I see the 1Ks are the blocking and load Rs. It would not be hard for me to install another 1K with the current 1Ks to give me 500R.
IMHO, using better active device will bring more benefits than any boutique passive parts :
http://www.diyaudio.com/forums/soli...-high-performance-source-follower-module.html
Patrick
http://www.diyaudio.com/forums/soli...-high-performance-source-follower-module.html
Patrick
I'm with you. I use standard metal film R where I can and basic carbons where I need more wattage.
Give me a better jFET than the 2SK170!!
It's not to be taken as a personal opinion. What I am meaning is that the wide spread thought is that using a linear psu will bring far better improvements that coating a wire with cotton, so I see that as kinda ironic.
Hi,
most attenuators present a highish resistance/impedance to the source signal.
Choose a pot/switch impedance that suits the source.
Most attenuators present a highish resistance/impedance to the receiver.
Many systems do not perform well if the receiver sees a highish source impedance.
There are ways to reduce the symptoms:
a.) Very low impedance attenuator.
b.) Very low capacitance between attenuator and receiver.
c.) Receiver designed in conjunction with the attenuator to avoid most of the performance shortfall.
d.) Adding a buffer after the attenuator to present a low source impedance to the receiver.
Adding a B1 or a DCB1 or any emitter or source follower effectively converts the highish output impedance of the attenuator to a low source impedance that can drive any sensible combination of receiver and cables.
most attenuators present a highish resistance/impedance to the source signal.
Choose a pot/switch impedance that suits the source.
Most attenuators present a highish resistance/impedance to the receiver.
Many systems do not perform well if the receiver sees a highish source impedance.
There are ways to reduce the symptoms:
a.) Very low impedance attenuator.
b.) Very low capacitance between attenuator and receiver.
c.) Receiver designed in conjunction with the attenuator to avoid most of the performance shortfall.
d.) Adding a buffer after the attenuator to present a low source impedance to the receiver.
Adding a B1 or a DCB1 or any emitter or source follower effectively converts the highish output impedance of the attenuator to a low source impedance that can drive any sensible combination of receiver and cables.
> Give me a better jFET than the 2SK170!!
Didn't you read
http://www.diyaudio.com/forums/soli...-high-performance-source-follower-module.html ?
2SK372V, of course.
Patrick
Didn't you read
http://www.diyaudio.com/forums/soli...-high-performance-source-follower-module.html ?
2SK372V, of course.
Patrick
Soo, except for the input capacitance, it's actually really nice? Why exactly haven't I ordered half a kilo of those already?
Some thoughts.
B1 needs output resistors, at least circa 200R to be generally safe against oscillations in typical arrangements. In one case reported in my DC threads someone went to 330 Ohm lowest in an installation after it actually acted up. High gm 2SK372V could lower B1's intrinsic Zo to 20R, that would be about 50% better than a straight out 7-8mA 2SK170BL. It would make only 8% difference when a say, 220R buffer resistor is employed though. High gm and process lends to 372V best low nvrtHz. It also showed good shallow shape in the 1/f noise pick up region in Scott's FFTs along with the 15mA 2SK369 (was it a high BL or V?). Scott estimated its superior for generation-recombination noise VS the random 2SK170 he compared and a BF862. Those things are great for a ribbon mic or low mc cart head amp component selection, but in a line level unity gain buffer? Most will hook up a 6Vpk-pk CDP. On the other hand, those noise gains in the absolute sense, bring along even higher Ciss, Crss with the extra Yfs. Specifically more than double that of 2SK170's. Its our experience that we lose some sense of transparency for higher than 20K pots in B1. That is based on enough feedback from builders. That ''limit'' would have to be advised lower than 10K log if with 2SK372V's capacitance most probably? If yes, getting a bit heavier for some sources. I/V SRPP tubed DACs and the like. Also don't know where that high frequency noise rise comes from for 369/372 in Scott's FFTs, absent in 170. If this is actually Igs related to some leakage due to their wafer process, maybe it can impact something negatively in a wide bandwidth application like B1? Mini package with flat face makes bonding better for good thermal contact tracking in differential amps or DCB1. The V variety brings dissipation (no different than TO-92 sinking issue would be if with 170V). But going V also means much easier to get enough offset due to ''wilder'' gm, for matching's imperfect accuracy. Also V varies wider in selection per batch. % accuracy there means naturally bigger differences in IDSS. From a small batch some may not get a good enough for offset quartet. All in all those were my qualms about migrating to some ultra low noise high IDSS jfets recommended. Something like back to square 1, plus sinking effort, and more difficult matching & sourcing in DC. I hope they are not well founded. Personally I would like a BF862 in TO-92 that people could pick up with fingers and solder for this one, as much available and cheap as the SMT. Both stronger and less capacitive than 170 and noise wise super adequate here. Alas, SMT only ever, for the patient soldering skilled ones.
