Getting one more digit from a 6.5 digit meter--without using GPIB

[RandallMcRee]

I "discovered" how to get that one more digit from my keithley 2015, on the front panel.

Pretty sure that this works for virtually all 6.5 digit keithleys. Post here if not.

Anyway, just set up the buffer to STORE a few readings. Hit ENTER. "* comes on for the time it takes to fill the buffer. Hit RECALL hit down twice and over once (you know the triangle keys). Voila, there is your average to 7 or 8 digits. Looks pretty stable on my meter.

Putting in a size of 1024 gets a nice noise reading (via STDDEV) on your "burning" reference.

Shorting the leads together and taking 1024 readings showed a noise floor of 760nV.  All this on the 10V range.

Hope this is useful to those with pro handled DMMs. 

[hwj-d]

Yes, it works ...
(playing around with some REF5020)

[Macbeth]

Hahaha. It just so happens I took my decent 6.5 digit stuff out of storage today. I just noticed this and decided to give it a spin... 



Who needs a 3458A? 

[hwj-d]

Who needs a 3458A? 

Yesss, give it to me, please ... 
If it is ok, i will give you my wonderfull k2k, ok? 

[IconicPCB]

Would this approach work with Keithley 196?

[tszaboo]

Makes me want to buy a 2015.

[hwj-d]

Makes me want to buy a 2015.

But, you know, this is not a dynamic display.
Also the 34461a has it in the statistics, as average, as you can see.

[tszaboo]

But the 2015 usually goes for 350 dollar on ebay, obviously used. The 34461a is 3x the price.

[hwj-d]

But the 2015 usually goes for 350 dollar on ebay, obviously used. The 34461a is 3x the price.

That's right. I love them both.
If you need and can get one, in spec or bring it back to, with fairly good display, take it.

[HighVoltage]

Pretty sure that this works for virtually all 6.5 digit keithleys. Post here if not.

Not working for the 2420-C

[bitseeker]

I assume it'll be the same for the 2700, since it's a cousin to the 2000 and 2015. I'll have to give it a try. It may not be a constantly updating value, but handy when you're not capturing data remotely.

[Andreas]

Hello,

branadic told me a trick that works on his PREMA.
Just use the math functions to suppress the leading zero and getting a further digit.

This works also on my K2000 if using the MxB (math) function key.
On the PREMA it works also with a factor 1.0
On my K2000 I have to enter either a factor 1000 or 0.001 to get the further digit.
(and a Display in KMxB or mMxB instead of Volt)
It makes sense to use a sliding average with that to futher reduce noise in slow mode.

With best regards

Andreas

[dr.diesel]

Works on a K2010 as well!

[egonotto]

Hello,

does not work with Prema 6030 .

Best regards
egonotto

[bastl_r]

A few years ago I did the same thing with my Kontron4066(Solartron 7066) to display amps with homemade shunts made from manganinribbon.

[tggzzz]

I can get an extra 5(?) digits on my Solartreon 7081 simply by outputting the values on the HPIB or RS232 interfaces.

Of course the extra digits are little better than random numbers

Numerical basics... If you have two numbers X and Y each with a +- uncertainty of 1%, what is the uncertainty in (X+Y) and (X-Y)? Hint: not 1%

Another hint: consider X=10 and Y=10.

[TERRA Operative]

I have occasionally been using the averaging function to get the extra digit on my Keysight 34461A, I always wondered if it was legit, but at least it gives me an idea of how close the (now second to) last digit was to flicking up or down.

[deepfryed]

I do occasionally average on the 34401A and get an extra digit on the 10v range, but mostly use the serial / gpib for any continuous measurements.

[tggzzz]

I have occasionally been using the averaging function to get the extra digit on my Keysight 34461A, I always wondered if it was legit, but at least it gives me an idea of how close the (now second to) last digit was to flicking up or down.

It can be valid, with some reservations.

Averaging can "remove" or "smooth out" random uncorrelated noise; N samples gives sqrt(N) reduction in noise.

If the noise isn't uncorrelated, then effects equivalent to mixing signals can occur. That "raises the noise floor".

[iMo]

The noise is one thing and "one more digit" the another one, imho.
The issue with the 399 ref based meters is - there is the "white" noise (like 3-5uVpp at the 10V range), random walk (like 5-10uVpp) and popcorn (random shifts 4-5uV). Not counting the ADC and math "noise". All together around 1-2ppm at 10V - the designers knew well why to set the resolution 10uV only (at 10V), ie. in case of the 34401A.

PS: and add up the TC of your meter as well, which is typically +/-5uV/C internal DMM temperature (at 10V range) unless your DMM compensates for that.. (the 34401A does not)..

[tszaboo]

But the 2015 usually goes for 350 dollar on ebay, obviously used. The 34461a is 3x the price.

I left this message 7 years ago. The meters are 7 year older now, and the price is now 600 dollar.

[trobbins]

PS: and add up the TC of your meter as well, which is typically +/-5uV/C internal DMM temperature (at 10V range) unless your DMM compensates for that.. (the 34401A does not)..

I recently repaired a 34401A with a new clone power transformer (a fried bridge rectifier had slowly baked the PT without the mains fuse blowing).  Of interest was that for 240Vac mains, there was a 13-14V margin between +/- 18V regulated rails (sitting at 18.8V) and the raw rectified/filtered voltages, forcing the 337 and 317 regs to dissipate a reasonable level of power internally, that is by its nature dependent on mains voltage.  To suppress that dissipation I am using a bucking transformer to provide 210Vac, as the meter happily works at 180Vac so there is margin for AC mains variation.  The meter is long out of cal, and changing the mains voltage by -12% may add more uncertainty due to a different internal temperature rise, but I'll live with that for now.

Simple testing of internal enclosure temp rise after turn-on showed it took nearly 2hrs to stabilise at nearly a +10C rise with the 210V mains change.  The general equipment accuracy spec is for +/-10% mains, and although I didn't measure internal temp rise at nominal 240V (or even at 264V for +10%), I'd be guessing HP must have included quite a few degC variation in the absolute accuracy specs.   The equipment specifications do make note of a 2hr warm-up for transfer standard use and some specs, but only a 1hr warm-up for some other specs.  Although this is not directly aligned with the topic of 'one more digit', it does increase the gulf between accuracy spec and resolution.

[iMo]

My experience with my 34401A is the difference between the ambient and the internal dmm temperatures is aprox 18C (240V). The meter requires much more than 2hours to "stabilize till the last 7th digit"  , 6-8hours is a minimum.. When powered off for a longer period even more..
PS: I glued the LM35 temp sensor on the aluminum shield over the ADC, powered by the outguard 5V and wired at a free pin on the rs232 connector (fyi - there are 2 free pins there, the second one I've been using as a 5V power output for the external MCU)..

[bdunham7]

I'd be guessing HP must have included quite a few degC variation in the absolute accuracy specs.   

Yes, there's going to be significant tempco going from 18C to 28C ambient, as well as going from 1 hour power-on to 3+ hours.  But based on mine and a few others I've seen, they'll be well within their published specs well before a full hour and over the entire 18-28C range and more.  If you're going for that "extra digit", I'd just set it up connected to a computer with the display turned off and leave it on more or less forever.