Parasitic capacitance of SMD resistors and their generated noise

[pavelm...@gmail.com]

Hello, colleagues
My question is about basic component that we all are working with - resistor, actually SMD 0402 type of them.
There are few technologies on the market to make them - thick film, thin film, metal film, and probably some more.
Resistors produced by different technologies has different performance.
Currently I am interested in three characteristics for 0402 size of resistors - parasitic parallel capacitance, parasitic series inductance, and generated noise, and their repeatability/stability/predictability in production. I am talking about frequency range from DC and up to 2 GHz.
Could you provide/reveal values of these properties of 0402 resistors made by each of above technologies? Any additional information will be appreciated.
Thank you

[Jeroen Belleman]

I measured parasitic capacitance and inductance for 1206 resistors
and got about 50fF and 500pH. I'd expect about the same for 0402,
because they have the same shape. Of course, pad layout and trace
width matters, as does the proximity of other conductors. I only
measured 1% metal film resistors.

Common wisdom says that only metal film should be used where noise
matters. I never made comparative measurements. In my field,
resistor cost is relatively negligible, so I tend to use metal
film almost everywhere.

Jeroen Belleman

[jla...@highlandsniptechnology.com]

On Mon, 07 Mar 2022 12:50:45 +0100, Jeroen Belleman
<jer...@nospam.please> wrote:

>On 2022-03-07 12:10, pavelm...@gmail.com wrote:
>> Hello, colleagues
>>
>> My question is about basic component that we all
>> are working with - resistor, actually SMD 0402 type of them. There
>> are few technologies on the market to make them - thick film, thin
>> film, metal film, and probably some more. Resistors produced by
>> different technologies has different performance. Currently I am
>> interested in three characteristics for 0402 size of resistors -
>> parasitic parallel capacitance, parasitic series inductance, and
>> generated noise, and their repeatability/stability/predictability in
>> production. I am talking about frequency range from DC and up to 2
>> GHz. Could you provide/reveal values of these properties of 0402
>> resistors made by each of above technologies? Any additional
>> information will be appreciated. Thank you
>>
>
>
>I measured parasitic capacitance and inductance for 1206 resistors
>and got about 50fF and 500pH. I'd expect about the same for 0402,
>because they have the same shape. Of course, pad layout and trace
>width matters, as does the proximity of other conductors. I only
>measured 1% metal film resistors.

The parasitics of various resistors are easily googled.

>
>Common wisdom says that only metal film should be used where noise
>matters. I never made comparative measurements. In my field,
>resistor cost is relatively negligible, so I tend to use metal
>film almost everywhere.
>
>Jeroen Belleman

Johnson noise must be the same regardless of the materials.

Thick-film resistors can have excess noise if there is voltage across
them, but it's hard to measure. I managed to measure some once, but it
took very high value resistors and high voltages.

High tempco resistors in a voltage divider or equivalent can make low
frequency (subsonic) noise from small temperature fluctuations.



--

I yam what I yam - Popeye

[Phil Hobbs]

Yup. Cermet, thick film, and anything with carbon in it are disastrous
for 1/f noise. They're all pretty well equivalent at high frequency.
("Thin film" is SMT-speak for "metal film".)

JL made some TDR measurements of (iirc) thin film resistors and showed
that the inductance was significantly reduced by mounting them upside
down, so that the resistive track was next to the board.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

[Rick C]

The inductance or the impedance? If the resistive layer has a smaller spacing from the board it would also impact the capacitance to the ground layer greatly impacting the impedance, probably much more than just the inductance.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

[Phil Hobbs]

As Kipling's Shipwrecked Mariner said to the Whale, "Not so, but far
otherwise." (At least in my business, i.e. ultrasensitive measurements,
that is.)

One time I was chasing my tail for about a day, trying to figure out why
my shiny new super low noise laser driver gizmo was exhibiting horrific
1/f noise on the spectrum analyzer. It turned out to be the classic
Tektronix 50-ohm 2-W BNC feedthrough terminator I was using--it was made
of cermet. Switching to metal film dropped the 1/f noise by about two
orders of magnitude IIRC. Not subtle at all.

