Hi millwood:
>in absolute terms how high/low is the THD for carbon resistors?<
I normally don't think that lumped THD offers much insight into the hows and whys of a circuit's performance, and in this case the relevent equations suggested that a certain level of modulation should have been going on, in addition to the THD. So by no means do I think that the THD readings paint the complete picture, and therefore I wouldn't want to attempt to extrapolate those into any kind of prediction of listening tests. Still, at the end of the day, the THD provided an easy, compact graph that I could at least look at and say "There, that's _something_ concrete."
On a 5kohm load, the line level preamp circuit in question had a best-case THD measurement of about 0.00015% at an output level of around 26Vrms for 20Hz and 1kHz. The THD for 20kHz was just a little worse - about 0.00017%, also at around 26~27Vrms. The clipping point was a little over 30Vrms. That's with a NFB network comprised of a 5kohm series return Vishay S102 metal-foil resistor and a 560ohm shunt to ground, again a Vishay S102.
Using 1/4W carbon-films resulted in reasonably similar distortion measurements up to about 9Vrms (where the distortion was about 0.00035%), but whereas the distortion of the metal-foil version kept dropping steadily after 9V, the distortion of the carbon took off upwards until it exceeded 0.003% at 30Vrms, shortly after which the circuit clipped. So while the first part of the THD curves looked sort of similar for both the metal-foil and carbon-film versions, the latter part looked totally different.
>if, for example, after the 30db hit, it is still below audible levels, I don't think anyone can make an argument that they can hear that (by definition).<
I'm trying to make the best possible circuit that I can, and if I can predict, calculate or observe some area of the performance that I feel unhappy with, I work on it until I'm happy with the results. If the outcome is audible, fine, but if not, it is still an intellectual and engineering challenge and I'm still going to do my best to overcome the problem.
>what happened to other cases where the carbon resistor didn't impair THD by 30db?<
I don't know if I fully understand your question, but if it is what I think it is, the primary difference was in the amount of voltage applied across the carbon-film resistor (in reality, I think that the issue was how much power was being dissipated through the resistor.)
>what is the mechanism in the carbon resistors that did the 30db damage?<
Equations and the calculator suggested that the temperature coefficient was figuring strongly. To test this theory, I subsequently changed the 1/4W carbon-film resistor to something that would have have a better tempco and be able to dissipate power better - i.e., two pieces of 1/2W metal film in parallel. Now the THD curves looked a lot closer to the Vishays, suggesting that the tempco was indeed partly to blame. Not quite as good performance as the S102s, mind you, but at least close enough that I no longer felt like rubbing my eyes and swearing.
regards, jonathan carr
>in absolute terms how high/low is the THD for carbon resistors?<
I normally don't think that lumped THD offers much insight into the hows and whys of a circuit's performance, and in this case the relevent equations suggested that a certain level of modulation should have been going on, in addition to the THD. So by no means do I think that the THD readings paint the complete picture, and therefore I wouldn't want to attempt to extrapolate those into any kind of prediction of listening tests. Still, at the end of the day, the THD provided an easy, compact graph that I could at least look at and say "There, that's _something_ concrete."
On a 5kohm load, the line level preamp circuit in question had a best-case THD measurement of about 0.00015% at an output level of around 26Vrms for 20Hz and 1kHz. The THD for 20kHz was just a little worse - about 0.00017%, also at around 26~27Vrms. The clipping point was a little over 30Vrms. That's with a NFB network comprised of a 5kohm series return Vishay S102 metal-foil resistor and a 560ohm shunt to ground, again a Vishay S102.
Using 1/4W carbon-films resulted in reasonably similar distortion measurements up to about 9Vrms (where the distortion was about 0.00035%), but whereas the distortion of the metal-foil version kept dropping steadily after 9V, the distortion of the carbon took off upwards until it exceeded 0.003% at 30Vrms, shortly after which the circuit clipped. So while the first part of the THD curves looked sort of similar for both the metal-foil and carbon-film versions, the latter part looked totally different.
>if, for example, after the 30db hit, it is still below audible levels, I don't think anyone can make an argument that they can hear that (by definition).<
I'm trying to make the best possible circuit that I can, and if I can predict, calculate or observe some area of the performance that I feel unhappy with, I work on it until I'm happy with the results. If the outcome is audible, fine, but if not, it is still an intellectual and engineering challenge and I'm still going to do my best to overcome the problem.
