linuxguru buffer redux and beyond - diyAudio
Go Back   Home > Forums > Source & Line > Digital Source

Digital Source Digital Players and Recorders: CD , SACD , Tape, Memory Card, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 22nd July 2013, 03:17 AM   #1
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Default linuxguru buffer redux and beyond

Back a while ago linuxguru posted a unity-gain buffer circuit that had some interesting features. See Discrete Opamp Open Design

I studied it and began some investigations as to how its performance (already exemplary for such a simple circuit) might be improved. I mentioned I would eventually get around to posting something. I had occasion to revisit and extend that work recently while looking into an I-V converter that needed a buffer. A schematic of an example is attached.

One of the things I most like about linuxguru's approach is the notion that the JFET input device(s) are basically running the show, but with the heavy lifting into loads done by the rest of the circuitry. Not only is the noise performance outstanding, with a little care mostly due to the JFET(s), but the latter operate at both nearly constant voltage and constant current --- that is, constant dissipation. So the sorts of second-order effects that often plague designs that look otherwise good "on paper" are sidestepped here.

Quite a few more parts than the original, but I think worthwhile. The particular example will drive +/- 20V peak to peak into 600 ohm loads. I have yet to build it, and note the caveats about a propensity to high-frequency oscillations involving one or more active devices. I'd recommend actually stacking the three BF862 in the Z plane as it were, using a bit of bus wire as a combination conductor and heatsink for the gate leads.

The +50V is a constraint from other parts of associated circuits, and can be reduced for other applications, as long as the positive output swing is accomodated. The surface mount DMOS needs to have some copper in the vicinity of the large drain lead connection to manage the dissipations. The circuit needs a little lumped input C for ideal frequency response, which may well be supplied by the source impedance.
Attached Images
File Type: jpg linuxguru-inspired buffer 07-21-13 001.jpg (227.6 KB, 264 views)
  Reply With Quote
Old 22nd July 2013, 05:30 PM   #2
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Default guide to the possibly perplexed

The correspondence between the presented circuit and the original may be somewhat obscured.

Q7 through Q11 are just forming a low-noise high power-supply-rejection ~39mA current source, a function performed in the original circuit by J2 and R2. This could probably be simplified without much loss of performance.

Q3 and neighboring parts are another current generator, and substitute for R2 in the original. The effect of the generator is much higher gain of the loop that detects changes in the output current and stabilizes things. Of course the input impedance at the gate of Q2 is quite high.

There is no great effort made in this circuit by itself to stabilize output voltage drift, as in the intended application of buffering the resistor-terminated output of a current conveyor, there will likely be an overall d.c. servo. However as remarked elsewhere the voltage drift of BF862 parts is fairly small near Idss. If it were to be further stabilized, a matched set or a single matched device could be used to control a current generator that would replace Q7 etc.

Note that the gate-source voltage drift of Q5 is a second-order effect, since the output conductance at the 862 drains is fairly small.
  Reply With Quote
Old 22nd July 2013, 05:57 PM   #3
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Quote:
Originally Posted by bcarso View Post
The correspondence between the presented circuit and the original may be somewhat obscured.

Q7 through Q11 are just forming a low-noise high power-supply-rejection ~39mA current source, a function performed in the original circuit by J2 and R2. This could probably be simplified without much loss of performance.
...
If it were to be further stabilized, a matched set or a single matched device could be used to control a current generator that would replace Q7 etc.
Thanks, Brad - your circuit elaborations are always very illuminating.

For the upper current source, how about a set of 3x BF862 (matched to the lower 3) in parallel, with their gates tied to source, and a depletion-mode Supertex DMOS nFET (or similar) as an upper cascode for all 3 JFETs? That should take care of both the Vds (breakdown) of the BF862s and PSRR.
  Reply With Quote
Old 22nd July 2013, 07:12 PM   #4
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Quote:
Originally Posted by linuxguru View Post
Thanks, Brad - your circuit elaborations are always very illuminating.

For the upper current source, how about a set of 3x BF862 (matched to the lower 3) in parallel, with their gates tied to source, and a depletion-mode Supertex DMOS nFET (or similar) as an upper cascode for all 3 JFETs? That should take care of both the Vds (breakdown) of the BF862s and PSRR.
That sounds like a very acceptable approach, and would eliminate some parts.

By this point in most circuits, we are usually out of the woods with respect to noise anyway, so my concern for the output noise induced by a noisy current generator will usually be unwarranted.

One of the applications of this stage could be as the unity-gain buffer in Sallen-Key filter sections, ones which would typically use opamp voltage followers. Doug Self spends a good deal of time stressing the importance of characterizing opamps for input common-mode distortion, in his book on active crossover networks, a book I have yet to acquire but was able to get some of the content from in online excerpts.

