Output transistor safe operating area

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have been trying to learn how to appropriately select appropriate output transistors for power amp designs to keep them within their maximum ratings, unfortunately with little success. From the material I have read on this site it seems to be an extraordinarily complicated matter with much disagreement! I have read the material at Semiconductor Safe Operating Area and while I understand the issues involved, the site doesn't seem to actually go into an actual process that a designer could use. I guess what I'm looking for is an "example problem" to show the thought process a designer goes through when making the calculations for a particular requirement. Does anyone have a reference for something like that?
 
I have been trying to learn how to appropriately select appropriate output transistors for power amp designs to keep them within their maximum ratings, unfortunately with little success. From the material I have read on this site it seems to be an extraordinarily complicated matter with much disagreement! I have read the material at Semiconductor Safe Operating Area and while I understand the issues involved, the site doesn't seem to actually go into an actual process that a designer could use. I guess what I'm looking for is an "example problem" to show the thought process a designer goes through when making the calculations for a particular requirement. Does anyone have a reference for something like that?

You are correct, this is never an easy problem. First, you have to decide what minimum impedance and worst phase angle the amplifier must operate into without having to go into protection. You next have to decide how agressive the protection will need to be to keep the transistots safe.

If you use modern transistors with good safe area, like MJL21193/4, a semi-reasonable rule of thumb for sizing the output stage is as follows: divide rated power into 8 ohms resistive load by 75 and round up to next integer. This is the number of output pairs to use. This is only a very rough suggestion and does not address the details of how much protection you need when this number of transistors is used. For this number of output pairs, the necessary protection will not have to be overly agressive for most reasonable speaker loads. This rule of thumb will get you to about 150W/8 ohms with two output pairs for an amplifier rated for continuous duty into 4 ohms if a heat sink is used that will never be allowed to get above 60C (which is the highest temperature you can keep your finger on indefinitely).

Again, this is a tough problem involving many tradeoffs, including risk, and I'm sorry I can't be more specific with details here. It would take a chapter in a book.

Cheers,
Bob
 
This rule of thumb will get you to about 150W/8 ohms with two output pairs for an amplifier rated for continuous duty into 4 ohms if a heat sink is used that will never be allowed to get above 60C (which is the highest temperature you can keep your finger on indefinitely).

Bob, I would suggest 4 pairs of MJL21193/94 for the power amp rated at 150W/8ohm, 300W/4ohm. I would take into account numerous speakers with impedance dips to 2ohm, and complex load, to feel "safe".

Regards,
 
Bob, I would suggest 4 pairs of MJL21193/94 for the power amp rated at 150W/8ohm, 300W/4ohm. I would take into account numerous speakers with impedance dips to 2ohm, and complex load, to feel "safe".

Regards,

No strong disagreement here, but four pairs for 150W is pretty generous. Remember, what I said is that it involves a tradeoff of protection scheme agressiveness vs number of pairs. I did not say that two pairs for a 150W amplifier did not require a protection circuit.

However, apart from SOA, four pairs for a 150W/8-ohm amplifier is nice because it reduces beta droop at high current and makes the amplifier more capable of delivering high current.

Remember, even driving a low-Z resistive load requires a certain amount of safe area, usually peaking at some power level below max (but not necessarily 1/3 power where average dissipation peaks). The introduction of phase angle not zero increases the needed amount of safe area for handling of a load that has a given minimum impedance (the phase angle is close to zero at the minimum impedance, so the worst case safe area need occurs at an impedance magnitude that is greater than the minimum impedance).

So, for example, if your worst case 4-ohm speaker has a minimum impedance that dips to 2 ohms (often Re of the woofer, but not always), the load impedance that imposes the greatest safe area need will have a magnitude greater than 2 ohms and will have a phase angle usually in the 40-60 degree range.

Again, one very rough rule of thumb emerges: If an amplifier has enough safe area for a 2 ohm resistive load at the worst-case power level, it probably has enough safe area for a 4-ohm-rated reactive load.

Keith Howard wrote a good article in Stereophile a year or two back that helps shed some light on this.

BTW, it is possible to use LTspice simulations with a variety of different loads to display the safe area excursions when a sinewave is applied.

Finally, a nearly pure low-z reactive load as sometimes presented by electrostatics can be difficult to deal with.

Cheers,
Bob
 
at 300W /4 R , that make 600W peak , alternatively on
each rail, supposing that the load is purely resistive...
with a load that would be reactive and dipping at the same time,
even 4 x 200W devices/rail would be at pain if the amp is pushed to
its limits for a long time, say a live performance, in a overheated stage,
that is, continous duty..
4 pairs are a minimum for such an amp...
 
