Is it really worth waiting for the 6.5 digit multimeter to warm up?

[dp]

I decided to get myself the answer to this question, with a small experiment. Not having a voltage or current reference I can rely upon, I opted for the 2-wire resistance measurement. I've simply shorted the inputs with a good cable, and waited several hours to be sure the cable is in thermal equilibrium with the instrument inputs (to avoid any contributions from the Seebeck effect to my experiment). Both my 6.5 digit multimeters have the trend display feature, so I simply used it and then captured the screen on a USB flash drive. At the beginning of the experiment I've pressed the Null button. Here's what I got on Agilent 34461A and Rigol DM3068:

[jpb]

The problem with selecting resistance measurement as a test is that you are continuously passing a current through the cable so there will be some heating effect albeit small for a cable - it may be that you're measuring how long the cable takes to come to thermal equilibrium.

Why not just measure the "zero" volts across a short? It would be noisy but should settle down. You could perhaps do the same for the volts across a battery - given that the meter impedance is very high the battery shouldn't discharge and change in voltage over a short time frame.

[bdivi]

I have measured the warm-up of my Keithely 2015. It starts cold 25 ppm off and then in one minutes gets within 10 ppm of the final value.

Good to know that the meter needs so little to stabilize.

[Vgkid]

My solartron 7065 takes almost 2 hours to become stable :/

[Lightages]

Even handheld meters are usually more accurate after a warmup. How long a warmup is needed for any instrument relates to how accurate you need your measurement. Only need 3.5 digits of accuracy? Just power on and use it. Need that measurement right to the spec of the instrument? Wait the recommended time.

[dp]

The problem with selecting resistance measurement as a test is that you are continuously passing a current through the cable so there will be some heating effect albeit small for a cable - it may be that you're measuring how long the cable takes to come to thermal equilibrium.

The thermal effects of a 1 mA current passing through a thick measurement cable are lost in the noise.

Why not just measure the "zero" volts across a short? It would be noisy but should settle down.

Because I'd get noisier graphs with no redeeming benefits. See below the actual results of this experiment:

[jpb]

The problem with selecting resistance measurement as a test is that you are continuously passing a current through the cable so there will be some heating effect albeit small for a cable - it may be that you're measuring how long the cable takes to come to thermal equilibrium.

The thermal effects of a 1 mA current passing through a thick measurement cable are lost in the noise.

The 1mA presumably passes through internal circuitry as well as part of the constant current source - certainly your resistance measurement curves show more warm up time than the voltage measurement ones. So yes, the power dissipation in the cable is negligible but the warm up time of the current source is significant. This is interesting in its own right.

Why not just measure the "zero" volts across a short? It would be noisy but should settle down.

Because I'd get noisier graphs with no redeeming benefits. See below the actual results of this experiment:

I don't know what you mean about "no redeeming benefits", the curve is a very different shape with a sharp knee around 12 minutes where the curve reaches roughly zero. The resistance curve is smoother but takes much longer to start to flatten off.

[dp]

I don't know what you mean about "no redeeming benefits", the curve is a very different shape with a sharp knee around 12 minutes where the curve reaches roughly zero. The resistance curve is smoother but takes much longer to start to flatten off.

To me, the Agilent curves look very similar, the differences being artefacts in the noise of the latter. The slight difference between the Rigol curves can be due to the fact that the instrument was "pre-warmed" from the previous experiment, so it took slightly less time to stabilise. In either case, it is obvious that we're not seeing the thermal effects of the 1 mA current used in the resistance measuring experiment.

[mimmus78]

I used to play with a LM399 reference I built some time ago so I notice my K2000 warm up in few minutes (maybe five) and the other puppy a K2001 takes at least 40 minutes to stabilize.

[TooOldForThis]

I couldn't find a warmup spec for my Agilent 34461A so I ran an experiment.  I connected the 34461A and a Rigol DM3068 to the same 6mV DC source.  Both meters were set to their lowest range (100mV for the Agilent, 200mV for the Rigol) and both meters were set to 100 PLC integration time. The Rigol unit had been running for 10 hours. The Agilent had been in standby for 20 hours.  I started collecting data as soon as the Agilent completed its bootup sequence.   The graph shows the difference between the two units in micro-volts vs time in seconds.   The Agilent starts 11uV low and slowly rises.   After 16 minutes the two meters are reading within 1uV of each other.  After 20 minutes they are within 0.5uV.  After 23 minutes the Agilent is fully stable.     

[amyk]

34461A specs are after 60 minutes warmup time, as mentioned in http://cp.literature.agilent.com/litweb/pdf/5991-1983EN.pdf

[KedasProbe]

I did a test to see if the DM3068 negative start of about -10µV is higher for other ranges.
So I did a cold start and set it to the 20V range instead of 0.2V,  so would the start error now be -1mV (100x -10µV)?


Answer: No, if you are about to use the 20V range, there isn't really a point to wait, the error isn't even visible on the display.

[nukie]

I used to play with a LM399 reference I built some time ago so I notice my K2000 warm up in few minutes (maybe five) and the other puppy a K2001 takes at least 40 minutes to stabilize.

Keithley 2000 generally requires a shorter warm uptime due to a more efficient power supply and lower power consumption. The older and aged units typically achieves stability within 5 minutes. The 2001 is not really a Keithley and not blood related.

[HiTech]

Both of my Datron 1065A units require 30 min warm up/stability period before I go trusting their readout as "absolute". Although for the most part they are spot on much sooner than that, I've just gotten into the habit of allowing all of my equipment a good warm-up period before taking measurements that I deem as needing to be "accurate". You need to remember that the aging rate of the components used in high end test equipment is a major factor in repeatability and stability.

[macboy]

Keithley 2000 generally requires a shorter warm uptime due to a more efficient power supply and lower power consumption. The older and aged units typically achieves stability within 5 minutes. The 2001 is not really a Keithley and not blood related.

I wish people would stop making this mistake. The 2001 is a true Keithley engineered and Keithley built instrument. The 2100 is not a Keithley-made instrument (Array M3600A apparently).

[TiN]

Since when 2001 is not a Keithley
My 2001 warm's up around 30 min, but it's not hard to get pretty close readings after 5 min already.
8.5 digit 2002 takes about 2 hours tho, and has neat internal temperature to calibrated temp delta measurement, which visible to user.

[mimmus78]

Thanks TiN for your confirmation about warm up time of K2001.
I really was thinking 40 minutes was too much for it.
I never could imagine than the K2002 would take much much more.

What I really was thinking is that than more accurate this stuff go,
than less warm up time will influence the accuracy ... wondering what
is the black magic recipe inside it.

[TiN]

Black magic pretty simple, it's tempco of all different parts inside meter. Yes, LM399 or LTZ1000 in meters is ovenized, but except voltage reference, there are resistors, capacitors, active parts all around measurement switches, function blocks, and all of them drift differently. It takes time to get all temps to stable level before you can get desired accuracy.
And instrument was calibrated on specific temperature, so you can see that as a single point on the temp graph. More off you from that sweet spot - less accuracy you have.
More precision gear just have that "in-spec" window smaller, so it's easier to miss 10ppm accuracy window on 8.5digit meter rather than 5000ppm on 4.5digit one

If you read carefully thru calibration manual on Model 2002, you can even note that presence of metal cover or scan card option affects calibration, so it's recommended to do calibration with
proper option installed