Turning fan off to reduce DMM voltage noise

[maxwell3e10]

I tried to unplug the small fan in Keysight 34461A DMM to see the effect on electrical noise. It gives a rather dramatic reduction of short-term voltage noise for shorted inputs on 100 mV DCV scale, see picture below. The black trace is taken immediately after the fan was unplugged and the case closed again, so the transient is due to meter coming to new thermal equilibrium. The internal sensor temperature went up from 27C to 31C.

It is not surprising that the noise would go down somewhat due to absence of air currents, but it seems rather dramatic. Presumably it would have similar effect in 34465 and 34470 meters, since they have very similar construction. Has anyone tried to turn off the fan in other DMMs to see the effect on the noise?
Edit: Checking Dave's old teardown videos, DMM7510 has plastic covers over all analog input circuitry, Keysight 34460-70 meters don't. Maybe one could design a little 3-d printed cover.

[HighVoltage]

I would not have expected such a big difference in electrical noise, with and without the fan.

On almost all my Agilent / Keysight gear with fan, I have inserted a 100 Ohm resistor in line with the fan, to slow it down and reduce the auditable noise. May be it also had an effect on electrical noise?

[niner_007]

try narrowing down by powering the fan from an external source, to see whether it's the thermal gradients or EMI

34461 uses an LM399, I expect few uV of noise from that in any case, and worse it's not shielded; in the 34470 the LTZ1000 is shielded, so you might see different results

[maginnovision]

If you really wanted to test this I would run it as such for much longer than a couple hours, maybe a week.

My Datron units run fanless stock. They are pretty stable right up until the heat overwhelms them. They then become unstable until they finally hit a plateau where they're able to get rid of any further heat through the case. I don't actually run them like this though. I have a fan blowing where the primary 5V regulator attaches to the steel case near where the transformer also connects. I suspect the 34461A will also suffer from some heat soak until the case is hot/warm enough to keep it equalized. At that point though you're cooking it all. Perhaps isolating the analog circuitry from the airflow isn't bad but it's not how it was designed so I would again recommend some longer term logging to compare to stock performance.

[guenthert]

  Which Datron model is that?  My Datron 1271 doesn't get more than just hand-warm.  The older HP 3456A however gets quite warm (only the first generation came with a fan).

  Now, I'd understand that manufacturers add a fan for test gear meant for industrial process control, like the Datron 1271 or HP3458A which might sit in enclosures which get quite hot, but not why it would be added for gear meant for metrology sites (like e.g. the Datron 1281 or Fluke 8508A) with their controlled environment or bench instruments, like the HP34401A which will be used in environments, not so tightly specified, but not too unpleasant either.  There are after all, only so many submarines and HP has made special military versions of their gear in the past.

[maginnovision]

I've got a few but the Datron 1082 gets the warmest. The units have the 5V regulator(the hottest one) and transformer bolted to the rear of the case so most of the heat should escape that way. Obviously if you have good airflow and temperature regulation for the room it's not a big deal since that's their ideal environment. I didn't have any issues when they were running in my office with AC. I've been running them in my garage and even on a cool day the lack of airflow eventually causes heat issues. I also have a Prema 5017 and it's fanless but if you keep airflow from the case(covered, in a cabinet, etc...) it will also heat up quite a lot. When you have a meter that has a fan you can bet it's not going to fair as well as these two which are made to be fanless.

[NoisyBoy]

That’s an interesting observation.  My 34461A sits on top of two late built 34401A in my home lab, while I love the graphing capabilities of the 34461A and its shorter depth, the silent aspect of the 34401A is golden.  Now you get me thinking about the electrical noise aspect when one compares the two platforms.

[maxwell3e10]

The effect is most dramatic on the 100 mV scale, it is less significant on 1V and barely noticeable on 10V scale. I suspect it would be similar for 34465 and 34470 meters. The noise for shorted inputs  does not depend so much on the voltage reference. In 34470 the LTZ1000 is shielded in a box, but the analog input circuitry is the same and is not shielded. On all precision Keithley meters (2001, 2002, 7550), there is a plastic cover over all analog electronics.

I also found this graph from OldNeurons, who first posted on the forum about this a few years ago. I think its clear from the plot that the effect is mostly thermal, not due to electrical interference from the fan.

Fan:
I suspect that the fan is creating some turbulences into the DMM resulting into non homogeneous temperature over sensitive components.
I am surprised that you don't see any difference with the sides covered or not. I ran yesterday another test from front panel, with the correct LF setting and manual recording of the DMM internal temperature. To cover the sides, I used some rectangular pieces of magnetic rubber.

Everything is summarized in the graph below.

The results are, for me, obvious, and best results are achieved with sides covered and fan off ...

Regards,
Jean

[TheSteve]

The 34460A doesn't have a fan - but it is also has poorer specifications. If the 34460A can run without a fan it is likely the 34461A can as well. That is if you don't mind your meter running nice and toasty. I tried my 34461A without a fan for a bit and it was silen,t but it also had a lot more drift with temp. I didn't do any electrical noise checks. I do know my 34461A is "quieter" when logging a 10 volt reference compared to my 34401A. However that could just be down to the LM399's in each.

