Safe Operating Area calculation for audio power amps

[Clifford Heath]

Phil Allison wrote:

"thermal runaway" was a process that took minutes, not ms.
** You were right the first time.
Thermal runaway and second breakdown are two quite different things.

Same basic problem, just occurring on the chip - and much faster.

You just need to realise that amp dissipation is always simply the DC
supply watts MINUS load watts taken over one cycle.

Yes, that's the easy part. It's getting the peak load current into a
reactive load that I wasn't sure of. I think it's just V/|Z| isn't it?

The load power is easy - just the rms current squared times the resistance.

Ok, that's nice, because it also works with a reactive load without
needing to worry about the inductance - only the resistive part
can dissipate the input power.

You should repeat this calculation for a variety of
power levels and plot the results to find the power level that creates the
most amp dissipation.

I've derived an expression for it, instead of a numerical search.
Shouldn't be too hard to simplify and plot now.

BTW In the ETI 480 - the BD139/140 driver transistors are MORE likely to
fail due to over current or SOA limits being exceeded than the outputs are.

Yes, I worried about that - the Hfe of the 2n3055 is so low.
I'm guessing that with MJ15003/4 that one could reduce the bias
current (increase the 470 Ohm resisters) through the BD139/140
to reduce the load on them, while still providing a enough drive
current.

The SC 480 is the better amp.

What transisters does SC480 use instead of BD139/140? Or does it
just do what I've suggested above? It's rated only 70W into 4 Ohms,
from a +-40V supply - this seems unnecessarily wasteful since you
only need a +-24V swing for 70W.

** Those dudes would not have the slightest idea.
Do any of *them* design or even repair guitar amps for a living ??

The bloke who actually sold me the speaker is a pro bass player,
and I believe he uses one of those speakers in his kit. But that
doesn't mean he has experience with more than this one setup.
He also wasn't there when I returned it.

** For bass guitar, the more cone area the better the result.

Ok, I'll see if I can adapt the cab.

Using four 100 dB /watt speakers will give 106 dB /watt efficiency
(anywhere on axis)

Bass is not really directional like this is it? The sonic wavelength
is just too long (13+metres at 25Hz).

Clifford Heath.

 

[Phil Allison]

"Clifford Heath"

Phil Allison wrote:

"thermal runaway" was a process that took minutes, not ms.
** You were right the first time.
Thermal runaway and second breakdown are two quite different things.

Same basic problem, just occurring on the chip - and much faster.

** They are quite different in terms of actual cause, nature and the
possible cures.

That's why they have different names !!

You just need to realise that amp dissipation is always simply the DC
supply watts MINUS load watts taken over one cycle.

Yes, that's the easy part. It's getting the peak load current into a
reactive load that I wasn't sure of. I think it's just V/|Z| isn't it?

** Just remember that "impedance" is the always ratio of the AC voltage to
resulting AC current in a device.

So an 8 ohm "impedance" draws the same current from a sine wave voltage
source as an 8 ohm resistor - the difference is the phase may not be the
same.

The load power is easy - just the rms current squared times the
resistance.

Ok, that's nice, because it also works with a reactive load without
needing to worry about the inductance - only the resistive part
can dissipate the input power.

** Of course.

So if the load is a pure reactance - all the dissipation is in the poor
amp. A load that was a 4 ohm reactance would make your amp dissipate 168
watts at full output - or 42 watts per device !!!

Transformer coupled loads can cause this to happen at low frequencies - so
require special design precautions.

You should repeat this calculation for a variety of power levels and plot
the results to find the power level that creates the most amp
dissipation.

I've derived an expression for it, instead of a numerical search.
Shouldn't be too hard to simplify and plot now.

** You will soon see how inefficient even class B amps are at low powers.

BTW In the ETI 480 - the BD139/140 driver transistors are MORE likely
to fail due to over current or SOA limits being exceeded than the outputs
are.

Yes, I worried about that - the Hfe of the 2n3055 is so low.

** Its gets very low at high currents - taking 1 amp base drive to reach
10 amps.

In the case of the ETI 480, the BD139/140s have more voltage than the
outputs most of the time.

In the case of the SC 480 - they remain almost the same.

I'm guessing that with MJ15003/4 that one could reduce the bias
current (increase the 470 Ohm resisters) through the BD139/140
to reduce the load on them, while still providing a enough drive
current.