It's interesting that with higher source impedances, for example 10k, it appears that the primary distortion mechanism is the variable input capacitance of the PMOS input device. The distortion predicted thus is still quite benign, but to reduce it further an additional BF862 can preface the PMOS gate, while sharing the drain connections of the other JFETs (a while back I also tried cascoding the PMOS drain, but it seemed intractable to stabilize). Additional current from the upstairs generator is needed, and a current gen for the new source as well. With a 10k resistive source and the added FET, about 2pF is needed to common at the new JFET gate to assure stability against oscillations. The circuit bandwidth looks to be -3dB at about 14MHz under those conditions.

I really do like this approach though. It appears to me hard to beat by anything else I've seen out there, including diamond buffers of various varieties. The adaptation to different loads with little change in performance is particularly appealing.
  Reply With Quote
Old 10th August 2013, 04:03 PM   #5
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Quote:
Originally Posted by bcarso View Post

By this point in most circuits, we are usually out of the woods with respect to noise anyway, so my concern for the output noise induced by a noisy current generator will usually be unwarranted.
Another idea (not really novel) for the upper current source - instead of a somewhat hard-to-find depletion-mode MOSFET, perhaps a jelly-bean LM317 or similar could be put to use as the upper cascode element. It does have a small additional current component through the ADJ pin, but that should be fairly small compared to the 3 x Idss of the JFETs, and will contribute only a small additional offset voltage at the output (which in any case can be taken care off by an outer DC servo).

Of course, it would help if that ADJ current is relatively constant and independent of the input voltage to the LM317.
  Reply With Quote
Old 10th August 2013, 06:05 PM   #6
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
I wonder if anyone has ever looked at the noise in the 317 adjust terminal current? The noise of the device as such used as a regulator is almost certainly strongly dominated by the bandgap reference, so it's possible not too much attention was given to the adjust input current.

I'll be learning more, empirically, about 862 voltage drift soon, as I'm to the breadboard point with a composite amplifier using one 862 and an LME49990. There is a lot of d.c. gain and a servo is essential --- but the question is, once the JFET bias is adjusted at operating temperature for roughly zero volts out of the servo, how much servo range is really needed to assure that it never becomes unable to correct the main amplifier?

In a more extravagant design I've considered a Peltier device, and even stabilizing at a cooler temperature. But this invites issues with condensation if carried too far. The noise advantages are slight unless the impedances are super high, like in a charge amplifier.
  Reply With Quote
Old 10th August 2013, 07:23 PM   #7
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
I did reduce the positive rail in the posted schematic to 40V with no degradation in performance predicted. R9 is the only change, to 28.0k (not very critical). This is a change complementary to the other parts of the I-V converter, which had a 50V positive rail.
  Reply With Quote
Old 10th September 2013, 04:57 AM   #8
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Link (for future reference) to a Kaneda discrete opamp variant which borrows some ideas from the CCB for the output stage:

Discrete Opamp Open Design

In principle, the VAS could use PMOS devices to improve the output swing capability, and the output stage could use NMOS devices like the 2N7002. With biasing changes, the LTP active load could also be PMOS - thus allowing an all-FET discrete opamp topology.

Last edited by linuxguru; 10th September 2013 at 05:06 AM. Reason: addendum
  Reply With Quote
Old 17th October 2013, 07:56 AM   #9
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Link (for reference) to a thread about a Class-A push-pull BJT buffer (by Ken Peter and Elvee) which uses related ideas derived from the c.1973 Taylor source-follower, which in turn is derived from the c.40s White follower:

four transistor emitter follower (diamond buffer) as power output stage?

(The clipping behaviour is a bit messy in simulation, but seems to scale well from mA to A of output-stage Iq.)

Last edited by linuxguru; 17th October 2013 at 08:05 AM. Reason: addendum
  Reply With Quote
Old 17th October 2013, 05:04 PM   #10
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Quote:
Originally Posted by linuxguru View Post
Link (for reference) to a thread about a Class-A push-pull BJT buffer (by Ken Peter and Elvee) which uses related ideas derived from the c.1973 Taylor source-follower, which in turn is derived from the c.40s White follower:

four transistor emitter follower (diamond buffer) as power output stage?

(The clipping behaviour is a bit messy in simulation, but seems to scale well from mA to A of output-stage Iq.)
I'm beginning to look favorably on "quasi-complementary" stages lately, especially if an overall power amp has a bridge output. The efforts made chasing complementarities in P and N devices, whether bipolar or FET, just seem to be ultimately limiting. Of course local error correction usually helps a lot.

I would like to see what Hansen is doing --- as I may have remarked, he believes he has single-handedly resurrected the diamond circuit for an output stage. Naturally the Stereophile writer doesn't know otherwise and dutifully reports this.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
DePalma redux taj Tubes / Valves 12 25th July 2010 05:40 PM
Maida Regulators Redux Eli Duttman Tubes / Valves 27 3rd June 2007 04:31 PM
Critical Q redux Dave Jones Multi-Way 18 20th February 2004 10:09 PM
gainclone cases redux Andrewbee Chip Amps 3 14th July 2003 06:46 PM
tube x-over noise redux jlg4104 Tubes / Valves 3 23rd December 2002 09:07 PM


New To Site? Need Help?

All times are GMT. The time now is 05:53 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2