Anything due to self-heating is something that LTSpice might require a lot of augmentation to model correctly.
LTSpice doesn't at the moment model lateral FETs correctly at low currents. However after the latest update of LTSpice this may soon change.
Along those lines, this is something I stumbled upon recently: https://www.fairchildsemi.com/application-notes/AN/AN-7533.pdf - any thoughts ... ?
That modeling technique will work in a limited context, but creates a lot of problems if you want to put it in a thermal context with other devices and have all the thermal time constants interact realistically. The RC string thermal subcircuit only fits a curve, it does not reflect the spatial nature of the thermal interface.
I've made several thermal subcircuits for transistors which work remarkably well although there are some flaws that I can't get around because I can only add self-heating signals to the transistors, I can't actually cause their model parameters to change with frequency. Still, it does most things right and I've found it to be accurate several times when testing my prototypes. I've even done some work on the idea of making a set of subcircuits and symbols for the thermal environment that click together on the schematic to pictorally and functionally simulate the connections between heatsinks and thermally coupled transistors, etc. You really can make symbols and models for heatsinks, insulators, convection, etc.
The problem is, thermal parameters for physical objects take time and effort to measure, and no one seems to really be interested in it.
I've made several thermal subcircuits for transistors which work remarkably well although there are some flaws that I can't get around because I can only add self-heating signals to the transistors, I can't actually cause their model parameters to change with frequency. Still, it does most things right and I've found it to be accurate several times when testing my prototypes. I've even done some work on the idea of making a set of subcircuits and symbols for the thermal environment that click together on the schematic to pictorally and functionally simulate the connections between heatsinks and thermally coupled transistors, etc. You really can make symbols and models for heatsinks, insulators, convection, etc.
The problem is, thermal parameters for physical objects take time and effort to measure, and no one seems to really be interested in it.
If one is evaluating Vc by measuring H3, this is not correct. Keeping measurement time short only addresses the mater of static resistance change by heating.
Only if you are measuring Vc by a dc measurement does keeping the measurement time short. In this case, you might put 100V dc on a 1 ppm/V resistor and look for a 100 ppm change in current on a pulse basis.
Cheers,
Bob
I'm confused. This is what Richard said:
"Here's some hard data:
Measured thin film THD with VC of 1ppm at +20dbu = .00019%
a 3ppm VC resistor gave .00056%."
Are you saying that this data was just from simulations by Doug Self and not measured data? Where is the disconnect here. Simulations by Self does not constitute "hard data".
Or are you saying that this is data measured by Self - in which my questions would still apply.
Cheers,
Bob
Yes that is what I suspect because these figures are exactly the ones found in a table of his book where it is said that they were calculated by LTspice.
Using B.Hofer formula comes close if you include H5 but not exactly ( 0.4ppm).
I wonder if at that low level round off errors in FFT would have an influence.
What is important is the confirmation of his models.
His thermal static and dynamic model is of course exact. The cubic relation between voltage and current ( i x power = k i³ ) is said in the literature to come ( in thin films) from deviation of vacancies concentration from equilibrium concentration.
The other non linearities of the relation current/voltage are coming (as I understand ) from scattering by impureties or non homogeneties, trapping and tunneling in the oxide layer surrounding grains. This oxyde layer is process dependent.
The tunneling through the ceramic base gaps is more for ultra thin films making islands separated by ceramic gaps acting as potential barrier for tunneling. I Wonder if thin film résistors are ultra thin island films.
The decrease of resistance with voltage increase ( as shown by B.Hofer) model makes sense if we consider than the potential difference is now higher at these potential barriers effectively decreasing them.
The reading of http://www.danbridge.com/Files/filelement30.pdf is rather confusing. There all the non linearities are taken as a cubic current voltage relation which is a thermal model.
I could not find an explanation of Voltage coef measurement other than 3th harmonic measurement that must englobe all the phenomena.
B.Hofer is claiming that above 1 or two Khz the dynamic thermal resistance becomes zero so the experiments in the above reference with 20khz should not be thermal.
