Current-dumping Class-AB Self Type-II EF O/P - diyAudio
Go Back   Home > Forums > Amplifiers > Solid State

Solid State Talk all about solid state amplification.

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 January 2009, 01:09 PM   #1
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Default Current-dumping Class-AB Self Type-II EF O/P

The following topology is an attempt to graft a current-dumping topology around a stock Class-AB Douglas Self Type-II Emitter-Follower Output Stage. While the original idea (Walker and Albinson, 1975) of current dumping was built around a Class-A stage driving a Class-B stage, there's nothing conceptually difficult about using the same idea to correct the non-linearity (crossover distortion, etc.) of a Class-AB stage as well.

The main idea behind this topology is to treat the LTP, VAS and the driver stage as a linear Class-A stage with gain, and the O/P stage as a separate Class-AB voltage-follower. The emitter-to-emitter resistance between the drivers of the Self Type-II stage is split into two equal resistances, and the Class-A output 'corrector current' is obtained from the mid-point of the two. A balanced reactive bridge (a la Quad 405) is driven by a Class-AB voltage-follower driving the low-impedance side of the bridge, and deriving its input from the high-impedance side, which is itself driven by the Class-A combination of the LTP, VAS and driver stages.

In the schematic shown, a standard CFP LTP (Q1-Q2-Q3-Q4) drives a tricked-up VAS pair (Q7-Q10) into a Wilson current mirror (Q6-Q8-Q9). It simulates well with good sonics at all swings, but there's nothing special about the input and VAS stage, and it could be replaced by anything else, including a bootstrapped VAS. The VAS drives a standard Self Type-II driver and output stage, as noted previously, with the Class-A output being derived from the junction of R21-R22, driving one leg of the balanced reactive bridge consisting of R27-L1-R19-C6. The closed-loop gain is set by the ratio (1 + R19/R18). Stability is obtained by a combination of C3-R8, C4-R12 and the bridge capacitance C6.

As shown, it is designed for 80W with a maximum output swing of +/- 36V with +/- 42V rails, but there's no reason why it won't work with different (higher or lower) rail voltages. The quiescent current in the output stage is not critical, and anything from 50 mA to 200 mA should be OK - the simulated value is around 100 mA.

Edit: This concept has been influenced significantly by the very clever work of our Bulgarian friend Widowmaker, with his DoGC and similar current-dumping topologies that he has posted from time to time on this forum.
Attached Images
File Type: jpg self_type2_quad.jpg (89.0 KB, 1534 views)
  Reply With Quote
Old 22nd January 2009, 01:14 PM   #2
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Here's the simulated THD20 FFT plot at 20W (~ +/- 18 V swing) into 8 ohms. H2 dominates at ~ -114 dB (~ 0.0002%).
Attached Images
File Type: jpg self_type2_quad_20w_fft.jpg (59.5 KB, 1132 views)
  Reply With Quote
Old 22nd January 2009, 03:21 PM   #3
diyAudio Moderator Emeritus
 
lumanauw's Avatar
 
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
Interesting
From a glance, it seems like ordinary 3 stage amp.

What happen (what will be different?) if we dismiss the bridge configuration of R27-L1-R19-C6?
  Reply With Quote
Old 22nd January 2009, 03:29 PM   #4
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
If you remove the balanced bridge (and just retain the feedback gain network, output inductor and Zobel) you'll find that distortion will go up by about 10..20 dB, other things being equal. The contribution of the Class-A stage to the output current is small (~10s of mA at maximum swing), but it plays a crucial role in correcting non-linearity, hence lowering distortion.
  Reply With Quote
Old 22nd January 2009, 04:27 PM   #5
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
are you saying that adding R27 + C6 changes this from a very conventional ClassAB amplifier into a current dumping amplifier?
__________________
regards Andrew T.
  Reply With Quote
Old 22nd January 2009, 05:06 PM   #6
diyAudio Moderator Emeritus
 
lumanauw's Avatar
 
Join Date: Oct 2002
Location: Bandung
Send a message via Yahoo to lumanauw
Hi, Linuxguru,

Do you have your version of ClassS (Sandman) power amp? This topology is also somekind of current dumping.
  Reply With Quote
Old 23rd January 2009, 04:15 AM   #7
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Andrew: Yes, more or less. Note that without current-cumping, we'd typically use a higher impedance feedback network than shown. That would in turn affect the compensation schema.

With current dumping, there's a loose constraint on the range of values R19 to attain bridge balance. If it's made too large, C6 will have to be made very small to maintain bridge balance. If C6 gets into the sub-10 pF range, it will be inconvenient to maintain it accurately because of stray capacitances from the tracks. Small values of C6 also adversely affect stability. That constrains it the the range of 10 pF to a few 100 pF, and accordingly constrains R19. We'd also not want to make R19 too small, which would increase its power dissipation substantially. The values shown are about par for the course, and there's some room for fine tweaking to further minimize distortion.

