Looking at the NF vs Rg plot of the datasheet of 2sk389, 2sk170 and 2sk117, it indicates that to get lowest noise figures, the signal source resistance should be around 10kohms. So doesn't it seem to mean that if you have an output impedance of say 100 ohms from your CD player and you are designing a low-noise preamp using jfets, you should put a resistor of 10kohms at the gate of the jfet to mimic a 10k source resistance ? This is counter-intuitive since that 10kohm will add voltage noise at the input that will be gained up by the preamp.
Let's phrase the question another way. What should I do at the input of the preamp to get the lowest noise performance based on the NF vs Rg plot from the datasheet ?
Thanks,
rlim
Let's phrase the question another way. What should I do at the input of the preamp to get the lowest noise performance based on the NF vs Rg plot from the datasheet ?
Thanks,
rlim
General:
> to get lowest noise figures, the signal source resistance should be around 10kohms. So doesn't it seem to mean that ... you should put a resistor of 10kohms at the gate of the jfet to mimic a 10k source resistance?
No.
It means that if you have a perfect transformer, you wind it to step-up or step-down so the amplifier "sees" 10K.
If you had a 100Ω pickup, and that perfect transformer, a 1:10 voltage ratio (1:100 impedance ratio) would give higher source voltage and thus better Signal/Noise ratio.
There are no perfect audio transformers.
Imperfect transformers are widely used for moving-coil phono and microphone inputs. MC phono (and also ribbon mikes) have such low-Z low-Voltage outputs that it is hard to make an amplifier work great directly, and the transformer's flaws may be less than the amplifier's flaws. Transformers are also handy for balancing and floating long lines in professional studios, even when the noise-resistance is not a big issue. However very-high-Gm low-R BJTs and JFETs have been replacing transformers in most applications. Ribbon mikes are still transformered, MC phono may be one or the other, and dynamic mike preamps are transformerless except when high isolation is needed, or last-dB noise is needed, or when transformer coloration is desired.
Specific:
> if you have an output impedance of say 100 ohms from your CD player and you are designing a low-noise preamp using jfets,
Then I bet you will find that you can ignore noise calculations, that a simple design will wind up at far lower noise than the CD noise level or the noise level of the post-DAC amp (if any).
Very roughly: home hi-fi CD players deliver 2VRMS, noise (digital grit) is 96dB lower, or 32 microVolts. Even a TL071 does 2µV. Generic 10/$1 JFETs are similar, if you keep them above 1mA or so. Yes, you may want to keep preamp noise "far" below source self-noise, but I would think that 2 µV is "far" below 32µV.
For "best" noise, use a 100Ω:10KΩ transformer. The CD-format peak signal level is transformed up to 20VRMS, the noise/grit level is at 320µV, and maybe your FET noise stays at 2µV. The 100Ω transformed to 10KΩ adds another 2µV. So you have 20VRMS signal, 320µV digital grit, 4µV source and device noise. This isn't an audible change in noise from the no-transformer situation. And you will probably have overload trouble on 20VRMS (28Vpk) signals.
> to get lowest noise figures, the signal source resistance should be around 10kohms. So doesn't it seem to mean that ... you should put a resistor of 10kohms at the gate of the jfet to mimic a 10k source resistance?
No.
It means that if you have a perfect transformer, you wind it to step-up or step-down so the amplifier "sees" 10K.
If you had a 100Ω pickup, and that perfect transformer, a 1:10 voltage ratio (1:100 impedance ratio) would give higher source voltage and thus better Signal/Noise ratio.
There are no perfect audio transformers.
Imperfect transformers are widely used for moving-coil phono and microphone inputs. MC phono (and also ribbon mikes) have such low-Z low-Voltage outputs that it is hard to make an amplifier work great directly, and the transformer's flaws may be less than the amplifier's flaws. Transformers are also handy for balancing and floating long lines in professional studios, even when the noise-resistance is not a big issue. However very-high-Gm low-R BJTs and JFETs have been replacing transformers in most applications. Ribbon mikes are still transformered, MC phono may be one or the other, and dynamic mike preamps are transformerless except when high isolation is needed, or last-dB noise is needed, or when transformer coloration is desired.
