Thermostating the HP-34401A meter

[iMo]

I've been thinking about the feasibility of keeping the internal temperature of my 34401 constant  .

As the first step I've isolated the internal frame from the outer aluminum case by a 3mm thick isolating material (an isolating sleeve put on the internal frame in the entire length) to avoid fast changes of the internal temperature (internal air flow).

My idea comes from the fact(s) that

a) the display filament (wired permanently on) takes 2-3 Watts off the transformer (6V AC, 10ohm when cold and maybe 20ohm resistance when hot),
b) the transformer gets pretty hot,
c) the total power consumption of the meter is "25 VA peak (10 W average)",

therefore I "do assume" I can keep the internal temperature at the constant level, say, ~35 degC by using the transformer and a load wired to the 6V AC (not the filament itself) as "the heater".

I've been usually using the display switched OFF while doing longer measurements so those 2-3 Watts could be redirected for the temperature regulation at that time.

Now, this is an idea, and while the technical solution with a temperature control seem to be rather feasible, I am not sure the 2-3Watts taken off the filament is enough for the stabilization.

Any experience with such an exercise?

[coromonadalix]

You will affect the reference voltage stability if you throw air in it,   you have to add to protection around the lm399 voltage reference.

Some people added a cage / a 3d molded casing or even somekind of foaming around it ....  on other meters.

Yes the xformer gets hot, but the meter was designed this way, and fiddling any or many things in it can affect its precision  ... 

The best scenario i've seen :  some people have removed the rear gray panel, it provide a small venting area, no more.


I think i saw you have measurement software,  try to measure a precision voltage for a few hours, 1hrs of warm up,  and try to cool down the inside from the rear with a small speed controlled fan, and do some checkups ...

And we did sugesst you on another thread, you can turn of the display with a command,  sure it wont cut the vfd filament ....  the xformer is designed to isolate many circuits from each other, but if you derive power from it you may add more problems ... this xformer cost around 100$-120$ from Keysight if it bust.

Or simply add a small push switch on the rear to cut the filament voltage from the xformer ?

Or add an thermal pad between the xformer and the frame to transfert some heat into it ??

[iMo]

Indeed, I do plan to measure the internal temperature (and to fuse the information into the serial data stream) to get some idea on the temperature inside (a pity HP had not simply put a sensor on the pcb, there are still some free ADC pins on the 80C196 left  ).
The regulation has not to be draconian such the transformer will smoke, just within the typical filament on/off current range.

PS: opening the rear bezel will just create internal airflow caused by the various temperature gradients, imho.
Therefore I've done the opposite - I've isolated completely the outer case from the internal frame, there is none free air gap/slot under the bezel visible when you look from the rear.
I cannot say how is the influence of that thermal isolation to the calibration as my references (ref01, lt1021-5/10, LM399) drift +/- 1-3ppm with longer measurements against the meter, such I cannot distinguish whether the drifts come from the meter or from the reference. A standard problem .

[coromonadalix]

have you seen  the qu1ck   34401a  oled project ?    leaking segment ....

You get a display without the heat ...  and surely you could add your thermal sensor on the board, or create a new board

He had a firmware with debug ---  serial output possibility

[Kleinstein]

The 34401 is limited accuracy anyway. There is not much to improve on the LM399 reference - maybe a cap to keep air currents constant (to reduce variations in thermal EMF.  The other critical part would be the resistor for the ADC - this would likely be the point to have a possible temperature sensor at.

The noise of the 34401 is also quite high - so it takes quite some times to measure to high resolution.

Chances are it would be easier to correct the small residual temperature effect instead of regulating the temperature.

[iMo]

That is an option too. The LM399 is pretty close under the aluminum shield, thus an isolation cap would not fit easily, imho.
I've been thinking to place the temp sensor (ie an LM35) from bottom side at the red mark (close to the resistor's package).

[Kleinstein]

The shield close to the LM399 reduces the need for a cap on the LM399.

The marked position for the sensor is an option. Another option would be the shield, to get a temperature representing a large area and thus with less chance for local variations.

To see how much the 2-3 W of possibly available heater power can do, the power consumption for the part to be regulated in normal operation is an important point.  The power consumption sets a limit on how much insulation is possible without getting to high a temperature.

So I kind of doubt it would work regulating to some 35 C, unless it's pretty cold outside. I would more expect 35 C internal temperature even in normal mode. So the temperature rise due to heater power would come on top of that - so more like 40-60 C.

[coromonadalix]

Not to be rude or anything,  what will be the overall gain with thinking this or that in the end, i do understand your points, but

Its a good meter, but not an absolute dmm like an 3458 with an ltz1000  or an Transmill 8.5 digit  etc ...


I had 3 of them, now 2, they are stacked on each other with the bumpers making space between the casings,  i dont have any problems with them, 1 had a new xformer (was put on the wrong voltage)  and 1 has a new china cloned vfd at 30$ usd

1 hr warmup and thats all.
 
Even with an thermal sensor in the central point of the dmm  Q:  will you get the results you want ??   Is it worth all the troubles ?, the 34401a  has a sucessor with better specs ...

[TheSteve]

What are you using as a voltage reference(assuming you are measuring voltage, not resistance) to verify the stability?
I recently logged a Fluke 732B reference with two 3458A's, a Datron 1281, 34401A and 34461A. The temperature stability of the 34401A was very respectable.

