Peer Critique needed for my new design

Hi bvp,

What can I say. It's a bit Barry Manilow.

Is that 2 electros directly in the signal path?
Have you considered bypassing that offset adj pot to gnd (on both sides)?
have you tried a cascode on the input stage to raise collector impedance? The high degen you have helps and with reducing Ccomp. This should be stable without those 22pF - (take off the feedback lead C and disable the input 100p and view a 10KHz sq wave) You can improve the PSRR and HF distortion both.

Your feedback lead C seems to be the dominant HF rolloff - would be better at the input RC which is 5 times higher.

The tail current sources could be better 2 tr unless you need the lights.

Just at first glance. Looks VERY stable (with zobel of course)

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
pinkmouse said:
Do you like Dan's designs, or are you just being diplomatic? ;)

Dan and I are very old friends, dating back to '76 or so. At that
time, as the Dayton-Wright rep, he did me the great service of
hauling an 800A around with which to demo the XG8 speakers -
no trivial effort.


/pass/: has a book full of Dan stories and Mike Wright stories
as told by Dan.
I've been putting a filter capacitor of 100uF across my CCS LEDs.

The 10 ohm base resistors on the output transistors are larger than my normal values of 3.3 ohms, down to 1 ohm for low voltage outputs with high hfe. I think the latest Leach uses 3.3 ohms.

I have been putting my DC offset bias on the feedback input of the differential, not at the input. My current amp uses a servo that nudges the two front-end constant current sources to get a DC null. Do you need the input resistors bias with the large 100uF capacitor on the feedback circuit?

If you've simmed this, try resistors similar to your output Re's to superimpose the output stage commutation currents on the rails and see how it affects THD, FFT. You don't want it hypersensitive to power supply hash or you'll be doubling up supply caps and transformer size chasing minute ( maybe 6dB) improvement when
a few low cost refinements can make maybe 20-30 db better isolation.


2004-12-12 8:07 pm
Thank You

Thank you all for your valuable inputs. I'm very grateful for that. With all of your feedback at once, I would like to say (if I missed any question, please forgive me and I'll get back to that)

1. I got the idea of the circuit from the Motorola Application Notes where I found from the Internet. I don't know why it is not available at Onsemi anymore. I haven't have a chance to see any Krell Amp yet. However, I would love to if I have a chance.
2. About the Zobel Network, yes, I included it in the second part - speaker protection ultilizing uPC1237HA.
3. Couple of electrolytic caps across the Leds recommendation sounds good to me. I'll add them in rightaway. I guess they just hold the voltage across leds (1.82V), don't they?
4. And I probalby don't need the 100uF in series with 1k-resistor in the feedback circuit. I'll find out in the lab soon.
5. Please introduce me with Dan's Design as you mentioned.
6. I wanted to load the pcb I made in pdf format, but its size is 450kB which is not allowable. Sorry for that.

Thanks again


2003-02-17 7:38 pm
i would definitely look for better vas and pre-driver transistors, if they're just what you have it should work but generally i would think the triple darlington should have decreasing ft from the fast pre to medium driver to slow output device, you have relatively fast output devices so the pre-driver in particular may benefit from being smaller and faster (@ 300 V, 20 W the mje340 has to be pretty slow, my data sheet doesn’t think ft is worth mentioning) - check out some of the vas Q suggestions in several other threads - low Vce, higher hfe, smaller die (= smaller Ccb,ib) pre-driver should present smaller, more linear load to vas

also i would consider raising driver current, with 12A peak output needing 100 mA base current with typ hfe ( 300mA if you use min hfe! ) and 4x power device base C to charge/discharge 30 mA driver bias may be light for low distortion with so little loop gain – not that high loop gain can fix a poorly commutating class B output
Nice to see someone else trying a triple darlington output like me :)

I'm looking for something better than MJE340/350 for vas and predrivers - with models as i'm still at the simulation stage. BD139/140 works well but I don't think these have a high enough Vceo for 50V rails. I tried MJE243/253 but these seemed to be very unstable in the design (although perhaps the OnSemi spice model is wrong - it wouldn't be the first one!) is my current attempt, not too dissimilar from your schematic. I'm probably exceeding the Vceo of the 546/556's there but I plan to add a cascode in there to 24v for each of the LTP's. Yes - the currents through the top and bottom half are different, this is to remove the 40mV output offset that seems to occur otherwise.