Resistors Used in an LM399 Circuit

[bsw_m]

I apologize, I may not have been very clear in my previous post. The wire together with the insulation has a diameter of 13µm, but the wire itself has a diameter of 2µm. In any case, winding such a resistor is not an easy task.

[Edwin G. Pettis]

I've never heard of nor seen a wire that fine from any wire source I know of, that is impressive indeed.  That would have been exceedingly difficult to draw out a wire that tiny even in Evanohm alloy.  0.000078" vs 0.0004", essentially five times thinner.  That would be roughly 20K ohms per foot for Evanohm alloy.  That still would have been about 50,000 feet of wire.

[EC8010]

I apologize, I may not have been very clear in my previous post. The wire together with the insulation has a diameter of 13µm, but the wire itself has a diameter of 2µm. In any case, winding such a resistor is not an easy task.

That's very fine wire. Some of the final generation of valves used tungsten wire 6um diameter for their control grids, and I imagine tungsten is very difficult to draw into a wire.

[iMo]

Several weeks back I repaired by myself a 1MegOhm wirewound resistor, thus nothing special.
The wire was not visible with my naked eye, I did it because I was feeling it when under strain upon taction. I made a shot under uscope, it was around 10um or less. Unfortunately it broke and about 30cm of it was 8kOhm thus my 1Meg is now by 8k lower.
Here is the shot
https://www.eevblog.com/forum/projects/how-do-they-wind-coils-with-ridiculous-gages-like-56awg/msg4910013/#msg4910013

[EC8010]

I've measured tungsten wires of about those dimension using a good micrometer, but I would expect resistance wire to be proper fragile, necessitating microscope. I don't imagine the postman delivers replacement parts very often where you are?

[bsw_m]

That would be roughly 20K ohms per foot for Evanohm alloy.  That still would have been about 50,000 feet of wire.

The wire that was used to wind the resistors shown above has a resistance of about 860kOhm/meter or about 260kOhm/foot (this is with a wire diameter of 2µm).
Which gives a more realistic 1160meters or 3.846 feet to get a resistance of 1GOhm.

[Kleinstein]

Making thin  turngsten wire is not easy, but there was a large demand for such wire for producing light bulbs and tubes.
Making wires from the resistor alloys may be a bit easier as they are likely less brittle.

The glass encapsulated micro-wires are different from normal wires as the are drawn together with the hot glass, more like making a special kind of glass fiber.

[iMo]

I've been thinking to use this kind of wires (ie from old wire wound resistors) for creating a divider for 399/1000 circuits. I mean to wind it on a small bobbin/former as a coil with a tap. Not an easy exercise though.. For a total resistance of the divider say 20k you would need couple of meters of such a wire..
The problem (among others) is the higher the resistivity the shorter the wire but the harder to place the tap precisely. I made my ww resistor for myADR1001 (110ohm) trying to fit 10V, I did it with a wire where the 110ohm was around 18cm (great) but I was not able to cut it such it fit, finally I used another wire where I needed around 1m (well, still ok..) of the wire and it was much easier (and I am off by 130uV).

PS: FYI - the ADR1001 has got all those difficult resistor dividers on chip and ovenized, what has been used for fitting the 10V at the output is the RISET resistor (around 110ohm, it sets zener's current) which "10V output tuning sensitivity" around those 110ohm is roughly +/- 250uV per -/+ 1ohm change..

[bsw_m]

The glass encapsulated micro-wires are different from normal wires as the are drawn together with the hot glass, more like making a special kind of glass fiber.

Absolutely correct!

[alm]

I've been thinking to use this kind of wires (ie from old wire wound resistors) for creating a divider for 399/1000 circuits. I mean to wind it on a small bobbin/former as a coil with a tap. Not an easy exercise though.. For a total resistance of the divider say 20k you would need couple of meters of such a wire..
The problem (among others) is the higher the resistivity the shorter the wire but the harder to place the tap precisely. I made my ww resistor for myADR1001 (110ohm) trying to fit 10V, I did it with a wire where the 110ohm was around 18cm (great) but I was not able to cut it such it fit, finally I used another wire where I needed around 1m (well, still ok..) of the wire and it was much easier (and I am off by 130uV).

