JFET Coupling Cap
Yes. You suggested 1uf earlier. Coupling cap values I saw in various JFET circuits range from 1uf to 220uf. A pSpice example showed how the frequency at which the JFET circuit shifted from DC gain to AC gain was a function of coupling cap value. A 1uf cap was in the middle of the audio frequency band. All source resistor JFET coupling cap examples were for amplifiers, not constant current aplications. One coupling cap explanation was that at higher AC frequencies it would shunt the AC signal thus raising the impedance of the JFET circuit. It seems to me that if the coupling cap is bypassing AC signals, then it is not functioning as a constant current device at higher frequencies. Doede, is 1uf still the value you want me to test or should I try a higher value?
I see a lot of effort in the forum to find a current production JFET to use in place of the 2SK170 which is made of unobtainium. It seems most of them need a Vg-s of more than 1V for 0.4ma. One JFET was less than 0.4ma at a Vg-s of 0v. The 2SK170 runs at 0.4ma with a Vg-s of .35-.40V and is very linear above a Vd-s of 0.5V. There is a current production version of this JFET by LINEAR INTEGRATED SYSTEMS called an LSK170 and is described as having low input capacitance.
Yes. You suggested 1uf earlier. Coupling cap values I saw in various JFET circuits range from 1uf to 220uf. A pSpice example showed how the frequency at which the JFET circuit shifted from DC gain to AC gain was a function of coupling cap value. A 1uf cap was in the middle of the audio frequency band. All source resistor JFET coupling cap examples were for amplifiers, not constant current aplications. One coupling cap explanation was that at higher AC frequencies it would shunt the AC signal thus raising the impedance of the JFET circuit. It seems to me that if the coupling cap is bypassing AC signals, then it is not functioning as a constant current device at higher frequencies. Doede, is 1uf still the value you want me to test or should I try a higher value?
I see a lot of effort in the forum to find a current production JFET to use in place of the 2SK170 which is made of unobtainium. It seems most of them need a Vg-s of more than 1V for 0.4ma. One JFET was less than 0.4ma at a Vg-s of 0v. The 2SK170 runs at 0.4ma with a Vg-s of .35-.40V and is very linear above a Vd-s of 0.5V. There is a current production version of this JFET by LINEAR INTEGRATED SYSTEMS called an LSK170 and is described as having low input capacitance.
The current of the lm334 various with temperature - that's why it is used as a temperature sensor. Internally it has 6 bipolar transistor junctions. I would expect it to have a different sound signature from a JFET CCS. The data sheet does not emphasize higher frequency performance. That is why we do listening tests.
don't expect that sole Jfet CCS is having lower tempco than shebang with tight voltage reference , which LM334 is ;
though - choice wise - voltage limit at pin 20 is more important than tempco and/or effective dynamic impedance of CCSink itself
Idss is 200-600uA ....... so , what you need is luck
400uA is smack in the middle of the range... a statistical average should hit our target precisely. Doede pointed out that 340uA is more like what the specs may indicate as average, but even so, it sounds like 400uA should be met with some ease given enough samples. 2N4338 hits an ideal spot, I think, for the application, the next parts up running into risks with Vgs(off). but we work with what's out there
having said that, heck, man... thank you for doing the ground work. it wasn't a lucky streak for sure - more like the opposite...
not sure where the 1uF comes from? never mind, it needs to be significantly higher as I mentioned in my earlier post... 10-100uF is more like it. I would suggest to find a compromise between low freq corner and size ... 10uF wil do for tests as well...
it can be compensated by adding just one diode and a resistor. the diode is in series with the LM334, so even the 0.65V forward bias on the diode would leave more than enough voltage for the LM334 to work (min. is 1V) - it seems to work in principle.
per Jung, about 110dB rejection at 40kHz - not bad at all (the single JFET CCS comes in at about the same but it's worse at lower frequencies). Still under 90dB as far as 200kHz, but the single JFET does better there. In other words, if you want to do a theoretical direct comparison, the s-JFET wins at higher frequencies while LM334 wins at the opposite ends of the spectrum. lab/setup measurements limitations may factor in.
