Sziklai pair power amplifier bias

On 03/14/2016 01:32 PM, John Larkin wrote:
> On Sun, 13 Mar 2016 20:23:27 -0500, M Philbrook
> <jamie_ka1lpa@charter.net> wrote:
>
>> In article <mcmurtrie-66898D.22594612032016@news.sonic.net>,
>> mcmurtrie@pixelmemory.us says...
>>>
>>> In article <C2BEy.28788$%_.1066@fx03.iad>,
>>>   bitrex <bitrex@de.lete.earthlink.net> wrote:
>>>
>>>> I was looking over the schematic for the Roland Jazz Chorus amplifier,
>>>> and the power amp section seems pretty standard, except for the fact
>>>> that the designers apparently chose to use a "PNP" configured Sziklai
>>>> pair as the bias voltage source, to set the quiescent current in the
>>>> output devices (biased into class AB I'd guess.)
>>>>
>>>> Bottom right:
>>>>
>>>> http://tinyurl.com/h4jnsg2
>>>>
>>>> Any ideas as to the reason behind this decision? Thermal stability?
>>>
>>> It looks like it's to regulate the current on R117 and set the bias.
>>> Maybe it's flawed in a way that sounds interesting.
>>
>> The combination looks like they are being used as the diode
>> multiplier. That config gives a little more than a normal forward
>> bias volts on trannies, around .7 to .8 volts.
>>
>> When things warm up abit, I do think the saturation level drops and
>> also reduces the bias in that circuit for Q20 and Q18 to compensate.
>>
>> Looks like a good place for it and a typical quasi output.
>>
>> Jamie
>
> For class AB bipolar amps (which are pretty much ancient history) a
> good scheme is to use biggish emitter resistors paralleled with power
> diodes. No thermal problems, no bias pots.
>
> Better is to use mosfets, each with its own closed-loop opamp to drive
> its gate.
>
>

Class A is still big with the audiophiles, even solid state.

Say 15 watts output, and the output transistors mounted on aluminum 
heatsinks the size of a brick

John Larkin wrote:

> 
> 
> For class AB bipolar amps (which are pretty much ancient history) a
> good scheme is to use biggish emitter resistors paralleled with power
> diodes. No thermal problems, no bias pots.
>

** Class AB bipolar amps are by far the most common both in use and manufacture. Millions of them are pouring out of China each year - both as individual products and incorporated inside items that include speakers. 

The alternative is class D amplifiers that, aside from single chip designs, remain a small minority. 

It is rare to see emitter resistors paralleled with diodes, as it creates issues with crossover distortion and the use of VI limiting. VI limiting relies on emitter resistors to accurately sense current in each BJT and this becomes impossible when there is a diode in parallel. 

The problems of biasing class AB stages were solved long ago.




...  Phil 

On Mon, 14 Mar 2016 18:52:51 -0700 (PDT), Phil Allison
<pallison49@gmail.com> wrote:

>John Larkin wrote:
>
>> 
>> 
>> For class AB bipolar amps (which are pretty much ancient history) a
>> good scheme is to use biggish emitter resistors paralleled with power
>> diodes. No thermal problems, no bias pots.
>>
>
>** Class AB bipolar amps are by far the most common both in use and manufacture. Millions of them are pouring out of China each year - both as individual products and incorporated inside items that include speakers. 
>
>The alternative is class D amplifiers that, aside from single chip designs, remain a small minority. 
>
>It is rare to see emitter resistors paralleled with diodes, as it creates issues with crossover distortion

AB sort of always has crossover distortion. Needs lots of feedback. 


> and the use of VI limiting. VI limiting relies on emitter resistors to accurately sense current in each BJT and this becomes impossible when there is a diode in parallel. 

Good point; it takes an added current sense resistor to fix that
glitch.

Another technique is to run the finals at zero quiescent current and
bleed in some of the driver power to the output, to ensure continuity
in the crossover region. 

