Maximum device dissipation

[smithy666]

I'm trying to figure out the maximum device dissipation for an output stage with 4 devices (2 +ve, 2-ve). Assuming supply rails of +/-48V and an 8R load (also assume resistive only), would the maximum peak device dissipation occur at 22V? On my calcs, the load current would be 2.75A, resulting in peak dissipation per device of 30.25W ( (48-22)*2.75A/2). If I was using a 0.4Deg/W heatsink, this would amount to a temperature rise of 25 degrees per pair. Is this the correct way of calculating max device dissipation? What other factors do I need to take account of?
Thanks

[jaycee]

I would suggest the following articles are worthy of a good read:

http://sound.westhost.com/patd.htm
http://sound.westhost.com/soa.htm

[smithy666]

Thanks. Very useful site. Back to do some more calcs...

[AndrewT]

Hi Smithy,
a number of other calcs to take into account.
1.) de-rate the device for elevated temperature.
if the Ts =30degC (almost cold) and Tc=40degC then de-rating for a plastic package is [150-40]/[150-25]=0.88. So a 100W device can dissipate a continuous 88W IF YOU CAN KEEP Tc=40degC.
for Tc<=87.5degC de-rating factor = 0.5 This allows a sink operating temperature upto about 75degC.
2.) de-rate the device for elevated voltage. You must consult the datsheet to see at what voltage the device starts to enter second breakdown. At Vrail=48V some devices are down to just 50% others are still up at 95%.
3.) you need to combine the two de-rating factors from above.
4.) de-rate the heatsink for reduce delta T.
Most manufacturers specify their sink dissipations for a sink temperature between 70 and 80Cdegrees above ambient. If your delta T (Tsink-Tambient) =40Cdegrees then the de-rating can be about 1.2 to 1.4 consult the manufacturer's datsheet on dissipations for this factor for their sinks.
5.) the speakers are reactive. I find that one can model the worst case fairly closely for first selection but doing the following.
Take Vrail and multiply by the peak current into your nominal load. This gives the peak dissipation into a 45degree phase angle load, moderate rather than severe. Now compare this peak dissipation to the de-rated power from 3 above.
6.) once some possible devices AND heatsink are identified, one can start to do the manual calculations, which are quite extensive, to ensure your devices never exceed the one shot dissipations that the manufacturer specifies/guarantees in the datasheet.

Amp design is not back of the fag packet stuff, it involves real arithmetic. Once the experts have done a few dozens of complete amp designs, they get quite good at instantly assessing requirements, us mere mortals have to go the long route until we can build up sufficient experience.

[jacco vermeulen]

Originally posted by smithy666
Is this the correct way of calculating max device dissipation? What other factors do I need to take account of?

Nope
=> a heatsink calculation requires RMS values.
=> max rail voltage due to mains surge and transformer regulation.

[unclejed613]

when i worked at APT in the '80s, i think i remember that a class AB amp actually dissipates more heat when it is running between 50 and 75% rated output power than at 100%

[richie00boy]

That is correct. Worst case in when the output voltage is half the rail voltage.

[AndrewT]

Hi Richie,
he is referring to the DC output voltage. That is hardly a valid AC requirement.

For AC signals the maximum average dissipation occurs when the output voltage is much closer to maximum than half voltage (50%) and higher than half power (70%). More like 75% to 80%.

Can someone tell us the actual % of maximum voltage to develop worst case dissipation for a sinusoidal waveform into a resistive load?

[KSTR]

A lot of usefull discussion about SOA and dissipation issues, also with reactive loads, can be found here:

Burr-Brown AB-39: POWER AMPLIFIER STRESS AND POWER HANDLING LIMITATIONS

APEX App Note 1: General Operating Considerations
(Sections 7.1 & 7.2)

APEX App Note 22: SOA and Load Lines

Those deal with power operational amplifiers but there is no difference in the output stage compared to discrete amplifiers.

The app note section at apexmicrotech is a good resource on a variety of topics relevant to power amplifier design, that's why I recommend it quite frequently.
http://eportal.apexmicrotech.com/mai.../app_notes.asp


For sinusodial AC into resistive loads, the maximum device dissipation appears when the peak AC output voltage reaches 63% of the supply voltage, to answer the specific question. And that's for class B (or AB with very low bias current) and freq>50Hz.

Regards, Klaus

[unclejed613]

ok, for a 100w/8r amp that would 1/2 of +/-60v, or +/-30v.
peak value would be 30v, so rms is 21.21v or 56w which would be 56%.

actually i think it's 75%, but i'll have to look it up.