** Watch out - the ETI 480 is notorious for being marginally stable.

Even minor changes could create parasitic oscillation problems.

The SC 480 is the better amp.

What transisters does SC480 use instead of BD139/140?

** MJE 340 / 350 - much more rugged that the BDs.

Or does it
just do what I've suggested above? It's rated only 70W into 4 Ohms,
from a +-40V supply - this seems unnecessarily wasteful since you
only need a +-24V swing for 70W.

** That 70 watt / 4ohms figure is due to the poor regulation of the power
transformer used in the project - the SC 480 amp is speced as ** 105
watts /4 ohms ** ( music power) ie when the DC rails do not drop under
load.

The two amps have almost the same efficiency.

Using four 100 dB /watt speakers will give 106 dB /watt efficiency
(anywhere on axis)

Bass is not really directional like this is it?

** Right - low bass is non directional, so the boost in dBs is heard at
any position. The mids and highs are directional - so the boost will be
restricted to listeners reasonably close to the axis. Largish cone drivers
are inherently quite directional at high frequencies.

Ever wondered why some Marshall cabs have a split slope front baffle ??




............... Phil

 

[Clifford Heath]

Phil Allison wrote:

Same basic problem, just occurring on the chip - and much faster.
** They are quite different in terms of actual cause

Why? As the temperature increases the Vbe drops, so that device or
part of the device, or the O/P stage as a whole, turns on harder,
which makes it hotter. Simple +ve feedback cycle based on temp and Vbe.

Totally different in terms of how you cure/avoid them of course.

** Watch out - the ETI 480 is notorious for being marginally stable.
Even minor changes could create parasitic oscillation problems.

I've read that the problems are because the Zobel network lacks the
usual series inductor. I plan to add one - is that all that's needed?

** MJE 340 / 350 - much more rugged that the BDs.

Perhaps I should find some.

Ever wondered why some Marshall cabs have a split slope front baffle ??

Ok, same deal for the curved horns I assume. I've used a parabolic
reflector to increase directional gain in a 40KHz sonar I built.
When the wavelength is only 7mm, you don't need a very big dish!
It was a central transmitter with two receivers 10cm apart, and could
resolve objects distance *and direction*, down to about 2-3 degrees.
Neat for a robot avoiding chair legs etc.

 

[Phil Allison]

"Clifford Heath"

Phil Allison wrote:

Same basic problem, just occurring on the chip - and much faster.
** They are quite different in terms of actual cause

Why?

** The cause of amp thermal runaway is bias instability in the design - a
true class B design has no output device bias and cannot runaway.

The cause of second breakdown is the inherent NTC of doped silicon in the
chip.

As the temperature increases the Vbe drops, so that device or
part of the device, or the O/P stage as a whole, turns on harder,
which makes it hotter. Simple +ve feedback cycle based on temp and Vbe.

** For a start - that is not runaway, merely bias creep.

Totally different in terms of how you cure/avoid them of course.

Repost :

" ** They are quite different in terms of actual cause, nature and the
possible cures. That's why they have different names !! "


** Now - do you want to parrot me again ???

** Watch out - the ETI 480 is notorious for being marginally stable.
Even minor changes could create parasitic oscillation problems.

I've read that the problems are because the Zobel network lacks the
usual series inductor. I plan to add one - is that all that's needed?

** The lack of that inductor is not the problem - parasitic problems will
not go away just by adding one.

An output inductor is used to isolate an amp from small capacitive loads
( 0.01uF to 0.1 uF ) that would cause oscillation if bridged across the
output.

** MJE 340 / 350 - much more rugged that the BDs.

Perhaps I should find some.

* Go read the previous warning again.

Ever wondered why some Marshall cabs have a split slope front baffle ??

Ok, same deal for the curved horns I assume.

** Horns are curved in order to make them work - nothing to do with
dispersion.



................ Phil

 

[Clifford Heath]

Phil Allison wrote:

** Now - do you want to parrot me again ???

I was only agreeing.

** The lack of that inductor is not the problem - parasitic problems will
not go away just by adding one.

Ouch. Is this the warning you mean when I suggested using MJE340/350?

** MJE 340 / 350 - much more rugged that the BDs.
Perhaps I should find some.
* Go read the previous warning again.

Or was it a different one? I guess the MJE340/350 are much faster too,
which could be a problem.