With music we use ~50% the RMS power, but has income class AB which is ~60%, I dimencion of according to the RMS power. Dips on the low impedance are peaks, I do not know if should include.
I know that care must be greater with professional audio amplifiers that are taken to the limit in operation.
 
at 300W /4 R , that make 600W peak , alternatively on
each rail, supposing that the load is purely resistive...
with a load that would be reactive and dipping at the same time,
even 4 x 200W devices/rail would be at pain if the amp is pushed to
its limits for a long time, say a live performance, in a overheated stage,
that is, continous duty..
4 pairs are a minimum for such an amp...

Bear in mind that the power delivered to the load is not the same as the power dissipated in the output stage.

Also bear in mind that I said that my rule of thumb was for when the heat sink never goes above 60C. In many cases, this can mean a very large heat sink, especially if the ambient is well above 25C.

Cheers,
Bob
 
However, apart from SOA, four pairs for a 150W/8-ohm amplifier is nice because it reduces beta droop at high current and makes the amplifier more capable of delivering high current.

Exactly. I made quite a deep study concerning simulations into complex loads, and 4 pairs seem (about 25mV at every Re of 0R22) to be a good solution, with total O/P stage idle current about 450mA.

Regards,
 
I have been trying to learn how to appropriately select appropriate output transistors for power amp designs to keep them within their maximum ratings, unfortunately with little success. From the material I have read on this site it seems to be an extraordinarily complicated matter with much disagreement! I have read the material at Semiconductor Safe Operating Area and while I understand the issues involved, the site doesn't seem to actually go into an actual process that a designer could use. I guess what I'm looking for is an "example problem" to show the thought process a designer goes through when making the calculations for a particular requirement. Does anyone have a reference for something like that?

One point to remember is the output transistors dissipate most with an AC signal that is 2/3rds of B+. This is to do with the function of voltage/current through the output transistors and speaker(s).
 
Bear in mind that the power delivered to the load is not the same as the power dissipated in the output stage.

Also bear in mind that I said that my rule of thumb was for when the heat sink never goes above 60C. In many cases, this can mean a very large heat sink, especially if the ambient is well above 25C.

Cheers,
Bob

right, bob, i was pointing the fact that these 600W
must pass across the (fully) conducting devices...
sure that power dissipation is another thing, but
the amp wll dissipate more than the 150w theorical, because
when the output voltage is at half the value (in rms term) of the
power supply , this latter will not be as collapsed that at full power,
making the 300w amp worst case dissipation the one of a 350 to 400
theorical one (one which would have fixed supply voltage for the
convenience of the maths)..

regards,

wahab
,
 
Below is a table generated from C# program to calculate power into output transistor and a speaker at a set inpout voltage. The B~+ used was 60 volts and speaker impedance was 4 ohms.

It can be seen from the table that there is 4 times more power into the speaker than the output transistors at full input voltage swing.
Yet at 2/3 the powers are equal.


Input voltage=0 Speaker power=0 Transistor power=0
Input voltage=1 Speaker power=0 Transistor power=0
Input voltage=2 Speaker power=0 Transistor power=0
Input voltage=3 Speaker power=0 Transistor power=0
Input voltage=4 Speaker power=1 Transistor power=0
Input voltage=5 Speaker power=2 Transistor power=23
Input voltage=6 Speaker power=3 Transistor power=29
Input voltage=7 Speaker power=4 Transistor power=33
Input voltage=8 Speaker power=6 Transistor power=36
Input voltage=9 Speaker power=8 Transistor power=53
Input voltage=10 Speaker power=10 Transistor power=60
Input voltage=11 Speaker power=13 Transistor power=64
Input voltage=12 Speaker power=15 Transistor power=67
Input voltage=13 Speaker power=18 Transistor power=80
Input voltage=14 Speaker power=22 Transistor power=87
Input voltage=15 Speaker power=25 Transistor power=91
Input voltage=16 Speaker power=29 Transistor power=94
Input voltage=17 Speaker power=33 Transistor power=105
Input voltage=18 Speaker power=37 Transistor power=110
Input voltage=19 Speaker power=41 Transistor power=113
Input voltage=20 Speaker power=46 Transistor power=116
Input voltage=21 Speaker power=51 Transistor power=125
Input voltage=22 Speaker power=56 Transistor power=129
Input voltage=23 Speaker power=62 Transistor power=132
Input voltage=24 Speaker power=67 Transistor power=134
Input voltage=25 Speaker power=74 Transistor power=142
Input voltage=26 Speaker power=80 Transistor power=145
Input voltage=27 Speaker power=86 Transistor power=147
Input voltage=28 Speaker power=93 Transistor power=149
Input voltage=29 Speaker power=100 Transistor power=154
Input voltage=30 Speaker power=107 Transistor power=157
Input voltage=31 Speaker power=114 Transistor power=157
Input voltage=32 Speaker power=123 Transistor power=159
Input voltage=33 Speaker power=130 Transistor power=162
Input voltage=34 Speaker power=139 Transistor power=165
Input voltage=35 Speaker power=147 Transistor power=164
Input voltage=36 Speaker power=155 Transistor power=165
Input voltage=37 Speaker power=165 Transistor power=167
Input voltage=38 Speaker power=174 Transistor power=168
Input voltage=39 Speaker power=183 Transistor power=167
Input voltage=40 Speaker power=193 Transistor power=167
Input voltage=41 Speaker power=203 Transistor power=168
Input voltage=42 Speaker power=214 Transistor power=167
Input voltage=43 Speaker power=224 Transistor power=166
Input voltage=44 Speaker power=235 Transistor power=164
Input voltage=45 Speaker power=246 Transistor power=164
Input voltage=46 Speaker power=257 Transistor power=162
Input voltage=47 Speaker power=268 Transistor power=161
Input voltage=48 Speaker power=280 Transistor power=159
Input voltage=49 Speaker power=292 Transistor power=157
Input voltage=50 Speaker power=305 Transistor power=155
Input voltage=51 Speaker power=318 Transistor power=151
Input voltage=52 Speaker power=329 Transistor power=148
Input voltage=53 Speaker power=342 Transistor power=145
Input voltage=54 Speaker power=355 Transistor power=143
Input voltage=55 Speaker power=369 Transistor power=138
Input voltage=56 Speaker power=384 Transistor power=134
Input voltage=57 Speaker power=397 Transistor power=130
Input voltage=58 Speaker power=411 Transistor power=126
Input voltage=59 Speaker power=425 Transistor power=121
 
