Voltage Distortion
While I was rereading parts of the Groner paper, I noticed that he quotes measured voltage distortion by type of resistor. He said that foil and wirewound resistors "showed distortion levels of -150 dBc" ...... "Metal film parts gave distortion levels of less than
-140 dBc at +20 dBu; with the expected 12 dB reduction this would provide adequate performance for the oscillator design." With a footnote that these measurements were taken on 1k resistors and a higher resistance typically has higher voltage coefficient.
The -12 dB voltage distortion in the first post was only from replacing a single resistor with a Quad of identical resistors.
Unfortunately, I don't have any voltage coefficient information about metal-Ox and thick film resistors. Texas Components does quote a voltage coefficient of <0.1 ppm/V which converts to less than -140 dB. That seems in line with Groner's measurements. I also don't have any spec from Susumu on their KRL, but it is a foil resistors, so it should be pretty good.
For me, the story is less about voltage distortion than power distortion while keeping an eye on current noise differences too.
While I was rereading parts of the Groner paper, I noticed that he quotes measured voltage distortion by type of resistor. He said that foil and wirewound resistors "showed distortion levels of -150 dBc" ...... "Metal film parts gave distortion levels of less than
-140 dBc at +20 dBu; with the expected 12 dB reduction this would provide adequate performance for the oscillator design." With a footnote that these measurements were taken on 1k resistors and a higher resistance typically has higher voltage coefficient.
The -12 dB voltage distortion in the first post was only from replacing a single resistor with a Quad of identical resistors.
Unfortunately, I don't have any voltage coefficient information about metal-Ox and thick film resistors. Texas Components does quote a voltage coefficient of <0.1 ppm/V which converts to less than -140 dB. That seems in line with Groner's measurements. I also don't have any spec from Susumu on their KRL, but it is a foil resistors, so it should be pretty good.
For me, the story is less about voltage distortion than power distortion while keeping an eye on current noise differences too.
Last edited:
If the main concern is voltage distortion, then 4 series resistors is a valid solution. In the less than 1 Ohm values, it may be hard to find resistors at 1/4 the value as a power resistors with good material properties, but it certainly is possible. The other concern would be heat dissipation to keep power distortion low. One approach to using 4 series resistors would be to do something like my Quad pcb, but on a one side clad board. Then you would have two of these boards in series on the circuit with 2 resistors on each pcb and the same heat sink area as I was using in my diy Quads.
That is correct but separating them is difficult. It is hard to increase the power in a resistor and keep the temperature constant especially on a dynamic basis. You end up with a complicated multi-physics problem. It is easier when the two effects are far apart. This is the case with the old light bulb stabilized oscillator the temperature coefficient is large but the voltage coefficient is what actually stabilizes the oscillator.
As for excess noise this is a definitive study. https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf
Trying to imagine what Vishay would charge for a four chip Z-foil resistor array, especially in a hermetically sealed oil-filled package such as their several VH- series just to get an even more ultra-stable resistor.
I know I couldn't afford to use them.
http://www.vishaypg.com/docs/63146/vhp203.pdf
I know I couldn't afford to use them.
http://www.vishaypg.com/docs/63146/vhp203.pdf
Last edited:
I agree. I quoted the same paper in the opening post to reference current noise. Although every resistor I would want measurements for is not in that paper, there are an amazing number of resistors commonly used in audio which made it an excellent learning experience for me.
I would think so too. Just trying to think of a semi-plausible way to "improve" performance in the face of multiple effects when violating Occam's Razor rule. I think the only criticism of bulk foil resistor performance I've read about is power coefficient, where a well-chosen wirewound resistor might be a slight improvement in certain cases. Trying to measure distortion effects below -140 db seems to mean chasing multiple effects when trying to build a practical application circuit.
How far do we have to go before hitting quantum noise? There's got to be a rock bottom somewhere!
--Damon
There are a few relevant references in Letters | Linear Audio
One concern with interpreting resistor temperature variation is the issue of temperature gradient within the part, and the measurement of a relevant 'temperature' with respect to ambient. smt packaged parts could likely have a significant temperature variation from solder tab to the mid-line if the part was being operated at a significant % of dissipation rating. TO-220 packages may well have a similar internal Rjc situation to that of a semiconductor die within.
