Audio Power Amplifier Design book- Douglas Self wants your opinions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
AX tech editor
Joined 2002
Paid Member
You lost me - this is about analog electronics, there's nothing like deciding and branching. There is no 1uS SOA limit, it's about how much a device can survive outside the SOA limits (that is, until the hot spots melt end short the device). That's well under 1uS.

Think non destructive SOA testing, lots of references about in the last 20 years.

Ohh no, not at all. 1uS is a limit that goes with say, 50V Vce and 10A Ic. If you limit it to a period of say 10mS you may dissipate say 40V Vce with say 8A (I'm just making up these numbers to show the principle).

So the SOA is basically three-dimensional: combinations of Vce, Ic and time duration determine whether it is 'safe' or not.
See for instance fig 1 here: Safe Operating Area calculations

So, if you measure say a particular Ic and Vce combination you need to decide how long you can allow that before activating the protection. Problem is that the Vce and Ic combinations change dynamically so you need some way to keep track of history and present and then decide that now is the time to activate the protection. Or not.
The switching speed of the actual activation is, in this context, pretty much irrelevant.

Jan
 
Think non destructive SOA testing, lots of references about in the last 20 years.

Please! Two sentences cannot be seperated by a comma unless the connecting words "and", "or", "but", "while" or "yet" are used after the comma.

For your sentence quoted above a semicolon would have been appropriate.

Poor punctuation makes your prose look illiterate very quickly indeed.

Please buy the following treatise; it's magnificent, and represents excellent value:

http://www.amazon.co.uk/Penguin-Gui...-Books/dp/0140513663/ref=cm_cr_pr_product_top
 
An analogue computer can be very fast. In some cases very much faster than a digital computer.

The IV detection is in effect an analogue computer. It can detect an event and action a response in very short timescales.

If a "history" is added into that analogue computer by say adding in an RC, then it can action a response that takes account of instantaneous values and a history weighting to protect the output stage in <<10us. I don't know whether it is fast enough to do <<1us.

Even though I know almost nothing about digital style protection systems, I'll wager a size-able bet that in this situation, analogue is both simpler and faster than digital for the same expenditure.
 
Ohh no, not at all. 1uS is a limit that goes with say, 50V Vce and 10A Ic. If you limit it to a period of say 10mS you may dissipate say 40V Vce with say 8A (I'm just making up these numbers to show the principle).

So the SOA is basically three-dimensional: combinations of Vce, Ic and time duration determine whether it is 'safe' or not.
See for instance fig 1 here: Safe Operating Area calculations

So, if you measure say a particular Ic and Vce combination you need to decide how long you can allow that before activating the protection. Problem is that the Vce and Ic combinations change dynamically so you need some way to keep track of history and present and then decide that now is the time to activate the protection. Or not.
The switching speed of the actual activation is, in this context, pretty much irrelevant.

Jan

That's about dynamically avoiding getting over the SOA limits, not pulling out from the dangerous zone. For avoiding the SOA limits, the speed requirements are much more relaxed. See my other messages above.

What is your slow SOA protection going to do in the case of a sudden load short, to 0.1 ohm?
 
Please! Two sentences cannot be seperated by a comma unless the connecting words "and", "or", "but", "while" or "yet" are used after the comma.

For your sentence quoted above a semicolon would have been appropriate.

Poor punctuation makes your prose look illiterate very quickly indeed.

Please buy the following treatise; it's magnificent, and represents excellent value:

http://www.amazon.co.uk/Penguin-Gui...-Books/dp/0140513663/ref=cm_cr_pr_product_top

Thank, you, for, your, comment.
 
AX tech editor
Joined 2002
Paid Member
What is your slow SOA protection going to do in the case of a sudden load short, to 0.1 ohm?

Why do you think it is slow; in what way?
There's nothing magically about a 0.1 ohms load.
The load - and the signal! - produce a combination of Vce and Ic for a certain duration and that is what the protection circuit should operate on.

jan
 
Why do you think it is slow; in what way?
There's nothing magically about a 0.1 ohms load.
The load - and the signal! - produce a combination of Vce and Ic for a certain duration and that is what the protection circuit should operate on.

I am referring specifically about the secondary breakdown part of the SOA curve, what you are saying applies to the thermal breakdown. See this reference, I think it will clarify the time dependence of FBSOA and RBSOA.
 
AX tech editor
Joined 2002
Paid Member
I am referring specifically about the secondary breakdown part of the SOA curve, what you are saying applies to the thermal breakdown. See this reference, I think it will clarify the time dependence of FBSOA and RBSOA.

