Discrete Opamp Open Design - Page 233 - diyAudio
Go Back   Home > Forums > Source & Line > Analog Line Level

Analog Line Level Preamplifiers , Passive Pre-amps, Crossovers, 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 7th December 2012, 10:24 PM   #2321
diyAudio Member
 
Join Date: Nov 2011
Location: Cooktown, Oz
I've been trying to meet at least the spirit of Marshy's requirement for less active devices without spoiling the performance of the #2060 JFET990. Presently, it rivals SW-OPA ... at least in SPICE world

A key factor in this performance is the 'pure' Cherry compensation which keeps the Q10/15 output bases at very high Z.

Taking the VAS cascode reference to the emitter of Q16 helps and gets rid of one diode. It shows slightly less THD and better stability too. This type of cascode, used in the i/p as per Guru Wurcer, is responsible for good CMR & CM range.

Other small changes are ..

- Q13 becomes D2. R1 value is one area where the SPICE models are inaccurate. I'm certain a diode connected BJT will have less voltage drop than 1n4148/914. But its just a resistor value and this is our tweak point for output offset.
- Q16 is shown as BC557C instead of the 2n4403 cos it gives 100dB from 90dB distortion products. I doubt the difference will be seen in a real amp.
- C9 across the bias BJT tidies up O/L slightly which is obviously an obsession of mine. I try to keep caps small to minimize O/L recovery time.

J5 + R17 is just a CCS and could be replaced with a 0.5mA device provided it could withstand full VCC - VEE. Suggestions for this anyone?
Attached Images
File Type: gif JFET990-9.gif (52.6 KB, 368 views)
  Reply With Quote
Old 8th December 2012, 02:02 AM   #2322
RNMarsh is offline RNMarsh  United States
diyAudio Member
 
RNMarsh's Avatar
 
Join Date: May 2012
Location: 2457 Cascade Trail; Cool, CA. 95614
Quote:
Originally Posted by kgrlee View Post
I've been trying to meet at least the spirit of Marshy's requirement for less active devices
J5 + R17 is just a CCS and could be replaced with a 0.5mA device provided it could withstand full VCC - VEE. Suggestions for this anyone?
I've been a fan of fet constant current 'diodes' since the '70's. The commercial ones (Siliconix) rated at 55volts are actually cascoded inside - thus the higher v rating- also have higher Zout and thats good for cmr. (for Q17, too) -RNMarshy

Last edited by RNMarsh; 8th December 2012 at 02:07 AM.
  Reply With Quote
Old 8th December 2012, 03:46 AM   #2323
diyAudio Member
 
Join Date: Nov 2011
Location: Cooktown, Oz
Quote:
Originally Posted by RNMarsh View Post
I've been a fan of fet constant current 'diodes' since the '70's. The commercial ones (Siliconix) rated at 55volts are actually cascoded inside - thus the higher v rating- also have higher Zout and thats good for cmr. (for Q17, too)
Thanks for this Mr. Marsh. It's certainly what's required to replace J5. I'm reminded I have some SPICE models for Siliconix diodes too.

Presently, Q17 and resistor allows flexibility in choice of i/p LTP currents but a Siliconix diode would be perfect for production.

I'm quite a Luddite and in da old days would round up on young engineers whose designs had more bits than the products they replaced. Today, with SM, it's probably worth counting solder joints rather than bits but I still feel parts count is a significant indicator of reliability especially for a DIY item.
_____________________

This is a segue into another variant of JFET990. While investigating the 'bootstrapped cascode', I came across Baxandall Super Pair which suggested the Great Guru originally developed it for high Ro .. just what's required for 'pure' Cherry.

This little change
  • gets rid of 2 more diodes and on preliminary tests
  • appears more stable than #2060 & #2321.
  • Also overload is even better though we are now at the stage of shoving more than +/- 15Vp down their throats.
In essence, Q16 is re-introduced as an emitter follower (as JE990) but w/o the need for an evil latching diode across the Miller cap. I apologise to GGB's disciples who want a more respectful description of the Master's work.