Specific:
> if you have an output impedance of say 100 ohms from your CD player and you are designing a low-noise preamp using jfets,
Then I bet you will find that you can ignore noise calculations, that a simple design will wind up at far lower noise than the CD noise level or the noise level of the post-DAC amp (if any).
Very roughly: home hi-fi CD players deliver 2VRMS, noise (digital grit) is 96dB lower, or 32 microVolts. Even a TL071 does 2µV. Generic 10/$1 JFETs are similar, if you keep them above 1mA or so. Yes, you may want to keep preamp noise "far" below source self-noise, but I would think that 2 µV is "far" below 32µV.
For "best" noise, use a 100Ω:10KΩ transformer. The CD-format peak signal level is transformed up to 20VRMS, the noise/grit level is at 320µV, and maybe your FET noise stays at 2µV. The 100Ω transformed to 10KΩ adds another 2µV. So you have 20VRMS signal, 320µV digital grit, 4µV source and device noise. This isn't an audible change in noise from the no-transformer situation. And you will probably have overload trouble on 20VRMS (28Vpk) signals.
No wonder Mr. Curl gets impatient
I have no idea what the last post was about and the only was about and the only bit of sanity in it was the remark about the Jfets noise be small enough to be irelevent at line level signals. Noise figure is really only useful as figure of merit in RF design.
What the curve means is that the noise contribution of the jfet refered to the input is less than the thermal noise of the source resistance for points on the curve below 0 dB. Look at the E sub N figures if youwant to know what the noise voltages are. The impedances in line level circuit are chiosen for bandwith and distotion. Low impedances for better bandwidth but not low enoungh to increase the distortion of the circuits driving them.
http://www.eutelsat.com/glossary/8_1_2.html
Noise Figure A method for quantifying the electrical noise generated by a practical device. The noise figure is the ratio of the noise power at the output of a device to the noise power at the input to the device, where the input noise temperature is equal to the reference temperature (290 K). The noise figure is usually expressed in decibels.
http://www.rf-amplifiers.com/index.php?topic=noise_figure
I have no idea what the last post was about and the only was about and the only bit of sanity in it was the remark about the Jfets noise be small enough to be irelevent at line level signals. Noise figure is really only useful as figure of merit in RF design.
What the curve means is that the noise contribution of the jfet refered to the input is less than the thermal noise of the source resistance for points on the curve below 0 dB. Look at the E sub N figures if youwant to know what the noise voltages are. The impedances in line level circuit are chiosen for bandwith and distotion. Low impedances for better bandwidth but not low enoungh to increase the distortion of the circuits driving them.
http://www.eutelsat.com/glossary/8_1_2.html
Noise Figure A method for quantifying the electrical noise generated by a practical device. The noise figure is the ratio of the noise power at the output of a device to the noise power at the input to the device, where the input noise temperature is equal to the reference temperature (290 K). The noise figure is usually expressed in decibels.
http://www.rf-amplifiers.com/index.php?topic=noise_figure
Both of the above threads are correct, but NF is almost useless with audio jfets. I know, I know, they still put it on the spec sheet, BUT it is a hangover from low noise bipolar transistors, where it really makes a difference. All of the fets that you mentioned are pretty quiet. This means that they have low voltage noise. If you are transformer coupled, then adding gain to the transformer will give you a better noise figure, because you are then operating at a higher input level. The tradeoff is the output impedance of the transformer, which means virtually NOTHING to both tubes and fets, BUT is VERY important with bipolar transistors. This is because bipolar transistors have an extra noise source in the base emitter junction that will go bonkers when you have too much input source impedance. I hope that this helps.
I agree with John, actually....
As my background is RF design, NF is an important concept there, as it also realizes that components further down stream in the RF chain can also contribute noise. In those situations, you need to know the gain, and NF, of every part in the chain.
I have never worried about NF at audio, period.
But.....having said that.......
"Never" is an awful long time, John. I suppose that if you put a 10 ohm gate damper into one those Toshiba JFETs that we both use, the noise contribution of that individual stage would go up 2 dB or so, compared to a dead short on the input.