>
>The problems of biasing class AB stages were solved long ago.
>

I like to compute (analog or with a uP) actual device power
dissipation and measure heatsink temp, then run a model to simulate
junction temperature. That is a much better way to protect (and push)
the transistors than a current limit.


-- 

John Larkin         Highland Technology, Inc
picosecond timing   precision measurement 

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

John Larkin wrote:
> Phil Allison
> 
>  
> > and the use of VI limiting. VI limiting relies on emitter resistors to accurately sense current in each BJT and this becomes impossible when there is a diode in parallel. 
> 
> Good point; it takes an added current sense resistor to fix that
> glitch.
>

** One is needed for each device and will add to conduction losses at full output.  Nobody does it. 
  
> 
> Another technique is to run the finals at zero quiescent current and
> bleed in some of the driver power to the output, to ensure continuity
> in the crossover region. 
> 

** Then you have a class B output stage. 

 The best and most famous of which is the Quad "Current Dumping" topology which eliminates crossover distortion and biasing issues completely. 


> >The problems of biasing class AB stages were solved long ago.
> >
> 
> I like to compute (analog or with a uP) actual device power
> dissipation and measure heatsink temp, then run a model to simulate
> junction temperature. That is a much better way to protect (and push)
> the transistors than a current limit.
> 

** VI limiting allows use of the available SOA of BJTs and often includes delayed action so brief excursions outside the published DC curves are catered for - important for coping with reactive loads. 

A thermal device that senses heatsink temp and takes appropriate action completes the job of protecting an output stage in case of long term overload or cooling failure. 

Crown Audio ( of the USA ) are the only amp maker know who went down the path of having an analogue computation of SOA built into each power channel in their "Microtech" & "Macrotech" series. Worked well enough but had no practical advantages over a well implemented VI limiter.  



....   Phil

On Wednesday, March 16, 2016 at 4:44:32 AM UTC+11, John Larkin wrote:
> On Mon, 14 Mar 2016 18:52:51 -0700 (PDT), Phil Allison
> <pallison49@gmail.com> wrote: 
> >John Larkin wrote:
>
> >> For class AB bipolar amps (which are pretty much ancient history) a
> >> good scheme is to use biggish emitter resistors paralleled with power
> >> diodes. No thermal problems, no bias pots.
> >
> >** Class AB bipolar amps are by far the most common both in use and manufacture. Millions of them are pouring out of China each year - both as individual products and incorporated inside items that include speakers. 
> >
> >The alternative is class D amplifiers that, aside from single chip designs, remain a small minority. 
> >
> >It is rare to see emitter resistors paralleled with diodes, as it creates issues with crossover distortion
> 
> AB sort of always has crossover distortion. Needs lots of feedback. 

"Sort of" means "very little" if you do your biasing right. Feedback helps, but if you can live with a bias current of about 100mA through a quiescent output stage for an 8 ohm load (and that's not a lot) you can go through crossover without getting any visible kink.
 
> > and the use of VI limiting. VI limiting relies on emitter resistors to accurately sense current in each BJT and this becomes impossible when there is a diode in parallel. 
> 
> Good point; it takes an added current sense resistor to fix that
> glitch.
> 
> Another technique is to run the finals at zero quiescent current and
> bleed in some of the driver power to the output, to ensure continuity
> in the crossover region. 

Not the cleverest way of doing it.

> >The problems of biasing class AB stages were solved long ago.
> 
> I like to compute (analog or with a uP) actual device power
> dissipation and measure heatsink temp, then run a model to simulate
> junction temperature. That is a much better way to protect (and push)
> the transistors than a current limit.

And a lot more complicated. Good analog multipliers are expensive, but uPs are cheap as dirt. Putting an A/D converter in the right place to digitise something that can be translated into output power isn't trivial, but won't complicate manufacture a lot. Putting a thermistor (or whatever) on the output heatsinks to measure their temperatures is dead expensive in mass-produced gear.

Measuring the Vbe of bipolar output devices avoids extra wiring, but you only know the base current at cross-over, which makes for an even more complicated sampling scheme.

-- 
Bill Sloman, Sydney