TEKTRONIX DMM4050 + Quad LT1021 stability...

[blackdog]

Hi,

There is much talk on this forum about the LTZ1000 (a) and several multimeters like the 3458a.
I want to show you what is possible, with a TEK DMM4050 end one of my voltage references.

This is a link to the topic of the voltage reference that is used for this test.
Page two and schematic version 1.3 is the last one.
https://www.eevblog.com/forum/projects/10v-reference-(i-did-it-my-way)-no1/

In my LAB i have 2x TEK DMM4050 and KeySight 2x 34461A and a 3458a, but no good script for the last one.
Both TEK DMM4050 multimeters are very stable and also a lot better than the KeySight 34461a multimeters.
The 34461a let u see 0,1 PPM or better resolution, but the temperature coefficient of these multimeters is not as good as that of the TEK DMM4050.
The TEK DMM4050 is "always on" 24/7/365.
I used the zero function one hour after being connected to the 10V reference.
The measurement is done at 100PLC.

OK, the drift picture.
The oven in my referent has about 60x reduction of the outside temperature, that is more than enough for the LT1021, it has a very flat temperature coefficient.



Even just now its still on 0.00000 afther almost 3 days.
Please ignore the error message, I had forgotten to remove it before I started to measure.



And fore everyone who like to play with excel => www.bramcam.nl/24H-11-Feb-2015-DMM4050-QuadOven-LT1021.csv

The two 34461a multimeters i own, are about 3 times as sensitive to temperature changes than the TEK DMM4050...
I'am happy with both instuments,  these different instrument's have good and bad points, nothing new about that...

I hope you find this interesting because it often is just about the interface of multimeters and almost never of the stability of the multimeters!

Kind regarts,
Blackdog

[TiN]

Looking forward for your tests. Let us know what help you need with 3458, I already uploaded some python apps to run datalogs with it's command set. Those are fairly easy to modify. I have scripts in Python for RaspberryPi + linux-gpib, and some LabView stuff for Windows.

[Andreas]

OK, the drift picture.
The oven in my referent has about 60x reduction of the outside temperature, that is more than enough for the LT1021, it has a very flat temperature coefficient.

Hello Blackdog,

did you measure the individual T.C. of your LT1021?
My ADC3 with a LT1021 (5V) has around +7.5 ppm/K.
(which is quite a lot).

A second candidate of LT1021 measured +6.5 ppm/K

And I am not surprised that the stability over time is good for the metal can housing.

With best regards

Andreas

[blackdog]

Hi Andreas,

I selected after about one year the best 4 LT1021 metal cans, and i have another 14 pieces on for a year 
I dit not make precies measurements on the temperature dependence, just lookt at the mesurements data i took for one year and the LAB temperature from that measurement.
Most of the metal cans are realy good and even better in the oven! About 42,7C

Kind regarts,
Blackdog

[jesuscf]

I am doing something similar with a couple of  heated LM4040AIZ5.0 references.  I was concerned with my stability results, but looking at blackdog's plots has given me a huge confidence boost.   I wish I had a Tek DMM4050 at home (a Fluke 8846a would do as well!).  So far, I have a BM869s and an old Fluke 45 available.  I shouldn't complaint too much: I am very happy with the BM869s.  The problem with the BM869s is that runs on a 9V battery which will not last with the multimeter on for several weeks.  One possible solution will be to power the BM869s with an external power supply.  But then, I don't think the BM869s is designed to run 24/7.

[acbern]

What makes the LTZ1000 so stable is the temperature regulation, that fact that its a burried zener and the hermetic case. Any other burried zener, if properly temperature controlled and hermetic, can be a very precise and stable solution over time and ambient temperatures.
Just think about the Fluke 732a or 5440A calibrator to mention just a few. 

[Andreas]

I was concerned with my stability results, but looking at blackdog's plots has given me a huge confidence boost.

Hello,

the LM4040 is a bandgap reference in plastic package.

the LT1021 of Blackdog a buried zener in hermetically package.