Is the instability a layout issue or something else (and can it be
fixed/avoided)?

 

[Phil Allison]

"Clifford Heath"

Phil Allison wrote:
** Now - do you want to parrot me again ???

I was only agreeing.

** The lack of that inductor is not the problem - parasitic problems
will not go away just by adding one.

Ouch. Is this the warning you mean when I suggested using MJE340/350?

** MJE 340 / 350 - much more rugged that the BDs.
Perhaps I should find some.
* Go read the previous warning again.

Or was it a different one? I guess the MJE340/350 are much faster too,
which could be a problem.

Is the instability a layout issue or something else (and can it be
fixed/avoided)?

** Please do not snip the conversations into little bit like this.

It compeltely destroys continuity, context and sense.

You are a very rude person.





.............. Phil

 

[Pooh Bear]

Clifford Heath wrote:

Phil Allison wrote:
** Says a pommie smartasre who has never bought those types from Dick
Smith Electronics ( who sold various fakes in the past and have now dropped
them) or Jaycar who sell "MOSPEC" brand or WES who still sell fakes

Where in Aus can you buy real ones?

In any country proper franchised distributors are the best source. Of course this
route isn't always open to hobbyists.

I've come across the dreadful MOSPEC brand MJs and they're awful. Sadly the
company still seems to be trading. http://www.mospec.com.tw About 1/2 the die
size of the genuine part and of questionable process type.

Why did you call Graham a smartasre?

Phil does that all the time.

He's no doubt having a go at me for not warning you of the various counterfeit
parts out there. Applies equally to Toshiba parts and others too. Just make sure
you get genuine On-Semi parts. If it says Motorola on the can it's probably fake
since Motorola sold off their discretes line years ago.

Google 'counterfeit MJ15003' and you'll get some useful links.


Graham

 

[Clifford Heath]

Phil Allison wrote:

** Please do not snip the conversations into little bit like this.

I'm sorry you find it hard to read my posts. I do my best to make
them readable. To me, that means removing all unnecessary context,
to reduce re-reading, leaving just enough so folk can work out
which bit I'm responding to. That last one became disorderly in
the process though, sorry.

When I suggested replacing BD139/140 with MJE340/350, you said I
should "go and read the previous warning again". I wanted to
confirm which warning you meant. Was it the one about parasitics?

Can anything else be done to avoid parasitics, or is it a problem
with the board layout causing some unwanted coupling somewhere?

Clifford Heath.

 

[Phil Allison]

"Clifford Heath"

Phil Allison wrote:
** Please do not snip the conversations into little bit like this.

I'm sorry you find it hard to read my posts. I do my best to make
them readable. To me, that means removing all unnecessary context,

** Never remove the context and never assume it is unnecessary .

Your posting style is ABOMINABLE !!!

When I suggested replacing BD139/140 with MJE340/350, you said I
should "go and read the previous warning again". I wanted to
confirm which warning you meant. Was it the one about parasitics?

** Follow the context - could it logically be any other one ?

Can anything else be done to avoid parasitics,

** Yes - build the SC 480, published Jan 2003





.............. Phil

 

[Brian Goldsmith]

"Pooh Bear" <rabbitsfriendsandrelations@hotmail.com> wrote


I've come across the dreadful MOSPEC brand MJs and they're awful. Sadly the
company still seems to be trading. http://www.mospec.com.tw About 1/2 the
die
size of the genuine part and of questionable process type.


****Whioch brings up the question,has there ever been a list of counterfeits
compiled and presented here or elsewhere?

Brian Goldsmith.

 

[Phil Allison]

"Pooh Bear"
Clifford Heath

Why did you call Graham a smartasre?

Phil does that all the time.

He's no doubt having a go at me for not warning you of the various
counterfeit
parts out there. Applies equally to Toshiba parts and others too. Just
make sure
you get genuine On-Semi parts. If it says Motorola on the can it's
probably fake
since Motorola sold off their discretes line years ago.

Google 'counterfeit MJ15003' and you'll get some useful links.

** Yep - and most of what you will find on MJ15003/4 s came from me.

WARNING !!!!