Custom heatsinking of individual resistors may be more applicable to larger leaded parts than using two in series. As indicated by the series smt custom heatsink mechanism, a hassle with series connections of parts can be an inability for heat to transfer away from the middle junction(s). Leaded parts that are not subject to high withstand voltage to ground could benefit from a form of heatsinking the non-conductive coating, although relying on the coating itself for electrical insulation would be a worry. Or just the practical design of a pcb could benefit from multi-layers and mulitple vias to better the thermal management of key parts.
One concern with interpreting resistor temperature variation is the issue of temperature gradient within the part, and the measurement of a relevant 'temperature' with respect to ambient. smt packaged parts could likely have a significant temperature variation from solder tab to the mid-line if the part was being operated at a significant % of dissipation rating. TO-220 packages may well have a similar internal Rjc situation to that of a semiconductor die within.
Custom heatsinking of individual resistors may be more applicable to larger leaded parts than using two in series. As indicated by the series smt custom heatsink mechanism, a hassle with series connections of parts can be an inability for heat to transfer away from the middle junction(s). Leaded parts that are not subject to high withstand voltage to ground could benefit from a form of heatsinking the non-conductive coating, although relying on the coating itself for electrical insulation would be a worry. Or just the practical design of a pcb could benefit from multi-layers and mulitple vias to better the thermal management of key parts.
Last edited:
Thanks for that link. I checked out the list and I will have some interesting reading to do.
I completely agree. I would add that not all smt designs are equal in this regard. I read somewhere that, in a typical smt resistor, 80% of the heat transfer is to the pcb trace because the encapsulation on top is a thermal insulator. Other designs work hard at improving heat transfer to both air and pcb. In the end, a careful designer has to look at each resistor design, in all types, and the layout for handling the power. I admit that I used to just pick the power rating and ignore the rest.
As anti is saying, I wonder if we wouldn't be better off using soldier mounting (vertical) for axial power resistors. The inductance might be a little higher, but I would expect the heat transfer from a vertical cylinder to be better than a horizontal one. Plus we would have more lead exposed and leads can do a good job of heat transfer. Finally, heatsinking of the body might be more easier.
When I was learning about electronics from my father as a kid, he told me that -30 dB was something you could hear, but very quiet. -70 dB was a great noise floor. For years I believed that. Then I heard the difference that -90 or -100 could make and was shocked.
I don't know where the bottom is, but I keep trying to learn.
Better convection may help with power dissipation, but will make very little difference to any thermal distortion because for that you need to consider the instantaneous temperature changes of the resistive element, and that is some distance in thermal terms from the surface temperature.lehmanhill said:
Some of us might be shocked if you can demonstrate that you can genuinely and reliably hear down there, especially when not in a specially prepared 'quiet room'.
Inspired by Samuel, who uses Beyschlag MELF in series-parallel in his low-distortion oscillator design :
http://www.diyaudio.com/forums/equi...-measurement-amp-ikoflexer-6.html#post5123470
Patrick
http://www.diyaudio.com/forums/equi...-measurement-amp-ikoflexer-6.html#post5123470
Patrick
In modular DAC's vertical mounted standard resistors can exhibit a differential thermocouple effect due to the heat flowing preferentially out the board end. Many standard semiconductor manufacturing materials are not pure metals and you can get 10-100uV/C in some cases from even bond wire terminations.
Patrick,
I should have given you credit for starting me on this configuration. A diyaudio friend pointed me to your posts and you pointed me to Groner. Thanks. It has been an interesting learning experience.
Jac
I would be shocked as well. I am not claiming to be able to hear -90 dB sounds. But the first time I heard a modern amp, not knowing any of it's specs, I was amazed how clear and detailed the music sounded. One of the characteristics of that amp was a noise floor about -90 dB. Since then, I have found some level of correlation between a low noise floor (among other things) and sweet sounding music. That's all.
Perhaps.
A plate resistor typically has a circa 500V rating requirement, which often dictates a 2W resistor at least. Maybe a 0.6W or 1W resistor with 350V rating is just acceptable. That quad of resistors look a bit like 0.6W style - so maybe that quad is more about spreading heat-dissipation, and economics of manufacture by using a common part on the pcb.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.