CONS?
  • Some careful balancing of input currents, current mirror resistors & R4 to get at least Vce=0.5V across important BJTs else the excellent O/L & THD is compromised
  • Re-introducing another gain stage may cause stability problems. I haven't done a full stability analysis & investigation. The Baxandall Super Pair thread has many caveats.
  • The current mirror has to use smaller resistors so now contributes more noise. But the BF862s & 20R resistors still dominate. May change if we run as much i/p current as SW-OPA. See Sam Groner's critique of Self's PA book for details.
Getting to the stage where I regret being a beach bum

ricardo.txt at #2306
Attached Images
File Type: gif JFET990-10.gif (52.1 KB, 345 views)
Attached Files
File Type: asc 990-10.asc (9.4 KB, 21 views)

Last edited by kgrlee; 8th December 2012 at 04:01 AM.
  Reply With Quote
Old 8th December 2012, 10:52 AM   #2324
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Quote:
Originally Posted by kgrlee View Post

This is a segue into another variant of JFET990. While investigating the 'bootstrapped cascode', I came across Baxandall Super Pair which suggested the Great Guru originally developed it for high Ro .. just what's required for 'pure' Cherry.

This little change
  • gets rid of 2 more diodes and on preliminary tests
  • appears more stable than #2060 & #2321.
  • Also overload is even better though we are now at the stage of shoving more than +/- 15Vp down their throats.
In essence, Q16 is re-introduced as an emitter follower (as JE990) but w/o the need for an evil latching diode across the Miller cap. I apologise to GGB's disciples who want a more respectful description of the Master's work.

CONS?
  • Some careful balancing of input currents, current mirror resistors & R4 to get at least Vce=0.5V across important BJTs else the excellent O/L & THD is compromised
  • Re-introducing another gain stage may cause stability problems. I haven't done a full stability analysis & investigation. The Baxandall Super Pair thread has many caveats.
  • The current mirror has to use smaller resistors so now contributes more noise. But the BF862s & 20R resistors still dominate. May change if we run as much i/p current as SW-OPA. See Sam Groner's critique of Self's PA book for details.
Getting to the stage where I regret being a beach bum

ricardo.txt at #2306
Baxandall (historically, properly attributed to the late Frank S. Boxall who invented it in ~1957) has drawbacks compared to the bootstrapped cascode (the latter also developed long before Hawksford's popularizations among audio workers). The bootstrapped cascode reduces base width modulation, Boxall does not. Transconductance overall is reduced, comparatively. For both topologies the effectiveness of the stage depends on how large the impedance in the main device's emitter is (higher is better for recovering more of the base current loss of the main device, as the returned current divides between the main emitter impedance and the external impedance).

Stability-against-oscillation issues, as you noted for Boxall, exist for both approaches (see my recent note on a mostly-air circuit using the bootstrapped cascode, which needed signficant R in series with the output collector for stability, and various remarks about stabilizing Boxall [-Larson-Baxandall-Shallow etc.] in the Baxandall Super Pair thread you referenced, and elsewhere (?) ).

An advantage of Boxall is the partial temperature compensation afforded by the opposing Vbe's. I stumbled on it way late by asking a what-if when using a complementary transistor for temp-comped bias, as to what would happen if something were done with the bias device's collector current? It was bad enough to discover that Larson at Tektronix and PJ Baxandall had stolen the idea when I was about 18, but even worse when I found the Boxall ref.
  Reply With Quote
Old 8th December 2012, 04:42 PM   #2325
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Quote:
Originally Posted by kgrlee View Post
... just what's required for 'pure' Cherry.

... Q16 is re-introduced as an emitter follower (as JE990) but w/o the need for an evil latching diode across the Miller cap.
Referencing the VAS Miller cap to the output node, a la AD797, is right on the money - at least, that's what LTSpice says.

Regarding the diode across the Miller cap, I thought it was a Baker antisaturation clamp for the VAS. The Blameless-style voltage-follower VAS seems to work fine without the diode, and getting rid of its non-linear junction capacitance helps reduce H3 quite a bit.

It also helps to use separate current sources for the LTP and VAS, to reduce the possibility of overload/clipping transients spilling over from the VAS into the LTP. I think Douglas Self went wrong in the Blameless by using a shared biasing scheme for both LTP and VAS current sources.
  Reply With Quote
Old 8th December 2012, 05:01 PM   #2326
dadod is offline dadod  Croatia
diyAudio Member
 
dadod's Avatar
 
Join Date: Apr 2006
Location: Zagreb
Quote:
Originally Posted by linuxguru View Post
I think Douglas Self went wrong in the Blameless by using a shared biasing scheme for both LTP and VAS current sources.