**** Unless you deal with AUTHORISED agents for Motorola you can never
know what you are getting ****


NOTE WELL :

Back in the late 1980s, I bought 100 x MJ15003/4s from a firm in
Melbourne, "Rod Irving Electronics". The devices were branded with the
Motorola "M" but looked suspicious - they were in the obsolete aluminium
package and looked dull and stained - certainly not new - despite having a
current date code inked on.

When I rang the firm, I was assured by the assistant manager the devices
came from a particular Motorola dealer in Australia.

Finding that story totally *incredible" - I rang the dealer alleged to
have supplied the devices and simply asked about any recent sales to Rod
Irving Electronics. No surprise - the computer database showed no such
sale in many years.

Back to Rod Irving Electronics - this time I spoke to the man himself .
When pressed he admitted he had really got them from another source - a
similar retail business in Melbourne - "Jock Ellis Electronics."

Smelling a great big * RAT * by now - I proceeded to do a few tests on
the devices, plus cut a couple open to look at the chips. The conclusion
was inescapable, the devices were really old stock MJ2955 and 2N3055s
rebadged.

I next contacted Motorola head office in Phoenix Arizona - about 2 am
Sydney time or 8 am there. There specialist TO3 bipolar power device sales
guy ( they have sales specialist for every product line !!!! ) was not
too surprised at the story and told me how to compare the inked on dates
with the embossed ones. When the two conflict by several years you know you
have a fake - for sure.

Next, the whole story was presented to Motorola Australia - who just DID
NOT WANT TO KNOW !!!!

Next, I took some samples to Jim Rowe at Electronics Australia magazine -
he was barely able to get anyone from Motorola Australia on the phone to
even discuss the topic - certainly no the then manager, a Mr Bevelacqua.
Nevertheless JR published some photos and a warning in the magazine.

Soon as that happened, JR let me know another dealer (Gary Johnston of
Jaycar) had rung him on the matter. That dealer was * LIVID * since he
could no longer sell the many hundreds of identical devices he had in stock
after the warning published in EA magazine. Turns out he had also bought
them from Jock Ellis Electronics who got them ( allegedly) from a warehouse
operator in Hong Kong.

Before returning my devices to Rod Irving for a full refund, I took the
liberty of using some acetone to remove the inked on markings. You can
guess why I did that.

This prompted a very hostile letter from Rod Irving's lawyer in Melbourne,
saying his client was now unable to get a refund from his supplier due to my
action.

My reply letter to the lawyer was terse and very pointed - as you might
imagine.

I got my refund cheque a week later.




.............. Phil

 

[Phil Allison]

"Brian Goldsmith"

"Pooh Bear"


I've come across the dreadful MOSPEC brand MJs and they're awful. Sadly
the
company still seems to be trading. http://www.mospec.com.tw About 1/2
the
die size of the genuine part and of questionable process type.


****Whioch brings up the question,has there ever been a list of
counterfeits
compiled and presented here or elsewhere?

** No point to that all - as new ones keep popping up all the time.

A purchaser simply HAS to know the source of supply is NOT tainted in
order to be sure of getting genuine devices.
BTW

MOSPEC devices are not fakes of Motorola - they are a licensee maker of
some MJ lines.

They seem however to be using older die designs than the current MJ ones and
doing the mounting in the package a bit less well. Had no real trouble with
them myself.





.............. Phil

 

[Pooh Bear]

Phil Allison wrote:

< snip >

A purchaser simply HAS to know the source of supply is NOT tainted in
order to be sure of getting genuine devices.
BTW

MOSPEC devices are not fakes of Motorola - they are a licensee maker of
some MJ lines.

That would perhaps explain why they haven't been driven out of business.

They seem however to be using older die designs than the current MJ ones and
doing the mounting in the package a bit less well. Had no real trouble with
them myself.

Well, I sure did. More accurately, our Asian subcontractor started using them
and we had absurdly high failure rates on short circuit test here.

I opened one up to look and the die size was barely 1/2 of a genuine Motorola
device ( I think it was actually still Motorola back then too ).

Luckily we caught the problem - replaced all the devices and instructed the
subbie never to touch MOSPEC again.

UK audio semiconductor specialist Profusion briefly distributed MOSPEC - then
dropped the line - no surprise to me at least.


What's your current take on MOSPEC Phil ? Opened any up recently ?


