Take my word.
2SK1058/2SJ162 is a much better choice than IRF's.
Lower Vgs (hence loss of headroom), lower and much more linear capacitances, negative tempco (self stabilising), .....
Just listen to those who built the UTHAiM and the XELF :
UTHAiM -- Just for Fun
The XEN XELF Headphone Buffer
They are not pin compatible to the verticals.
You have been warned.
Cheers,
Patrick
2SK1058/2SJ162 is a much better choice than IRF's.
Lower Vgs (hence loss of headroom), lower and much more linear capacitances, negative tempco (self stabilising), .....
Just listen to those who built the UTHAiM and the XELF :
UTHAiM -- Just for Fun
The XEN XELF Headphone Buffer
They are not pin compatible to the verticals.
You have been warned.
Cheers,
Patrick
The 2SK1058/2SJ162 are Renesas/Hitachi MOSFET with a G-S-D pin pattern different from the Toshiba and IRF G-D-S. It is in TO-3P package which is too big for most headphone amp board using TO-220 size output device including the one that Zung has.
There go my 20 bucks of IRF. Oh well...
While we're at it, in the XEN XELF thread, you said you also tried the Exicon, but you didn't say which one you like. Care to comment?
Good point, plus 3x the Crss.
But I have another thing in mind: I have a blown Marantz 15 with a needlessly complicated output stage including 2 GE 1134 automotive bulbs as protections. I'm thinking those L-MOS could be a simple bolt-in fix. Of course, the flip side is I'll most likely loose all those period-correct s***ty SS sound of the 60's.
> but you didn't say which one you like.
Exicon. But more expensive.
> It is in TO-3P package which is too big for most headphone amp.
Easily done off board.
> Good point, plus 3x the Crss.
Do the comparison at Vds = 10V..
And I prefer a constant Crss than one that changes with Vds.
But as long as you are happy with IRF's, that is all that matters.
Just ignore what I said.
Patrick
Exicon. But more expensive.
> It is in TO-3P package which is too big for most headphone amp.
Easily done off board.
> Good point, plus 3x the Crss.
Do the comparison at Vds = 10V..
And I prefer a constant Crss than one that changes with Vds.
But as long as you are happy with IRF's, that is all that matters.
Just ignore what I said.
Patrick
Haha, it is unfair for you to say that!> But as long as you are happy with IRF's, that is all that matters.
Just ignore what I said.
Patrick
I was perfectly happy with my E19 headphone amp with K2381/J407 output. But knowing what you said and tested were right, I finally break down and ordered the 2SK2013/2SJ313 for replacement. They were used in many Pass Lab DIY project with good results.
I have 12 pairs of 2SK1058/2SJ162 from jackinnj. I considered using them for the headphone amp outboard, but decide to do a little thermal measurement first. I increased the PS from 15 to 21 volts. I up the bias current from 45 (factory) to 60 mA. The K2381/J407 and everything else stay within a good temperature margin. So I decided to order the 2SK2013/2SJ313 instead.
Last edited:
Thank you for all the files. I am a mechanical engineer in training. With these files, I can try doing what I dare not try myself.
Can you please start a thread in the HEADPHONE SYSTEM forum? I believe this discussion will be of interest to many for its low cost, high payoff, high mod potential nature. The LTspice work you've done will be very helpful to many.
Last edited:
It would make sense if the project is sustainable, at least for a short while.
My second kit is supposed to come in any day now, we'll see if the input Fets are well matched (or not). The shop brags about "996 pieces available ".
Will it make sense to replace the 2SK246/2SJ103 with K170/J74? While a headphone amp is not as noise sensitive as a phono stage, a lower noise floor may still be audible, particularly with high sensitivity headphone. How important is the noise spec of these input devices?
punkydawgs sells matched and generic K170/J74 at e-bay. Not cheap, but still reasonable.
It is not a performance question, rather a "cost effectiveness" question. 2 matched quad of K170/J74 from punkydawgs will set you back $150 which is almost 5 times of the original kit. I know the answer are:
- It depends.
- Can you hear it yourself?
Back around 2015, I bought one of the JC-2 clone kits that were discussed on the thread here at that time. As John Curl said, the case was nice, most of the parts were nice, and at ~$150 it looked like a great experiment.
When it arrived, it sounded ok but changed sound quality over time. By time, I mean after turning on. There was also some temperature variation to the sound.
