Hi all.
I have build quite a bit with BJTs and opamps over the time, but I feel it's time to play with JFets, now that I'm in the planning phase of a preamp.
I struggle to understand noise figure on JFETs. My local supplier have the following:
2N5459: Noise figure: 3,0dB af 1KHz; Rg = 1mohm
BF245A: Noise figure: 1,5dB at 100 MHz; RG =1kohm
2SK246: No noise figure in datasheet
2SK117: Noise figure: 1dB at 1khz; Rg = 1 kohm
The 2N5459 is quite cheep, while the 2SK117 is expensive, so what is enough? I'll need normal line-in and a RIAA for MM.
I have build quite a bit with BJTs and opamps over the time, but I feel it's time to play with JFets, now that I'm in the planning phase of a preamp.
I struggle to understand noise figure on JFETs. My local supplier have the following:
2N5459: Noise figure: 3,0dB af 1KHz; Rg = 1mohm
BF245A: Noise figure: 1,5dB at 100 MHz; RG =1kohm
2SK246: No noise figure in datasheet
2SK117: Noise figure: 1dB at 1khz; Rg = 1 kohm
The 2N5459 is quite cheep, while the 2SK117 is expensive, so what is enough? I'll need normal line-in and a RIAA for MM.
One pragmatic way to choose is to buy 100 of each; for those part-numbers which are too expensive to buy 100 -- cross them off the list. Then when the units arrive at your lab, test each of the part-numbers for matching and for noise. The lowest noise, best matching units are your winners. And you've already got 100 of them so you can start prototyping immediately.
The noise figure is by definition the ratio expressed in dB of the noise of the signal source and the amplifier together to the noise of the signal source alone, assuming that the signal source produces thermal noise at a temperature of 290 K.
Assuming that 1 mohm is a typo for 1 Mohm, 3 dB related to 1 Mohm means that the FET is as noisy as a 1 Mohm resistor, which is quite bad.
A spec at 100 MHz doesn't tell you much about the noise at audio frequencies, as FETs have excess noise at low frequencies.
1 dB related to 1 kohm at 1 kHz would mean a noise voltage equivalent to about 260 ohm, which is perfectly suitable for MM RIAA.
Aren't there any typical graphs in the datasheets that show noise densities versus frequency?
Assuming that 1 mohm is a typo for 1 Mohm, 3 dB related to 1 Mohm means that the FET is as noisy as a 1 Mohm resistor, which is quite bad.
A spec at 100 MHz doesn't tell you much about the noise at audio frequencies, as FETs have excess noise at low frequencies.
1 dB related to 1 kohm at 1 kHz would mean a noise voltage equivalent to about 260 ohm, which is perfectly suitable for MM RIAA.
Aren't there any typical graphs in the datasheets that show noise densities versus frequency?
Fet specs are confusing and subject to large variations. The NF has too many dependencies to be useful; the only figures I look at is the noise voltage vs frequency, if it's available and if I think I can trust it. It looks like this, and the lowest curve/item is about right for audio purpose:
To keep it simple, the ones to get are the obsolete 2SK170: there's a current group buy you may want to check out. Alternatively, the LSK170 available at the store are also good.
To keep it simple, the ones to get are the obsolete 2SK170: there's a current group buy you may want to check out. Alternatively, the LSK170 available at the store are also good.
NJFETs for Source Follower Applications
More FET noise measurements (for EUVL)
http://www.waynekirkwood.com/images...ns_Danyuk_Electronic_Design_April_19_2013.pdf
http://www.angelfire.com/az3/dimitri/images/AX_Dec2018_pp56-58.pdf
Completely spoiled for choices (as long as you do not want P-JFET).
No need to pay for unobtanium.
And 2SK209 = 2SK117; 2SK208 = 2SK246.
If you can still find, then 2SJ106 = 2SJ103.
And I normally buy 100x a time.
Good for matching, fewer rejects.
Patrick
More FET noise measurements (for EUVL)
http://www.waynekirkwood.com/images...ns_Danyuk_Electronic_Design_April_19_2013.pdf
http://www.angelfire.com/az3/dimitri/images/AX_Dec2018_pp56-58.pdf
Completely spoiled for choices (as long as you do not want P-JFET).
No need to pay for unobtanium.
And 2SK209 = 2SK117; 2SK208 = 2SK246.
If you can still find, then 2SJ106 = 2SJ103.
And I normally buy 100x a time.
Good for matching, fewer rejects.
