Circuit design of branded electrometers

[Kirill V.]

How to keep the current stable when the load changes.
And how do you manage to keep the current constant? After all, sooner or later, the linear voltage will not be able to grow further?

The current does not depend on the load because the capacitor resistance for DC is very large, this resistance becomes the internal resistance of the source. This source is very close to ideal
This source works periodically
I have book DEVICES FOR MEASUREMENT OF LOW VOLTAGES AND CURRENTS electronically

[strannik2039]

Calibrators are ЕК1-6 and НК4-1 old. Especially ЕК1-6 (1979). From them the DC source is formed in two ways. Up to 0 .1- 1nA is a source of constant voltage U and a set of resistances R. Further according to formula I = U/R. Low limits are provided by linear voltage generator and differentiators. Current is equal in this chain to I=C ∙∆U / ∆t it simply from the theory. The device generates different output voltage and time intervals, at the end of which output voltage is disconnected and differentiators are discharged. If during this time the device could not be removed, everything is repeated with intervals for discharge. It 's in general about the principle of work.

[strannik2039]

Hello. As promised, the book of Alexandrov, Spynishnikov "Electronic DC galvanometers." In addition, part of the description is ЕК1-6 during current reproduction.
https://drive.google.com/drive/folders/1vome9_QqLaweis-r045eWefezxteb9ju?usp=sharing

[Kirill V.]

I also want to show a sample of engineering art - this is the input unit of the V7-30 electrometer
https://radikal.ru/lfp/s15.radikal.ru/i189/1104/d6/a0c2f7456d8e.jpg/htm
The input stage is made on two MOSFETs. Soviet industry did not produce pairs of such transistors and single transistors were selected in pairs manually. These are gold-plated cases in the photo, and you can also see two bipolar transistors in black cases that protect the input from overvoltage.
More photo of this electrometer
http://kxk.ru/dustyattic/v20_700390_16.php

[strannik2039]

Hello. Old lamp light. I like these indicators. According to the MNIPI album, this voltmeter was released in 1978. But the design of MNIPI (Minsk Research Instrumentation Institute) cannot be repeated because of their input switches. This product was done by the institute itself. If you look towards KEITHLEY 617 for example, we see the herkon switches that can be bought now. And so, on the principle of work, the instruments are little different. Voltage drop across resistor due to low current flow and feedback.
As for pairs of transistors, there were also such. In В7Э-42 for example stood hybrid assembly НТ002.
On Google disk I showed the diagram of the input device of the electrometer В7Э-42

[strannik2039]

Hello. As a continuation of the theme of В7-45, I created a folder in which I showed a pair of photos of the remote 1Е-15А dynamic capacitor unit and the measurement limit, and patents for dynamic capacitor and input device (Шуклин and ВХОДНОЕ УСТРОЙСТВО). https://drive.google.com/open?id=1-YYb--vVrcen3OqlZ5mA3dBsl4o4uw_b
Of course, it is not possible to repeat this at home. But here at the limit of measurement I would like to stop. Today, manufacturers are seeking to switch to so-called "long scales." It 's beautiful, I don 't argue. But if you look deeper, how true is that? Let 's present a web device with 0.000000Е-12А readings. Looks beautifully. But, the difference between 1E-12A and 1E-18A is 1000000. And the measurement error at a given limit 1Е-12А say 0.1% is 0.001 of the limit. So we 're better than 0.001Е-12А. But everything else seems to be in doubt.
Now let 's look at the 1Е-15А range and the error of 0.1%. Now we have 0.001Е-15А. Here 's a truer statement. Of course it 's just math. But all this is to the fact that when developing the instrument, it is necessary to decide on the circuitry of the electrometric amplifier.