>
> High tempco resistors in a voltage divider or equivalent can make low
> frequency (subsonic) noise from small temperature fluctuations.

Yup. Random temperature drift has roughly a 1/f**2 noise PSD.

The LIGO folks published a study on noise in resistors, IIRC, but I'm
not laying my hands on it at the moment.

[jla...@highlandsniptechnology.com]

Was there voltage across that terminator? Was it making unusually
large amounts of Johnson noise?

[jla...@highlandsniptechnology.com]

Regular cermets, but the geometric effects are the same.

You can buy resistors without wrap-around end caps, intended to solder
element down.

[Phil Hobbs]

Yes, I was pumping a couple of hundred milliamps into it. 1/f noise
arises from conductance fluctuations, so it doesn't appear unless
there's current flowing. As you say, Johnson noise depends only on the
resistance and the temperature, but of course it also assumes thermal
equilibrium, which stops applying when the power gets turned on.

Cheers

[Gerhard Hoffmann]

Am 07.03.22 um 16:31 schrieb jla...@highlandsniptechnology.com:

> Johnson noise must be the same regardless of the materials.
>
> Thick-film resistors can have excess noise if there is voltage across
> them, but it's hard to measure. I managed to measure some once, but it
> took very high value resistors and high voltages.
>
> High tempco resistors in a voltage divider or equivalent can make low
> frequency (subsonic) noise from small temperature fluctuations.

I have a version of Win's AOE3 ribbon microphone amplifier,
not differential but only single-ended with an ugly coupling cap
and 16 transistors instead of 64.

That has a switchable ~6 Ohm resistor to get "calibration noise".
It is easy to see.

<
https://www.flickr.com/photos/137684711@N07/44311358915/in/album-72157662535945536/
> and left / right arrows

My 10 * 2 * ADA4898 amplifier uses 60R as a 1nV/rtHz normal.


Cheers, Gerhard

[John Larkin]

On Mon, 7 Mar 2022 12:35:32 -0500, Phil Hobbs

It's probably too late, but it would be interesting to measure the 1/f
noise vs current. It might be square law, namely thermal.

Thinfilms have much lower tempcos than cermets. There could be
micro/localized thermal effects too, like at grain boundaries. Laser
trimming can create horrors.

I found it difficult to measure excess noise in cermets, and it was in
the ballpark of the Johnson noise. These were megohm range values, so
self-heating was negligable.

Then it was hard to find 10M sorts of thinfilms.

--

If a man will begin with certainties, he shall end with doubts,
but if he will be content to begin with doubts he shall end in certainties.
Francis Bacon

[LM]

On Mon, 07 Mar 2022 09:21:55 -0800, jla...@highlandsniptechnology.com
wrote:

Digikey had several types of resistors.

Carbon combosition
Carbon film
Ceramic
Metal Element
Metal Film
Metal Foil
Thich Film
Thin Film
Wirewound

Last time I checked, there was only thin and thick film. That is not
much of a choice.

[Phil Hobbs]

Nah, definitely linear, and definitely 1/f not 1/f**2. A constant-rate
drift notionally has a spectrum that goes as 1/f**4, and a random-walk
drift ('brown noise') goes as 1/f**2. That's what you usually see go
away when you put styrofoam on top of the resistor.

It's quite possible that the conductivity fluctuations depend on
temperature, but I don't recall hearing that mentioned.

>
> Thinfilms have much lower tempcos than cermets. There could be
> micro/localized thermal effects too, like at grain boundaries. Laser
> trimming can create horrors.
>
> I found it difficult to measure excess noise in cermets, and it was in
> the ballpark of the Johnson noise. These were megohm range values, so
> self-heating was negligable.
>
> Then it was hard to find 10M sorts of thinfilms.

Well, believe me, this example was the very furthest thing from subtle.

Cheers

Phil Hobbs

[Phil Hobbs]

To an engineering approximation, all metal / 'thin film' resistors
exhibit only Johnson noise.