However, in the real world, the average guy here will not have a perfect, zero ohm source. The use of a gate damper, when judisciously chosen, will not have a staggering effect on noise contribution. In fact, in a lot of "real world" applications, its effect may almost be totally swamped out. Not completely, but almost.
Mind you, I am not talking about doing something obviously ridiculous, like putting a 1K gate damper in a MC phono preamp.
Jocko
As my background is RF design, NF is an important concept there, as it also realizes that components further down stream in the RF chain can also contribute noise. In those situations, you need to know the gain, and NF, of every part in the chain.
I have never worried about NF at audio, period.
But.....having said that.......
"Never" is an awful long time, John. I suppose that if you put a 10 ohm gate damper into one those Toshiba JFETs that we both use, the noise contribution of that individual stage would go up 2 dB or so, compared to a dead short on the input.
However, in the real world, the average guy here will not have a perfect, zero ohm source. The use of a gate damper, when judisciously chosen, will not have a staggering effect on noise contribution. In fact, in a lot of "real world" applications, its effect may almost be totally swamped out. Not completely, but almost.
Mind you, I am not talking about doing something obviously ridiculous, like putting a 1K gate damper in a MC phono preamp.
Jocko
Hi,
Good, I know pretty well what Jocko means and what John means.
The net result of these gatestopper/gridstopper craze is that I see an awful lot of people in both transistor based and tube based designs using values that are quite likely to not not only damp gates but also severely reduce bandwidth for no reason.
My advice is, if you don't have a scope and you don't experience oscillation, keep gatestopper/gridstopper values to a minimum if not you pay twice: you lose bandwidth and information and instead of gaining S/N you'll lose it even more.
With CDP replay, curtailing frequency response is one way to mask a problem but when it hits way below the 20Kc mark it also becomes a problem when countermeasures are overdone.
Concluding I'd recommend tackling RF right at the source trying to prevent it from entering the audiocircuit in the first place.
Cheers,
Good, I know pretty well what Jocko means and what John means.
The net result of these gatestopper/gridstopper craze is that I see an awful lot of people in both transistor based and tube based designs using values that are quite likely to not not only damp gates but also severely reduce bandwidth for no reason.
My advice is, if you don't have a scope and you don't experience oscillation, keep gatestopper/gridstopper values to a minimum if not you pay twice: you lose bandwidth and information and instead of gaining S/N you'll lose it even more.
With CDP replay, curtailing frequency response is one way to mask a problem but when it hits way below the 20Kc mark it also becomes a problem when countermeasures are overdone.
Concluding I'd recommend tackling RF right at the source trying to prevent it from entering the audiocircuit in the first place.
Cheers,
Horses for courses
30 years ago we didn't have cell phones and 1Ghz PCs putting RFI into everything.
I think I will take a hit of a few db in noise for the RF suppresion and lack of ringing from high Q resonate circuits at RF frequencies. It is hardly a secret that good RF design techniques pay pretty good dividends even in audio frequency circuits.
Throw digital audio into the equation and you had better be designing for what goes on at RF frequencies with equal importance to how the circuit performs at audio frequencies. RFI will not go away just because you ignor thinking about it. It gets worse the more RF spectrum devices go into use every day.
30 years ago we didn't have cell phones and 1Ghz PCs putting RFI into everything.
I think I will take a hit of a few db in noise for the RF suppresion and lack of ringing from high Q resonate circuits at RF frequencies. It is hardly a secret that good RF design techniques pay pretty good dividends even in audio frequency circuits.
Throw digital audio into the equation and you had better be designing for what goes on at RF frequencies with equal importance to how the circuit performs at audio frequencies. RFI will not go away just because you ignor thinking about it. It gets worse the more RF spectrum devices go into use every day.
Hi,
So, since we all are so paranoid about RFI we should all use stoppers to reduce audio bandwidth to reduce BW to 15Kc now?
Com'on Fred, you know what I mean, I see all too many circuits with values of gridstoppers that make using HQ OPTs in tube circuits absolutely a mere waste of money, not to mention the noise penalty.
Let's get a proper balance and some perspective here, shall we?
That's all I have to say about it, plus I'll send Jocko and you a recipe for bolognese sause that's neither plaster nor H2O + tomatoes.
Deal?