Yes. Thermal breakdown as a result of exceeding the SOA. Good paper.
As I said: SOA is basically three-dimensional, defined by Vce, Ic and time duration.
So what did you mean with that 'fast' and 0.1 ohm thing?

jan
 
Yes. Thermal breakdown as a result of exceeding the SOA. Good paper.
As I said: SOA is basically three-dimensional, defined by Vce, Ic and time duration.
So what did you mean with that 'fast' and 0.1 ohm thing?

If the device enters the secondary breakdown area, it will self destruct in (more or less) 100nS. This is not the case for the thermal breakdown, the time constants are much larger there. So you may say that while thermal breakdown is a slow process, and whatever method (analog or digital) will do, the secondary breakdown tolerance is zero - and very fast electronics is required for protection.

If the conditions at the load short incept throws the output devices in secondary breakdown, then your slow solution won't do it's job.

But I concede that a 100% failsafe protection solution against secondary breakdown in audio amplifiers, analog or digital, is in 99.9% of the cases an overkill. Industrial power modules are a different story.
 
But I concede that a 100% failsafe protection solution against secondary breakdown in audio amplifiers, analog or digital, is in 99.9% of the cases an overkill. Industrial power modules are a different story.
If your supa dupa SOA protection matches the DC SOA limits, you usually have some time to react. Alas, this often gives insufficient VA capability into reactive loads.

If you enter the time limited SOA regions, you are living dangerously.

Can yus SOA protection experts, eg Jan.didden, tell us whether their 'real life' experience used the DC limits?
________________________

Just a reminder that 'instantaneous protection' that limit drive to the output devices (or indeed anything earlier) with multiple slope or even steep single slope protection exhibits nasty 'snap crackle pop' noises when triggered on real speakers.

The last of the in-house designed LEAK amps disconnected the speaker for a few seconds when triggered like Jan's designs.
 
Last edited:
One of the most advanced MCU's today, an ARM R4 running at 80-160MHz, the TMS570 from TI (costs anywhere between $5 and $20 in 1000's), is doing 12bit sample + hold and full conversions in 600nS. See the data sheet at pp. 112.

ARM's R range is for safety critical applications, looks not to be an optimal choice to run tandem CPUs here.

Not fast enough, even before considering the time required to compute and actuate. I would appreciate if you or Defiant could indicate a proper MCU for this application, I can't find any.
Feel free to drop me a mail, this looks to be going considerably OT. Certainly if you need response in well under a microsecond, a 600nS ADC isn't going to cut it.

P.S. "A/D Conversion complete" is always an interrupt.
For your chosen uC, sure. However if you're really short of CPU cycles and can't tolerate excess latency then you'd find an SoC that uses DMA to transfer the ADC contents into memory, hopefully not impacting the CPU's operation. Or alternatively you'd use a co-processor such as comes with the LPC43XX range from NXP.
 
AX tech editor
Joined 2002
Paid Member
Can yus SOA protection experts, eg Jan.didden, tell us whether their 'real life' experience used the DC limits?

Depends how you mean 'using the DC limits'.

Secondary breakdown rears it head especially with high Vce/high Ic combinations and this typically arises from reactive (speaker) loads. If you mean by 'DC limits' the limitations with a (DC) resistive load, you get into trouble with reactive loads.

OTOH if you mean modelling the protection to the DC SOA curve you throw away a lot of your output stage capability. The highest power levels in audio will appear somewhere in the mid range, and your output stage can dissipate more at those frequencies than at DC.

Jan
 
AX tech editor
Joined 2002
Paid Member
If the device enters the secondary breakdown area, it will self destruct in (more or less) 100nS.

The protection should be designed to prevent the amp to end up in 2nd breakdown. That means you need to 'keep an eye' on the development of the Vce/Ic combination over time and pull the plug when you get close to the danger zone. This takes time, and it is very simple to time-bias this process. If, for instance, you are in a situation where you need to activate the protection after 10mS, you can bias it to 9mS which gives you a leisurly 1mS to activate the protection. There's really not any time pressure here.

What makes the design of such a system interesting is that all of this is dynamic. The Vce/Ic combinatios change continually, and the allowed SOA also varies with the current heat sink temperature. These are interesting issues :)

Jan
 
AX tech editor
Joined 2002
Paid Member
Here is part of the protection system I did in 1984 or thereabouts, attempting to model the SOA. It's rather complex and if I look at it today I'm not sure I still understand what I was doing at the time. But it DID work as designed :)
 

Attachments

  • prot.jpg
    prot.jpg
    182.7 KB · Views: 149
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.