[pavelm...@gmail.com]

Hello, Everybody
I am excited to read the discussion about the influence of resistor technology on noise. It is very interesting and helpful.
Though one of the original questions was about how 0402 resistor's parallel capacitance depends on the resistor's technology. Vishay's application paper (freqresp.pdf) talks about comparison of thin film resistors with different termination styles (wrap, flip ship) and their capacitance. But they are all made with thin film technology.
They mentioned that for thin film "normal" 0402 resistors with wrap termination this C may be expected to be about 40fF. This is, of course, the resistor's own parasitic C, as I would like to believe, without taking into account the PCB and pads.
Does anybody of you have a similar number for a comparable thick film 0402 resistor with wrap terminals? I couldn't google it, to my surprise.
Thank you for your responses

[whit3rd]

On Monday, March 7, 2022 at 11:47:16 AM UTC-8, Phil Hobbs wrote:
> John Larkin wrote:
> > On Mon, 7 Mar 2022 12:35:32 -0500, Phil Hobbs
> > <pcdhSpamM...@electrooptical.net> wrote:
> >
> >> jla...@highlandsniptechnology.com wrote:
> >>> On Mon, 7 Mar 2022 12:03:17 -0500, Phil Hobbs
> >>> <pcdhSpamM...@electrooptical.net> wrote:

> >>>> ...gizmo was exhibiting horrific

> >>>> 1/f noise on the spectrum analyzer. It turned out to be the classic
> >>>> Tektronix 50-ohm 2-W BNC feedthrough terminator I was using--it was made
> >>>> of cermet. Switching to metal film dropped the 1/f noise by about two
> >>>> orders of magnitude IIRC. Not subtle at all.

> > It's probably too late, but it would be interesting to measure the 1/f
> > noise vs current. It might be square law, namely thermal.

> Nah, definitely linear, and definitely 1/f not 1/f**2. A constant-rate
> drift notionally has a spectrum that goes as 1/f**4, and a random-walk
> drift ('brown noise') goes as 1/f**2.

Carbon resistors are semi-metals, so they get recombination noise; probably
the 'cermet' is a similar material. For metal film, though, it's terribly difficult to
get a thin and long continuous metal path for high resistance, AND keep it from
oxidizing and changing value. So, carbon (carbon film?) still is the solution for high-ohms
items.

[John Larkin]

On Mon, 7 Mar 2022 13:09:19 -0800 (PST), whit3rd <whi...@gmail.com>
wrote:

I have some 1T ohm surface-mount cermet resistors.

[Phil Hobbs]

whit3rd wrote:
> On Monday, March 7, 2022 at 11:47:16 AM UTC-8, Phil Hobbs wrote:
>> John Larkin wrote:
>>> On Mon, 7 Mar 2022 12:35:32 -0500, Phil Hobbs
>>> <pcdhSpamM...@electrooptical.net> wrote:
>>>
>>>> jla...@highlandsniptechnology.com wrote:
>>>>> On Mon, 7 Mar 2022 12:03:17 -0500, Phil Hobbs
>>>>> <pcdhSpamM...@electrooptical.net> wrote:
>
>>>>>> ...gizmo was exhibiting horrific 1/f noise on the spectrum
>>>>>> analyzer. It turned out to be the classic Tektronix 50-ohm
>>>>>> 2-W BNC feedthrough terminator I was using--it was made of
>>>>>> cermet. Switching to metal film dropped the 1/f noise by
>>>>>> about two orders of magnitude IIRC. Not subtle at all.
>
>>> It's probably too late, but it would be interesting to measure
>>> the 1/f noise vs current. It might be square law, namely
>>> thermal.
>
>> Nah, definitely linear, and definitely 1/f not 1/f**2. A
>> constant-rate drift notionally has a spectrum that goes as 1/f**4,
>> and a random-walk drift ('brown noise') goes as 1/f**2.
>
> Carbon resistors are semi-metals, so they get recombination noise;

Not sure what you mean by that, exactly. To get recombination noise you
need minority carriers, no?

