope leans itself nicely for touch . pinch to zoom etc. especially the newer version that has a couple of rotary knobs is really nice.

For a mulitmeter .. i don;t know. you can show some stats on the display if you want. but chart recording ? whenever i needed recording it was to plot behavior of a device over long time ( like battery charge / discharge ) , or pulsed behavior ( like a cellphone going from deep sleep to sleep to contacting the tower , to sending a message and going back to deep sleep. those are complex charts. not handy on a tiny screen. you want that data in a computer.
Same thing if you make a setup where you have an input multiplexer in front of your dmm. like when characterizing a D/A convertor. step ,measure, step , measure. you do that on mulitple devices, then run statistics between devices. none of that can be done on the machine itself. you need a data dump in excel or matlab.

I have a standard, that calls for a temperature measurement, when the temperature settled, eg: It doesn't change more than 1K over 30 minutes. The output is the max value.
You can do average energy usage measurements for low power devices by just running it from the DMM. Onboard statistics gives you directly the end result. Only need to do a screen capture, and there is less possibilities for mistakes.
This is just two examples why it is useful.

yes for such simple things.

 i guess i am more used to do device characterisation. Test devices if they follow spec. Sweep over different supply voltages, temperature ranges and do cross correlation with what the big testers see. or monitor a system (board test). those things invariably collect so much data that needs post processing anyway. Like do the inl/dnl of a DAC's. Do this over 25 devices on min , nominal and max VCC and from -25 to +125 degree in steps of 25 degree. such setups need to be driven as they can run for hours. board in the oven. 25 chips in sockets , input mulitiplexer , dmm , power supplies and oven all under software control. data spit into excel and immediately the spread plotted. those are typical things i used to do. not possible with the on board graphing stuff.

my criteria for a good multimeter are different
when it comes to measuring :
- stable and fast reading for volts and ampere and ohms
- predictable behavior (locking ranges. autoranging is the work of the devil in certain cases)
- high input impedance on volts
- low burden voltage on current
- remotability

-troubleshooting functions ( diode/zener/led test , capacitor meter , inductance meter , continuity , low ohms measurements (4wire) . touch-hold. min/max, temperature probe )
for these functions i don;t need top notch performance. i have a true lrc bridge if i need to characterize a cap or inductor.
troubleshooting functions are for handheld meters. a bench machine needs to excel in VOA , drift , stability and digits.

[nctnico]

troubleshooting functions are for handheld meters. a bench machine needs to excel in VOA , drift , stability and digits.

That is nonsense IMHO. A handheld DMM sucks big time for bench use because it is always out of batteries, switches off just when you want to read it and the display is in the wrong orientation. Fortunately there are good bench DMMs for daily 'troubleshooting' use. Don't fall in the trap thinking that a bench DMM is a high precission instrument by definition. It isn't.

[tszaboo]

IMHO that is mostly for people who still think they need to have buttons. On most touchscreen enabled test equipment I hardly use the buttons but some people insist on having buttons. MicSig for example offers versions of their tablet oscilloscopes which have buttons nowadays after starting out with button-less designs; it seems there is a large enough market for it.

mixsig has gotten it right. an oscilloscope leans itself nicely for touch . pinch to zoom etc. especially the newer version that has a couple of rotary knobs is really nice.

For a mulitmeter .. i don;t know. you can show some stats on the display if you want. but chart recording ? whenever i needed recording it was to plot behavior of a device over long time ( like battery charge / discharge ) , or pulsed behavior ( like a cellphone going from deep sleep to sleep to contacting the tower , to sending a message and going back to deep sleep. those are complex charts. not handy on a tiny screen. you want that data in a computer.
Same thing if you make a setup where you have an input multiplexer in front of your dmm. like when characterizing a D/A convertor. step ,measure, step , measure. you do that on mulitple devices, then run statistics between devices. none of that can be done on the machine itself. you need a data dump in excel or matlab.

I have a standard, that calls for a temperature measurement, when the temperature settled, eg: It doesn't change more than 1K over 30 minutes. The output is the max value.
You can do average energy usage measurements for low power devices by just running it from the DMM. Onboard statistics gives you directly the end result. Only need to do a screen capture, and there is less possibilities for mistakes.
This is just two examples why it is useful.

yes for such simple things.

 i guess i am more used to do device characterisation. Test devices if they follow spec. Sweep over different supply voltages, temperature ranges and do cross correlation with what the big testers see. or monitor a system (board test). those things invariably collect so much data that needs post processing anyway. Like do the inl/dnl of a DAC's. Do this over 25 devices on min , nominal and max VCC and from -25 to +125 degree in steps of 25 degree. such setups need to be driven as they can run for hours. board in the oven. 25 chips in sockets , input mulitiplexer , dmm , power supplies and oven all under software control. data spit into excel and immediately the spread plotted. those are typical things i used to do. not possible with the on board graphing stuff.

my criteria for a good multimeter are different
when it comes to measuring :
- stable and fast reading for volts and ampere and ohms
- predictable behavior (locking ranges. autoranging is the work of the devil in certain cases)
- high input impedance on volts
- low burden voltage on current
- remotability

-troubleshooting functions ( diode/zener/led test , capacitor meter , inductance meter , continuity , low ohms measurements (4wire) . touch-hold. min/max, temperature probe )
for these functions i don;t need top notch performance. i have a true lrc bridge if i need to characterize a cap or inductor.
troubleshooting functions are for handheld meters. a bench machine needs to excel in VOA , drift , stability and digits.

They clearly didn't made this with production testing in mind. There are no jacks on the back, trigger and complete is missing, and it is very slow for those sort of measurements.
In fact probably they probably limited the remote interface some unexplained way.

Like they wrote this for the EDUx:
Remote interface: USB, LAN standard
Programming language: SCPI-1994.0, IEEE-488.2

While the 3446x
LXI (rev 1.4)10/100Base-T Ethernet (Sockets, VXI-11 protocol, Web user interface) (Optional on 34460A)
USB USB 2.0 (USB-TMC 488 & MTP protocol)
Language SCPI-1999, IEEE-488.2, 34401A compatible

The old SCPI interface and the lack of specification for LAN doesn't increase my confidence.

[Kleinstein]

Given that the µC on the display side look quite powerful, it is really odd to have only a limited SCPI interface. Chances are this is taken over from the old version of the 34450 that also only supports the 1994 verion.
This meter is really not one build for performence. Many of the limitation may be from the DMM chip set - so it may be hard to change on some key paramters without a new design. Looks like the old 34450 is from around 2003 (at least thats the date of the EMC compilance tests - there may have been a predecessor), so it is a pretty old design. With most functions inside a special DMM chip more like a extended handheld than a cut down 6 digit meter.

[madires]

Nothing odd, just business as usual:
We need an educational bench DMM for x bucks. Since it's 200 bucks less than our standard model it can't have the same features. Remove some! And don't make it too obvious! It should look like a new design.