Oh No! I got infection.

[Vtile]

It seems that I finally got the volt-nuttery infection. I figured out today that I measure just a bit of the temperature response of those 1N823 buried zeners I have had in my component box some time now. Good excuse to play with my new toy Keithley 197A "microvolt DMM" (true mystery unit, from local pawnshop) soon I noticed that I need to see finer detail in both current and voltage. *Meh* now many many hours later. I know quite a bit on the TC response of those zeners and invented (like everyone else) the breathing temperature sensor (the reference zener in ambient temperature without current trimming). I ended up doing differential voltage measurements with my trusty (fleabay find from some years ago) Meratester seen on photo. In the evening it were on 1.5mV range, resulting about 3,75uV resolution for drift. Nice apparatus in this kind of work (FET-chopper 100 Meg input), it were funny to follow the needle to go back and forth and see those steps/pops while temperature did change in one direction.    ... Next step, programmable oven and some automated test procedure me thinks.  The calculator had a small program to beep in certain interwals with certain tone to switch the lightbulb on and off for consistency.

Later I changed the current control of the "reference" zener from decade box (trashbin find) to double precision potentiometer controlled (the DUT were already in this way [trashbin find, technical school equipment]) and then fixed the airflow problem with a piece of tissue paper on top of the "reference" zener. 

The zero T.C is close to 4,7250 mA while 4.4444 mA had better for close to room temperature operation point. Funny to see how "nonlinear" it is.

Is there any cure for this illness?

[Conrad Hoffman]

No, the illness is tellurium copper terminal.

[Vtile]

No, the illness is tellurium copper terminal.

Sounds serious. I need to build 10X pre-amp to that Meratronic.. Got  today a few OP07s (were on the shelf on local shed) to work with.

[Vtile]

I have been reading Conrads excellent mini-metrology articles. Figuring out if I do Kelvin-Varley divider or something else, the KVD is so tempting (and would be useful), but the damn rotary switches .... there is none in the "cheap end" of product that would suffice as far as I have been searching in a few catalogs. Conrads method to use the jumper wire kind of system is clever, but I have this fixation to rotary swithces. I could use the same divider topology as Krohn-Hite DC standard, but again low end decades will be a serious problem with cheaper than should swithes.

Now what I buffer each decate with sufficient type OPA and add offset trim as in "calibrated pot" scheme from days now gone. Then I could use 10k resistor in each decade as the buffer prevents them to see each other electrically. I wouldn't need to go smaller and smaller values in each decade and became a victim of the contact resistance fluctuations. This solution naturally restricts the divider to only low voltages inside the operation range of the buffers, but the design only needs cheap 1:12 swithces. The offset error adds up though. Maybe I need to make some calculations. 

Any jellybean OPAs one could consider, I'm not after sub-ppm stability with this (even on my dreams), more of a working reference. FET or OP07 style bipolar or chopper ... low drift per deg.C is important and long term stability even more I suppose.

PS. If I do 6 decade divider I could separate only 4 first ones and leave the last two passive give or take one decade. 10k -> 1k then one mOhm fluctuation in contact resistance (a dream.. I suppose it is counted in tens minimum) would be only 0.1ppm non-linearity on the last decade.

Bubble did burst, see below. 

[MosherIV]

the KVD is so tempting (and would be useful), but the damn rotary switches .... there is none in the "cheap end" of product that would suffice as far as I have been searching in a few catalogs.

Have you looked in ebay?
I found a number of vendors selling dual gang rotary switch - quality - do not know.
eg
http://www.ebay.co.uk/itm/Band-Channael-Rotary-Switch-2P11T-2-Pole-11-Position-Dual-Deck-E8I-/182406735547?epid=1487137967&hash=item2a7849e6bb:g:t6kAAOSw5cNYaRy-

Now what I buffer each decate with sufficient type OPA and add offset trim as in "calibrated pot" scheme from days now gone. Then I could use 10k resistor in each decade as the buffer prevents them to see each other electrically. I wouldn't need to go smaller and smaller values in each decade and became a victim of the contact resistance fluctuations. This solution naturally restricts the divider to only low voltages inside the operation range of the buffers

Once you deviate away from the ratio of one decade to the next - you no longer have a KVD (you do realize that dont you?).

The genious of the KVD is that each decade seeing a perfect decade divide down (do the maths on resistor parallel and series combination and you will see what I mean). It also means that ALL the resistors MUST be exact match value wise and temp co AND each decade ration MUST be exact.

Good luck with your recovery 

[Vtile]

the KVD is so tempting (and would be useful), but the damn rotary switches .... there is none in the "cheap end" of product that would suffice as far as I have been searching in a few catalogs.