Exactly. I made quite a deep study concerning simulations into complex loads, and 4 pairs seem (about 25mV at every Re of 0R22) to be a good solution, with total O/P stage idle current about 450mA.

Regards,

Hi Pavel,

Yes, I agree that more pairs in parallel, optimally biased, improves sound quality because it increases the class A region of operation and decreases crossover distortion. A related benefit that also reduces distortion with real program material is that there is less thermal swing in each transistor for a given output power, reducing memory distortion and mis-biasing from thermal mis-tracking. Keeping that 25 mV under real-world junction temperature swings is not that easy.

Cheers,
Bob
 
It can be seen from the table that there is 4 times more power into the speaker than the output transistors at full input voltage swing.
Yet at 2/3 the powers are equal.

Hi nigelwright7557

You must be talking about the power split due to operation in push-pull, but power is divided into high and medium frequencies, for low frequencies the current is almost the peak of the sinusoid. Is in low frequency that PSU is more required in amplifiers also

Bob,

I'm using MJL-3281/1302 for 100W rms 4R that is what the recommend manufacturer for audio amplifiers, the datasheet shows me a SOA 200W, but for 25°C, is validity for amplification class AB?
Also think that caution should be greater in terms of dissipation due to plastic encapsulation
 
Analysis similar like this is needed. One may be surprised.

Regards,
 

Attachments

  • soa.PNG
    soa.PNG
    33.5 KB · Views: 531
I would suggest the use of excel in the design process.
It is easy to generate the reactive load line for a certain load phase angle, load magnitude and supply.
The main problem with SOA and protection is that the chip temperature is averaged from the peak power ( so the signal) in a not well defined way and the limit to use in the design process is the temperature of the chip.

By plotting the derated ( depending on heatsink and max temp allowed) power dissipation and the second breakdown limit ( from data sheet) on the same graph as the reactive load line, you can apply some overrun allowance in the power dissipation and find graphically a protection limit that fits the constraints . You can easely change design parameters like heatsink thermal resistance, max temp, % of overrun and so on.
See the figure coming from a real design and made with excel.

JPV
 

Attachments

  • load line.doc
    86 KB · Views: 149
Hi guys, good thread, could I join the discussion? ;)


I always designing OP stage like following:

- First I calculate min. resistance: 0.63×Re (eg. Rmin=2.5r when using 4r speaker)
- The I calculate max. peak current (Ohm' law)
- Then I calculate current values at given phase angles (90 deg., 105 deg., 120 deg., ... 165 deg.)
- I calculate the maximum Vce voltages at the calculated current load cases
- I calculate the voltage values at 45 deg. phase shift (135 deg., 150 deg., ... 210 deg.)
- I draw a table including the calculated current/voltage values
- I derating the 100mS SOA (look datasheet) from Tc=25C to Tc=70C.

Then examining the derated SOA graph and the I/V table it's easy to decide how many pairs are needed.


Anyone disagree? :p
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.