B1 needs output resistors, at least circa 200R to be generally safe against oscillations in typical arrangements. In one case reported in my DC threads someone went to 330 Ohm lowest in an installation after it actually acted up. High gm 2SK372V could lower B1's intrinsic Zo to 20R, that would be about 50% better than a straight out 7-8mA 2SK170BL. It would make only 8% difference when a say, 220R buffer resistor is employed though. High gm and process lends to 372V best low nvrtHz. It also showed good shallow shape in the 1/f noise pick up region in Scott's FFTs along with the 15mA 2SK369 (was it a high BL or V?). Scott estimated its superior for generation-recombination noise VS the random 2SK170 he compared and a BF862. Those things are great for a ribbon mic or low mc cart head amp component selection, but in a line level unity gain buffer? Most will hook up a 6Vpk-pk CDP. On the other hand, those noise gains in the absolute sense, bring along even higher Ciss, Crss with the extra Yfs. Specifically more than double that of 2SK170's. Its our experience that we lose some sense of transparency for higher than 20K pots in B1. That is based on enough feedback from builders. That ''limit'' would have to be advised lower than 10K log if with 2SK372V's capacitance most probably? If yes, getting a bit heavier for some sources. I/V SRPP tubed DACs and the like. Also don't know where that high frequency noise rise comes from for 369/372 in Scott's FFTs, absent in 170. If this is actually Igs related to some leakage due to their wafer process, maybe it can impact something negatively in a wide bandwidth application like B1? Mini package with flat face makes bonding better for good thermal contact tracking in differential amps or DCB1. The V variety brings dissipation (no different than TO-92 sinking issue would be if with 170V). But going V also means much easier to get enough offset due to ''wilder'' gm, for matching's imperfect accuracy. Also V varies wider in selection per batch. % accuracy there means naturally bigger differences in IDSS. From a small batch some may not get a good enough for offset quartet. All in all those were my qualms about migrating to some ultra low noise high IDSS jfets recommended. Something like back to square 1, plus sinking effort, and more difficult matching & sourcing in DC. I hope they are not well founded. Personally I would like a BF862 in TO-92 that people could pick up with fingers and solder for this one, as much available and cheap as the SMT. Both stronger and less capacitive than 170 and noise wise super adequate here. Alas, SMT only ever, for the patient soldering skilled ones.
I dare to disgree.
Firstly I do not understand why you need a output resistor. I have been using such circuits for almost 6 years in Sallen Keys and what not, and I never use an output resistor. Nor I ever had any oscillations. If you are really worried about oscillations, the one place I would put a resistor is at the gate, not at the source. But, only my personally experience.
Secondly, I have been using V grade for years. They do not lead to higher DC offset, if anything lower. My published measurements in the link above confirmed that once again. Also go to the original blowtorch thread and see what John Curl says about V & BL grades.
Thirdly, if you argue that BF862 is a better JFET (and everyone knows I use BF862 a lot myself), then you must also know that BF862 has just as high transconductance as 2SK372, just as high Idss as V grade, and factor of 5 lower capacitance, or in other word much more susceptible to oscillation.
IMHO, to have real meaningful discussions, you really have to try 2SK372 in this application yourself.
It only costs you 2USD a pair, unmatched.
Patrick
Firstly I do not understand why you need a output resistor. I have been using such circuits for almost 6 years in Sallen Keys and what not, and I never use an output resistor. Nor I ever had any oscillations. If you are really worried about oscillations, the one place I would put a resistor is at the gate, not at the source. But, only my personally experience.
Secondly, I have been using V grade for years. They do not lead to higher DC offset, if anything lower. My published measurements in the link above confirmed that once again. Also go to the original blowtorch thread and see what John Curl says about V & BL grades.
Thirdly, if you argue that BF862 is a better JFET (and everyone knows I use BF862 a lot myself), then you must also know that BF862 has just as high transconductance as 2SK372, just as high Idss as V grade, and factor of 5 lower capacitance, or in other word much more susceptible to oscillation.
IMHO, to have real meaningful discussions, you really have to try 2SK372 in this application yourself.
It only costs you 2USD a pair, unmatched.
Patrick
Last edited:
I was thinking specifically about the B1. Not about Jfet merits in general. It has been acting up with even with 220R into some cabling and load occasions, its been reported. I did not talk about the source buffer/gate stopper resistor at all. It should be there no doubt.