Note that in the schematic shown, just removing R27 not only increases the distortion, but also the character of the distortion, making H3 dominant over H2. If R27 is retained but C6 is removed, the circuit becomes difficult to stabilize. If both R27 and C6 are removed, distortion and instability both increase. Removing R27 and C6 will force changes to both the feedback network as well as compensation schema with recover stability, but distortion will remain higher regardless.

Lumanauw: I played around with Class S for a short while, but found that current-dumping is able to produce produce comparable or lower distortion with much lower circuit complexity. In simulations, I found that the nature of the high-gain Class-A stage employed is critical to maintain stability in Class-S, and not every Class-A stage works well or has good sonics in Class-S. In particular, there is a H3-dominance in most topologies, even when the Class-A stage simulates well on its own. IMHO, current dumping is easier to tweak for good sonics as well as stability.

However, I'll revisit Class-S (or its Technics variant) in simulation eventually - there is scope for reactive bridging there also, which may aid stability and/or distortion.
  Reply With Quote
Old 23rd January 2009, 05:12 AM   #8
GK is offline GK  Australia
Account disabled at member's request
 
Join Date: Jan 2006
Oh man, have you learnt to simulate loop gain and phase margin yet? These THD-20 figures you are posting would be fanciful even if a pure class A output stage was tacked on.

Cheers,
Glen
  Reply With Quote
Old 23rd January 2009, 05:20 AM   #9
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Glen, why don't you try to understand the Walker and Albinson patent and publications, instead of spouting gibberish? It's fairly simple to verify the LTSpice FFT numbers I posted, and you'll get something similar with a wide variety of BJTs - there's no fancy ultra-linear boutique device required, except for the VAS. You can even substitute the input stage and VAS with something very simple, and you'll still get the 10..20 dB improvement I'm claiming, as compared the version(s) without current dumping.

Edit: With an LME49860 as the Class-A gain stage, and a tricked-up current-dumping Class-A voltage-follower that I'd prefer not to disclose for a few months, I'm already at -160 dB THD20 at 1W into 8 ohms. Even with ordinary diamond buffers as the output stage, you can get to -130..-140 dB without any difficulty. The Class-AB numbers shown are not difficult at all.
  Reply With Quote
Old 23rd January 2009, 07:42 AM   #10
GK is offline GK  Australia
Account disabled at member's request
 
Join Date: Jan 2006
Quote:
Originally posted by linuxguru
Glen, why don't you try to understand the Walker and Albinson patent and publications, instead of spouting gibberish? It's fairly simple to verify the LTSpice FFT numbers I posted, and you'll get something similar with a wide variety of BJTs - there's no fancy ultra-linear boutique device required, except for the VAS. You can even substitute the input stage and VAS with something very simple, and you'll still get the 10..20 dB improvement I'm claiming, as compared the version(s) without current dumping.

Edit: With an LME49860 as the Class-A gain stage, and a tricked-up current-dumping Class-A voltage-follower that I'd prefer not to disclose for a few months, I'm already at -160 dB THD20 at 1W into 8 ohms. Even with ordinary diamond buffers as the output stage, you can get to -130..-140 dB without any difficulty. The Class-AB numbers shown are not difficult at all.


LOL!

The answer to my question is obviously “no”.

I didn’t say anything about the validity of the current dumping technique and you can retain the view that I am talking gibberish if you like. I’ll continue to outline the basic lack of understanding of amplifier frequency compensation on your part anyway:

I have “verified” your simulation results, and the problem with your method of obtaining ultra-low THD figures in SPICE is exactly the same as the last time I “verified” that Mullard inspired amplifier you LTspiced and quoted totally unrealistic (in the real world) THD-20 figures for.

This time the first thing I did was to enter your circuit into LTspice exactly as per the schematic in your opening post, but with the omission of the current dumping network (see attachment).
The output stage was optimally biased with ~50mV emitter-emitter at the output devices (Andy C’s models).

I then ran a transient analysis and THD analysis at 20kHz. THD-20 came in at 0.0013% and, not surprisingly, the amplifier could not even drive (without oscillating) the 10R/100n Zobel network without an isolation inductor.

I then ran a loop gain analysis (see attachment). The unity loop gain frequency is over 6MHz with a phase margin a couple of degrees! There is no way in a million years this circuit will work in real life.

The loop gain at 20kHz is therefore optimistically high and the THD optimistically low.
To return realistic THD-20 figures (which would actually be about 20dB higher) you need to revise the frequency compensation for a unity loop gain frequency at least 6 times lower.
Attached Images
File Type: jpg duff.jpg (81.3 KB, 1064 views)
  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
Commercially available current dumping amps? Key Solid State 30 23rd May 2009 10:19 PM
current dumping for gainclones steven344 Chip Amps 2 20th November 2005 11:42 AM
Current Dumping Amp KYMcJerry Solid State 12 24th August 2005 06:23 AM
Current Dumping. The way forward ! 405man Solid State 18 16th June 2005 12:05 PM


New To Site? Need Help?

All times are GMT. The time now is 12:00 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