>what happened to other cases where the carbon resistor didn't impair THD by 30db?<
I don't know if I fully understand your question, but if it is what I think it is, the primary difference was in the amount of voltage applied across the carbon-film resistor (in reality, I think that the issue was how much power was being dissipated through the resistor.)
>what is the mechanism in the carbon resistors that did the 30db damage?<
Equations and the calculator suggested that the temperature coefficient was figuring strongly. To test this theory, I subsequently changed the 1/4W carbon-film resistor to something that would have have a better tempco and be able to dissipate power better - i.e., two pieces of 1/2W metal film in parallel. Now the THD curves looked a lot closer to the Vishays, suggesting that the tempco was indeed partly to blame. Not quite as good performance as the S102s, mind you, but at least close enough that I no longer felt like rubbing my eyes and swearing.
regards, jonathan carr
IanHarvey said:
it is hard to imagine a case where the value of a resistor will cause distortion.
On the other hand, the non-resistor attributes of a resistor, like inductance for a wire-wound resistor, may contribute to distortion. but it is hard to imagine a carbon resistor will have such a trait.
Either way, it sounds quite interesting and I look forward to more info from Jonathan on this subject.
jcarr said:
first of all, the 30db is between 0.003% and 0.00015%, right? I am not sure if any one of us can hear the difference between 0003% and 0.00015% thd but that is a differernt story.
it seems to me that the value of the carbon resistor is accurate as both circuits clip'd at around 30v.
also, the carbon resistor is working well within its power rating (0.18w at 30vrms for the 5k resistor). I suppose that the resistor is reasonably sized and it is hard to imagine that it would have trouble dissipating that little power and react, thermally, to 20khz current flow.
I am struggling to understand and explain the increase in thd, and how/why beefier resistors cured it. so there may be other reasons that will cause measurable impact from resistors. except that I don't know why that would be the case but I will take your experiment as proof of such an impact.
thanks for sharing.
I totally agree. With that low of a level of THD its way out of normal hearing resolution.Just look at the THD of even a good driver?I realize that perfection is the goal but what i have found out is that maybe 1% of the population can really tell the difference in the last 10% of sound quality.I know this is vague but as i have ppl over to hear my system and help with A/B testing (free beer helps) as the sound gets more refined the A/B testing with a given number of ppl gets more iffy and the % of agreement drops.
ron
ron
Hugh
There's a SilmicII to replace it, I recently had some samples of these from Elna Japan.
Just buying the things is the hard part!
This is, I feel, part of the reason for the cynicism prevalant in threads like these - the Elna capacitor MOQ's are for two thousand parts minimum here in the UK, so the investment this represents to anyone wishing to use them is VAST. To a reseller a range of maybe 20 values represents an outlay of maybe six thousand pounds, which may sit on the shelf for years.
It's not therefore surprising that an audio grade capacitor like a Silmic that I can buy for 0.12 ea., then sells for ten times that, as one would have to spend several hundred pounds per value required.
There are very real differences between ALL electronic components, not just audio grade ones and these differences can matter, under ideal circumstances.
How they matter is down to specific circuits and application - understand this and you may find enlightenment. There is no magic though - Black Gates or whatever are not some magic cure-all for every audio disease, they will even make things worse in certain cases.
Andy.
There's a SilmicII to replace it, I recently had some samples of these from Elna Japan.
Just buying the things is the hard part!
This is, I feel, part of the reason for the cynicism prevalant in threads like these - the Elna capacitor MOQ's are for two thousand parts minimum here in the UK, so the investment this represents to anyone wishing to use them is VAST. To a reseller a range of maybe 20 values represents an outlay of maybe six thousand pounds, which may sit on the shelf for years.
It's not therefore surprising that an audio grade capacitor like a Silmic that I can buy for 0.12 ea., then sells for ten times that, as one would have to spend several hundred pounds per value required.
There are very real differences between ALL electronic components, not just audio grade ones and these differences can matter, under ideal circumstances.
How they matter is down to specific circuits and application - understand this and you may find enlightenment. There is no magic though - Black Gates or whatever are not some magic cure-all for every audio disease, they will even make things worse in certain cases.