Depends on the type of wire. I believe common types for precision wirewound resistors are either Manganin or Evanohm and similar brands. Manganin has to be treated very carefully in terms of stress and heat otherwise you risk shifting the tempco, but you can solder it. The tempco is generally not very low (in the order of 10 ppm/K best case). Evanohm and similar alloys from other manufacturers is superior to Manganin in both heat tolerance and temperature coefficient, but can't be soldered with any degree of reliability and has to be welded. Read anything by Edwin Pettis since he is (one of) the industry expert on precision wirewound resistors.

He wrote a really interesting two part article together with another person that won't help you make your own resistors but gives a fascinating insight into the precision wirewound industry:
The last half-century: Wirewound resistors Part one
The last half-century: Wirewound resistors Part two

[iMo]

Interesting reading, thanks. I would prefer Manganin as it is solderable and easy to get from old ww resistors, and when wound as the "divider" on the same bobbin (a bobbin with 3 legs) the TC could cancel, imho.

You may wind the R1, then make the tap (almost anywhere around the expected R1 value), and then do wind the R2, where you will finetune the length of the R2 such it fits the "ratio R1/R2" (by cutting the wire slowly down, soldering, measuring, and so on until it fits).

The absolute value of R1+R2 does not matter here. I think it is doable..

[Kleinstein]

NiCr based resistors are expected to have better long term stability (comes with better tolerance to temperature) than copper based alloys (e.g maganin) too. Manganin is lower specific resistance and may have a lower thermal EMF. Together with the easier conacts this makes it attractive for low value resistors and shunts.

[iMo]

How to check the wire is Manganin or something else? Imagine you have a bunch of ww resistors from 1k - 1Meg.

[Kleinstein]

How to check the wire is Manganin or something else? Imagine you have a bunch of ww resistors from 1k - 1Meg.

To do this non destructive could be a bit challanging. One point may be the 2nd oder TC. AFAIK this is not changing much with the aging details on manganin (though it may change with the exact alloy, like different manganin variants). I would expect a high 2nd order effect for manganin than NiCr. 

Another chances could be from the magnetic properties: NiCr should be slightly magnetic (maybe just paramagnetic).  Still the connecting wires may make this test tricky if they are magnetic.

Another point could be to get the resistor types and find data sheets that tell the alloy. More open designs may show the way the wire is connected: Manganin is usually soldered, while NiCr is welded in some way. NiCr is a bit more modern, but there is also also older CuNi (konstantan), which is not good with a rather high thermal EMF.

For the lower resistance range there is manganin wire avaiable. I got some from Ebay that worked out quite well:  TC well below 10 ppm/K and very low thermal EMF (I can hardly tell it appart from pure copper or give the sign). Just the resistance is a bit higher than nominal (may be some 5% on the diameter).

[alm]

Interesting reading, thanks. I would prefer Manganin as it is solderable and easy to get from old ww resistors, and when wound as the "divider" on the same bobbin (a bobbin with 3 legs) the TC could cancel, imho.

As a first order approximation this is correct, but note that this only applies if the heating on both sides is the same. If you have a 10:1 voltage divider, then one leg will dissipate 9x more heat than the other, so it will reach a higher temperature.

[Edwin G. Pettis]

Hi, after doing some research, there are no alloys that are anywhere near the resistance per meter (or foot) that you have 'calculated'.

Tungsten is over 150 times lower in resistance per unit length than Evanohm, Tungsten, the smallest diameter wire I could find was 0,000,3" diameter, if it could be drawn to 2um diameter would be 17984 ohm per meter while Evanohm, at the same diameter, would be 65.616,8 ohm per meter.  That is using the most common form of Evanohm 800 alloy, there are also a 825 and 875 alloy variants of Evanohm which would increase the resistance a little bit further.  The TCR of Tungsten is +45 PPM/°C.

There are absolutely no pure metal or alloy that gets anywhere near 860K/m in a wire form, even at 2um diameter, even 1um wouldn't get you there even if a wire could be drawn that tiny (Gold is the only candidate here and its resistance is far too low).

While I am not directly disputing your measurement of the wire's diameter (how accurate?), the resistance per meter and calculated  length are not possible.

[bsw_m]

I disassembled a simpler resistor of 1Mohm nominal. It is wound with a wire with a diameter (without glass insulation) of ~50μm my mistake, wire diameter 5um. Photo of the wire under a microscope (optical magnification 200x), resistance of the resistive element and its dimensions (in metric system) below in the photo. If necessary, I can try to take a meter of wire and measure its resistance.

