You have no idea what I do and what kind of resistors I'm using on a daily basis. Yourself, got any complaints from your customers or fan club that the 2 penny metal film resistors in the junk box are not good enough for audio?
Otherwise, with each of your reply, I can feel the love in the air. And BTW, speak for yourself, rather than "we".
Hi JPV,
It does indeed worry me that I find my measurements to be inconsistent with what Bruce's paper suggests. He is indeed a brilliant analog designer (and a good friend for at least 35 years). He is the guy who first turned me on to the use of the state variable architecture for extremely low distortion oscillators.
Cheers,
Bob
I don't think absolute precision is the real point; it's the fact that a very high precision resistor is stable enough to hold that precision under a wide range of temperature, aging, and hopefully power and voltage coefficient and other known non-linearities.
Metal film resistors can be readily obtained at 25 ppm and lower; Vishay/Dale PTF series goes down to 5 ppm, for example. The cost for higher stability metal film resistors is not too high so I generally use those when my tiny budget for bulk metal foil is (almost always) exceeded. 1% or better Wirewounds shouldn't be neglected either; I see bulk foil resistors as miniaturized wirewounds.
At any rate, I feel like I've established a perspective for resistor types of known quality that allows me to select the better/best parts based on good engineering principles at reasonable cost. While it might be argued that resistors are already "good enough", distortion is getting so low in modern circuit designs that I feel it's worthwhile to start going after the usually overlooked passive components and hopefully clean up their tiny contributions as well.
Last edited:
There's a few problems here.
The quote is a piece of speculation ("may result"), that is unsubstantiated, unreferenced, by an anonymous writer, in a commercial vendor's handout.
It also appears to be incorrect on a few points.
Bruce Hofer, Bob Cordell and the resistor manufacturer's data all indicate that "thin metal film... resistors" have a low coefficient. Your reference claims the opposite.
Even the claim about the ceramic substrate looks dubious.
Check Kyocera data sheets for the surface finish on typical substrates, they are reasonably smooth, obviously depends on exactly what scale you are interested in.
So the reference looks a lot like claims for almost mystical "quantum" effects by cable vendors and the like.
But Barth looks reputable, so lets run the numbers to do a quick plausibility check.
First for classical conduction, the "or current flow across the ceramic valleys" bit
Resistivity of commercial alumina resistor substrate at room temperature is > 1e12 ohm m.
Resistivity of nichrome is about 1e-6
So about 1e18 difference, 1 part in 1000 000 000 000 000 000, just to make it obvious.
There will be a multiplier because of thicker substrate and shape factors but no surface unevenness could possibly produce a valley where contribution of the ceramic would be even remotely detectable, not for the resistor at any tolerable temperature.
Now, "quantum tunnel" conduction requires such specific conditions that it is much less likely than classical conduction.
So the previous estimate of (some multiplier) of 1e-18 looks quite likely by comparison.
Take a reasonable maximum potential difference in a thin film resistor as 10 V/mm, so 10 uV/nm.
A typical Metal-Insulator-Metal tunnel diode has the insulator a few nm thick and in the order of a volt.
So the potential is one ten thousandth to a hundred thousandth less.
The current flow has an exponential factor with a constant of -2.
So at least e-20 000 less. Where e is base of Ln = (2.718 etc), not the power of 10 notation.
I think there is a additional linear multiplier that makes it even worse.
Even if you assume some surface unevenness that multiplies the potential across the barrier, it looks extraordinarily unlikely.
Been a while since I did quantum mechanics at Uni but if I am even approximately close then you wouldn't see one electron tunnel in the lifetime of the universe.
So if you learned this in school I am very interested to know where.
Do you have a reference to a text book?
If you can find a reputable peer-reviewed paper that supports the claim then that will spin my head even more than Circ Soleil did.
I will send you the price of a ticket just to have that same fun sensation once more.
Best wishes
David
Last edited:
Please have a look at this paper from Vishay.
http://www.ieee.li/pdf/essay/basics_of_linear_fixed_resistors.pdf
The 3th harmonic for metal film is< -110db which can be in the ~area of B. Hofer paper. The carbon composite is < 60 db
JPV
Not being an EE or physicist I don't know anything technical about quantum mechanics as to electron flow, but being involved in manufacturing my whole adult life I would expect that there is a specification listing the allowable rms value of surface finish of any of those ceramic substrates. This is typical in so many instances where surface finish affect the final quality of suitability of an item. In typical metal manufacturing an optical comparator is used to confirm the final finish or at any stage in production. I imagine the same must be taking place in the manufacturing of the ceramic substrate used in these devices spoken of here. If tunneling is a known phenomena at a certain peak valley value in a substrate I would expect that value would be used to make sure you stay out of that range of roughness, who would want to create millions of failure prone parts due to a poor substrate, even if they would never truly fail? Now if we are looking for a glueon or some quarks in a detector I imagine things get rather crazy, but in normal electron flow is any of this an issue?
