davidsrsb,
IMHO, you are correct. The performance is more important than the 'pretty' factor.
It's important to keep the components as close as possible to the active devices.
IMHO, you are correct. The performance is more important than the 'pretty' factor.
It's important to keep the components as close as possible to the active devices.
Actually one resistor problem is non-uniform heating! If one end is placed closer to say a heatsink than the other end there will be a thermocouple voltage produced at the connections of the resistor!
Other than that close is good, but when doing an accurate circuit analysis be sure to include RLC models for the interconnection system!
Other than that close is good, but when doing an accurate circuit analysis be sure to include RLC models for the interconnection system!
It is really simple any two dissimilar metal when they touch and there is a temperature differential create a thermo electric voltage. If the resistor is uniformly heated one end will counteract the other end. if it is not there will be a small voltage generated across the resistor. It's on the web somewhere about low voltage T & M relay contacts.
And how do you make your resistors? Many are thin metal film on ceramic core, steel endcaps and copper leads. Wirewound are all metallic contact.
Or you could just tale a resistor hook it up to a DVM and apply a soldering iron to one end.
This one is true, and it can be a problem when processing µV or nV DC voltages.
But let's be honest: where in an audio amplifier would a parasitic voltage of even 1 full µV be troublesome?
Audio deals with AC signals (although some people claim they are able to hear DC), and the worst such a voltage could do is upset the bias.
Except of course that such an effect would be a drop in a sea of other sources of drift, tempco's, etc
I just did the experiment with several resistors. There was a change in resistance. There was no d.c. voltage produced. A thermocouple produces a voltage.
Resistors I have measured show 200 uv typical some go way beyond that few a bit less. So how many resistors do you use? What is the gain following? What do you think the temperature differential does to noise?
Not sure how you did it. I used 1/4W 1% metal film resistors of the type used in much audio gear. At 100 ohms the typical offset was 200 - 400 uv with the iron tip on one end. Placing it on the other end would yield the inverse voltage.
200µV seems huge. That's something like ten times what you would get for 1°C delta with the infamous kovar/copper couple.
Are you sure you don't apply the temperature delta to other parts of the circuit? With thermocouple effects, such distinctions can be tricky.
Anyway, the temperature differential is not important to noise. The absolute temperature of a resistor is significant, and if the two halves of a resistor are at different temperatures, their average must be taken into account.
I do understand and I will experiment with more resistors.
We older guys learn something new every day.
To be honest though .... who would place a circuit which is sensitive to temperature variations next to something which produces enough heat to melt solder?
Thanks simon7000!
We older guys learn something new every day.
To be honest though .... who would place a circuit which is sensitive to temperature variations next to something which produces enough heat to melt solder?
Thanks simon7000!
Lets talk about the main global feedback resistor! OK 400 UV would be extreme, unless someone finds an "Audiophile" special then all bets are off as to how bad it could be!
So if we place the resistor with one end near the heatsink we might have a 50 UV offset that changes with volume.
Could you hear that? Probably not. Is it real simple to turn the resistor sideways to completely avoid the problem? Of course! Then there is no longer the golden ears issue of if you could hear it!
So the conclusion is that good layout is good layout and the more issues you are aware of the better a job you can do.
I have made some tests with various resistors of different technologies and origins. All 100 ohm.
The heating method I used was crude and poorly defined quantitatively: apply the tip of a soldering iron to one side.
But I took great care no to heat eventual parasitic thermocouple junctions.
I estimate that I could reach temperature delta between the two ends of tens of degrees C, at least, perhaps over 100°C.
As expected, the voltages generated were tiny: max 20 to 30µV, and most of the times such extremes were reached for a very short time, during the heating period.
But.....
So far I had tested only resistors constructed with end caps, both carbon and metal film.
In the end, I tried one more, this time with a flush body, the ceramic being metallized at the ends.
I think this construction was used by Sfernice, among others.
I had a big surprise: even before any heating, there was already a shift of almost 100µV.
As I approached the soldering iron, without even touching it, the readings went mad, and after the contact was made, it went to several millivolts.
It was a metal-film, 1%, from a reputable manufacturer...
All the other, including the crappy carbon ones from China showed only negligible effects.
I think I'll make further tests....
The heating method I used was crude and poorly defined quantitatively: apply the tip of a soldering iron to one side.
But I took great care no to heat eventual parasitic thermocouple junctions.
I estimate that I could reach temperature delta between the two ends of tens of degrees C, at least, perhaps over 100°C.
As expected, the voltages generated were tiny: max 20 to 30µV, and most of the times such extremes were reached for a very short time, during the heating period.
But.....
So far I had tested only resistors constructed with end caps, both carbon and metal film.
In the end, I tried one more, this time with a flush body, the ceramic being metallized at the ends.
I think this construction was used by Sfernice, among others.
I had a big surprise: even before any heating, there was already a shift of almost 100µV.
As I approached the soldering iron, without even touching it, the readings went mad, and after the contact was made, it went to several millivolts.
It was a metal-film, 1%, from a reputable manufacturer...
All the other, including the crappy carbon ones from China showed only negligible effects.
I think I'll make further tests....
Actually some of the cheap Xicon resistors showed better results on some of my tests that much more expensive resistors!
The actual formula for the resistance material is sort of a cooking secret and different manufacturers have different results.
One of the things you may find is that some resistors that have unusual results are sometimes touted as sounding different and therefore better!
The actual formula for the resistance material is sort of a cooking secret and different manufacturers have different results.
One of the things you may find is that some resistors that have unusual results are sometimes touted as sounding different and therefore better!
Why am I not surprised at this?
Temperature differentials across a PCB do generate thermocouple voltages across resistors, but this is swamped by what this would do to a discrete transistor LTP. Remember 25mV/C
I made some systematic tests on a large number of resistors, both from my stock and from various boards I have in my junk.
There are very few bad apples.
Here is the summary:
-The construction without metal end caps (metallized ceramic) is far from being systematically synonymous with poor performance, but all the worst offenders belong to that category.
-Some older metal-oxide precision types (as opposed to more recent metal-film) have a moderate but measurable thermal voltage.
-Recent, metal-film with end caps have no thermal effect whatsoever.
Here is the pic of the three worst offenders:
The green one is the absolute worst, followed closely by the dark one. Both came from old HP equipement.
The third one is intermediate. I think it was made by Vishay. Could someone confirm this (and also identify the origin of the other two)?
I didn't give quantitative data, because the heating was, well ... not very quantitative.
There are very few bad apples.
Here is the summary:
-The construction without metal end caps (metallized ceramic) is far from being systematically synonymous with poor performance, but all the worst offenders belong to that category.
-Some older metal-oxide precision types (as opposed to more recent metal-film) have a moderate but measurable thermal voltage.
-Recent, metal-film with end caps have no thermal effect whatsoever.
Here is the pic of the three worst offenders:
The green one is the absolute worst, followed closely by the dark one. Both came from old HP equipement.
The third one is intermediate. I think it was made by Vishay. Could someone confirm this (and also identify the origin of the other two)?
I didn't give quantitative data, because the heating was, well ... not very quantitative.
Attachments
Now all you need to do is wait a few years and retest some of the same samples to see what age has to do with the issue!
(Good work!)
(Good work!)
The third one is a Vishay bulk metal foil, really more of a miniaturized wirewound than a metal film. It should be extremely stable compared to most metal films, let alone other types.
- Status
- Not open for further replies.