P.S. Please note, this wire is not drawn out. It is casted in glass insulated.

[iMo]

..There are absolutely no pure metal or alloy that gets anywhere near 860K/m in a wire form, even at 2um diameter, even 1um wouldn't get you there even if a wire could be drawn that tiny (Gold is the only candidate here and its resistance is far too low)..

The resistors are old Soviet era resistors. Soviets were known for creating special exotic alloys unknown for outside world, and because the number of western spies in USSR R&D institutes was (and still is) by several orders of magnitude lower than vice versa, it could be you just face such a mystery alloy.. 

[iMo]

Interesting reading, thanks. I would prefer Manganin as it is solderable and easy to get from old ww resistors, and when wound as the "divider" on the same bobbin (a bobbin with 3 legs) the TC could cancel, imho.

As a first order approximation this is correct, but note that this only applies if the heating on both sides is the same. If you have a 10:1 voltage divider, then one leg will dissipate 9x more heat than the other, so it will reach a higher temperature.

I talk here aprox 1:2 ratio (like in a typical LM399 circuit with 7V->10V opamp amplifier).

[Kleinstein]

Looking at the wire inside the glass can overestimate the wire diameter, as the glass acts as a kind of magnifier.

[bsw_m]

No problem, the wire diameter was measured after removing the glass. Wire diameter 50um, my mistake, wire diameter 5um.
The resistance of the 10cm long wire was also measured and it was 6.92kOhm. It wasn't really very easy, but I got through it.

Sorry, I got a messed up wire diameter calculation (I'll double-check and recalculate everything after I get some rest)

Double-checked myself. The real diameter of this wire is 5µm.

[MegaVolt]

The Russian wiki has a table with resistance per meter.
https://ru.wikipedia.org/wiki/%D0%9C%D0%B8%D0%BA%D1%80%D0%BE%D0%BF%D1%80%D0%BE%D0%B2%D0%BE%D0%B4

1 GΩ hermetic resistor design:

[bsw_m]

The Russian wiki has a table with resistance per meter.
https://ru.wikipedia.org/wiki/%D0%9C%D0%B8%D0%BA%D1%80%D0%BE%D0%BF%D1%80%D0%BE%D0%B2%D0%BE%D0%B4

The link has very incomplete data.
Already in the 60's, casted microwires with a core diameter of 0.1µm were produced.
And another example, a 5µm diameter microwire made of 53kxgs alloy will have resistance >102kOhm/meter.

[EC8010]

And another example, a 5µm diameter microwire made of 53kxgs alloy will have resistance >102kOhm/meter.

102 kohm per metre sounds an awful lot. So I took a hard look at the photographs. Knowing how difficult it is to take photographs through a microscope, I assumed that most of the turns in the "wire under microscope" shot were out of focus and invisible, but that the two adjacent turns on the right hand side represent the winding pitch of 20um. The active length of the resistor is about 18mm, so it has 900 turns. For 5um wire and 1M, that would need a rho of 170 compared to the 43 of manganin, which seems perfectly possible. The huge gain from winding single layer is that you don't need lacquered wire, so you can use the available space more efficiently.

Using the photograph of the dissected LT450 resistor earlier in this thread (and an LT450 that I could measure), I estimated its internal bobbin size and calculated what wire diameter and length would achieve 1M in the volume available. I assumed 8um thick lacquer and needed 411m of 15um manganin wire. That was 2.4 kohm per metre. Oh, and 20,465 turns.

[mawyatt]

Interesting reading, thanks. I would prefer Manganin as it is solderable and easy to get from old ww resistors, and when wound as the "divider" on the same bobbin (a bobbin with 3 legs) the TC could cancel, imho.

As a first order approximation this is correct, but note that this only applies if the heating on both sides is the same. If you have a 10:1 voltage divider, then one leg will dissipate 9x more heat than the other, so it will reach a higher temperature.

In the concept that iMo proposed, making a continuous WW resistor and tapping the appropriate point for the divider, the power is dissipated equally along the physical resistor because each incremental segment sees the same current and thus the same dissipation per length because of constant effective resistance per length of the resistor.

So if our understanding is correct, then iMo's concept wouldn't experience any change in voltage division ratio due to temperature, either ambient or induced (self heating). Of course in reality there will be some variations in this due to resistor non-uniformity, especially at the tap and ends.

Best,