[coppercone2]

whats going to happen now, external fan driven by a belt or a blower mod?

or is it the airflow on high impedance stuff causing problems?

hard to imagine the turbulence is causing problems

[Gyro]

whats going to happen now, external fan driven by a belt or a blower mod?

or is it the airflow on high impedance stuff causing problems?

hard to imagine the turbulence is causing problems

Actually, it's very easy to demonstrate the turbulence causing problems. The LM399 datasheet*, for instance, specifically mentions low frequency noise effects caused by draughts across its (Kovar) leads, due to thermocouple effects. Other sensitive circuit nodes will suffer the same problems due to thermocouple effects in turbulent airflow.

Thermal EMFs need to be allowed to stabilize.


EDIT: *Actually it's the LM399 Application note... NS AN161.

[HighVoltage]

Both of my 34470A are with Keysight to fix the ACAL problems, so I can not test those.
However, I could test my 34465A with and without fan.

What data do you need for comparison?

100mV scale, 100 NPLC, any other settings?
3 hours running with shorted input with fan and then 3 hours with the fan disconnected?

[Dr. Frank]

If the 100mV and 1V ranges are affected only, but not the 10V range, then it has nothing to do with the references, but with the x10, x100 DC amplifiers, or the signal path including the jacks in front of the amplifiers.

I'm also no fan of the fan, especially inside the 3458A, where the air is also blowing across the LTZ1000A circuits solder joints, which is NOT recommended by LT.
Same goes for the reference of the 34470A, only the solder joints of the LTZ itself are protected.

Maybe I'll do some experiments with my 3458A.

Frank

[IanJ]

I had noticed how noisy my 34461A in comparison to the 3458A whilst logging the two together running of the same source.........I'll do some experiments tonite with the fan. Very interesting.

Ian.

[maxwell3e10]

What data do you need for comparison?

100mV scale, 100 NPLC, any other settings?
3 hours running with shorted input with fan and then 3 hours with the fan disconnected?

Yes, that is the setting I was using. The other thing one can try based on OldNeurons data (which I haven't yet) is to cover the side vents of the meter.

[IanJ]

Hi,

FWIW, cheap n cheerful 30min 10Vdc test on the 34461A.

Red = W/shop temperature
Yellow = 3458A (for reference)
Aqua = 34461A
Both DMM's connected to same 10V source.

A = 34461A Fan on, case closed
B = 34461A Fan off, case closed
C = 34461A Fan off, case open
D = 34461A Fan off, case closed

[Mr. Scram]

I always felt that introducing moving parts into a solid state machine was taking a step back but this is more significant than I'd imagined.

[Kleinstein]

At the µV level a fan can have quite some effect. However it is not clear whether the fan will improve things or cause more problems: With a fan one ideally has a relatively steady flow of air. There may be some turbulence, but mainly shorter time scale.
Without a fan hot air can build up and than flow off in turbulent convection with variations on the longer time scale. There is also a good chance that it takes longer to reach thermal equilibrium.

If mainly effecting the 100 mV range and tested with a short, the reference and also the ADC itself should not be the critical part, but more like the input amplifier, protection and input switching.

Besides turning off the fan all together, it may also help to change the heat / air flow. Some sheet metal to keep out the air flow, but still alow heat flow could be a good idea.

I would definitely not run the 3458 without the fan - there are some parts (e.g. hard to get high speed comparators and parts of the AC board) that run rather hot even with the fan. These may overheat without extra cooling. The speed of the air flow can have a small effect on the readings - this is why they suggest cleaning the air filter before critical measurements / calibration.

[niner_007]

Both of my 34470A are with Keysight to fix the ACAL problems, so I can not test those.

oh wow, is that so they can diagnose the problem, or only your units are affected, what about mine lol? is it software or hardware, hmmm?

[iMo]

Perhaps off topic (as I do not have the fan here )..

https://www.eevblog.com/forum/projects/thermostating-the-hp-34401a-meter/msg2619108/#msg2619108

[maxwell3e10]

I used Kapton tape to build an analog electronics air cover for the 34461A.


It works rather well. The noise level (for 100 mV scale, 100PLC) is now 25 nV rms, with fan or no fan. That is a factor of 4 reduction from the original noise level of 100 nVrms with fan on. The start-up from cold is a little faster now with fan on, see picture below. The internal temperature is 27C, the same as before any modifications.

So, putting an air flow cover over electronics really helps with low-level voltage measurements. It should also help on current and resistance ranges that often use small voltages. In this way Keithley knows what they are doing in always putting a plastic box over sensitive electronics in their precision DMMs.

[MegaVolt]

I wonder if it is possible to localize the component that makes the greatest contribution and isolate only it?

[Kleinstein]

One could look for sensitive parts with a local probe, either a small extra heater or a nozzle with fresh air. Chances are one would find some temperature sensitive parts. There is still the uncertainty how much changing the air flow would effect that part. At least one could identify parts that are not that sensitive.

Ideally the designers have done a similar test with a prototype and if needed added more shielding, or otherwise improved really bad spots already.

[maxwell3e10]

It seems there could be many components that can cause 100nV level noise due to thermocouple effects, anywhere there is a contact between dissimilar metals. For example, front-rear terminal switch, input relays, etc. If one looks at DMM7510 teardown video, Keithley has one big plastic box around all analog input section,  plus extra smaller boxes over some other components, in addition to the box over voltage reference.

[coppercone2]

I wonder if you can put little fins and stuff on the chips and dividers between them to direct airflow in a favorable way