Linuxguru, RNMarsh has kindly offered to test your FET990 samples.
He's one of the few people in the known universe that can reliably measure THD20k below 1ppm.
I've sent you a message about how we should organise this.
It would be really nice to do this and establish an important bridge between the 'real world' and SPICE world.
OK, I've built a few 1st-gen FET990 modules so far - I'm awaiting an EBay transistor tester which will allow Idss matching of SOT23 JFETs more easily. I'll make an additional pair of the 1st-gen modules with matched devices shortly.
Preliminary 2nd-gen layout (with the bootstrapped VAS clamp, Diamond Cascode O/P stage, etc.) is ready, just holding off on board fabrication to see if anything else can be incorporated.
At the moment, I have Toshiba 2sk209 FETs for the input LTP - will eventually go with BF862 (higher Yfs) once I can find a reliable local source. Are these EBay BF862s, claimed to be manufactured by Zetex, authentic?
http://www.ebay.com/itm/10PCS-BF862-JFET-N-CHAN-20V-SOT-23-ZETEX-NEW-GOOD-QUALITY-R1-/300880906571
Edit: Link above appears not to work, please search for BF862 in the EBay search box.
For the VAS, there seems to be no SOT23 PNP that's much better than the ubiquitous BC856B - any suggestions? The low Ccb alternatives are BF550 and BF821 - both possibly with much lower gain than the BC556B (not sure if the VAS beta is relevant - the VAS emitter-follower/buffer can be a high-beta device like BC859c).
Last edited:
One candidate output-stage enhancement is a low-distortion buffer from Kenpeter and Elvee on this board, called Ken's Diamond (actually, not a diamond buffer - it has more in common with the Taylor source-follower):
http://www.diyaudio.com/forums/soli...-buffer-power-output-stage-9.html#post2056889
It uses 6 BJTs (2 more than the Diamond Cascode) and a voltage reference, but gets rid of one CCS (typically a JFET). That also frees up the rail voltage limits (now limited to about +/-15V because the JFETs that are best suited for CCS duty are limited to 30V Vds). Board area is likely to be of the same order as (maybe a bit higher than) a Diamond Cascode because 6-pin duals can be used for 4 of the BJTs.
Let me run it through the sims for a while and see how it goes.
Shipping charges to India from Mouser/Digikey are steep - it would be great to have a reseller who can ship cheaply from North America by first-class mail or similar.
For the EBay tester, please search for "M328 Transistor Tester" in the EBay search box (deep links into EBay don't seem to work now).
Yup ebay is a total waste these days, it times out on me.
I can be a re-seller that will ship post to you. Make me an offer, you can probably use my paypal account to transfer funds. I have BF862 is my stock. Have you looked at KSC1845/KSA992 and the smt versions fjv992/1845?
I have a pending Mouser order. I wait til I get to $200 as shipping is free for me.
I can be a re-seller that will ship post to you. Make me an offer, you can probably use my paypal account to transfer funds. I have BF862 is my stock. Have you looked at KSC1845/KSA992 and the smt versions fjv992/1845?
I have a pending Mouser order. I wait til I get to $200 as shipping is free for me.
Thanks, I'll PM you later regarding the BF862. I've used the original NEC 2sc1845/a992 extensively, but had no idea until now about the Fairchild SMD versions - will check them out.
This is a simulation schematic of a simplified 4-BJT Ken's Diamond as the output stage - it's derived from the 6-BJT Ken's Diamond, but the CFPs at the output have been reduced to single BJTs, thus reducing the BJTs in the output stage to 4. The voltage reference has also been opportunistically obtained from the VAS CCS without any additional device.
It seems to simulate with a few dB lower H3 than the Diamond Cascode at similar output stage bias of around 8 mA. It still needs further analysis of stability, phase margin, overdrive and clipping behaviour, PSRR, etc., but it looks to be a viable contender for a Class-A output stage for the FET990.
It seems to simulate with a few dB lower H3 than the Diamond Cascode at similar output stage bias of around 8 mA. It still needs further analysis of stability, phase margin, overdrive and clipping behaviour, PSRR, etc., but it looks to be a viable contender for a Class-A output stage for the FET990.
Attachments
I've laid out a board with to92 thru hole parts but I have some questions. Do the input jfets require shielding around the gate pads? Some jfet input opamps require this. Also are catching diodes on the output necessary as part of the diamond compensation. They are not on any of Scott's schematics but appears on others. Also moving the offset null to the resistor leads of the positive current source requires many more parts than putting it where Scott originally had it. Just the things I'm wrestling with. Thanks Ray
Hi Ray,
This was designed by SW and so should use his schematic and his Ok on any changes or additions (like diodes on output). Others' mods/changes to it can have the others' pcb made. I would like to use the design exactly as Scott created it. Then others designs can have their pcb for it made.
The ring around the inputs (jFETs) is a good idea so we can try using it as a guard ring.
Thx--- I have been waiting a long time for the thru-hole version.
-RNMarsh
This was designed by SW and so should use his schematic and his Ok on any changes or additions (like diodes on output). Others' mods/changes to it can have the others' pcb made. I would like to use the design exactly as Scott created it. Then others designs can have their pcb for it made.
The ring around the inputs (jFETs) is a good idea so we can try using it as a guard ring.
Thx--- I have been waiting a long time for the thru-hole version.
-RNMarsh
Last edited: