Interesting app note on JFET noise

TheFettler said:
Those of you 'in the know' will doubtless have already read this, but I thought it was of general interest:


Low Noise JFETS - Superior Performance to Bipolars?

http://www.vishay.com/document/70599/70599.pdf

ray

yes, I have seen similiar curves, but for older devices, in the big Siliconix books on JFETs. I recommend reading that book. It gives lots of insights in the various types of JFETs: long-channel, short-channel ...

To see better comparisons for JFETs and bipolars, noise-wise, see the famous Motchenbacher/Fitchen: Low Noise Electronics.

Only the Japanese are able to produce JFETs with noise as low as 1nV/Sqrrt(Hz), and these are only the 2SK170 and its siblings. On the other side there are hundreds of bipolars, who are able to, at 1mAmps, these are old like 2N4033, or standard, like 2SC2240, or high-tech, like MAT-02.

The main difference in noise is, that there is virtually no current noise in JFETs, making them optimally suited for high-impedance source. When designing a line stage with bipolars in the input, one should avoid pots over 10k, whereas with JFETs, one could go up to 100k and above. (considering only noise - bandwidth and miller caps are a different subject).

For moving magnet cartridge input stages (they normally have app. 300 Ohms resistance and app. 1k impedance), you can use both types equally good, if you confine yourself to 250uAmps collector current in the input stage, when using bipolars.

Even for moving coils, you can use both. The lowest noise MC phono stage, the Linn Linto with >80dB signal/noise ratio, has a bipolar input. But a 2SK170 would do also it in practical use.

Though, getting substantially below 1nV/Sqrrt(Hz) is difficult. For ultra low output moving coils with output at 0.1mV like the big Allaerts or the big Audio Note, only transformers will do.

regards,
Hartmut
 
The big big disadvantage is the supply of JFET's. Can be hard to get and the part has

http://www.mcmelectronics.com/welcome.jhtml
http://www.ceitron.com/

Jfets are grouped for Idss by part number. Matched differential pairs are available for low offset. They are no harder to design with than BJTs. This has been discussed on the forum several times. Maybe some people should read the forum instead of just posting........

H.H.

P.S. "Only the Japanese are able to produce JFETs with noise as low as 1nV/Sqrrt(Hz)" Oh really?................

http://www.interfet.com/pdf/DS_IFN147.pdf
 
How noisy are Hexfets?

By comparison, how noisy are hexfets like IRF610 that Nelson Pass uses in his low level amps? I think this is interesting because they use millions of parallel cells and so I expect some of the noise should statistically cancel because the individual contributions of each cell are uncorrelated, i.e. random in phase with respect to each other.

GP.
 
Re: How noisy are Hexfets?

Circlotron said:
By comparison, how noisy are hexfets like IRF610 that Nelson Pass uses in his low level amps? I think this is interesting because they use millions of parallel cells and so I expect some of the noise should statistically cancel because the individual contributions of each cell are uncorrelated, i.e. random in phase with respect to each other.

GP.

Ciclotron,

MOSFETS are a lot more noisy than JFETs. You can use them in the inputs of line stages only, if you have the volume pot behind the line stage, thus attenuating the noise when attenuating the music.

For MM or even MC inputs, they are absolutely no-no.

But you can you use them everywhere in intermediate (read VAS) or output stages of amplifying blocks, just because the noise of the input is dominant in most cases.

Have a look at the Pass Ono/Xono phono stages, JFETs are dominating that field.

regards,
Hartmut