Patrick
Take a look at the NSVJ3557 and NSVJ5908 (dual), NF=1dB, CPH5905 (15V Max)
If you like cascode, CPH6904
NSVJ3557 = 2SK3557
Sanyo's answer to BF862. Not as low noise. Very low Vds rating.
When biased at Idss, can only take even less Vds.
Only practical when cascoded.
NSVJ5908 = dual 2SK3557
Equally low Vds rating. No specs on matching.
I personally prefer 2SK209 and 2SK2145.
More versatile, and more practical Idss classes.
Very good Vds rating.
Patrick
Sanyo's answer to BF862. Not as low noise. Very low Vds rating.
When biased at Idss, can only take even less Vds.
Only practical when cascoded.
NSVJ5908 = dual 2SK3557
Equally low Vds rating. No specs on matching.
I personally prefer 2SK209 and 2SK2145.
More versatile, and more practical Idss classes.
Very good Vds rating.
Patrick
Thanks for all the inputs, and apologies to the OP for a light hijacking.
Let's push the boundaries a bit more: I'm working on a preamp based on the Toshiba 2SK146(dual)/2SK147(single), cascoded with a triode on top, and followed by a passive RIAA EQ. I have 5 duals in my stash, already broke one, so I'm looking for spares, if possible:
Let's push the boundaries a bit more: I'm working on a preamp based on the Toshiba 2SK146(dual)/2SK147(single), cascoded with a triode on top, and followed by a passive RIAA EQ. I have 5 duals in my stash, already broke one, so I'm looking for spares, if possible:
- N-channel
- <1nV/SQRT(Hz) (the K146 has about .73nV@5mA, and its 1/F corner freq. is about 400Hz)
- Ideal Idss: about 10mA, working Id is 5mA
- Yfs about 40mS
Attachments
Last edited:
Have a look at the LSK389 - B grade should fit.
If you can get them, the BF862 also a great device with <1nV/rt Hz and low 1/f. Dimitri did some measurements about 8yrs ago IIRC and it came out tops. VDC is 20 V though but you can always cascode.
If you can get them, the BF862 also a great device with <1nV/rt Hz and low 1/f. Dimitri did some measurements about 8yrs ago IIRC and it came out tops. VDC is 20 V though but you can always cascode.
I prefer 2SK372, 2SK369.
More FET noise measurements (for EUVL)
A simplified universal differential or single ended phono preamp
But if you need thermal tracking and hence duals, then my choice goes to 4x 2SK2145Y in parallel.
But the two source pins are tied together.
Patrick
More FET noise measurements (for EUVL)
A simplified universal differential or single ended phono preamp
But if you need thermal tracking and hence duals, then my choice goes to 4x 2SK2145Y in parallel.
But the two source pins are tied together.
Patrick
Here is a link to Dimitri’s measurements
Measurements Rate SMT Low-Voltage n-JFETs Under Consistent Conditions | Electronic Design
Measurements Rate SMT Low-Voltage n-JFETs Under Consistent Conditions | Electronic Design
Not really a JFET, but behaves much like one: DN2540.
http://ww1.microchip.com/downloads/en/DeviceDoc/DN2540%20B060313.pdf
It's actually a depletion mode N-mosfet, is cheap as chips and no worries about voltage breakdown (400V). It is a MOSFET, though, so best don't solder to it when other gear is connected to the circuit.
I don't have noise measurements or noise specs for DN2540, but have built some amplifiers with it and found it to be very quiet. Run at about 5 to 20mA, and cascode it to eliminate g-d capacitance variations with voltage. With a cascode and source degeneration resistor (about 300 ohms) to set its current, it will have extremelt high input impedance limited only by bias resistor and is capable of quite low distortion, too. (Running at higher Vds might work rather than a cascode, I didn't though).
http://ww1.microchip.com/downloads/en/DeviceDoc/DN2540%20B060313.pdf
It's actually a depletion mode N-mosfet, is cheap as chips and no worries about voltage breakdown (400V). It is a MOSFET, though, so best don't solder to it when other gear is connected to the circuit.
I don't have noise measurements or noise specs for DN2540, but have built some amplifiers with it and found it to be very quiet. Run at about 5 to 20mA, and cascode it to eliminate g-d capacitance variations with voltage. With a cascode and source degeneration resistor (about 300 ohms) to set its current, it will have extremelt high input impedance limited only by bias resistor and is capable of quite low distortion, too. (Running at higher Vds might work rather than a cascode, I didn't though).
PF5102 on the graph is process 51 jfet, I guess that J111 family is similar? Any links for the noise measurement of J112/113?