[iMo]

I do not claim issues with the meter. My primary intention has been to elaborate an idea of keeping the internal temperature of the meter stable (whatever the temperature will be) provided you have got a free source of 6Vx400-600mA of manageable heat inside the meter (when the display is off including the filament).

@Kleinstein: let us assume the sensor (TO92 package) will be glued to the aluminum shield. For an initial experiment with measuring the temperature it may work fine, but the ground of the sensor will be the outguard ground, however.. Thus I need to put a good insulator beneath the sensor - like beryllium oxide ceramics or a piece of diamond
The same apply when the sensor will be glued on the bottom as depicted above..
How could be the isolation issue critical (provided I am not going to mess with 1000V measurements)? Any capacitive coupling (2-3pF)?

[iMo]

Temperature sensor installation.

[iMo]

The first ~hour (Tamb aprox 26degC). The measurement started a couple minutes after the switching meter on.

[iMo]

The second hour, still rising..
The fourth hour, still rising (44C now), the ambient temp is stable (26.3C), hopefully we reach an equilibrium soon (the ambient temp will certainly decrease during the night)

sleve = sleeve

[iMo]

Now, provided we will reach 44.5C internal temperature with 10W heat (I can see at the mains socket) at 26.3C ambient, the

deltaT/1W = (44.5-26.3) / 10W = 1.82C/W

With, say, 0-3W heat stolen from the filament, the regulation range will be

Temp_reg = 3W * 1.82C/W = 5.46C

provided my calculation is somehow correct.. 

[3roomlab]

estimating heating parameters using your data in LTspice
I think 44C is too hot. but maybe it is normal

(edit : the full spec of 34401a is 10w, this requires the thermal resistance to be 0.9R to reach 44.5C in 15k seconds)

[iMo]

After 5.5h we are reaching equilibrium at 44.1935C..
LM399 is happy
Not sure the other chips and tantalum caps are happy too 

[iMo]

The above "time to achieve the equilibrium" was most probably influenced by the rising ambient (I cannot measure with higher precision yet). So my estimate is the time till the settling is shorter.

The important finding is the aprox 1.79C/W coefficient (in this setup) which gives the potential temperature regulation range.

[iMo]

FYI - below a few measurements WITHOUT the insulation SLEEVE.
The temperature at the shield "settled" at 42.8C (display off, Tamb=26C).

[iMo]

As the results without the sleeve are _much_ worse, I've doublechecked the sensor is still glued at the shield (ok) and I've put the sleeve back.

Here is the new measurement after 1hour with the insulation sleeve ON, display off, Tamb=25.8C.

It looks the fluctuations (of what?) without the sleeve are pretty high.

PS: added the new measurement with the sleeve after 2hours - almost no turbulence..

.. after 3h20mins, Tamb=25.9, opened window, gentle breeze in the room..

[COSMOS2K]

I've been thinking about the feasibility of keeping the internal temperature of my 34401 constant  .

As the first step I've isolated the internal frame from the outer aluminum case by a 3mm thick isolating material (an isolating sleeve put on the internal frame in the entire length) to avoid fast changes of the internal temperature (internal air flow).

My idea comes from the fact(s) that

a) the display filament (wired permanently on) takes 2-3 Watts off the transformer (6V AC, 10ohm when cold and maybe 20ohm resistance when hot),
b) the transformer gets pretty hot,
c) the total power consumption of the meter is "25 VA peak (10 W average)",

therefore I "do assume" I can keep the internal temperature at the constant level, say, ~35 degC by using the transformer and a load wired to the 6V AC (not the filament itself) as "the heater".

I've been usually using the display switched OFF while doing longer measurements so those 2-3 Watts could be redirected for the temperature regulation at that time.

Now, this is an idea, and while the technical solution with a temperature control seem to be rather feasible, I am not sure the 2-3Watts taken off the filament is enough for the stabilization.

Any experience with such an exercise?

Hello:
In my laboratory I have 4 multimeters of 6 ½ digits 1x HP3456A, 2x Fluke 8505A and 2x Agilent 34401A. The last to be acquired were 34401a, had symptoms of instability (without taking into account their calibration), the trafo heated a lot of one of them, in the other it remained a few degrees above room temperature, between 8 ~ 10. I really surprised me.

The excess of temperature was because the previous owner changed the capacitors of the source without realizing that they had to be low esr, when changing the temperature of the trafo under a lot, it remains like the other unit.

The voltage reference has its own oven, it can be improved by putting a sponge (preferable black) surrounding it but that will not improve the accuracy or stability, only that it will stabilize faster and the possible air currents will be harder to act About it when the multimeter is without the lid.

The complete thermal stability of the whole set implies that both the resistors and U101, U102 and precision components are at the temperature of the interior of the multimeter, if it heated to heatter more and more. If you look at the xformer, it is far from the tension reference for this reason, but with an ambient temperature of 22° the xformer of which I have in the laboratory is about 30º more or less, until the resistances of the dividers do not Have that temperature the multimeter will not measure correctly.

If you do not have a stable reference voltage to test it, or even having it you want to see how the temperature acts on the divisor resistors connects a 4-wire short circuit and put the multimeter to measure ohm to 4 threads and have the null and leave it so until Stabilize, you will see the effect of the micro-ohms (more on the right) as as time passes the value of the measure, of course this with the multimeter with the cover set.

Sorry for my English

Greetings.