Graham

 

[Phil Allison]

"Pooh Bear

Phil Allison wrote:

A purchaser simply HAS to know the source of supply is NOT tainted in
order to be sure of getting genuine devices.
BTW

MOSPEC devices are not fakes of Motorola - they are a licensee maker of
some MJ lines.

That would perhaps explain why they haven't been driven out of business.


They seem however to be using older die designs than the current MJ ones
and
doing the mounting in the package a bit less well. Had no real trouble
with
them myself.

Well, I sure did.

** But you told us the devices likely came from within India.

MOSPEC is a Taiwan based maker.

More accurately, our Asian subcontractor started using them
and we had absurdly high failure rates on short circuit test here.

I opened one up to look and the die size was barely 1/2 of a genuine
Motorola
device ( I think it was actually still Motorola back then too ).

Luckily we caught the problem - replaced all the devices and instructed
the
subbie never to touch MOSPEC again.

UK audio semiconductor specialist Profusion briefly distributed MOSPEC -
then
dropped the line - no surprise to me at least.

What's your current take on MOSPEC Phil ? Opened any up recently ?

** Taiwan made ones or Indian made fakes of MOSPEC - like your mob
bought ??

( The recent DSE MJ..... fakes were thought to be from China or
India. )





.............. Phil

 

[Pooh Bear]

Phil Allison wrote:

"Pooh Bear
Phil Allison wrote:

A purchaser simply HAS to know the source of supply is NOT tainted in
order to be sure of getting genuine devices.
BTW

MOSPEC devices are not fakes of Motorola - they are a licensee maker of
some MJ lines.

That would perhaps explain why they haven't been driven out of business.


They seem however to be using older die designs than the current MJ ones
and
doing the mounting in the package a bit less well. Had no real trouble
with
them myself.

Well, I sure did.

** But you told us the devices likely came from within India.

That was the best information available to me at the time. Our sub-contractor
was somewhat vague about their origin. I was just glad to get rid of them. Maybe
they meant they came from an Indian distributor ? Working with Indians often
gives rise to such puzzles.

I see that Indian counterfeits are indeed mentioned in the info that's available
online. Seemingly another issue unless these were Indian counterfeits of genuine
Mospec parts ????

MOSPEC is a Taiwan based maker.

Indeed. I have recently looked into this matter again - and you're right.

More accurately, our Asian subcontractor started using them
and we had absurdly high failure rates on short circuit test here.

I opened one up to look and the die size was barely 1/2 of a genuine
Motorola
device ( I think it was actually still Motorola back then too ).

Luckily we caught the problem - replaced all the devices and instructed
the
subbie never to touch MOSPEC again.

UK audio semiconductor specialist Profusion briefly distributed MOSPEC -
then
dropped the line - no surprise to me at least.

What's your current take on MOSPEC Phil ? Opened any up recently ?

** Taiwan made ones or Indian made fakes of MOSPEC - like your mob
bought ??

( The recent DSE MJ..... fakes were thought to be from China or
India. )

Unknown now. It was a long time ago when I came across the problem devices and
none have survived that I could now re-analyse. Must be easily 6 yrs ago -
probably more.

To repeat though. Have you taken the lid off a real MOSPEC MJ to compare die
size. ? Would be interested to know the result and any general comments you
might have about MOSPEC parts.


Graham

 

[Phil Allison]

"Pooh Bear"

Phil Allison wrote:

To repeat though. Have you taken the lid off a real MOSPEC MJ to compare
die
size. ? Would be interested to know the result and any general comments
you
might have about MOSPEC parts.

** Try reading what I posted already:


" They seem however to be using older die designs than the current MJ ones
and doing the mounting in the package a bit less well. "


Now - how do you reckon I know that ??



.............. Phil

 

[Pooh Bear]

Phil Allison wrote:

"Pooh Bear"
Phil Allison wrote:

To repeat though. Have you taken the lid off a real MOSPEC MJ to compare
die
size. ? Would be interested to know the result and any general comments
you
might have about MOSPEC parts.


** Try reading what I posted already:

" They seem however to be using older die designs than the current MJ ones
and doing the mounting in the package a bit less well. "

Now - how do you reckon I know that ??

I did get that bit. I'm surprised that the die design has significantly changed
on a mature device like that though.