Have you looked in ebay?
I found a number of vendors selling dual gang rotary switch - quality - do not know.
eg
http://www.ebay.co.uk/itm/Band-Channael-Rotary-Switch-2P11T-2-Pole-11-Position-Dual-Deck-E8I-/182406735547?epid=1487137967&hash=item2a7849e6bb:g:t6kAAOSw5cNYaRy-

Now what I buffer each decate with sufficient type OPA and add offset trim as in "calibrated pot" scheme from days now gone. Then I could use 10k resistor in each decade as the buffer prevents them to see each other electrically. I wouldn't need to go smaller and smaller values in each decade and became a victim of the contact resistance fluctuations. This solution naturally restricts the divider to only low voltages inside the operation range of the buffers

Once you deviate away from the ratio of one decade to the next - you no longer have a KVD (you do realize that dont you?).

The genious of the KVD is that each decade seeing a perfect decade divide down (do the maths on resistor parallel and series combination and you will see what I mean). It also means that ALL the resistors MUST be exact match value wise and temp co AND each decade ration MUST be exact.

Good luck with your recovery 

Ah, dang you are correct. I should have put something on the paper first, would have saved a fruitless thinking time.  My idea just results of series multiplicators with 1 to 0.1 ratio selection, so no use for this use. Edit. PS. Unless one creates more complex and more unstable and harder to calibrate adjust circuit than KVD.

[tggzzz]

It seems that I finally got the volt-nuttery infection.
...
Is there any cure for this illness?

Yes and no, in the sense that methadone is a cure for heroin addiction.

If you just want to see a 1 followed by as many zeros as possible, timenuttery might be cheaper.

There are Yahoo support groups for voltnuts and timenuts, and of course there's EEVBlog's very own support network.

[Vtile]

It is serious ilness, I must resist and only make a 10V reference (PS. 3x actually) and a buffered pot (PS. array actually) for toying around. I think that is about the way it starts.

The resistor matching is pretty relaxing job, I will make the KVD in form or in another sooner or later.. Hmm... Maybe I should cook my own from pure copper or silver buttons and some TLC.   ..NO! Those eBay switches are too mystery, should have atleast model number and datasheet available.

[Vtile]

Progress.. My prototype reference seems to be stable enough to allow me to sort rest of the 1N813 diodes and examine the resistors etc. for the final references. This obviously is grude build for just getting somewhat temperature stable and short term drift free reference for testing purposes to tighten the tolerances. The blue component is CCCP vintage from VEF (State Electrotechnical Factory) at Riga, Latvia.. random ceramic capacitor, it seems to work as supply filter. The circuit is the same as here: https://www.eevblog.com/forum/metrology/old-fashioned-zener-10v-reference/ except there is no RC filter as I did not have any good enough capacitors to put to that part of the circuit.



For a curiosity I hooked up my Fluke/Philips 3390A Combiscope to the reference above. My atm. best meter Keithley 197A were also hooked to the 10.00... volt reference point. Pretty noisy, I figured in digital mode. Turned to analog and started to see ringing.



Ah. That analog screen is somehow magical especially on photos and usefull in this case as digital mode refused to tricker in continuous mode.


In digital mode it only shows up barely on stop mode or photo.



The source ... Keithley 197A or a combination of all four systems (PSU, DMM, Scope and reference circuit) as when I removed the Keithley, then scope did show almost clean line with only random noise.

I think I 'need' another Oscope with 500uV/Div resolution? ..and fix the reticle/grid illumination.

[schmitt trigger]

By all means, fix the reticle illumination IMMEDIATELY!

Otherwise, your mind will deteriorate and you will have an urgent desire to purchase a brand new oscilloscope.
You will have nightmares, you will have hallucinations.
Until one day, innocently browsing the internet you'll find an oscilloscope you cannot resist! The road to perdition leads from there.

But to tell you the truth, it is a sweet, sweet road. 

[MosherIV]

That noise looks suspciously like conducted smps noise through the PSU.

Filter the PSU into your reference circuit.

I have similar problems with my General Resistance DAS56A
https://www.eevblog.com/forum/metrology/general-resistance-das56a/msg1062492/#msg1062492

[Vtile]

That might be some SMPS leaking indeed,  I need to investigate it further, as the noise from the Keithley should be on 3.5uV range (1 meg load). It is such an old design that the filtering of its supply propably doesn't meet todays needs with dirty main. The PSU that runs the actual reference is Thandar (TTi) linear and results "clean" output of the reference in my measuring accuracy (and without much time used for investigation). Anyway I'll need to solve the source for future.  ...Reading your link now.

PS. Spend some time scoping more.. When Keithley is not connected the ringing goes almost unnoticed, I only did find it when I know what to look for. That humble (oldish) DSO side if the scope doesn't see it at all (user error in triggering propably). It definedly is from the mains 26.7 kHz, I also start to suspect I have poor ground connection on that floor. I need to do more measuring and elimination in the house.

By all means, fix the reticle illumination IMMEDIATELY!

