Class H amp - stability issies - 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 12th October 2007, 05:33 PM   #1
wg_ski is offline wg_ski  United States
diyAudio Member
 
Join Date: Oct 2007
Default Class H amp - stability issies

I've been working on a class H amplifier for the past couple weekends and run into some stability problems. The only unusual thing about this one is that the output stage is (or more correctly, was) QC. The basic amp running directly off the low rail is stable. But adding the commutators makes the fake PNP side go compeltely nuts - even with it permanently switched to the high rail. Is this just not possible/not a good idea? I've tried overcomp, converting from Miller to lead/lag comp, every combination imaginable on the zobel, intentionally slowing the whole OPS down, and nothing shuts it up. Popping in some MJ15016's and doing a little re-routing on the board and it works PERFECTLY. The original objective was to use up a couple dozen 2N3773's and a high current 60-30-0-30-60 tranny that weren't going into anything important. Anyone who has messed with these amps have any ideas before I give up and find some more spare PNPs for the other three channels?
  Reply With Quote
Old 13th October 2007, 07:56 AM   #2
diyAudio Member
 
Workhorse's Avatar
 
Join Date: Dec 2004
Default Re: Class H amp - stability issies

Quote:
Originally posted by wg_ski
I've been working on a class H amplifier for the past couple weekends and run into some stability problems. The only unusual thing about this one is that the output stage is (or more correctly, was) QC. The basic amp running directly off the low rail is stable. But adding the commutators makes the fake PNP side go compeltely nuts - even with it permanently switched to the high rail. Is this just not possible/not a good idea? I've tried overcomp, converting from Miller to lead/lag comp, every combination imaginable on the zobel, intentionally slowing the whole OPS down, and nothing shuts it up. Popping in some MJ15016's and doing a little re-routing on the board and it works PERFECTLY. The original objective was to use up a couple dozen 2N3773's and a high current 60-30-0-30-60 tranny that weren't going into anything important. Anyone who has messed with these amps have any ideas before I give up and find some more spare PNPs for the other three channels?
Post your schematic, then only you could expect some help, look at this thread
The Class - H Amplifier
QC Class-H is not a big deal!
  Reply With Quote
Old 15th October 2007, 08:31 PM   #3
wg_ski is offline wg_ski  United States
diyAudio Member
 
Join Date: Oct 2007
Default Re: Re: Class H amp - stability issies

Quote:
Originally posted by Workhorse


Post your schematic, then only you could expect some help, look at this thread
The Class - H Amplifier
QC Class-H is not a big deal!
That's the first thing that came up searching 'class H'. I know it shouldn't be a "big deal", which I was confused as to why it misbehaved so badly. And then behaves normally after converting to full complementary in a fit of rage (because nothing else worked). There's nothing in the circuit that hasn't been done before, but probably not in this combination. Schematic attached (I hope - I don't think it took the first time and I stared over).
Attached Files
File Type: zip schem_zip.zip (98.0 KB, 132 views)
  Reply With Quote
Old 16th October 2007, 04:20 AM   #4
diyAudio Member
 
Workhorse's Avatar
 
Join Date: Dec 2004
Remove both 68pF from C-B junction of VAS, also remove 27pF from collector node to opamp input and then connect 2 X 470pF from each base of predriver stage to ground and insert another 22ohm in the collector path of lower side 2N3773 driver and connect the CFB 1N4007 diode through it.

The Class-H mechanism in your schematic is direct copy of QSC......its not very reliable though.......The LM311 is very sensitive to local noise of switching itself. Sometimes during hard switching it usually false trigger many times, i have seen this in many amps.
Attached Images
File Type: jpg qch.jpg (10.8 KB, 298 views)
  Reply With Quote
Old 16th October 2007, 05:20 AM   #5
Eva is offline Eva  Spain
diyAudio Member
 
Eva's Avatar
 
Join Date: Oct 2003
Location: Near the sea
Send a message via MSN to Eva
Bouncing in commercial class H amplifiers? Really?
Good old audio analog designers going partially switchmode and showing their lack of skills? )

I think that some hysteresis would be a good idea. Controlled dV/dt switching is also a good idea (then phase lead in the reference signal is required).

I also recommend powering everything except the output stage from independently filtered high voltage rails because switching produces (disturbing) ripple in the power rails (It should be locally filtered with CLC pi filters anyway...)
__________________
I use to feel like the small child in The Emperor's New Clothes tale
  Reply With Quote
Old 16th October 2007, 05:26 AM   #6
diyAudio Member
 
Workhorse's Avatar
 
Join Date: Dec 2004
Quote:
Originally posted by Eva
Bouncing in commercial class H amplifiers? Really?
Good old audio analog designers going partially switchmode and showing their lack of skills? )

I think that some hysteresis would be a good idea. Controlled dV/dt switching is also a good idea (then phase lead in the reference signal is required).

I also recommend powering everything except the output stage from independently filtered high voltage rails because switching produces (disturbing) ripple in the power rails (It should be locally filtered with CLC pi filters anyway...)
Evita If you look at switching waveforms of QSC RMX series, they exhibit severe bouncing and sometimes false triggering of rails....

Meanwhile in my version of Class-H this type of behaviour is non-existant....... its free from bouncing and false triggering...

