Yes, both ends of the resistor must be at the same temperature.
Orientation of the PCB will change the relative end temperatures.
I did, half the fun of these discussions is to reread my old textbooks.
When I find I understood and remembered them correctly then I am pleasantly reassured.
And when I didn't understand them well at the time then to look with fresh eyes and understand them better is even more fun.
But the details of the tunnel effect in conductors are not simple, in fact there was a Nobel Prize in it.
So if I have made a mistake then I wouldn't feel foolish, instead I look forward to learn more from you.
For reference, first book I found to reread was Klaus Ziock's on QM pp 33 to 37.
If you think the numbers themselves are incorrect then please provide better estimates and we can redo the calculations.
And if this is an attempt to be dismissive of my simple calculations then it's your mistake, I prefer the back of envelopes
But more seriously, I think a reasonable, quick approximation is not an inferior substitute for a complicated analysis but an essential tool to master a topic.
R.D. Middlebrook has some brilliant work on this.
Now this one leaves me bemused, what do you perceive is my hidden objective? I'm really curious.
I just offered you a nice cash reward for even one reference that would prove my estimate incorrect.
I'd say that is pretty keen to have my work checked and learn.
Even their titles and authors' names?
That looks like a particularly transparent excuse.
Just to keep this a little on topic, I will check Felix Zandman's book tomorrow.
If anyone is authoritative on the subject then surely it's the founder and chief scientist of Vishay.
Best wishes
David
Last edited:
And, yet, how often are there seen resistors close to and pointed at the heat sink.... one end hotter than the other. Rather than parallel with the heat sink. Plus, convection air flow direction etc.
THx-RNMarsh
Richard,
That would take critical thinking on the layout designers part and you know how often that happens anywhere it seems. If a computer can kick something out that seems to be about what you get. Your still using your human brain, wish more would do that. Turn a resistor 90 degrees, what a concept.
That would take critical thinking on the layout designers part and you know how often that happens anywhere it seems. If a computer can kick something out that seems to be about what you get. Your still using your human brain, wish more would do that. Turn a resistor 90 degrees, what a concept.
Hi Andrew,
If both ends of the resistor are not at the same temperature, what is it that happens that we are concerned about?
Is it a dc potential difference across the resistor due to thermocouple effect,
or noise, or something more sinister like nonlinearity?
For each degree C of endcap temperature differential, how much badness happens (whatever that may be)?
Cheers,
Bob
It would be very simple to test a resistor for thermocouple effect. Punch a small hole in some card stock just large enough for a resistor to fit through. Place a resistor half way in the hole.
Connect the leads of the resistor to a voltmeter say in the mV range. While heating one end of the resistor shoot the other end with cold spray and see if a potential develops.
Connect the leads of the resistor to a voltmeter say in the mV range. While heating one end of the resistor shoot the other end with cold spray and see if a potential develops.
The references I found all say Nichrome is non-magnetic but I found no quantitative data.
Another Nickel alloy with the same structure is austenitic stainless steel and this is practically non magnetic.
It is likely they will behave the same because the physics should be similar, but it is extremely difficult to work this out just from theory.
Do you have any data on this?
The end caps of many (most?) resistors are ferrous, I suspect that would make more difference.
The purist would prefer no ferrous at all but Zandman doesn't mention it and I can find very little information.
So I suspect it's trivial.
I think the industry has mostly moved to Nichrome (or minor variation).
Is there any information that the Dales are different?
I think I found where the mistake has arisen.
Zandman does mention tunnel effect in thick film resistors made from particulates, then he discusses comparatively thinner resistors but still made similarly.
When he does discuss metal thin films then there is no mention of tunnel effect.
Presumably he would do so if it was relevant since he mentions it elsewhere and the book is extremely detailed.
So it looks like my calculations stand but Walter and his quote were closer to the truth than I credited.
Best wishes
David
Last edited:
The Dale do not have pressed on end cap-leads. Further reading.... they are not more than 1ppm or less for common values and wattages as would be used in pre and power amps.
Here's some hard data:
Measured thin film THD with VC of 1ppm at +20dbu = .00019%
a 3ppm VC resistor gave .00056%.
In practice, you can divide the VC by the number of R's iin series.
Doug Self has published distortion measurements on a variety of thru hole and sm types before.
THx-RNMarsh
Here's some hard data:
Measured thin film THD with VC of 1ppm at +20dbu = .00019%
a 3ppm VC resistor gave .00056%.
In practice, you can divide the VC by the number of R's iin series.
Doug Self has published distortion measurements on a variety of thru hole and sm types before.
THx-RNMarsh
If all the metals used in the resistor were identical, then there would be no badness.
It's the different metals that introduces the cold junction vs hot junction thermocouple effect.
Some metal combinations will be very different from other combinations. That's how they choose the metals to make thermocouples.
Could the (uncontrolled) thermocouple create a non linear effect?
Resistor reference site -
Here is a site which covers many issues of interest on resistors. A one stop shop for information and references related to resistors.
www.gmsystems.com/resistors.html
THx-RNMarsh
Here is a site which covers many issues of interest on resistors. A one stop shop for information and references related to resistors.
www.gmsystems.com/resistors.html
THx-RNMarsh
Hi Richard,
Thanks for posting this link. I took a quick look at their resistor tutorial,
and note from their section on voltage coefficient they state the following typical ranges for voltco for three common types of resistors:
Carbon composition: 500 ppm/V
Carbon film: 5-30 ppm/V
Metal film: 0.05 - 10 ppm/V
These are pretty large ranges. I'll take a look see where some of the resistors I've checked fall in this range. So far I have not been able to see the voltco on a metal film, but I think my results so far for carbon composition and carbon film are toward the low end of these ranges.