I had specific qualms about matching practicality for V in DCB1 in my previous post. Specifically in DC configuration you got -6.5mV from a most tightly bonded near in IDSS pair. And you got a stash of 372V. Most report 0.5-2mV in the just inserted side by side, not on sink 2SK170BLs in Mez and Hypno builds. Good statistic since there are many documented. How practical it will V be for the average builder who looks for 4 matched pairs? What is the minimum qty you recommend to produce 4 pairs for less than 5mV in DC configurations? No DC problem in original B1 of course.
I have some high gm 2SK170V and 2SK363BL, I don't reserve them for line stages, but for head amps where I think its their primary destination. JC would use V for BT not for just everything. That used to be personally built with hand picked precision as a statement product.
I never argued that BF862 is a best Jfet, I argued it would be cheap, current, and better suited in this application especially if it was ever available in TO-92 so we could see it too. Here its not an MC cart head amp fed from 10R Zo coils. It will have a pot before, and the BF862 would let that, and its gate resistor to be comfortably suitable, or a ferrite bead better if had legs, when having the gm to give an even nicer THD profile than higher Ciss when the input level goes up in transients, which is the subtle ''voicing'' and different fingerprint of B1 VS a JLH style buffer, without some BJT aid.
All in all I don't argue the use of NOS high gm jfets like you do as a no no or demerited. Can have points in the hands of a designer in a specific system. But you is you, a designer and tester, and you know if you got oscillations with no or minimal resistive output buffering, what is your best pot value, and what is your best supersonic square wave in what system. B1 is immensely popular for ready pcb builders, plug n' play. Don't know how they would handle it with V grade. I would recommend to anybody that wants to try the 372V to kindly ask you provide your special sink as a must.
I find 372V ''specialized'' in a B1 style application, bottom line of my thinking.
Best regards.
I had specific qualms about matching practicality for V in DCB1 in my previous post. Specifically in DC configuration you got -6.5mV from a most tightly bonded near in IDSS pair. And you got a stash of 372V. Most report 0.5-2mV in the just inserted side by side, not on sink 2SK170BLs in Mez and Hypno builds. Good statistic since there are many documented. How practical it will V be for the average builder who looks for 4 matched pairs? What is the minimum qty you recommend to produce 4 pairs for less than 5mV in DC configurations? No DC problem in original B1 of course.
I have some high gm 2SK170V and 2SK363BL, I don't reserve them for line stages, but for head amps where I think its their primary destination. JC would use V for BT not for just everything. That used to be personally built with hand picked precision as a statement product.
I never argued that BF862 is a best Jfet, I argued it would be cheap, current, and better suited in this application especially if it was ever available in TO-92 so we could see it too. Here its not an MC cart head amp fed from 10R Zo coils. It will have a pot before, and the BF862 would let that, and its gate resistor to be comfortably suitable, or a ferrite bead better if had legs, when having the gm to give an even nicer THD profile than higher Ciss when the input level goes up in transients, which is the subtle ''voicing'' and different fingerprint of B1 VS a JLH style buffer, without some BJT aid.
All in all I don't argue the use of NOS high gm jfets like you do as a no no or demerited. Can have points in the hands of a designer in a specific system. But you is you, a designer and tester, and you know if you got oscillations with no or minimal resistive output buffering, what is your best pot value, and what is your best supersonic square wave in what system. B1 is immensely popular for ready pcb builders, plug n' play. Don't know how they would handle it with V grade. I would recommend to anybody that wants to try the 372V to kindly ask you provide your special sink as a must.
I find 372V ''specialized'' in a B1 style application, bottom line of my thinking.
Best regards.
My remarks to oscillation referred to the source follower (or B1) with 2SK170, up to 10 pairs in parallel with no gate stoppers.
And I really never had any oscillations. I do now and then use a gate resistor of 100R just to sleep well.
The 6.5mV was, as mentioned obtained with a reject pair of 0.3mA Idss difference.
If you buy 50, you will reject about 6, and the rest matched to 0.1mA or below, which means < 2mV offset.
There is no magic in this, DC offset = (Idss delta)/Yfs.
The lower the Yfs (as in 2SK170), the higher the offset.
I matched 100 and half of it was paired to >0.05 mA.
Please take my word and give it a try first.
I'll send you a couple of heatsinks for free, if you would take the plunge.
Patrick
And I really never had any oscillations. I do now and then use a gate resistor of 100R just to sleep well.
The 6.5mV was, as mentioned obtained with a reject pair of 0.3mA Idss difference.
If you buy 50, you will reject about 6, and the rest matched to 0.1mA or below, which means < 2mV offset.
There is no magic in this, DC offset = (Idss delta)/Yfs.
The lower the Yfs (as in 2SK170), the higher the offset.
I matched 100 and half of it was paired to >0.05 mA.
Please take my word and give it a try first.
I'll send you a couple of heatsinks for free, if you would take the plunge.
Patrick