If you choose to filter the rails individually such as individual caps for low and high rails, then they exhibit a sort of pumping which i had seen in Phonic XP3000, but if you choose series stack capacitors, the power rails donot show pumping effect to that extent.
Try to clip a QSC PL325 and you will see that the adjacent rails were also falsely triggered..........
  Reply With Quote
Old 16th October 2007, 05:47 AM   #7
Account Disabled
 
Join Date: Apr 2007
Quote:
Originally posted by Eva
I think that some hysteresis would be a good idea. Controlled dV/dt switching is also a good idea (then phase lead in the reference signal is required).
In PL6.0 we have implemented Hysteresis mode in order to over come false triggerring and the slope is based on predictive switching
  Reply With Quote
Old 16th October 2007, 04:12 PM   #8
wg_ski is offline wg_ski  United States
diyAudio Member
 
Join Date: Oct 2007
Quote:
Originally posted by Workhorse
Remove both 68pF from C-B junction of VAS, also remove 27pF from collector node to opamp input and then connect 2 X 470pF from each base of predriver stage to ground and insert another 22ohm in the collector path of lower side 2N3773 driver and connect the CFB 1N4007 diode through it.

The Class-H mechanism in your schematic is direct copy of QSC......its not very reliable though.......The LM311 is very sensitive to local noise of switching itself. Sometimes during hard switching it usually false trigger many times, i have seen this in many amps.
IOW, compensate like a BGW500 :-). I guess pure lag comp works better with this type of o/p stage? I originally left the 27pF in place when I was experimenting, and no dice. I'll try building up another channel this week and hook up the CFP as you suggest - I hadn't tried that in combination with .001's to gnd off the VAS yet. The last QC I did had only 2 o/p's in parallel and the only thing that bothered it was insufficient supply bypassing. Which is why I was suspecting too much impedance through the commutators somehow mucking with things. More outputs in parallel would only make the parasitics worse. Does lag compensation make the o/p stage less susceptible to nonzero supply side impedance?

The commutators seem to be reasonably well behaved in the board that I've stuck PNPs in. But I haven't hit it with 20k sines or square waves yet - switching seems to be ok with music or pure bass. But I am planning to take things a bit further - like trying to bridge into 2 ohms and see where it craps out. And to see if the larger IRFP2907's can be made to switch fast enough - for subwoofer duty at least. There's a 3-tier in the PCB layout phase. Would the QSC switch have better noise immunity if it were powered off a completely independent 15V supply? I was considering this for 'the big one' for a number of reasons.
  Reply With Quote
Old 16th October 2007, 07:51 PM   #9
TOINO is offline TOINO  Portugal
diyAudio Member
 
TOINO's Avatar
 
Join Date: Oct 2004
Location: Coimbra
Quote:
Originally posted by Eva
Good old audio analog designers going partially switchmode and showing their lack of skills? )

I think that some hysteresis would be a good idea. Controlled dV/dt switching is also a good idea (then phase lead in the reference signal is required).

I also recommend powering everything except the output stage from independently filtered high voltage rails because switching produces (disturbing) ripple in the power rails (It should be locally filtered with CLC pi filters anyway...)

Quote:
Originally posted by Workhorse


Evita If you look at switching waveforms of QSC RMX series, they exhibit severe bouncing and sometimes false triggering of rails....

Meanwhile in my version of Class-H this type of behaviour is non-existant....... its free from bouncing and false triggering...

If you choose to filter the rails individually such as individual caps for low and high rails, then they exhibit a sort of pumping which i had seen in Phonic XP3000, but if you choose series stack capacitors, the power rails donot show pumping effect to that extent.
Try to clip a QSC PL325 and you will see that the adjacent rails were also falsely triggered..........



Come on people… if you take a closer look to the switching circuit, you will see some hysteresis done by the 470K resistor between output and + input of LM311; also the 47pF capacitor between bases of gate drivers and mosfet drain controls the switching dV/dt in collaboration with resistor and diode from output comparator. Also the positive driver could supply the gate with a voltage superior to the voltage rail. All this functions done with great simplicity and without special floating supplies.

This gate driver circuit is sensitive to external noise as any other circuit. Good layout is mandatory here or in any other circuit. If the main supply is not enough for the job, as it seems usual nowadays, don’t blame the switching circuit; class-H is a heavy compromise anyway.

False triggering could happen with any circuit but with this class-H arrangement could be catastrophic if happen with both switches simultaneously.
Mutual exclusive switching arrangement is prudent here…
__________________
Yo no creo en brujas pero que las hay... las hay (Cervantes)
  Reply With Quote
Old 18th October 2007, 03:28 PM   #10
wg_ski is offline wg_ski  United States
diyAudio Member
 
Join Date: Oct 2007
Default Still unstable...

Quote:
Originally posted by Workhorse
Remove both 68pF from C-B junction of VAS, also remove 27pF from collector node to opamp input and then connect 2 X 470pF from each base of predriver stage to ground and insert another 22ohm in the collector path of lower side 2N3773 driver and connect the CFB 1N4007 diode through it.

Connected the CFP in this manner, and then tried the lag comp (again). Without the 27pF cap, it's globally unstable, but this mode is nondestructive. With or without, with either the 68 pF Miller caps or 500pF on each side the gnd, the local loop on the negative side is still unstable above about -10V peak unless the rail is connected *directly* to the power suppy. Shorting emitter to base on the driver and rebiasing *will* fix it, but I'd wonder about how well it would drive 1 ohm loads that way. With the whole OPS switched to true complementary, it will work with the triple. So I'll build up some new boards that way so I can get on with what I intended to use this thing for.

Thanks for the help anyway, I had nothing to lose by trying a few more things.
  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
Amp stability unclear Khron Solid State 30 3rd April 2009 06:45 AM
Class A SE power and driver transistor choice for stability and performance ionomolo Solid State 3 16th February 2009 09:09 PM
LF357 stability richwalters Chip Amps 3 15th March 2005 05:38 AM
stability djdamix Solid State 3 7th January 2005 07:30 AM
Aleph-X stability stefan Pass Labs 2 30th October 2002 09:00 PM


New To Site? Need Help?

All times are GMT. The time now is 05:50 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