Can I assume you're happy with die size then ? If anything, it seems that semi
makers are keen to reduce die size. I've heard apochryphally that the die in a
2SC5200 is smaller than the 2SC3281 for example despite the outwardly similar
spec.


Graham

 

[Phil Allison]

"Pooh Bear"

Phil Allison wrote:

To repeat though. Have you taken the lid off a real MOSPEC MJ to compare
die size. ? Would be interested to know the result and any general
comments
you might have about MOSPEC parts.


** Try reading what I posted already:

" They seem however to be using older die designs than the current MJ
ones
and doing the mounting in the package a bit less well. "

Now - how do you reckon I know that ??


I did get that bit.

** But like most things - it did not sink in ??

I'm surprised that the die design has significantly changed
on a mature device like that though.

** Yawn - shows how far out of touch you are.

Can I assume you're happy with die size then ?

** Same size and design as all MJ15003/4s sold up till the late 1990s.

If anything, it seems that semi makers are keen to reduce die size.

** Die size is not a reliable guide to power rating or overall quality.

I have a collection of 2N3055s where the variation in area spans a range of
5:1. The smallest is the Motorola one - known to be a good example of the
beast.

I've heard apochryphally that the die in a
2SC5200 is smaller than the 2SC3281 for example despite the outwardly
similar
spec.

** Quite possible - IME.

There was a big change in the MJ15024/25 chips a few years back -
Motorola initially used the new number, MJ21193/4 for the revised design.
Then, as they often do, reverted back to the familiar number. You can buy
both numbers now - exactly the same insides however. MJ21193/4s will cost
you more though.

The MJ21193/4 chip are identical to the ones inside the plastic
JL21193/4s - plastic packs cannot normally use chips designed for TO3
packs - that is how the new design came about. Now, one size fits all


Over the last 20 years, Motorola has sold the MJ15003 device badged as a
2N3773, MJ 802 and others. Similarly the MJ15003 has be badged as an MJ4502
and several others - with a price hike attached !!

Seems Motorola figure that any device that equals or exceeds the basic
specs of another can be sold as that other !! Plays havoc with trying to
match up new and old devices in parallel sets - "your problem" says
Motorola.

Genuine MJ802/4502 pairs will give lower THD figures in amps than the later
MJ15003/4s - or if the rebadged ones are fitted. Lots of traps for the
innocent.

BTW

If you follow EW magazine, you will be aware of Doug Self's " blameless "
amplifier design that nobody who has tried can duplicate the same
vanishingly low THD figures.

In the fine print of the article, Doug says he used devices from his
private stash of old (and therefore genuine) MJ802/4502s in the prototypes.


QED.




............... Phil

 

[Pooh Bear]

Clifford Heath wrote:

Phil Allison wrote:

BTW In the ETI 480 - the BD139/140 driver transistors are MORE likely to
fail due to over current or SOA limits being exceeded than the outputs are.

Yes, I worried about that - the Hfe of the 2n3055 is so low.
I'm guessing that with MJ15003/4 that one could reduce the bias
current (increase the 470 Ohm resisters) through the BD139/140
to reduce the load on them, while still providing a enough drive
current.

Just wanted to pop in here and add my comment.

Phil's absolutely right that the driver dissipation is another issue that needs to
be addressed. I should have mentioned that myself earlier.

Can you fit a TO-220 style device in place of the TO-216 ? BD's ? That opens up
your options considerably.


Mind you - beware of possible side effects ! I recall the popular Linsley Hood
amplifer that many of my friends built as kits. Depending on whether you got
Motorola devices or RCA it might or might not self-oscillate itself to destruction
due to the different fT of the respective devices. The design was clearly
marginally stable.


Graham

 

[Phil Allison]

"Pooh Bear"

Phil Allison wrote:

BTW In the ETI 480 - the BD139/140 driver transistors are MORE
likely to
fail due to over current or SOA limits being exceeded than the outputs
are.

Just wanted to pop in here and add my comment.

Phil's absolutely right that the driver dissipation is another issue that
needs to
be addressed. I should have mentioned that myself earlier.

** Check out this site - the ETI 480 is one of the amps featured and you
will find the schematic in the list of zip files.

http://www.alphalink.com.au/~cambie/


Why not have a go at simulating it ???

Try to determine the life expectancy of the BD139/140s when the output is
shorted ;-)




.............. Phil