For a feedback resistor, it is good to select a low tempco one. That resistor has the full output voltage across it, and with low frequencies and/or strong musical bursts that could heat the resistor on peaks enough that it slightly changes its value.
If it changes its value over the signal cycle that means distortion.
So get a low tempco one, or take a physically large resistor that heats less quickly.
Advanced: if you have a feedback with say 20k and 1k resistors, if you make that from 21 identical resistors of 1k, it can be shown that the temperature effects cancel and thermal distortion is absent.
Jan
If it changes its value over the signal cycle that means distortion.
So get a low tempco one, or take a physically large resistor that heats less quickly.
Advanced: if you have a feedback with say 20k and 1k resistors, if you make that from 21 identical resistors of 1k, it can be shown that the temperature effects cancel and thermal distortion is absent.
Jan
Last edited:
Please note that the CMF55 in that document is the E grade (25ppm). I believe the D grade (100ppm) will not perform as good! But I agree that it doesn't appear to have much to do with audio, as many low noise resistors are imo not 'good sounding'. (The Susumu RG seems to have good review in subjective perceived performance).
In designing, instead of trying to use low noise resistors, I try to use low value resistors and opt for good sounding types. The biggest resistance is usually in the feedback path, which is around 10k-11k for Voltage Feedback Amplifiers.
... can we agree that from a scientific point of view it seems like that people who are able to listen differences between cheap but good resistors and high end ones are hearing things ?
I would change "hearing things" to "experiencing confirmation bias", but the end result is pretty much the same.
Last edited:
Ahhh... I missed that. Thanks for pointing it out.
I use mostly 'H' (50ppm), but that will still narrow the difference a bit.
Cheers,
Jeff.
I would change "hearing things" to "experiencing confirmation bias", but the end result is pretty much the same.
Thanks !
I generally agree with spaceistheplace (SIP) with the noted exceptions:
TI[TE] YR1 and YR2 - not known
Vishay MMA0204 and MBB0207 - Magnetic
Holco H4 / H8 - Magnetic new, non-magnetic old
Dale / Vishay CMF and RN - non-magnetic
Dale / Vishay PTF - not known
Welwyn RC55Y - magnetic
Use the largest size which can fit and choose best precision and tempco in more critical locations (unless also used as a fuse to protect other components).
For those cost conscious Vishay Beyschlag seems the best deal going for a 50ppm 1% resistor.
https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf
Good research, sadly the author could have done a favor to everyone
by listing the resistors in low to high noise order. Due to screen color
temp variations and duplicate colors used it's challenging to figure out
which were the best, worst / better than, worse than resistors.
http://conradhoffman.com/papers_lib/TI_Noise_Prec_Resistors.pdf
Boltzman's constand keep coming up over and over again in the
literature. Good job Boltzman!
Cheers,
TI[TE] YR1 and YR2 - not known
Vishay MMA0204 and MBB0207 - Magnetic
Holco H4 / H8 - Magnetic new, non-magnetic old
Dale / Vishay CMF and RN - non-magnetic
Dale / Vishay PTF - not known
Welwyn RC55Y - magnetic
Use the largest size which can fit and choose best precision and tempco in more critical locations (unless also used as a fuse to protect other components).
For those cost conscious Vishay Beyschlag seems the best deal going for a 50ppm 1% resistor.
https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf
Good research, sadly the author could have done a favor to everyone
by listing the resistors in low to high noise order. Due to screen color
temp variations and duplicate colors used it's challenging to figure out
which were the best, worst / better than, worse than resistors.
http://conradhoffman.com/papers_lib/TI_Noise_Prec_Resistors.pdf
Boltzman's constand keep coming up over and over again in the
literature. Good job Boltzman!
Cheers,
Last edited:
Another issue that always seems to be left out of the equation is how the
resistors are oriented relative to the thermal source. The resistors should be
parallel to the thermal source to keep equal temperature along the resistive element.
Mounting resistors perpendicular to thermal source results in differences
in temperature along the resistive element leading to more noise and more
distortion--oh my.