JPV
yes, ---- I was pointing out that to you.... to seperate the heating effect from non-linearities in materials, you must keep time short (less than .5 sec) or use a pulsed signal.... I like the pulsed as it relates closer to the music signal. But the precision needed in measurement isnt good (enough?).
D.Self published measurements on many brands of resistors -- metal film. And the Dale th was included as were other sm types. It backs up my info re. Dale and sm being not as good. I dont recall the publication at the moment... I might find it around here, somewhere.
D.Self also goes into modeling of the VC et al in his A.PA Design. In the 6th edition it starts on page39. He seperates the mechanisms into 1st, 2nd and 3rd order distortions. Might give some additional info to help you find what you want to know about.
From his inspection.... most all the well known Japanese brands use CF type resistors.
THx-RNMarsh
Last edited:
Bob would it work to try and keep the body of a resistor at a constant temp for the Vc test.
Is it important this test be carried out at room temp?
it appears, from multiple sources, that the higher wattage parts have worse VC ! This makes series of smaller wattage more useful idea.
Also, see S. Groner work.
BTW -- there was a VC forum here (DIYAudio) started 17 March 2010... called 'Voltage Coefficient in Feedback Resistor'.
THx-RNMarsh
Also, see S. Groner work.
BTW -- there was a VC forum here (DIYAudio) started 17 March 2010... called 'Voltage Coefficient in Feedback Resistor'.
THx-RNMarsh
Last edited:
Hi Rchard,
Thanks, I'll look into Doug's material on Vc in his sixth edition. Am I correct in understanding that there was some other publication - not one of his books - where he went more deeply into his resistor investigations?
I'm still wrestling with separating out the thermal and voltage H3 distortions in measurement. As mentioned earlier, at levels of H3 in the -120dB to -130dB range with 11 V peak and maybe 1.5mA peak on a 6.8k 1/4 watt resistor, I'm not yet convinced that thermal distortions are not making a contribution to the measured H3. We are talking about peak power on the order of 10-20mW.
Cheers,
Bob
I'm not sure that keeping the body of the resistor at constant temperature would help in the H3 test, especially at 1kHz. Not sure how much local heat sinking effect the ceramic provides to the thin film and the time constant of the thermal.
It might not be really important to do the test at room temp. Did you have in mind putting the resistor in an ice water bath or something like that?
Cheers,
Bob
Interesting. That would be non-intuitive to me. Also, I use a 2W 6.81k metal film as my reference, thinking that it would have lower Vc due to larger film geometries.
Cheers,
Bob
Just letting everyone know of the latest developments in MOSFET subthreshold modeling:
http://www.diyaudio.com/forums/software-tools/266655-better-power-fet-models-ltspice.html
http://www.diyaudio.com/forums/software-tools/266655-better-power-fet-models-ltspice.html
Temp controlled cooling or heating. Ovens are easier to do. I've seen some that use a power transistor for the heating element and a current source for sensing temperature. An op amp for a proportional controller. Bits of Styrofoam like those packing peanuts can be used for insulation.
You can use the critical point of boiling water where the temperature remains constant during the phase change of liquid to vapor. I don't remember if this also works at the liquid to solid phase at freezing point. Water is strange stuff.
Nice work,
! Have you thought of making your efforts available for others to play around with - I would be interested in the model parameters changing with frequency thing ... always like looking into chewy things ... That is my thinking as well...... but I've read this about 3 times --- I'll have to triple check this and why.
-RNM
This looks incorrect for the usual thin metal film resistors.
Metallic conduction is by electrons, practically by definition.
I don't see how vacancies are involved, where in the literature?
Do you mean for Tantalum Nitride?
I replied to Walter that I don't think this occurs.
Found a foolish error in my estimate, left out a square root but it makes no material difference to the conclusion that no electrons tunnel in metal films of the thinness used in resistors.
Consider that the film is about 100 nM thick and tunnel distance is only about 1 nM.
Do you have a reference?
Best wishes
David
tunneling in island films is not through the thikness of the film but through the separation between islands of thin film.
see The Physical Properties of Thin Metal Films. B . K . Jones and G . P . Zhigal'skii