Before making a fool of yourself, brush up your quantum mechanics and also the numbers you are using in your napkin calculations.
I'm not ready to discuss with you anything that it is obviously tainted by your hidden agenda. Like your claim of audio amplifiers being non minimum phase systems, instead of stubbornly trying to prove anything that doesn't match your understanding as wrong, you would be much better spending your time in finding the knowledge and understanding you lack.
Text books from my university are not in the public domain, sorry.
Please have a look at this paper from Vishay.
http://www.ieee.li/pdf/essay/basics_of_linear_fixed_resistors.pdf
The 3th harmonic for metal film is< -110db which can be in the ~area of B. Hofer paper. The carbon composite is < 60 db
Am I wrong, or the topic was the "voltage coefficient" effect on the nonlinearity, not the overall linearity that obviously has much many contributors?
If you can prove that the "voltage coefficient" is responsible for the 110dB 3rd harmonic of metal film resistors, then please do so. If memory serves, the overall nonlinearities in metal film resistors are due:
a) terminations (transitions between two metals)
b) current crowding in thin film conduction (inhomegenities in very thin film lead to non-homogenous current conduction, which in turn triggers non-linearities at the material grain boundaries, under high injection). Is also dependent of the substrate quality, since the substrate roughness trigger nucleation in the film deposition process, leading to an non fully conformal/uniform metal layer.
c) thermal coefficient effects extended in the KHz range (the low pass filter is a single pole, therefore the effect decreases exponentially with frequency; if the thermal coefficient is high, the effect at KHz could be still measurable)
d) voltage coefficient (a non issue around -110dB, make that -140 to -150 dB)
e) Others, but with contributions close to the background noise of the Universe).
Many years ago, I studied up on resistors and what caused their distortion beyond lead metal/construction, thermal and voltage coeff. Some film formulations contain non-linear minerals.... ferrous. The formulations vary in content depending on resistor value range. So, you cant say a tested low distortion resistor necessarily applied to all values.
Beyond that is the temp coeff. The resistors are 'doped' with material to alter the TC and depending on the TC desired for that vaue, may also be a ferrous material. The doping material varies from type to type but - again- you cant say that all (zero) temp resistors are better than the 'undoped' ones which have higher TC.
Fortunately, the lowest distortion over the widest range of common values has been worked out with formulations and lead material/attachment et al. Its all done. But who is it? Its Dale. (Recently absorbed by Vishey.) Their thru-hole are low cost and have all the issues sorted for lowest distortion.
Now many of the sm resistors available have been tested and found their formulas are generally not as good as the thru-hole Dale type. So, A-P and others still use the thru-hole Dales (and a few others).
THx-RNMarsh
Beyond that is the temp coeff. The resistors are 'doped' with material to alter the TC and depending on the TC desired for that vaue, may also be a ferrous material. The doping material varies from type to type but - again- you cant say that all (zero) temp resistors are better than the 'undoped' ones which have higher TC.
Fortunately, the lowest distortion over the widest range of common values has been worked out with formulations and lead material/attachment et al. Its all done. But who is it? Its Dale. (Recently absorbed by Vishey.) Their thru-hole are low cost and have all the issues sorted for lowest distortion.
Now many of the sm resistors available have been tested and found their formulas are generally not as good as the thru-hole Dale type. So, A-P and others still use the thru-hole Dales (and a few others).
THx-RNMarsh
Last edited:
How about less of a "yus guys are all morons" attitude and being more helpful.
Dave has asked a reasonable question. I too would like to know the textbooks your claims come from. Dave is likely to go out & buy them as he's interested in learning .. as most of us are.
Please have a look at this paper from Vishay.
http://www.ieee.li/pdf/essay/basics_of_linear_fixed_resistors.pdf
The 3th harmonic for metal film is< -110db which can be in the ~area of B. Hofer paper. The carbon composite is < 60 db
JPV
Hi JPV,
Thanks for bringing this paper to my attention.
I read it, but did not see much helpful in regard to resistor voltage nonlinearity.
There was something about nonlinearity and 3rd harmonic on page 4, but there was no discussion of the voltage levels use for test. Similarly, in the table on page 5, where some numbers regarding nonlinearity are shown, there is no mention of the voltage applied.
I'm doing some more tests at a higher voltage level across the resistor (8-10V rms) using a 2X booster with an LM4562 on the output of my THD analyzer. This may help to get more definitive results, especially for the carbon film device, by getting some H3 further above the noise. I want to get more reliable data on how the distortion falls with applied level.