> probably the 'cermet' is a similar material.

Cermets aren't homogeneous--as the name implies, they're ceramic/metal
composites.

> For metal film, though, it's terribly difficult to get a thin and
> long continuous metal path for high resistance,

Why so? I used to make conducting films of 100 angstroms or so. Atoms
are pretty small, and using sputtering as opposed to directional
evaporation will make the film follow even rough substrates pretty well.

> AND keep it from oxidizing and changing value. > So, carbon (carbon
> film?) still is the solution for high-ohms items.

It's certainly true that it's harder to make very high resistances out
of very low resistance materials, and there are lots of low-precision
applications where 1/f noise is not a serious issue--overvoltage
protection, for instance.

However, carbon resistors are seriously nonlinear at high voltages--the
resistance of old style Allen-Bradley carbon comps was allowed to drop
by a quarter at their upper voltage limit.

[Phil Allison]

jla...@highlandsniptechnology.com wrote:
================================

>
> Johnson noise must be the same regardless of the materials.
>

** A simple fact overwhealmed by persistent myth to the contrary.

> Thick-film resistors can have excess noise if there is voltage across
> them, but it's hard to measure.

** No it isn't.

> I managed to measure some once, but it
> took very high value resistors and high voltages.

** Drivel.

Easy enough to characterise carbon film, carbon comp, metal glaze & MF types with values of say 100K and few DC volts.
JL does not underdstand audio.


...... Phil

[Phil Allison]

Phil Hobbs wrote:
==============

>
> However, carbon resistors are seriously nonlinear at high voltages--the
> resistance of old style Allen-Bradley carbon comps was allowed to drop
> by a quarter at their upper voltage limit.
>

** Really ????
That is massively ambiguous.

CC resistors can drop in value if they run hot and over time.
Mostly they slowly drift high, over a period of decades.

Crazy to crazy that " non linearity".


..... Phil

[John Larkin]

On Mon, 7 Mar 2022 13:56:36 -0800 (PST), Phil Allison
<palli...@gmail.com> wrote:

> jla...@highlandsniptechnology.com wrote:
>================================
>>
>> Johnson noise must be the same regardless of the materials.
>>
> ** A simple fact overwhealmed by persistent myth to the contrary.
>
>> Thick-film resistors can have excess noise if there is voltage across
>> them, but it's hard to measure.
>
>** No it isn't.
>

https://tinyurl.com/y89rmtk3

First paragraph.

>> I managed to measure some once, but it
>> took very high value resistors and high voltages.
>
>** Drivel.
>
>Easy enough to characterise carbon film, carbon comp, metal glaze & MF types with values of say 100K and few DC volts.

Got some measurements?

>JL does not underdstand audio.

Well, my hearing is terrible, especially below 10 Hz.

[John Larkin]

google resistor voltage coefficient

Carbons can be ballpark 100 PPM/volt.

[Phil Allison]

John Larkin wrote:
================
> >>
> >> Johnson noise must be the same regardless of the materials.
> >>
> > ** A simple fact overwhealmed by persistent myth to the contrary.
> >
> >> Thick-film resistors can have excess noise if there is voltage across
> >> them, but it's hard to measure.
> >
> >** No it isn't.
> >
> https://tinyurl.com/y89rmtk3
>
> First paragraph.

** Yawnnnnnnn ....

How pathetic.

> >> I managed to measure some once, but it
> >> took very high value resistors and high voltages.
> >
> >** Drivel.
> >
> >Easy enough to characterise carbon film, carbon comp, metal glaze & MF types with values of say 100K and few DC volts.
>
> Got some measurements?

** Published them years ago.

The excess noise as voltage was applied ( up to 20V) was marked, except for MF types

> >JL does not underdstand audio.

> Well, my hearing is terrible,

** And all your other senses too.

> especially below 10 Hz.


** Not audio.

Wot a tedious fuckhead.