Last edited:
I may be wrong, but I didn't see much effort here to "replace" the 2SK170 (not sure why we'd try?...). The work is in finding the right part(s) to do the job. It may be a FET (and then the 2SK170 may work, but there are a few aspects working against it, only one them being sourcing it), or a packaged current source/sink, or a separately biased multiple parts CCS...
agreed on the reality check. I don't see much tempco plots on the cut sheets for the JFETs we're considering, so I guess unless one of us measures their creep we can't know how they stack against each other when things warm up.
also agreed - if what you mean is that the voltage at pin 20 (2.4V) is the tight, exclusionary variable in this. Tempco and dynamic impedance (etc.) are luxuries to be considered after weeding out totally unworkable avenues.
it's during ungodly hours where I'm at - so I'll look at this again in the morning -, but it seems that there's a 10:1 ratio between R2 and R1, so I assume if you're calculating R1 (0.066/Id), R2 follows right along
Last edited:
Coupling Cap Installed
I installed a 22uf coupling cap around the JFET source resistors. My experience is that brand new components usually sound ok for about the first 20 minutes or so and then get variously ugly sounding for the next 100-300 hours. Capacitors tend to take the longest time to break-in.
Initially the coupling cap provided no improvement in sq and very little difference in sound. Now the DAC is sounding ugly. I'm going to put the JFET boards with coupling cap in my breadboard and cook them for at least a week with test signals before trying them again.
The adjustable JFET CCS was amazing sounding without the cap and it is recommended to the Forum as one workable improvement over the 6k resistor. Meanwhile, I hope others are testing additional CCS devices and circuits in their DACs instead of just tossing around technical discussions.
I installed a 22uf coupling cap around the JFET source resistors. My experience is that brand new components usually sound ok for about the first 20 minutes or so and then get variously ugly sounding for the next 100-300 hours. Capacitors tend to take the longest time to break-in.
Initially the coupling cap provided no improvement in sq and very little difference in sound. Now the DAC is sounding ugly. I'm going to put the JFET boards with coupling cap in my breadboard and cook them for at least a week with test signals before trying them again.
The adjustable JFET CCS was amazing sounding without the cap and it is recommended to the Forum as one workable improvement over the 6k resistor. Meanwhile, I hope others are testing additional CCS devices and circuits in their DACs instead of just tossing around technical discussions.
can anyone explain this for me?,
cant seem to get my head round it
what is the max ouput currents for the mono dddac's?
surely the data sheet specs is for regular stereo mode?
Thanks again for breaking the trail Carlsor. your results sound very promising.
I'm planning to give this a try and have various bits in the post at the moment. Including, J502's, LM334's, sot-23 (for J201) and soic-8 adapter boards (for LM334) which I hope will make nice drop in modules.
It's going to take me a few weeks to catch up.
Last edited:
Hi James, looking at the data sheet, I agree with Rax. as a ballpark figure R2 is factor 10 times the value of R1. In our case R1 was calculated to be 165Ohm (0,4mA/2.4V, depending on what C one uses), so R2 needs to be 1.65KOhm (in this combination).
It's assumed that the forwarding drop across the diode is 0.65V, which is importent to look for. It's also suggested to use 1% resistors, to reduce error %.
Ideal, thanks
The 1:10 ratio for the R1 and R2 made sense to me, but I wasn't clear whether it was just a bolt-on extra to the existing R1 or needed a new formula.
I'll get a few 1N457 diodes and 0.1% 1k65R resistors ordered and see how I get on
Wow Guys! Seems like you are all making some progress with pin 20! I wonder if this new tweak is suitable for my 11-Decks setup with Shunt regs? Is Heat dissipation a concern?
Doede, I have captured the complete 11 DAC boards and in detail of a DAC board by itself, and also Mainboard without DAC in hi res, raw. I will later capture the complete 11-Decks as single unit like the one captured by my iPhone. Now where can I upload?
Cheers,
Chanh
Doede, I have captured the complete 11 DAC boards and in detail of a DAC board by itself, and also Mainboard without DAC in hi res, raw. I will later capture the complete 11-Decks as single unit like the one captured by my iPhone. Now where can I upload?
Cheers,
Chanh