Andy.
jcarr said:Hi millwood:
[snip]On a 5kohm load, the line level preamp circuit in question had a best-case THD measurement of about 0.00015% at an output level of around 26Vrms for 20Hz and 1kHz. The THD for 20kHz was just a little worse - about 0.00017%, also at around 26~27Vrms. The clipping point was a little over 30Vrms. That's with a NFB network comprised of a 5kohm series return Vishay S102 metal-foil resistor and a 560ohm shunt to ground, again a Vishay S102.
Using 1/4W carbon-films resulted in reasonably similar distortion measurements up to about 9Vrms (where the distortion was about 0.00035%), but whereas the distortion of the metal-foil version kept dropping steadily after 9V, the distortion of the carbon took off upwards until it exceeded 0.003% at 30Vrms, shortly after which the circuit clipped. So while the first part of the THD curves looked sort of similar for both the metal-foil and carbon-film versions, the latter part looked totally different.
[snip]Equations and the calculator suggested that the temperature coefficient was figuring strongly. To test this theory, I subsequently changed the 1/4W carbon-film resistor to something that would have have a better tempco and be able to dissipate power better - i.e., two pieces of 1/2W metal film in parallel. Now the THD curves looked a lot closer to the Vishays, suggesting that the tempco was indeed partly to blame. Not quite as good performance as the S102s, mind you, but at least close enough that I no longer felt like rubbing my eyes and swearing.
regards, jonathan carr
Jonathan,
What about the higher noise figure of the carbons vs the metal foils? Would it be possible that at the low levels, they were similar because the noise was the dominant factor? Then you raise the level, improve the S/N ratio of the circuit, but since the carbon noise is larger it doesn't climb that far out of the noise as the metals do, so to speak?
Jan Didden
DrewP
Your questions,
Logical reasons for me to spend my money on them instaed of buying CD's?
Let me explain
I'm ok with spending money on decent caps like GE 41L series, but are their any reason to pay x10 more for Hovland?
Power resistors like caddock MP and MK series i think is worth it and they don't come much better than that, small footprint, non-inductive, decent precision, stable and with heatsinks they could handle alot of power. But I don't see any reason for searching hard to find exotic and crazy expensive hand made wire wound power resistors.
Low noise Opamp's like BB and AD, I do care for, because the specifications is a lot better.
And so on..
"Some people put resources into buying or building constant current sources for every stage, some into ultra low noise regulation, some into only valve rectification, some battery power supplies, some battery biasing"
I could agree with this, because this is design decisions and not "trying to be cool useing the weirdest exotic components"
this is what I think, maybe their is a reason to use them, I really don't know.
The subject seams sensetive to some so I have just decided to order some of thoose BG's that everyone seams to defend and check it out for myself. Some other things that you guys recomend me to try?
Magnus
Your questions,
Logical reasons for me to spend my money on them instaed of buying CD's?
Let me explain
I'm ok with spending money on decent caps like GE 41L series, but are their any reason to pay x10 more for Hovland?
Power resistors like caddock MP and MK series i think is worth it and they don't come much better than that, small footprint, non-inductive, decent precision, stable and with heatsinks they could handle alot of power. But I don't see any reason for searching hard to find exotic and crazy expensive hand made wire wound power resistors.
Low noise Opamp's like BB and AD, I do care for, because the specifications is a lot better.
And so on..
"Some people put resources into buying or building constant current sources for every stage, some into ultra low noise regulation, some into only valve rectification, some battery power supplies, some battery biasing"
I could agree with this, because this is design decisions and not "trying to be cool useing the weirdest exotic components"
this is what I think, maybe their is a reason to use them, I really don't know.
The subject seams sensetive to some so I have just decided to order some of thoose BG's that everyone seams to defend and check it out for myself. Some other things that you guys recomend me to try?
Magnus
what is interesting is that there is no shortage of people who swear by carbon resistors or carbon composition resistors (which have the worst noise performance) because they sound better than film types, 🙂.
I guess what sounds good/better depends critically which forum you visit and what birds congregate at that forum, 🙂.
I guess what sounds good/better depends critically which forum you visit and what birds congregate at that forum, 🙂.
millwood said:
Parasitic inductance or capacitance is linear, and does not therefore produce distortion; it will only cause such things as wobbles in the frequency response. (I justified this in more detail on the recent loudspeaker cables thread).
Thanks, Jonathan, for the info on the preamp - real measurements and hard facts are so much more persuasive. For instance, we might conjecture that carbon resistors change slightly in value depending on applied voltage; if this is the case it's not too hard to build a test circuit to measure it. One could then publish results for every resistor under the sun without any accusations of subjectivity creeping it.