..... Phil

[Phil Allison]

John Larkin bullshitted as usual wrote:
================================

> > Phil Hobbs wrote:
> >==============
> >>
> >> However, carbon resistors are seriously nonlinear at high voltages--the
> >> resistance of old style Allen-Bradley carbon comps was allowed to drop
> >> by a quarter at their upper voltage limit.
> >>
> >
> >** Really ????
>
> > That is massively ambiguous.
> >
> > CC resistors can drop in value if they run hot and over time.
> > Mostly they slowly drift high, over a period of decades.
> >
> > Crazy to crazy that " non linearity".
> >

> google resistor voltage coefficient

** Google " arrogant fuckwit " - see yourself described.

> Carbons can be ballpark 100 PPM/volt.

** But are in fact not.


..... Phil

[John Larkin]

Comps can be 200.

In a tube amp with, say, a 50 volt p-p swing, that could be 10000 PPM
gain change, 1% distortion. That's almost enough to hear.

[John Larkin]

I'm having fun. Are you?

[whit3rd]

On Monday, March 7, 2022 at 1:27:34 PM UTC-8, Phil Hobbs wrote:

> whit3rd wrote:

> > Carbon resistors are semi-metals, so they get recombination noise;

> Not sure what you mean by that, exactly. To get recombination noise you
> need minority carriers, no?

> > probably the 'cermet' is a similar material.

> Cermets aren't homogeneous--as the name implies, they're ceramic/metal
> composites.

Semi-metals are semiconductors, but at or near their intrinsic temperatures.
A bunch of metal shavings in a tube, if shaken, has enough Shottky rectification
at the contact points to be an RF detector (coherer was the antique radio term).
Ceramics or metal oxides, with metal, are going to have minority carriers.
At each metal connection, you get carrier injection.

> > For metal film, though, it's terribly difficult to get a thin and
> > long continuous metal path for high resistance,

> Why so? I used to make conducting films of 100 angstroms or so. Atoms
> are pretty small, and using sputtering as opposed to directional
> evaporation will make the film follow even rough substrates pretty well.

> > AND keep it from oxidizing and changing value. > So, carbon (carbon
> > film?) still is the solution for high-ohms items.

Carbon resistors, especially in presence of ozone, drift to higher values
because the carbon slowly turns to CO or CO2. Metals also grow oxide skins
(some more quickly than others). Substrate, metal, and glaze (or at least lacquer) plus
two solderable connections... a complete surface mount resistor packs a lot of
engineering into that little package.

[Phil Allison]

John Larkin bullshitted as usual wrote:
================================
>

> >> > Phil Hobbs wrote bullshit too:
> >> >==================

> >> >>
> >> >> However, carbon resistors are seriously nonlinear at high voltages--the
> >> >> resistance of old style Allen-Bradley carbon comps was allowed to drop
> >> >> by a quarter at their upper voltage limit.
> >> >>
> >> >
> >> >** Really ????
> >>
> >> > That is massively ambiguous.
> >> >
> >> > CC resistors can drop in value if they run hot and over time.
> >> > Mostly they slowly drift high, over a period of decades.
> >> >
> >> > Crazy to crazy that " non linearity".
> >> >
> >
> >> google resistor voltage coefficient
> >
> >** Google " arrogant fuckwit " - see yourself described.
> >
> >
> >> Carbons can be ballpark 100 PPM/volt.
> >
> >** But are in fact not.
> >
>

> Comps can be 200.


** Bullshit

> In a tube amp with, say, a 50 volt p-p swing, that could be 10000 PPM
> gain change, 1% distortion.

** Made up, absolute crap !!!!
-----------------------------------------------
-----------------------------------------------

Just to refute JL's rampant insanity - I carried a test:
With a sine wave at 1KHz feeding a small toroidal supply tranny in reverse.
Max available = 250Vrms at 0.05% THD.

A pair of MF resistors ( 100k and 1k ) in series showed only residual THD across the 1k.
Change the 100k to a ( very old ) 0.5W CC and the THD reading was then 0.33%
Drop the input level to 50V rms and the reading was 0.07%.