On the same board I built the 2X buffer, I used the other half of the 4562 to build an 11X non-inverting amplifier with the DUT as the feedback resistor. This should be a good sanity check, since in this configuration the H3 is compressive, whereas in the 11:1 attenuator arrangement H3 is expansive.
I am testing at 1kHz, but plan to test at 500Hz and 2kHz to see if any thermal distortion is creeping into the measurements.
Cheers,
Bob
The data sheet specifies extra digits in the model number system ('143') for "Non-magnetic". The leads in the standard version appear to be copper, so can we assume that magnetic end caps are used where '143' is not specified, or is this referring to the metal film, itself?
At 5ppm/V (according to the data sheet) those resistors may very well qualify as crap. Take a look at these and note they are SMD down to 0402 in size (1mm x 0.5mm). I would be though very surprised if they would meet the spec'd voltage coefficient of 0.1ppm/V in the small sized resistors, but rather in the larger ones (from 1206 and up).
Or these, adding to the already impressive specs above a 5ppm/C tempco and .02% tolerance.
may be? could be. I am going by measured results only. wanna be? I am sure there are others as good. but at what cost and availability. The Dales are outstanding for use in the best audio for 1% tolerance parts. Assume (!) magnetic end caps? Why all the speculation?
Whats with all the second guessing. If you dont trust what i told you as fact.... I started using them when LLNL engineers/scientists evaluated a variety of mfr and learned from their independant research and tests Dale mfilm were the best and so they are the store stocked item for R&D.
Use them without hesitation and enjoy the results. (BTW -- Audio Precision also uses Dale th in thier analog section...... sm elsewhere).
I would buy anything from Vishey, if required/needed, as they represent the best in their field. Which is why they bought Dale to cover th (thru-hole) met-film resistor markets.
Thx-RNMarsh
Whats with all the second guessing. If you dont trust what i told you as fact.... I started using them when LLNL engineers/scientists evaluated a variety of mfr and learned from their independant research and tests Dale mfilm were the best and so they are the store stocked item for R&D.
Use them without hesitation and enjoy the results. (BTW -- Audio Precision also uses Dale th in thier analog section...... sm elsewhere).
I would buy anything from Vishey, if required/needed, as they represent the best in their field. Which is why they bought Dale to cover th (thru-hole) met-film resistor markets.
Thx-RNMarsh
Last edited:
Hi Richard,
I looked at the RN60 datasheet, and these resistors are definitely not crap. Having said that, and in the context of the voltage coefficient discussion here, 5 ppm/V is quite a large voltage coefficient. I'm thinking there are 2 possibilities here. First of all, the datasheet number could be a mistake or they just threw out a very conservative number, perhaps just in response to an industry or mil spec requirement. They might be very, very much better in reality.
Secondly, if they really are several ppm/V this does not make them crap for many applications, where they may excel in other regards. However, in that case they aren't great for critical analog applications that require low voltage nonlinearity.
I'm guessing that if AP uses these in their analog circuits (not the least bit surprized through-hole) then they are likely much, much better than 5 ppm/V and that AP themselves verified this.
Cheers,
Bob
One might consider 1ppm as a max level of VC. But as noted before, there is not a single Vc number for all mf resistor values. it depends on the value. So, you get shown a number that represents the entire max range of VC values.
Anyway, for 1ppm with the Dale RN type at say 2ppm would only require two in series to get 1ppm. If your's were 1ppm then two would leave you with 0.5ppm. But, I think TC will be a larger issue and especially if you are not careful of thermal differences at each end (thermocouple effect). Position in the air flow will be critical if you want to achieve the full potential of the TC spec.
And, if pulsed power were an app (music might qualify) then R changes under pulsed conditions vary by type.
THx-RNMarsh
Anyway, for 1ppm with the Dale RN type at say 2ppm would only require two in series to get 1ppm. If your's were 1ppm then two would leave you with 0.5ppm. But, I think TC will be a larger issue and especially if you are not careful of thermal differences at each end (thermocouple effect). Position in the air flow will be critical if you want to achieve the full potential of the TC spec.
And, if pulsed power were an app (music might qualify) then R changes under pulsed conditions vary by type.
THx-RNMarsh
Last edited:
View attachment VFR_TN107.pdf
The bulk foil and wire wound are extreamly stable. And you can add a counter winding for the ww type to cancel inductance. But they are made of nichrome-nickle/chrom alloys ---- hardly the best if you want lowest distortion.
Which brings you back to the best MF types... like Dale.
TH-RNMarsh