Cheers
IH
if it is indeed resistor noise that caused thd spike, one would expect this to be contact noise as it is dependent on voltage (thus current) across the resistor, and 2) resistor type. If it were thermal noise, it wouldn't have matter if Jonathan had used a metal / carbon resistor.
still, it is interesting to see how high of a spike it is.
still, it is interesting to see how high of a spike it is.
Hi Magnus,
I work in WLA, live in Brentwood, so we must be neighbors.
I am currently building a gainclone with "boutique" parts. Using Jensen 4 pole electrolytic and Blackgate caps. Resistors will either be Caddocks, Riken, Tantalums or Resista's depending on what I end up picking.
Maybe we could get together some time and compare projects? I know a few other audio guys in the area, but nobody else into DIY stuff.
BTW, rest of my system is a EVS modded SA-14, Stan Warren modded Adcom 5200, DIY transformer volume control using SB tx102's, and Tyler Acoustics Taylo Reference monitors. All cables and PC's are DIY.
Randy
I work in WLA, live in Brentwood, so we must be neighbors.
I am currently building a gainclone with "boutique" parts. Using Jensen 4 pole electrolytic and Blackgate caps. Resistors will either be Caddocks, Riken, Tantalums or Resista's depending on what I end up picking.
Maybe we could get together some time and compare projects? I know a few other audio guys in the area, but nobody else into DIY stuff.
BTW, rest of my system is a EVS modded SA-14, Stan Warren modded Adcom 5200, DIY transformer volume control using SB tx102's, and Tyler Acoustics Taylo Reference monitors. All cables and PC's are DIY.
Randy
Hi,
I bet we are neighbors when we live on Montana and 20th
, and soon less neighbors moving down to Euclid and Montana.
Sure we could, would be fun to compair your boutique GC to one of my cheapo parrallel almost for free GC and see how much better yours are and if it would be worth the bucks to start using boutique parts.
/Magnus
I bet we are neighbors when we live on Montana and 20th
, and soon less neighbors moving down to Euclid and Montana.Sure we could, would be fun to compair your boutique GC to one of my cheapo parrallel almost for free GC and see how much better yours are and if it would be worth the bucks to start using boutique parts.
/Magnus
NoKnowledge said:Hi,
I bet we are neighbors when we live on Montana and 20th
Sure we could, would be fun to compair your boutique GC to one of my cheapo parrallel almost for free GC and see how much better yours are and if it would be worth the bucks to start using boutique parts.
/Magnus
I live on Pesquera, which is in the hills north of Sunset, off of Westridge, off of Mandeville canyon. 20th is close, Euclid as, but not far either.
I hope to finish my gainclone in the next couple of weeks, but then it will need a few weeks to break in. I'll let you know when it's ready.
Randy
OK let me know when it is ready, my cheapo stuff doesn't need to break in, most of my components are prebreaked in on Salvation army
Adjö Senapsfrö
/Magnus
Adjö Senapsfrö
/Magnus
Hi,
Swedish?
The first word, adjö, sounds like the french adieu.
The other one will require some first grade parts to figure out....assuming I got the first one right...
Cheers,😉
Swedish?
The first word, adjö, sounds like the french adieu.
The other one will require some first grade parts to figure out....assuming I got the first one right...
Cheers,😉
Yes Swedish!
Your guess is right!
senapsfrö = mustard seed, so I guess have thoose first grade parts here!😉
/Magnus
Your guess is right!
senapsfrö = mustard seed, so I guess have thoose first grade parts here!😉
/Magnus
IanHarvey said:
I agree on this. It is interesting to note though, that the usual
models for dielectric absorption (DA) in capacitors are linear,
yet a lot of people insist DA causes distorsion. It has never
become clear to me if they use the term distorsion more liberally
or if they do not accept the DA models used even by Marsh&Jung.
All resistors are voltage dependent. Quoting some figures from
the fact pages in the Elfa catalogue we have:
carbon comp. 200 - 500 ppm/V
carbon film <100 ppm/V
metal film c:a 1 ppm/V
metal oxide < 10 ppm/V
wire wound c:a 1 ppm/V
They also say (not surprisingly) that this dependency
causes distorsion and that it is often stated as a ratio in dB
between the signal and the third overtone. I have yet to see
such a figure in a datasheet though. (Elfa don't even have
datasheets for the resistors they sell. 🙁 ).