Otherwise, your mind will deteriorate and you will have an urgent desire to purchase a brand new oscilloscope.
You will have nightmares, you will have hallucinations.
Until one day, innocently browsing the internet you'll find an oscilloscope you cannot resist! The road to perdition leads from there.

But to tell you the truth, it is a sweet, sweet road. 

I'm not sure of the color, blue maybe?   

[schmitt trigger]

It is a Yellow Brick Road......

Speaking about your noise problem.....something that has worked for me many times to remove equipment PSU noise: is to "float" the offending piece of equipment, to prevent its switching noise from being conducted thru the AC plug's safety ground back to measuring instrument safety ground.

[Vtile]

Yellow submarine. 

I did do more measurements.... I did fire up my HP 6632A System DC supply for a first time I don't know how many years. It sounds like a fighter jett and have pita UI for quick electronics twidling. Anyway it should be good quality equipment.


Dang... Still there.  Hmm, I have Tandar still turned on, but not connected to anything else than mains socket.
https://www.eevblog.com/forum/metrology/oh-no!-i-got-infection/?action=dlattach;attach=333158

Uh.. It might be not be ... How the heck.. HP is turned Off, but signal is from HPs output terminals.
https://www.eevblog.com/forum/metrology/oh-no!-i-got-infection/?action=dlattach;attach=333160
Yep it definedly is the Tandar supply.  Tandar is out of socket.
https://www.eevblog.com/forum/metrology/oh-no!-i-got-infection/?action=dlattach;attach=333162

Some more repair work ahead, before my voyage to the ppm DC world can continue. First check - bridges and caps... But that is another fairy tale to different section. HP doesn't look partially clean either, but It didn't have any load and I can not remember the specs.

PS. Opened the Thandar and what I did found, the logic 5V / 5A DC part (as this is triple supply ~11 kg ) is definedly a switchmode supply after the (I suppose ) isolation transformer there is so many tuning and power coils that it can't be any other type than that, side cutters are my friend tomorrow then it is only a matter of a new powerswitch to turn off the noisy part, when not needed (=99.95% of times). Problem solved for now .. I hope.

[Vtile]

For a long time I did push this project forward today. Just finding zero TC current point for another 1n823 zener. Now interesting part were that I noticed that there is effect in which direction you heat the zener, as there is a small bulge dU~10uV in the TC it seems, but I'm not sure if it is actually the zener or seebeck effect from the test leads. 

[David Hess]

the KVD is so tempting (and would be useful), but the damn rotary switches .... there is none in the "cheap end" of product that would suffice as far as I have been searching in a few catalogs.

What about using relays?  Then the switches can be cheap or implemented digitally.

[texaspyro]

No, the illness is tellurium copper terminal.

Google "tellurium breath"... fun stuff.

[Vtile]

the KVD is so tempting (and would be useful), but the damn rotary switches .... there is none in the "cheap end" of product that would suffice as far as I have been searching in a few catalogs.

What about using relays?  Then the switches can be cheap or implemented digitally.

IIRC the relay approach weren't actually "that" much cheaper as the amount needed is pretty huge.    Not a big deal, I bite the bullet and get the parts I "need" when that time comes. For now I think I survive with three Bourns 10-turn potentiometers and a 1/1k dial. When I got the references build I start to look the KVD with new eyes.

[Vtile]

No, the illness is tellurium copper terminal.

Google "tellurium breath"... fun stuff.

I don't know if it is fun, but at least it is interesting. 

PS. That prototype reference in a breadboard did survive ~5 months surprisingly well, I would have thought it would be out of range next week or so, but no the drift were in a few hundred thousands ppms compared to my Keithley 197A after 5 months, but it didn't survive (needed some part re-seating and adjustment) without some damages when I moved my "lab" from room to another.

[Vtile]

Oh, satisfying. I finally did get my new prototype ready, as it had been on the drawing board so long.

All the resistors are standard 0.1% & 25ppm type. OP07CP and handpicked Motorola 1N823. The basic circuit is the same servo type as before.



Even more satisfying is that the dV is in last digit of my Keithley 197A at dT=10 (19..31 deg.C), which makes with quick calculus 1ppm per deg.C     (A wishful thinking I know) The absolutes and long term characteristics aren't probably anything, but average.


... I need 6½ digit meter to verify it better.

[Vgkid]

That circuit is satisfying to look at.

[Vtile]

Indeed. 

[Zucca]

Vtile that circuit is more an architectonic masterpiece.

[Vtile]

Vtile that circuit is more an architectonic masterpiece.

Thank you.

It is the byproduct of the enormous amount of resistor I had to use to get the obscure ZeroTC current value for the buried temperature compensated zener and then the gain set for the OP.

..But manhattan style ref. board looks nice. 

[Conrad Hoffman]

That's nice. I usually grind out isolated pads with a burr, but the small pieces of circuit board are way better.