So from the plastic package alone you will have around 10-15 ppm humidity change over the year.
The ageing rate of a bandgap is also a factor higher than that of a buried zener.
And also the noise of a low power bandgap is much higher.

The question is if you can measure this with your instrument resolution.
And I think that the Brymen has also a low power bandgap as reference.

So for deciding wether your instruments or your references drift it might be
a advantage of having some (pre-aged) LM399 references.

With best regards

Andreas

[jesuscf]

Thanks Andreas.  Here is a bit more information to see what the experts think/advice.

I built the circuit about six weeks ago at the beginning of January 2016, just for fun and also to quickly check how much different instruments I had access to at work agree to each other.  That includes two FLuke 45, one Tek DMM4020, one HP 34401a, one Rigol  3058E (brand new with recent calibration certificate), and the BM869s I just bought for personal use at home.  The LM4040 was/is heated with a very simple controller made of a difference amplifier and a LM335 temperature sensor to around 50C.  The voltage measurement of the reference with all the instruments but one were more or less in agreement within a few micro volts from each other (in retrospective, I should have taken a picture of the setup!).  One Fluke 45 (the one I am using now and the only one I can borrow for a longer time) was/is about 2 mV off compared to the other instruments.  With that done, I thought it would be interesting to see how the temperature stabilized reference would change over time.

I didn't have much hope of high stability, since as Andreas mentioned, the LM4040 is a band gap reference in a TO92 plastic package.   This is what I have observed so far: for the last four to five weeks the reference voltage has been flipping between 4.9974V and 4.9975V as measured by the Fluke 45, or between 4.99979V and 4.99984V as measured by the BM869s.  That is why I wish I have access to a bench top 6 or 6 1/2 digit multimeter like the DMM4050, so I could take automatic measurements periodically and display nice looking results as blackdog has done.

Hello,

the LM4040 is a bandgap reference in plastic package.

the LT1021 of Blackdog a buried zener in hermetically package.

So from the plastic package alone you will have around 10-15 ppm humidity change over the year.
The ageing rate of a bandgap is also a factor higher than that of a buried zener.
And also the noise of a low power bandgap is much higher.

The question is if you can measure this with your instrument resolution.
And I think that the Brymen has also a low power bandgap as reference.

So for deciding wether your instruments or your references drift it might be
a advantage of having some (pre-aged) LM399 references.

With best regards

Andreas

[quarks]

In my collection my Tek DMM4050 (or Fluke 8046A) is one of my favorites. It is very stable and accurate. Even compared to my much more expensive meters it is a very good performer and for the price it is in my opinion still one of the very best DMMs you can buy today.

[blackdog]

Hi DiligentMinds.com,

I hope you find this picture informational 



All meters have the same settings.

Kind regarts,
Bram

[quarks]

My Tek DMM4050 and Fluke 8846A also show around 1.2 mV

[Gertjan]

This is what I get from my one week old 8846A:


http://www.miedema.dyndns.org/fmpics/EEVblog/IMG_4401%20%208842A%20input%20bias-2000pix.jpg

So I guess this is pretty much standard.......


Hi Blackdog, 
I noticed in your graph that there seems to be a step after the thermionic event.
After the cool down and heat up of the night the average voltage seems to be a bit higher.
can you comment on that effect?

And after three days your meter is  still on zero (delta)
Is the effect occurring every day, but after three days it is still just under the threshold of the last digit? (3x 0,2=0,6ppm...)

Regards, Gertjan.

[Micke]

Well, quite old thread, discovered it today, so I did the input bias test on my FLUKE 8846A today...   
After warm up, and ambient temperature of 21C, the voltage is approx 1.13mV (=113pA bias current)
I am very pleased with this meter, very stable. At work we have a FLUKE 5520A calibrator, so I have during a year done many tests to see drift etc.
The 8846A is really good at DC [24ppm base accuracy]  (mainly because of the sealed thin film resistors I guess). But at AC, my Agilent 34410A (not 34401A) really outperforms the 8846A, since this multimeter does not use conventional RMS converter chip, but uses direct sampling of the AC signal instead. The 34410A AC specifications are very conservative!