So, I measured it. Like shown on post 300 in this thread, I found some high frequency junk that obviously wasn't simple harmonic distortion. Looking higher with an RF spectrum analyzer, I found tones well above 100 KHz. That looked like instability to me. These tones created IMD in the audio band.
It was kind of eye opening how these tones caused distortion in the audio band. That shouldn't have been terribly surprising since most transistors are less linear well above the audio band, and the amount of open loop gain available at those frequencies meant that the loop feedback wouldn't help much at reducing them if they were regular non-linearities, either.
It also gave some insight on how high frequency distortion that I thought might not matter did affect the sound. But, that part is a discussion for another time and place.
(I should mention here that I didn't save any test instrument displays. I didn't think I'd ever need them. So, you can either take me at my word for what I observed, or just dismiss it all. Up to you. I thought that this thread had died by the time I finished all the work.)
The next step was to simulate the circuit. I did find a schematic from the vendor, but I also traced the circuit on the PCB to be sure. From my research, there apparently was more than one "exact" clone design out there.
My example used pairs of 2SK170BL's and 2SJ74BL's at the input. The output devices were a ZTX550 and a ZTX450. Not exactly the same as the design published in the Audio Amateur.
I also noticed another deviation.
In the original JC-2, I guess the amplification section was potted in some kind of sealed module that may have been used in a number of applications by Levinson. The feedback network was external to this module. Obviously, that made it possible to customize the gain and all for the desired application.
In my example, the feedback network did not have the compensation capacitor that is often placed in parallel with the feedback resistor connected between the amp output and the amp inverting input. More on that in a bit.
Anyway, an LTSPICE simulation found that the phase margin for the JC-2 clone was a whopping 12 degrees, at least for the nominal SPICE model parts. (The gain margin was 23 dB, which is fine) That phase margin probably changed with temperature as the parts warmed up. All in all, the circuit was marginally stable and showed it. Those tones weren't non-linearities.
Changing the Miller compensation caps from the supplied 10 pF to 20 F, as suggested earlier, improved things, but not a lot. The phase margin went to 17 degrees.
However, adding 10 pF across the feedback resistor improved the phase margin to around 80 degrees.
Now, there was far less junk seen in the upper audio band. A lot of the frigglies above 100 KHz were gone, too. And, the preamp sounded much, much better.
Gee, now the circuit worked as intended. I guess the moral of the story is that if you're going to clone something, really copy it properly unless you are prepared to and capable of making suitable changes.
One addendum... Since I was already deep into this, I decided to see what could be done with other devices and improvements suggested by LTSPICE simulation.
I removed the input JFETs and measured them on a curve tracer. If these parts are fake Toshiba devices, whoever copied them did a marvelous job of matching the curves. They should sell the parts on the open market. They were well matched, too. Ultimately, the JFETs were replaced with 2SJ103BL's and 2SK246BL's based on a zillion LTSPICE simulations. They are a bit noisier, but in this application the difference is minor.
The output devices were changed to 2SA912's and KSC1845's. Again, this was based on LTSPICE simulations.
I also redistributed the current so that the output devices ran more current and the input JFETs less.
No cap across the feedback resistor because it wasn't needed. The Miller compensation caps were left at 20 pF. The predicted phase margin was about 65 degrees with 27 dB of gain margin.
While this worked great on the bench, it didn't sound much better. Maybe a little.
Ultimately, I changed the rectifier diodes to GI SRP100 series types, which have a much gentler and faster recovery characteristic. The high frequency frigglies almost disappeared. Adding proper "snubber" RC networks across the transformer secondary connections got rid of most of the remainder. Some were just coming in through the power line - a line filter helped with that. Improving the current sources used to supply the shunt regulators on the board helped, too.
Now, it all sounded much better and measured well.
One last thing...
I mentioned a lot of test equipment above. That all was available in the lab at work (non-audio job). But, now, I could make all the same measurements at home with a Peak DCA-75, an Analog Discovery 2, and a QuantAsylum QA-401. I could even easily measure the closed loop gain and phase margin, too. That's real progress on the DIY front.
When it arrived, it sounded ok but changed sound quality over time. By time, I mean after turning on. There was also some temperature variation to the sound.