Hi,
Not that I mind, but after about 14 months of forum activity it dawns on me that this is a cyclic event...
No one ever learned anything and those who actually did do the hands on job keep quiet about it.
In the meantime the oldtimers, just have to rehash the old stuff.
Was it Fred Dieckmann who first used the Deja Vu description?
Yes, Marsh and Jung did a good analysis on caps but how many have actually read the papers?
Not many from what I read.
And no, there's no way you can do without coupling caps just as you can't do without a PS so, why not go for the best?
Cheers,😉
Not that I mind, but after about 14 months of forum activity it dawns on me that this is a cyclic event...
No one ever learned anything and those who actually did do the hands on job keep quiet about it.
In the meantime the oldtimers, just have to rehash the old stuff.
Was it Fred Dieckmann who first used the Deja Vu description?
Yes, Marsh and Jung did a good analysis on caps but how many have actually read the papers?
Not many from what I read.
And no, there's no way you can do without coupling caps just as you can't do without a PS so, why not go for the best?
Cheers,😉
fdegrove said:
"where he built
the celebrated
mechanical cow
in which the queen
hid herself
in order to seduce the bull
her perverse passion
and from that
unnatural union
was born
the minotaur
a man
with the head and neck
of a bull "
Yeah? I don't think so.
Enjoy.
Christer:
>They also say that this dependency causes distorsion and that it is often stated as a ratio in dB between the signal and the third overtone.<
This sounds right. I am fairly sure that the prime distortion component that I was looking at from the carbon resistors was third-order.
Jan:
>What about the higher noise figure of the carbons vs the metal foils? Would it be possible that at the low levels, they were similar because the noise was the dominant factor? Then you raise the level, improve the S/N ratio of the circuit, but since the carbon noise is larger it doesn't climb that far out of the noise as the metals do?<
Seems plausible that this could contribute something, but I wouldn't know how large the magnitude would be.
It wasn't as though the carbons and metal-films showed different distortion curves altogether. Out to about 9Vrms, the respective distortion figures dropped at reasonably similar rates. But at higher output voltages, the results looked like they came from different circuits. I would be willing to entertain the possibility that the 0.00035% at 9Vrms reading implied some kind of noise threshold.
The particular circuit that I was using was quite handy for this kind of testing where you want to study voltage dependencies (it had a leading and trailing slew rate of 340V/us, and could swing full power out to past 1MHz). But I've since disassembled it (there were things about it that I didn't like), and the line-level circuits that I have now aren't designed to swing that kind of voltage.
But I will certainly intend to make a circuit with similar or greater output voltage capability in the future, and at that time, I may choose to look at resistors again.
Jan, MM sounds really interesting. But right now my hands are full with a complex board layout project. Would it be OK if I emailed you back in a week or so? I'd appreciate the extra time.
best regards, jonathan carr
>They also say that this dependency causes distorsion and that it is often stated as a ratio in dB between the signal and the third overtone.<
This sounds right. I am fairly sure that the prime distortion component that I was looking at from the carbon resistors was third-order.
Jan:
>What about the higher noise figure of the carbons vs the metal foils? Would it be possible that at the low levels, they were similar because the noise was the dominant factor? Then you raise the level, improve the S/N ratio of the circuit, but since the carbon noise is larger it doesn't climb that far out of the noise as the metals do?<
Seems plausible that this could contribute something, but I wouldn't know how large the magnitude would be.
It wasn't as though the carbons and metal-films showed different distortion curves altogether. Out to about 9Vrms, the respective distortion figures dropped at reasonably similar rates. But at higher output voltages, the results looked like they came from different circuits. I would be willing to entertain the possibility that the 0.00035% at 9Vrms reading implied some kind of noise threshold.
The particular circuit that I was using was quite handy for this kind of testing where you want to study voltage dependencies (it had a leading and trailing slew rate of 340V/us, and could swing full power out to past 1MHz). But I've since disassembled it (there were things about it that I didn't like), and the line-level circuits that I have now aren't designed to swing that kind of voltage.
But I will certainly intend to make a circuit with similar or greater output voltage capability in the future, and at that time, I may choose to look at resistors again.
Jan, MM sounds really interesting. But right now my hands are full with a complex board layout project. Would it be OK if I emailed you back in a week or so? I'd appreciate the extra time.
best regards, jonathan carr
- Status
- Not open for further replies.