I'm testing at 1kHz and am currently trying to verify whether my results are are being influenced by thermal distortion on the 1/4 W carbon film resistor. I suspect that there is less likelihood that temperature distortions are involved in measurements of the 1/2 W carbon composition resistor due to the larger thermal mass of its resistive element.
I'm looking at 500Hz and 2kHz measurements to separate out whether thermal distortions are influencing the results. I'm finding that drawing conclusions from H3 distortion differences in the -120 to -140 dB range can be tricky and difficult, since there are many potential sources of nonlinearity down there.
Stay tuned.
Cheers,
Bob
hi Bob,
The maximum range given is because it includes the whole range of values and the whole range of wattage... That is what gives it the spread indicated. I suspect the thinness of the film causes it to heat quickly and thus TC will modulate/change the R more than VC will. Especially, for pulsed or non-continuous and transient signals. Music may fall somewhere between steady state signals and transient.
What is your reason to test carbon film instead of thin metal film? Carbon film has higher TC and VC. What R value and wattage are you testing?
THx-RNMarsh
The maximum range given is because it includes the whole range of values and the whole range of wattage... That is what gives it the spread indicated. I suspect the thinness of the film causes it to heat quickly and thus TC will modulate/change the R more than VC will. Especially, for pulsed or non-continuous and transient signals. Music may fall somewhere between steady state signals and transient.
What is your reason to test carbon film instead of thin metal film? Carbon film has higher TC and VC. What R value and wattage are you testing?
THx-RNMarsh
Last edited:
Hi Richard,
If I understand correctly, you measured 0.00019% H3 for a thin film (metal film) resistor whose Vc was 1ppm/V when 7.75 V rms was across it. Is that correct?
Bruce's equation states that H3 ppm is Vc ppm / 5.9. In this case a 1 ppm/V resistor would exhibit H3 of 7.75/5.9, or about 1.3 ppm H3, corresponding to 0.00013%, which is in reasonable agreement with your result of 0.00019%.
I am implicitly assuming the voltage to be used in Bruce's equation is the rms voltage. If peak voltage is used instead, the correspondence with your measurement is almost perfect.
How did you know that the resistor had a Vc of 1 ppm/V?
What was the value of the resistor?
At what frequency did you make this measurement?
What was the power dissipation rating of the resistor?
Was it a through-hole resistor?
How did you verify that the 0.00019% distortion was not thermal in origin? At these low distortion levels we cannot rule out thermal distortion at even 1kHz.
This is good to have an independent measurement of resistor voltage distortion so that we can compare measurement results.
Cheers,
Bob
Hi Richard,
I started with carbon composition so that I could confidently see the Vc effect at an early stage of the game, not being sure how much sensitivity I would need in the distortion-measuring setup to see it.
My measurements were conducted on a 6.8k 1/2 watt carbon composition resistor, with 4V rms across it. The DUT resistor was connected as the series element in an 11:1 attenuator with a 681-ohm 2 watt metal film shunt resistor. For reference and sanity check, a 6.81k 2 watt metal film resistor was used in place of the DUT resistor.
With this setup, when there is 4V rms across the DUT, there is 0.4V rms at the input to the THD analyzer. Measurements were conducted at 1kHz. The distortion residual output from the THD analyzer was fed to a spectrum analyzer with 1Hz bandwidth. H3 was then viewed on the spectrum analyzer. The measurement floor was between -130dB and -140dB.
Measurements were made at lower voltages as well, to see the effect.
I next tried a 1/4 watt carbon film resistor and found that the H3 was low enough that I did not have a lot of measurement margin at only 4V across the DUT. I did not yet measure a metal film resistor because I assumed its Vc would be buried in the noise with this setup and only 4V rms across it.
I am now running some tests with about 9V across the DUT by having built a 6dB amplifier (built with an LM4562 running at +/-17V) at the oscillator output of my THD analyzer.
Cheers,
Bob
The time to measure VC must be small to prevent/minimize resistance change due to heating.
Here's the MIL version of VC tests:
View attachment std202mthd309.pdf
THx-RNMarsh
Here's the MIL version of VC tests:
View attachment std202mthd309.pdf
THx-RNMarsh
These figures were not measured but calculated by Self using Ltspice and a model not explained. The figures given are THD. If you add the 5th harmonic you come even closer to this estimate, therefore I suspect he used B.Hofer non linear model. Lt spice is not that accurate in harmonic analysis using FFT
I wouldn't be so certain of that - the various settings need to be correct for an appropriate precision to be achieved; I've fallen in the trap many times of getting fuzzy numbers because I was sloppy with getting the parameters right ... LTspice doesn't hold your hand, or give suggestions - the user needs to know what's going on ...
I downloaded the datasheet.
It tells us that they are either wire, or oxide.
Not metal film.
Tc is stated as +-300ppm/C i.e. 33C degrees gives a maximum of +-1% of value.
The 10r and 20r that I have are a film, almost certainly an oxide film.
I have >1000 and use them for many purposes. Dummy load, mains current limiter, soft start resistance, etc.