sorenb

Can't argue that, and you may be right, that it is rather the change in current than the CCS itself that I am hearing Though I did try to match the existing current as close as possible. Anyway part of the fun doing DIY is also trying out stuff, and maybe realize that it didn't matter that much. For now In rgrds to the current in the output stage I haven't seen you recommend a sweetspot, other than commenting that one might 'get away' with 18mA ...is 18mA the sweet spot ?

Having read the SRX-revisited thread by JimL and the referenced articles by W.Jung I decided to replace the current sources in the one Dynalo I've built. From the 1st article by Jung, it seems as if that the CCS design used in the Dynalo has indeed a very good performance. However as JimL has pointed out the cascoded CCSs seems to offer even better performance. Looking at the very simple LM334 with one resistor demonstrates 20db better rejection than the LED CCS although the performance rapidly degrades past 10kHz; the cascoded LM344/2N5486 however maintain the performance above 10kHz. So, based on that I decided to try it out. First I have implemented the non cascoded LM334. At first hand it sounds like an improvement. However, not night and day difference; the sound is a bit more refined/detailed, and the soundstage as gained a bit more depth.

good one Pars ) No, did my measurement with open input .... inputs grounded its within 4mV (+O - -O) I haven't put the servos back in, but spend some time listning. My Sennheiser HD800 have never sounded better ) Done my listning using my Tact Mill. mkIII, and the Dynalo is clearly better than hooking up the HD800 directly to the speaker terminals ....much better. Thnx Kevin and Pars!

Ok. I'll leave them as is, and give a listen to hear wheather the effort was worth it ( I guess it was ;-) Many, many thanks Kevin for making and sharing your design. Also thanks to Spritzer, mypasswordis, GrindingThud, gepardcv, Pars, JimL, mwl168, and others sharing/contributing to the thread. Pics, or i didn't happen:

Finished two sets of Dynalo this weekend and two sigma22 psu's. The offset seems to drift a lot; without the servo's it's within a range of 40mV on all boards - fluktuate the full range within a couple of min or so.With the servo in place it's within a range of 0,5mV Is that normal behavoir? Using +/- 18V for rails I get approx 22-28mA through the output transistors across the four boards. Would that be running those trans too hard? should I lower r38/39?

I guess the "It's been pretty well discussed, but the 369 is the 'replacement' for the 2SK147." settles it - I have bunch of 147/72. Thnx Pars

can the lsk389 be replaced by a pair of Toshiba 2SJ72-BL? I have 100 pairs of j72/k147 lying around for no use.

Can you elaborate a bit on that Birgir? As far as I understand the load from kgst is less than kgsshv, or do i miss something?

During the weekend I did a complete re-build of my beloved Lambda's. Heavily inspired by 4070 I decided to do a closed housing. The coloration from the housing was a bit hard to avoid so the combination is actually the best combo that even out the coloration. Also the pads had been rebuilt, and certainly removes any etch. The result is outstanding, no one can tell the difference between Rick Astley's "Never gonna give up on you" and Ludvig van's Ninth's ... Let's call 'em 007mkIII, as they are everything the 007 will never become ;o) The sound is "sick as a brick" ;o) (http://inhabitat.com/lego-headphones-bring-new-life-to-old-toys/lego1/?extend=1)

AFAIK there has been at least on run on KGSSHV PCBs from Seeds, has any reported problems with these?