250V rms = 705 V p-p.
50V rms = 141V p-p.

FYI

One cannot test the * lineartiy * using DC voltages.
Cos the damn things heat and drop value by several %.



..... Phil

[John S]

Yes! I learned that the hard way. Decided I was gonna make my own HV
divider from a string of 22meg CC resistors. I was astonished by the
change in resistance with voltage. So I made a new scale for the meter.

[jla...@highlandsniptechnology.com]

I said "comps", meaning carbon composition. We were talking about
carbon resistors.

[jla...@highlandsniptechnology.com]

In a divider made from a string of identical resistors, the
nonlinearity cancels. That's why the best dividers are all on the same
substrate.

[John Walliker]

The long-term resistance shift doesn't behave that way at high voltages,
because the ones at the top of the string suffer more corona discharge which
can therefore erode them faster. I found this out the hard way.

John

[jla...@highlandsniptechnology.com]

I guess it should be coated or potted or something.

Ww have a potential customer who wants a 1500 volt power supply
programmable and stable to 1 PPM. That could get interesting.

Some sort of shield could prevent corona. But 1500v isn't bad.

[John S]

No, I don't think so. The resistors all have a negative voltage
coefficient and they add in a string.

[jla...@highlandsniptechnology.com]

Given 10 identical resistors in series, each drops 1/10 of the total
voltage.

[John S]

Yes. But the resistance at rated voltage is much lower than it is at a
couple of volts.

[Phil Hobbs]

But in a divider with 25% matched nonlinearity, you'd have
1*0.75 / N*0.75 = 1 / N regardless.

Of course since the nonlinearity is caused by hot filaments beginning to
form inside the resistive element, one wouldn't expect them to be
particularly repeatable or time-invariant.

[Jan Panteltje]

On a sunny day (Tue, 08 Mar 2022 07:39:29 -0800) it happened
jla...@highlandsniptechnology.com wrote in
<5ute2hlgqblmtf9l7...@4ax.com>:

>I guess it should be coated or potted or something.
>
>Ww have a potential customer who wants a 1500 volt power supply
>programmable and stable to 1 PPM. That could get interesting.
>
>Some sort of shield could prevent corona. But 1500v isn't bad.

Not just corona, any moisure in the air would screw things up, condensation on your PCB too.
Done a lot of HV CRT stuff, foucus was for example about 4 kV, done with a large carbon
wound resistor in the antique sets,,,
1 ppm is pushing it.
https://www.wagneronline.com.au/high-voltage-tv-focus-resistors/television-parts/service-repair-parts/1fcr150k-3714/679/pd/

[whit3rd]

On Tuesday, March 8, 2022 at 7:39:44 AM UTC-8, jla...@highlandsniptechnology.com wrote:
> On Tue, 8 Mar 2022 03:46:03 -0800 (PST), John Walliker
> <jrwal...@gmail.com> wrote:

[about carbon resistors]

> >The long-term resistance shift doesn't behave that way at high voltages,
> >because the ones at the top of the string suffer more corona discharge which
> >can therefore erode them faster. I found this out the hard way.

> I guess it should be coated or potted or something.
>
> Ww have a potential customer who wants a 1500 volt power supply
> programmable and stable to 1 PPM. That could get interesting.
>
> Some sort of shield could prevent corona. But 1500v isn't bad.

It's about the same as focus voltages in old CRT displays; I've replaced
a lot of focus resistors.

To prevent corona around a carbon resistor, you might consider getting the
resistor sealed in vacuum (glass tube package). At 45 kV auto ignition
wiring just had to go to 8mm diameter silicone insulation: the corona in
silicone is harder to start than corona in air, so... you build up the
insulator to fill the high-field volume entirely.

Victoreen and Ohmite make vacuum packages : <https://www.ebay.com/itm/123046764521>

[Phil Allison]

John Shithead wrote:

==================

>
> >>> In a divider made from a string of identical resistors, the
> >>> nonlinearity cancels. That's why the best dividers are all on the same
> >>> substrate.
> >>>
> >>
> >> No, I don't think so. The resistors all have a negative voltage
> >> coefficient and they add in a string.
> >
> > Given 10 identical resistors in series, each drops 1/10 of the total
> > voltage.
>
> Yes.

** Game ove.,

> But the resistance at rated voltage is much lower than it is at a
> couple of volts.

** You awake ?

At any applied voltage, all the series Rs have the *same* voltage drop.
The 10:1 ratio is unchanging it they all have the same deviations.


...... Phil

[John Doe]

John S wrote:

> jla...@highlandsniptechnology.com wrote:

>> Given 10 identical resistors in series, each drops 1/10 of the total
>> voltage.

> Yes. But the resistance at rated voltage is much lower than it is at a
> couple of volts.

Weird... The current is the same. But as the voltage drops, the total
remaining resistance decreases too, moving through the divider. Should be
simple but... That's why we depend on rules.

[Anthony William Sloman]

> >>> Yes! I learned that the hard way. Decided I was gonna make my own HV
> >>> divider from a string of 22meg CC resistors. I was astonished by the
> >>> change in resistance with voltage. So I made a new scale for the meter.
> >>
> >> In a divider made from a string of identical resistors, the
> >> nonlinearity cancels. That's why the best dividers are all on the same
> >> substrate.
> >
> > No, I don't think so. The resistors all have a negative voltage
> > coefficient and they add in a string.
> >
> But in a divider with 25% matched nonlinearity, you'd have
> 1*0.75 / N*0.75 = 1 / N regardless.
>
> Of course since the nonlinearity is caused by hot filaments beginning to
> form inside the resistive element, one wouldn't expect them to be
> particularly repeatable or time-invariant.

Actually, once a hot channel has formed it can be pretty stable.

When there is just enough channeling to cause temperate gradients across the current path, these can move the current path around and can give you an unstable resistance. We saw this with NTC thermistors at one place where I worked - when the automated test gear tried to dissipate about a milliwatt in the thermistor the resistance - displayed to six significant figures - never showed the same last two digits for any length of time.

Measuring the resistance on a higher resistance range (with less power dissipated in the thermistor) gave a stable result, but at lower resolution.

--
Bill Sloman, Sydney

[Jan Panteltje]

On a sunny day (Tue, 8 Mar 2022 12:56:46 -0800 (PST)) it happened whit3rd
<whi...@gmail.com> wrote in
<9168331e-8253-4e0a...@googlegroups.com>:

>To prevent corona around a carbon resistor, you might consider getting the
>resistor sealed in vacuum (glass tube package). At 45 kV auto ignition
>wiring just had to go to 8mm diameter silicone insulation: the corona in
>silicone is harder to start than corona in air, so... you build up the
>insulator to fill the high-field volume entirely.
>
>Victoreen and Ohmite make vacuum packages : <https://www.ebay.com/itm/123046764521>

Very nice
This morning i ws tinking abiut different physics to measue g=high voltages
Icame upo (in my mind that is( with a varant f teh elctrometer

2 charged plates, on on ground the other on the HV,
and then the ground one mounted on a piezo like weight sensor
The attraction between the plates would perhaps be measurable, and no current load

--------- HV plate

---------- GND plate
///////// -> weight sensor
============= PCB (or whatever

Accuracy. probably bad


For my small PMT this setup works fine up to a kV or so:
http://panteltje.com/pub/PMT_HV_supply_componet_side_img_3180.jpg
HV resistor divider is on the far left.

http://panteltje.com/pub/PMT_HV_supply_with_regulator_img_3175.jpg

is stabilized of course:
http://www.panteltje.com/pub/PMT_regulated_power_supply_diagram_img_3182.jpg
filtered PWM rom a PIC micro controls it.
Been working fine for 11 years now. no idea how accurate, few volt perhaps.
use a precision opamp?

I have an other one for a bigger PMT with a lot more voltage:
http://panteltje.com/panteltje/pic/sc_pic/
also PIC controlled

1 ppm ..no but shoud be possible? 1kV 1mV sigh.. ripple.... ?
LOL

Do not move near it!