[strannik2039]

I mentioned the University of Tartu and the surname Anso for a reason above. In principle, there are many articles in the publications of the university. But there is work in which he summarized everything and defended his thesis.
The work is called:
 "DEVELOPMENT OF BASIC DESIGN OF SMALL CURRENT METERS WITH HIGH SPEED RESPONSE ON THE BASIS OF HIGH-OHM MEASURING RESISTORS"      (РАЗРАБОТКА  ОСНОВ КОНСТРУИРОВАНИЯ ИЗМЕРИТЕЛЕЙ МАЛЫХ ТОКОВ С ПОВЫШЕННЫМ БЫСТРОДЕЙСТВИЕМ НА БАЗЕ ВЫСОКООМНЫХ ИЗМЕРИТЕЛЬНЫХ РЕЗИСТОРОВ)
https://drive.google.com/open?id=1_lvb3UCbdrooXv3b7DbDMHjtKGJyHnHO

To be honest, there were doubts whether this work should be posted on the forum. But I hope she 's useful to anyone.

[strannik2039]

Hello. Something came silence on this branch of the forum. Did I so scare the forum with my photos of old Soviet equipment? Yes, I repeat such a house not to do, but it is not necessary to do it. You can look at the Pomona website and find the connectors, and the cable. https://www.pomonaelectronics.com/products/connectors/rf

[Kirill V.]

The silence came because forum participants know nothing but Keithley instruments and a single source of information - low level measurements handbook.
And personally I am engaged in reading Soviet books and designing my future electrometric buffer for a DMM based on the AD549
I suggest that we continue to frighten local residents with Soviet devices. They already know our bears, but electrometers are unlikely
I would very much like to discuss why manufacturers use printed circuit boards to conduct input circuits. What own current can be generated by the Board material and what is the influence of environmental factors. But I know it's impossible, because those who know won't say anything

[MegaVolt]

And personally I am engaged in reading Soviet books and designing my future electrometric buffer for a DMM based on the AD549
I suggest that we continue to frighten local residents with Soviet devices. They already know our bears, but electrometers are unlikely

I think many will be grateful for the new well-structured information on how to make electrometers correctly.

[Kirill V.]

Yes, this is probably the only place where such devices are discussed and where there are unique photos.
After the collapse of the USSR, a huge number of unique devices were destroyed for scrap. Nevertheless, some concerned people have made their contribution to preserving knowledge about this equipment - detailed photos and documentation. And we are talking about very ancient equipment
I don't see anyone here offering to look at ancient devices or talk about their circuitry. Where are the photos of Cary Instruments electrometers? Old Takeda Ricken? English, GDR, FRG devices? Yokogawa? Or at least links?
Why didn't anyone mention vacuum tubes in floating grid mode? These old tubes had a bias current less than that of modern op-amps. Attoampere range currents were measured back in the 60s, but no one mentioned these techniques
Where are the books on Electrometry in English? Where are the links to the IEEE articles?
Only guys from Belarus offered information and unique photos

[strannik2039]

Hello forumites. Thank you Kirill V. for the kind words.
I apologize for absence. The tooth decided to show its character. He had to explain who was in charge. By the way, dentistry is one of the directions of electrometrics. But that 's the way it is.
Now let 's get back to the boards. To be honest, there 's no great choice. If we take the same KEITHLEY 617 for example, we see Teflon racks in important places.

 https://www.google.com/search?rlz=1C1ASUM_enBY845BY845&q=KEITHLEY 617&tbm=isch&source=univ&sa=X&ved=2ahUKEwja-evl6sjoAhVoiYsKHeFzCkUQsAR6BAgJEAE

If you look at 6517, https://xdevs.com/fix/k6517a_u2/ 's where we see the protectors on the board to prevent leaks.

And if you look at the board В7-49    https://drive.google.com/open?id=1Xy1Bgb5u_dvljgYzZz7XzpQNcl9yN7VY  , we will not see varnish in the area of the input device of the electrometer at all. This is done to reduce leakage currents across the varnish coating. And since the switch is of its own design, neither props nor Teflon racks are required.

[Kirill V.]

Wow, you never cease to amaze with your riches
Why does B2987A not have Teflon?
https://xdevs.com/doc/HP_Agilent_Keysight/B2987A/img/bem_eopen.jpg
And if it's really so cool, why haven't they done it before? That's what worries me

[razvan784]

Why does B2987A not have Teflon?

