Jfet BOZ

[needtubes]

Would the 2SK117 or 2SK246 be acceptable as replacements for the 2SK170? I ask because Tech-DIY seems to be out of 2SK170 but has these...

[Fenris]

After reading the whole thread, I decided to try my hand at this design, but with a headphone amp in mind. I combined a JFET BOZ with a B1 and a mini Beast With a Thousand JFETs, plus a pinch of tube knowledge.

I've attached a schematic. It's a dual power supply design with a JFET gain stage and a JFET CCS on the Drain. There are three LEDs (2v drop each) to bring the value of the source resistor down to a reasonable value. This then goes into a B1 buffer but with five pairs of JFETs paralleled. Finally, a resistor provides feedback to minimize distortion and reduce gain down to 2. Right now the prototype proof of concept is playing fine on two 9v batteries. I built it with what I had laying around so the 20K resistor is only 18K and two of the 1meg resistors are 67K and the 600 ohms is 560, but those values are probably not too critical. It starts to distort pretty early, probably because I only used a single pair of JFETs on the output, so I only have few milliamps sourced through the CCS (I'm testing with hard to drive 34 ohm Sennheiser HD497s). I figure with five pairs in parallel I should be able to have 25-30 ma available which should drive most any reasonable headphone.

I'm wondering if the source resistor is the right size, it seems right for a load of about 5ma. The feedback seems to work, but could it be improved? Any thoughts on the overall design? I was wondering if I should bypass the LEDs with a capacitor to reduce the AC impedance, although it's probably only about 30-50 ohms with the old red LEDs I'm using. Constructive feedback is always appreciated.

[juma]

Hi Fenris,
I suppose you intend to use 2sk170 JFETs or similar ?
If it's so, it is unrealistic to expect from your SE output stage (B1) to drive 34 Ohms headphones (even with feedback).
Also, it is unclear why do you need the extra gain - few hundreds of mV will drive your headphones louder than you wish.
With low Z headphones you need your headamp to be capable of delivering the appropriate current into the load - search for DAO headamp, that will suit your headphones much, much better.

[Fenris]

Yes, 2SK170s. The idea was to be flexible - some gain for high impedance headphones (gain of 2 should be sufficient and compensate for the non-unity gain of the source follower) and to have enough current for lower impedance headphones, which is why I would parallel 5 pairs on the output. Efficient? No, but I've got 15 pairs of matched JFETs sitting around without a home and an empty Altoids tin that needs a purpose in life.

[juma]

Yes, 5 paralleled pairs changes the perspective. Try taking the feedback after the 1uF cap (Schade connection - puts gain JFET into the tride-like mode).

[qusp]

i do not really think B1 is good for headphones, 5 parallel or not the output impedance is too high no? b1 buffer sounds very good, but its about as far from the ideal buffer as you can get ie. very high input z and very low output z. i could be wrong of course, but this also matches info i recieved when asking the same questions

[Fenris]

Actually, high input impedance and low output impedance are traits of an ideal buffer. It's just that with only one pair of output devices the output impedance was only about 50 ohms - great for most line level stuff, but not really low enough for a pair of headphones, and definitely not low enough for 32 ohm headphones. When you parallel more (i.e. the Beast with 1000 JFETs), you drop the output impedance enough to drive output devices.

[qusp]

yeah but only 5? i wouldnt have thought that enough to raise the current to a sufficient level, or lower outputz to a low enough level. i suppose 5 will mean 35-50ma, enough, but not ideal imo. why not drop some lu1014d in there or a touch more well measured feedback. ooops bad word

[Fenris]

The "average" audio op-amp used for most outputs like the 2132 or 627 have about 20ma of current capability with a 600 ohm load. Two and a half times the current capability will give five times the output power and the overall output impedance will be around 10 ohms. This probably isn't the "ideal" or "ultimate" solution, but I'm willing to try something unusual and unique. I could drop the 1014 in there, but I've already got enough 2sk170s that need a good home.

[jacco vermeulen]

Originally Posted by Fenris

When you parallel more

For my headphone set i'd already be content with a 100 JFET beasty.