[jla...@highlandsniptechnology.com]

It might get expensive.

>
>Do not move near it!
>
>
>
>

Nice stuff. MMBD5004S/1SS398TE85LF is a dual 400v diode in SOT23,
which can shorten a C-W multiplier string.

https://www.dropbox.com/s/jfx8it5acjwg4je/Z206_CWmult.JPG?raw=1

https://www.dropbox.com/s/2pgawxq6a0xnqxz/DRQ_11.asc?dl=0

[whit3rd]

On Tuesday, March 8, 2022 at 11:45:48 PM UTC-8, Jan Panteltje wrote:

> This morning i ws tinking abiut different physics to measue g=high voltages
> Icame upo (in my mind that is( with a varant f teh elctrometer
>
> 2 charged plates, on on ground the other on the HV,
> and then the ground one mounted on a piezo like weight sensor
> The attraction between the plates would perhaps be measurable, and no current load
>
> --------- HV plate
>
> ---------- GND plate
> ///////// -> weight sensor
> ============= PCB (or whatever
>
> Accuracy. probably bad

That's a variant of traditional electrostatic voltmeter design with a vane
and plates, which typically is done in rotary fashion with a hairspring for
return force... and requires jeweled bearings and level adjustment...
it's a fiddly nuisance of an instrument, but does pretty well at
sorting out kilovolts without many picoamps of leakage.
I suspect the weight sensor will respond to every bit of building shake
in the wind...

> For my small PMT this setup works fine up to a kV or so:
> http://panteltje.com/pub/PMT_HV_supply_componet_side_img_3180.jpg

Yow. So many parts just to replace a single vacuum tube rectifier...

> HV resistor divider is on the far left.

And then the long string of low-V resistors...

It'd almost be easier to meter with a field mill (motorized vane
changing capacitance-to-electrode, and AC amplification
of the resulting current). An interesting variant would be
to program a second electrode with some smart PWM
and balance the pulses from the known-voltage electrode
and the unknown-voltage electrode.

[Phil Hobbs]

The classical method is the vibrating-reed electrometer.

[Jan Panteltje]

On a sunny day (Wed, 9 Mar 2022 13:35:31 -0500) it happened Phil Hobbs
<pcdhSpamM...@electrooptical.net> wrote in
<dc533e1c-b0f1-5bca...@electrooptical.net>:

Thanks, bit of googling found a nice wikipedia article:
https://en.wikipedia.org/wiki/Electrometer

[whit3rd]

On Wednesday, March 9, 2022 at 10:35:47 AM UTC-8, Phil Hobbs wrote:

> Jan Panteltje wrote:

> > This morning i ws tinking abiut different physics to measue g=high voltages
> > Icame upo (in my mind that is( with a varant f teh elctrometer
> >
> > 2 charged plates, on on ground the other on the HV,
> > and then the ground one mounted on a piezo like weight sensor
> > The attraction between the plates would perhaps be measurable, and no current load
> >
> > --------- HV plate
> >
> > ---------- GND plate
> > ///////// -> weight sensor
> > ============= PCB (or whatever
> >
> > Accuracy. probably bad

> > 1 ppm ..no but shoud be possible? 1kV 1mV sigh.. ripple.... ?
> > LOL
> >
> > Do not move near it!

> The classical method is the vibrating-reed electrometer.

That's also a 'do not move near it' solution. I scavenged a head unit
from an old Cary 31 vibrating-reed electrometer, and it's got 10lbs of cast iron to
keep acoustic input to a minimum; there were selected tubes
inside to deal with the microphonic feedback, and anyone who
could afford it (circa 1970) was swapping in new sockets and nuvistor
tubes, 'cuz microphonics.

[Phil Hobbs]

"Doctor, doctor, it hurts when I go like this!"

"So don't go like that."

I used to have a handheld electrometer with an analog dial. It looked a
bit like an old CdS photographic light meter, and worked fine.