Always is fun to simulate an SOA curve

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
It would be even more fun to measure one, but I'm lacking a Tek P6203 current probe for my AM5030 :/

At least I can see this amp I'm testing is going to catch fire :devily:

None of the first three lines should go above the light blue line, even though they all do for a great portion of their cycles. One pair of 2SA1943/2SC5200 just doesn't cut it.
 

Attachments

  • soa_ed.png
    soa_ed.png
    7.8 KB · Views: 168
The single shot apply when they are far enough apart that the junction has time to cool back down to the Value at which the chosen Tc was valid.
The Tj to Tc timeconstant is very short, so it indicates fast heating when the transient passes through and fast cooling after the transient has passed through.

Now look at the continuous rating that would apply to DC. This requires a continuous DC current combined with a continuous DC voltage. This is not music.

Somewhere between the two cases above is our average music level and regular peaks occur that are considerably above the average. These peaks might well suit the 100ms SOA at the chosen Tc.

The rare transient peaks that exceed the regular peaks could be suited to the 10ms and even the 1ms SOAs.

The very high Ic combined with a very high Vce can only happen with reactive loads and these only draw worst case peak currents on very fast transients, either stopping or starting.

The worst case conditions for Ppk during audio reproduction cannot be continuous sinewave test signals nor can they occur when a stream of prolonged high energy music signals are present, they must be associated with transients.
This allows one to use the one shot SOA provided the designer can model duration of worst case signals to event repetition for signals at -0dB and a range of other lower than full power signals.
 
I agree with your theory. But I don't know numerous things. For example, what is the cooling time of Tj to Tc? What we don't want is for Tj to ever momentarily exceed 150℃. Toshiba doesn't provide a thermal coefficient for that. As you say, we know what the limits are for a single shot from a starting Tc.

As for transients, the output's peak dissipation is fully realizable with steady state signals and a reactive load.

Some fun reading: Heavy Load: How Loudspeakers Torture Amplifiers | Stereophile.com

I'll post some code a bit later so you can play with it.
 
As a point for thought is the 100mS graph line that Toshiba supplies.http://akizukidenshi.com/download/2SC5200.pdf

That's 10Hz if you call 100mS a period. Is the time down from that area enough to bring Tj back down without elevating Tc for the next period? I don't have any actual reactive loads to test that, but would love to if I did.

Frequency of the signal is meaningless if we think of heat as a pulse width with time above and amount in balance with the time below and amount to produce an average which itself is the DC line that is already there being the result.

About time for some linear regression to get to the bottom of this :)
 
sim for LTspice and plot file for gnuplot attached. Use LTspice to export V(n001)-V(n006) and Ic(Q5)

You may have a point about cool-down time. Eventhough the peak output dissipation is 488W in the example, average is 116W, but it really depends where in the plot it is above to see how excessive (weighted) the peak really is. I'm not sure yet how to calculate that.

Here's the SOA box
Code:
#; 2SC5200 SOA limit data to draw the SOA box
#; Ic	Vce
15	3
15	10
3	50
0.04	230
0.03	230

and the 100mS one
Code:
#; 2SC5200 SOA (continuous, 100Ms) limit data to draw the SOA box
#; Ic	Vce
15	3
15	20
7	43
0.066	230
0.03	230
 

Attachments

  • test.asc.txt
    6.9 KB · Views: 39
  • SOA.plt.txt
    993 bytes · Views: 41
  • MJE15030.LIB.txt
    972 bytes · Views: 39
  • MJE15031.LIB.txt
    968 bytes · Views: 35
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.