Agreed
  Reply With Quote
Old 8th December 2012, 05:22 PM   #2327
RNMarsh is offline RNMarsh  United States
diyAudio Member
 
RNMarsh's Avatar
 
Join Date: May 2012
Location: 2457 Cascade Trail; Cool, CA. 95614
If you sweep the values of Rs (R3/4) do you get a null in any harmonic..... especially at 1/Yfs value?
  Reply With Quote
Old 8th December 2012, 08:44 PM   #2328
PHEONIX is offline PHEONIX  Australia
diyAudio Member
 
Join Date: Jul 2004
Location: Australia
Quote:
Originally Posted by bcarso View Post
Baxandall (historically, properly attributed to the late Frank S. Boxall who invented it in ~1957) has drawbacks compared to the bootstrapped cascode (the latter also developed long before Hawksford's popularizations among audio workers). The bootstrapped cascode reduces base width modulation, Boxall does not. Transconductance overall is reduced, comparatively. For both topologies the effectiveness of the stage depends on how large the impedance in the main device's emitter is (higher is better for recovering more of the base current loss of the main device, as the returned current divides between the main emitter impedance and the external impedance).

Stability-against-oscillation issues, as you noted for Boxall, exist for both approaches (see my recent note on a mostly-air circuit using the bootstrapped cascode, which needed signficant R in series with the output collector for stability, and various remarks about stabilizing Boxall [-Larson-Baxandall-Shallow etc.] in the Baxandall Super Pair thread you referenced, and elsewhere (?) ).

An advantage of Boxall is the partial temperature compensation afforded by the opposing Vbe's. I stumbled on it way late by asking a what-if when using a complementary transistor for temp-comped bias, as to what would happen if something were done with the bias device's collector current? It was bad enough to discover that Larson at Tektronix and PJ Baxandall had stolen the idea when I was about 18, but even worse when I found the Boxall ref.
Hello Brad ,

The Boxall in practice achieves lower THD figures thats why its appealing to use.
  Reply With Quote
Old 8th December 2012, 09:07 PM   #2329
bcarso is offline bcarso  United States
diyAudio Member
 
Join Date: May 2005
Location: Canoga Park, California
Quote:
Originally Posted by PHEONIX View Post
Hello Brad ,

The Boxall in practice achieves lower THD figures thats why its appealing to use.
If one drives Boxall with a high-Z source, like the collector of another transistor, to essentially remove the base-width modulation effect on Vbe, the performance can be fantastic. See the Danyuk reference in Walt's bibliography for example.

As simply a gain stage with an emitter R in the output device, although there can be some cancellation of even-order due to the complementary devices, I was a little disappointed.

Both stages look not-too-elaborate at first glimpse, but the details of biasing are one of the necessary complexities for both.
  Reply With Quote
Old 8th December 2012, 09:16 PM   #2330
diyAudio Member
 
Join Date: Nov 2011
Location: Cooktown, Oz
Quote:
Originally Posted by linuxguru View Post
Referencing the VAS Miller cap to the output node, a la AD797, is right on the money - at least, that's what LTSpice says.
Works in real life too. Especially for xover distortion. It's no longer a Miller cap but Prof. Edward Cherry's scheme. Baxandall & Self now promote their TMC bastardised scheme which is Miller / Cherry but this doesn't give the full advantages of pure Cherry.

Quote:
.. the diode across the Miller cap, ... The Blameless-style voltage-follower VAS seems to work fine without the diode, and getting rid of its non-linear junction capacitance helps reduce H3 quite a bit.
Works without the diode in Self's power amps cos the output compound emitter-followers/CFPs. Not with the single Class B stage here and in JE990. Horrible O/L.

Most of this project has been 'how to get rid of evil Cob effects without introducing latching' It's key to getting SOTA SW-OPA performance from a stone age circuit.

Quote:
It also helps to use separate current sources for the LTP and VAS, to reduce the possibility of overload/clipping transients spilling over from the VAS into the LTP. I think Douglas Self went wrong in the Blameless by using a shared biasing scheme for both LTP and VAS current sources.
Jensen claims this in his AES paper. I don't see any evidence in my sims and don't recall any problems from Jurassic REAL life.

Anyone got evidence? Either sim or REAL life

#2323 was a long evolution from JE990+Scott's FET i/p. Each change was tested ala #2306. I don't do fashion changes. Only in response to something which needs improving.

If I get rid of another 2 BJTs I'll meet Marshy's requirements for a production OPA OK. I'm cheating by not counting CCS diodes as actives.
  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
Discrete OPAMP audio-gd Vendor's Bazaar 27 20th September 2012 04:02 PM
discrete opamp help blackpowderaudio Parts 0 16th December 2009 03:46 PM
THAT transistor headphone amp (250ma discrete opamp) design sanity check. Russ White Headphone Systems 19 13th December 2007 12:52 PM


New To Site? Need Help?

All times are GMT. The time now is 07:41 PM.


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