So, since we all are so paranoid about RFI we should all use stoppers to reduce audio bandwidth to reduce BW to 15Kc now?
Com'on Fred, you know what I mean, I see all too many circuits with values of gridstoppers that make using HQ OPTs in tube circuits absolutely a mere waste of money, not to mention the noise penalty.
Let's get a proper balance and some perspective here, shall we?
That's all I have to say about it, plus I'll send Jocko and you a recipe for bolognese sause that's neither plaster nor H2O + tomatoes.
Deal?
Hi,
Don't know how severe the problem is at your end_ it's been a while since I visited your lovely country_ but it seems the problem is much less severe here in Europe.
I can have a cellphone and run a PC simultaneously, it doesn't seem to upset the audio system, nor does one upset the other.
Not that I don't prefer listening to the sysrem with the whole shebang shut down, but other than my own gear? No, no one else's stuff seems to interfere in any audible way.
Maybe some EU regulation took care of it at the source?
Cheers,
Don't know how severe the problem is at your end_ it's been a while since I visited your lovely country_ but it seems the problem is much less severe here in Europe.
I can have a cellphone and run a PC simultaneously, it doesn't seem to upset the audio system, nor does one upset the other.
Not that I don't prefer listening to the sysrem with the whole shebang shut down, but other than my own gear? No, no one else's stuff seems to interfere in any audible way.
Maybe some EU regulation took care of it at the source?
Cheers,
maximum acceptable resistor at input ?
Thanks everyone for your replies. So what is considered an acceptable maximum resistor at the input in order to keep noise levels in the weeds ? In particular, I am referring to a volume pot that's in front of the preamp, and if you have a 50k pot, there is going to be potentially 12.5k equivalent input resistance if you are mid-scale. Am I thinking about this correctly ?
Thanks again,
Richard
Thanks everyone for your replies. So what is considered an acceptable maximum resistor at the input in order to keep noise levels in the weeds ? In particular, I am referring to a volume pot that's in front of the preamp, and if you have a 50k pot, there is going to be potentially 12.5k equivalent input resistance if you are mid-scale. Am I thinking about this correctly ?
Thanks again,
Richard
It depends on what is in front of it. The odds are, the noise contribution of the preceeding stage.......let's say that is a phono stage, will dominate the noise level of the system.
If it is a CDP, which should have a lower noise contribution, then, yes, if it is too large when compared to the driving source impedance, it may raise the noise somewhat.
While it is easy to say that the lower the better, at some point lowering it will not make it any quieter. It really does depend on the application.
You may want to consider how changing that volume control will effect the frequency respsonse of the circuit.
Engineering is a series of compromises, and it is up to the designer to decide what parameters take priority. Any change made to maximize one, will most likely degrade another.
So...........
I normally use a 10K pot, just to keep the maximum resistance down to "an acceptable level" In which case, the highest R to ground would be 2.5K
Assuming zero ohm source impedance, which most likely, will never be the case.
One last digression:
Gate damper:
The one time that I did leave it out........by accident....the circuit oscillated. So, one went in, pronto. Really had nothing to do with RF immunity.
I prefer JFETs because of their inherent advantge over BJTs in this respect.
Jocko
If it is a CDP, which should have a lower noise contribution, then, yes, if it is too large when compared to the driving source impedance, it may raise the noise somewhat.
While it is easy to say that the lower the better, at some point lowering it will not make it any quieter. It really does depend on the application.
You may want to consider how changing that volume control will effect the frequency respsonse of the circuit.
Engineering is a series of compromises, and it is up to the designer to decide what parameters take priority. Any change made to maximize one, will most likely degrade another.
So...........
I normally use a 10K pot, just to keep the maximum resistance down to "an acceptable level" In which case, the highest R to ground would be 2.5K
Assuming zero ohm source impedance, which most likely, will never be the case.
One last digression:
Gate damper:
The one time that I did leave it out........by accident....the circuit oscillated. So, one went in, pronto. Really had nothing to do with RF immunity.
I prefer JFETs because of their inherent advantge over BJTs in this respect.
Jocko
Eye to eye on RFI
"Com'on Fred, you know what I mean"
Guess not..... I have never seen a gate stopper used limit the bandwidth to audio frequencies.