Look carefully at the board. It's not the usual FR-4 material. It looks more like something used for microwave PCBs. It might even be teflon-based
Also, leakage is drastically reduced by driven guard rings.
This being said, the B2987A is not really a metrology-grade instrument and is not marketed as such. It's marketed as a "Femto/Picoammeter", it is quite clearly specified with a 3fA uncertainty on its lowest range, and therefore you shouldn't expect it to be comparable to a metrology-grade device that measures attoamps The construction is good enough for 3 fA it seems.

[strannik2039]

Hello. This electrometer is positioned as a 10Е-17А current meter (0.01 fA minimum measurement resolution). It is difficult for me to judge such a statement without having an instrument.
But I can quote lines from the В7-45 description. If the input resistance of the electrometric unit is not less than 10Е+16 Ohm and the input capacity is not more than 30pF the equivalent input resistance of the voltmeter is more 10Е+19 Ohm, and the input capacity is not more than 0. 6pF taking into account the negative feedback. All these values are provided by sapphire insulators. And this electrometer measures the currents from the 10Е-17А.

I think that if the designers B2987A put teflon insulators and make props in the board in responsible places and applied electrometric Reed relay, such an electrometer would have more credibility.

[Kirill V.]

Keithley book says that the effective noise bandwidth of an analog panel meter is 1 Hz. Why?

[Grandchuck]

For higher frequencies the meter damping and inertia hold the pointer still.

[Kirill V.]

Qualitatively, yes, but quantitatively? Do all analog panel meters give 1 Hz bandwidth? Where did they get this number?

[Kleinstein]

Qualitatively, yes, but quantitatively? Do all analog panel meters give 1 Hz bandwidth? Where did they get this number?

The speed of the analog meter depends on the design and size. Larger meter movements are usually slower.
Depending on the movement one may have to add extra damping to avoid longer ringing. This can also slow down the movement.
1Hz bandwidth is kind of convenient for a reading the movement, so it needs some 5 seconds to be fully settled.

[Kirill V.]

Does this mean that the analog meter provides unsurpassed noise filtering? Only the low frequency part of the flicker noise can affect the reading?

[Kleinstein]

The analog meters have no especially good noise filtering. It just does not make much sense to make them much faster, as the eye is not good in reading it so fast and there can be mechanical overshoot. The eye / brain can do some extra noise filtering that can be pretty good and easier than with digital output. 

One can also implement filtering in digital meters and this can even be done better, though it may need some time to get used to it.
A digital filter can be FIR type and thus settles faster compared to the BW, especially if it is about settling to high accuracy (e.g. 0.1 % or better).

An electrometer tends to be slow / noisy anyway when it comes to the lowest ranges. For this reason digital instruments often also have some extra filtering. The final settling of DA related background may take quite long anyway - more like connect, have a coffee and than read the result. That new graphical LCD interface can make absolute sense for a electrometer - unless directly read to a computer.

[Kirill V.]

Keithley book says that the noise bandwidth of an amplifier is equal to pi/2 of its true bandwidth. Why?

[Kleinstein]

The ratio if noise bandwidth to -3dB bandwidth depends on the filter type. By definition they are equal for an ideal brick-wall filter. The ratios for the the simple filter types (e.g. 1 st. order, 2nd order) can be calculated with "simple" math / integration. Real world filters may not follow the basic simple types and can be sightly off the ideal number. Unless one wants to measure noise, one usually does not care about the exact noise BW, though it can still help to have a low noise BW for a given response time.

[David Hess]

Keithley book says that the noise bandwidth of an amplifier is equal to pi/2 of its true bandwidth. Why?

Noise leaks in above the filter cutoff frequency so the noise bandwidth of a single pole roll-off filter is 1.6 times the -3 dB bandwidth.

[Kirill V.]

So, the idea of the output stage. I purposely abandoned the Keithley circuitry because there will be no external feedback circuits and no need for large load currents. The load will be a multimeter with an input resistance of 10 MOhm.

Here there is a short-circuit protection of the load and if the output stage must sink current, the protection is obtained unintentionally, since the current generator on VT2 and VT4 can not give a greater current. If the output stage must be a current source, then VT3 is switched on and the short-circuit current is limited.
Can anyone criticize this?