Cheers,
resistors are oriented relative to the thermal source. The resistors should be
parallel to the thermal source to keep equal temperature along the resistive element.
Mounting resistors perpendicular to thermal source results in differences
in temperature along the resistive element leading to more noise and more
distortion--oh my.
Cheers,
Zoom the graphics to 300% and the difference in stylization becomes clearer.
Are you sure Welwyn RC55Y are magnetic?
Datasheet says:
"Construction
A metal film is deposited onto a high quality ceramic former.
Nickel-plated steel caps are force fitted to the former and
termination wires are welded to the caps.
The resistor is adjusted to value by a helical cut in the film
and the body is protected with a specially formulated epoxy
coating.
Terminations
Material Solder coated copper wire"
Anybody does this? It's about the only real technical issue discussed here that has an observable impact.
BTW I was wondering why anyone would start a thread 'Resistor opinion'. Why would my 'opinion' be of any interest? Why not ask for facts and figures about resistors?
Jan
That goes in the right way, but only partly. Your set of four is equivalent to 20k but each gets 1/4th the power, but compared to the 1k to ground (in this example) the 20k's still change 5 times as much in resistance value over temperature than the 1k.
In the case of 20 x 1k + 1k, which comes from Bruce Hofer btw, the lead designer of Audio precision at the time, all 1k's see exactly the same change and that means that the 20k to 1k ration is exactly maintained despite the resistor change.
For example, let us say that each resistor changes 1% due to temp modulation (it will be much less, but just for discussion). That means that the 20k becomes say 20k+1% = 20k200, and the 1k becomes 1k+1% = 1k01. But, and this is the crux, the ratio which was 20/1 to begin with, is now 20.200/1.01 which is equal to - 20/1! So the distortion-causing temperature modulation is exactly cancelled. Neat, huh?
Jan
Last edited:
Yes, Jan, really.
Ha, ha, I see you answered it and I saw that after I replied.
Well your explanation is good also.
Bob Pease maybe? Nuts, now I have to think about where i found that.
I tried to explain it. Someone who was doing resistor research
discussed it. I remember reading it, but can't recall the source.
It has to do with the resistor being evenly heated, or the same
temperature from end to end. Perhaps different temperatures
cause more distortion or noise due to thermal issues.
Isn't that why we want lower temp co with resistors?
So imagine a higher temp co resistor with one end close
to a heat sink and the other end farther away from the
heat sink - the resistor now had various temperature with
in it. Yadda, yadda, and yadda.
If I find it (the article) again I will post. Maybe Ed or someone knows it,
George is pretty good at finding stuff too.
Cheers,
Last edited:
It's still 4x better thermally, so the tracking delta is notably smaller between Rf and Rg (and, IIRC any voltage terms are usually V^2, so that 1/2 peak voltage is great) and much easier to package with less stray capacitance.
They're all along a spectrum of compromises vs spec needs.
This is a case of gooder vs gooder-er in terms of matching tempcos.Also the thermal gradients from everything else on the PCB are going to swamp out the Rf vs Rg imbalance at some point -- obviously depending on layout/case/etc
Usually tempco spec is similar for the same resistor type. So 20k or 1k they may have the same spec, say 1%/K, so no different, the gain is still the same either one 20k or 20x1k. But I think this is funny. Talking about the effect of minutes changes only to say at the end that the resistor's sound differences is BS, FUD.
Microscopically, resistors have (a) Voltage coeff (b) Power/thermal coeff (c) stray capacitance (d) inductance (e) Johnson's noise (f) current and shot noise... But how this individually will affect perception is still uncertain, let alone together. So I relied on ears and experimented with series and parallel networks. With parallel network, the coefficients should be more or less halved, but the resistance is twice increased (Johnson's noise!). For feedback, I compared 0.6W size Rf: 11k, 2x 22k and 3x 33k. I preferred 2x 22k. 3x 33k was most of the time unacceptable (noisy).