BTW We have RFI regulation requirements
here as well. I believe the US first established them in fact. In fact I did EMI compliance engineering as part of my telecom jobs. Having seen tricks like spread spectrum and knowing how sensitive audio equipment is to RFI, I don't really call it paranoia. Think of it as sound engineering practices.
FRed dIeckmann
"Com'on Fred, you know what I mean"
Guess not..... I have never seen a gate stopper used limit the bandwidth to audio frequencies.
BTW We have RFI regulation requirements
here as well. I believe the US first established them in fact. In fact I did EMI compliance engineering as part of my telecom jobs. Having seen tricks like spread spectrum and knowing how sensitive audio equipment is to RFI, I don't really call it paranoia. Think of it as sound engineering practices.
FRed dIeckmann
To go back to the original question
To understand more simply the responses to your original question, you need to understand what you have done, if you add the 10k you suggested.
In effect all you will have done is raise the noise level of the source, in order to make the JFET look better.
It's not too hard to see that this won't work, this is why NF can be a confusing parameter for beginners.
Andy.
To understand more simply the responses to your original question, you need to understand what you have done, if you add the 10k you suggested.
In effect all you will have done is raise the noise level of the source, in order to make the JFET look better.
It's not too hard to see that this won't work, this is why NF can be a confusing parameter for beginners.
Andy.
I agree entirely with PRR. You can only get a noise improvement if you manage to increase the source impedance to 10kohm without reducing the source's own signal to noise ratio, and the only way to do that is with an (ideal) transformer. Even then, the improvement is negligible in a line input stage.
By the way, noise figure of an amplifier is the ratio between the noise of the signal source plus the noise of the amplifier and the noise of the signal source alone, assuming that the source has a noise temperature of 290K (that is, that its noise is equal to the thermal noise of (the real part of) its output impedance at 290K). Values below 0dB are impossible. Values below 3.01dB indicate that the amplifier generates less noise than the source.
I think that, if interpreted well, noise figure can be a very useful figure of merit whenever you have a source generating thermal noise with a temperature near 290K. It shows you exactly how far your circuit is from perfection. Good audio examples of sources generating thermal noise with a temperature near 290K are a gramophone cartridge not playing any records or a dynamic microphone in a perfectly silent environment.
However, CD players and cartridges of gramophones which do play records generate more noise than just thermal noise, making the noise figure rather useless.
By the way, noise figure of an amplifier is the ratio between the noise of the signal source plus the noise of the amplifier and the noise of the signal source alone, assuming that the source has a noise temperature of 290K (that is, that its noise is equal to the thermal noise of (the real part of) its output impedance at 290K). Values below 0dB are impossible. Values below 3.01dB indicate that the amplifier generates less noise than the source.
I think that, if interpreted well, noise figure can be a very useful figure of merit whenever you have a source generating thermal noise with a temperature near 290K. It shows you exactly how far your circuit is from perfection. Good audio examples of sources generating thermal noise with a temperature near 290K are a gramophone cartridge not playing any records or a dynamic microphone in a perfectly silent environment.
However, CD players and cartridges of gramophones which do play records generate more noise than just thermal noise, making the noise figure rather useless.
Maybe Grey or Steve can explain it.
"I agree entirely with PRR. You can only get a noise improvement if you manage to increase the source impedance to 10kohm without reducing the source's own signal to noise ratio, and the only way to do that is with an (ideal) transformer. Even then, the improvement is negligible in a line input stage."
You both as confused as a barking bird. This is total nonsense. Go do some reading, both of you......
"I agree entirely with PRR. You can only get a noise improvement if you manage to increase the source impedance to 10kohm without reducing the source's own signal to noise ratio, and the only way to do that is with an (ideal) transformer. Even then, the improvement is negligible in a line input stage."
You both as confused as a barking bird. This is total nonsense. Go do some reading, both of you......
that one was Curls.........
http://www.diyaudio.com/forums/showthread.php?postid=280822#post280822
Take it up with John Curl. Maybe he can explain it to you.......... or maybe not.
http://www.diyaudio.com/forums/showthread.php?postid=280822#post280822
Take it up with John Curl. Maybe he can explain it to you.......... or maybe not.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.