So, I measured it. Like shown on post 300 in this thread, I found some high frequency junk that obviously wasn't simple harmonic distortion. Looking higher with an RF spectrum analyzer, I found tones well above 100 KHz. That looked like instability to me. These tones created IMD in the audio band.
It was kind of eye opening how these tones caused distortion in the audio band. That shouldn't have been terribly surprising since most transistors are less linear well above the audio band, and the amount of open loop gain available at those frequencies meant that the loop feedback wouldn't help much at reducing them if they were regular non-linearities, either.
It also gave some insight on how high frequency distortion that I thought might not matter did affect the sound. But, that part is a discussion for another time and place.
(I should mention here that I didn't save any test instrument displays. I didn't think I'd ever need them. So, you can either take me at my word for what I observed, or just dismiss it all. Up to you. I thought that this thread had died by the time I finished all the work.)
The next step was to simulate the circuit. I did find a schematic from the vendor, but I also traced the circuit on the PCB to be sure. From my research, there apparently was more than one "exact" clone design out there.
My example used pairs of 2SK170BL's and 2SJ74BL's at the input. The output devices were a ZTX550 and a ZTX450. Not exactly the same as the design published in the Audio Amateur.
I also noticed another deviation.
In the original JC-2, I guess the amplification section was potted in some kind of sealed module that may have been used in a number of applications by Levinson. The feedback network was external to this module. Obviously, that made it possible to customize the gain and all for the desired application.
In my example, the feedback network did not have the compensation capacitor that is often placed in parallel with the feedback resistor connected between the amp output and the amp inverting input. More on that in a bit.
Anyway, an LTSPICE simulation found that the phase margin for the JC-2 clone was a whopping 12 degrees, at least for the nominal SPICE model parts. (The gain margin was 23 dB, which is fine) That phase margin probably changed with temperature as the parts warmed up. All in all, the circuit was marginally stable and showed it. Those tones weren't non-linearities.
Changing the Miller compensation caps from the supplied 10 pF to 20 F, as suggested earlier, improved things, but not a lot. The phase margin went to 17 degrees.
However, adding 10 pF across the feedback resistor improved the phase margin to around 80 degrees.
Now, there was far less junk seen in the upper audio band. A lot of the frigglies above 100 KHz were gone, too. And, the preamp sounded much, much better.
Gee, now the circuit worked as intended. I guess the moral of the story is that if you're going to clone something, really copy it properly unless you are prepared to and capable of making suitable changes.
One addendum... Since I was already deep into this, I decided to see what could be done with other devices and improvements suggested by LTSPICE simulation.
I removed the input JFETs and measured them on a curve tracer. If these parts are fake Toshiba devices, whoever copied them did a marvelous job of matching the curves. They should sell the parts on the open market. They were well matched, too. Ultimately, the JFETs were replaced with 2SJ103BL's and 2SK246BL's based on a zillion LTSPICE simulations. They are a bit noisier, but in this application the difference is minor.
The output devices were changed to 2SA912's and KSC1845's. Again, this was based on LTSPICE simulations.
I also redistributed the current so that the output devices ran more current and the input JFETs less.
No cap across the feedback resistor because it wasn't needed. The Miller compensation caps were left at 20 pF. The predicted phase margin was about 65 degrees with 27 dB of gain margin.
While this worked great on the bench, it didn't sound much better. Maybe a little.
Ultimately, I changed the rectifier diodes to GI SRP100 series types, which have a much gentler and faster recovery characteristic. The high frequency frigglies almost disappeared. Adding proper "snubber" RC networks across the transformer secondary connections got rid of most of the remainder. Some were just coming in through the power line - a line filter helped with that. Improving the current sources used to supply the shunt regulators on the board helped, too.
Now, it all sounded much better and measured well.
One last thing...
I mentioned a lot of test equipment above. That all was available in the lab at work (non-audio job). But, now, I could make all the same measurements at home with a Peak DCA-75, an Analog Discovery 2, and a QuantAsylum QA-401. I could even easily measure the closed loop gain and phase margin, too. That's real progress on the DIY front.
Last edited:
The clone JC-2 that I got several years ago would sound only fair for the first couple of hours of operation,
until it warmed up and the (used RFE) yellow French electrolytic capacitors in the power supply reformed.
After several years of regular use they no longer need the burn in.
until it warmed up and the (used RFE) yellow French electrolytic capacitors in the power supply reformed.
After several years of regular use they no longer need the burn in.
Last edited:
