'Typical music' relating to SOA?

Status
Not open for further replies.
Jump to Latest
#21
Everything I read here discusses SOA from a power dissipation VS time perspective.

If the output devices are BJT's and some high Gm mosfets, there exists the possibility of secondary breakdown, which IS the limiting factor in many BJT's especially when operated near their max Vce ratings.

Secondary breakdown is an event where a large portion of the current through a device flows through a small localized area of the device die. This leads to rapid heating of the affected area and a resulting avalanche breakdown creating a collector to emitter short. This is why the SOA curves on a BJT are not simply related to dissipation and time. It is possible to pop a BJT that is operated within, but close to the SOA limits with a short burst of serious bass. The On Semi MJW21195 / MJW21196 come to mind.

Also, reproducing music that is recorded clipped like saturated guitars and the overdone bass in some EDM through an amplifier that is operated in a linear manner usually results in much higher device dissipation than playing a guitar through an amplifier cranked to full clipping since the output devices are operating as switches in the second case.

I might be known as Tubelab, but I have blown up far more silicon than vacuum tubes over the years. I would err on the side of caution, the margin of error depends on the intended use.

Brief explanation here:

https://en.wikipedia.org/wiki/Safe_operating_area
 
#23
tubelab.com said:
Everything I read here discusses SOA from a power dissipation VS time perspective.

If the output devices are BJT's and some high Gm mosfets, there exists the possibility of secondary breakdown, which IS the limiting factor in many BJT's especially when operated near their max Vce ratings.

Secondary breakdown is an event where a large portion of the current through a device flows through a small localized area of the device die. This leads to rapid heating of the affected area and a resulting avalanche breakdown creating a collector to emitter short. This is why the SOA curves on a BJT are not simply related to dissipation and time. It is possible to pop a BJT that is operated within, but close to the SOA limits with a short burst of serious bass. The On Semi MJW21195 / MJW21196 come to mind.

Also, reproducing music that is recorded clipped like saturated guitars and the overdone bass in some EDM through an amplifier that is operated in a linear manner usually results in much higher device dissipation than playing a guitar through an amplifier cranked to full clipping since the output devices are operating as switches in the second case.

I might be known as Tubelab, but I have blown up far more silicon than vacuum tubes over the years. I would err on the side of caution, the margin of error depends on the intended use.

Brief explanation here:

https://en.wikipedia.org/wiki/Safe_operating_area
Click to expand...

Well you may have noticed the word 'SOA' in the thread title 😉
In fact the SOA is what I am worried about; that is why we were discussing duty cycle and such.

I can model the amp's dissipation to the SOA limits; the black spot is how to relate 'typical' music power (and thus dissipation) demands to the various SOA limits.

an
 
#27
jan.didden said:
Well you may have noticed the word 'SOA' in the thread title 😉
In fact the SOA is what I am worried about; that is why we were discussing duty cycle and such.I can model the amp's dissipation to the SOA limits; the black spot is how to relate 'typical' music power (and thus dissipation) demands to the various SOA limits.
an
Click to expand...
Crunch guitar is square waves, 50% duty cycle on each output transistor. Compressed pop albums, hit max a lot of the time.
Rhianna Shut Up & drive makes my 1961 design Schober SS amp squeal most of the time with overdrive (IM distortion).
Cannon shots & bass drum hits, I don't have a memory scope to measure. My aural spectrum analyzer (ears) say they have a significant bass response, so probably they have significant energy at 100 ms period.
ZZ Top Afterburner CD has my reference bass drum track for speaker tests. Sharp Dressed Man is bout mm 140 (beats per minute) each bass drum hit 50-100 ms long, there is your peak current draw of a typical loud rock album.
 
Last edited:
#29
Sharp Dressed Man is bout mm 100 (beats per minute) each bass drum hit 50-100 ms long, there is your peak current draw of a typical loud rock album.
Click to expand...

You just reminded me of an event that took place in the early 90's.

I got a job at Motorola in 1973 testing and fixing two way radios (HT-220's) as they came off the assembly line. One of the perks was free silicon (within reasonable limits) just by filling out a sample request form. We had an audio club and we built, and blew up, solid state HiFi amps, and guitar amps. Many were based on the designs from SWTPC, including every version of the Tiger amps.

I worked my way up from radio tech to development engineer without a college degree, but in 1990 they offered to pay for me to attend college and get a degree. By then I was making tube amps, but I still had a collection of boards and parts left over from all the Tiger amps.

In one of my last college classes we actually had to design and build a project. The teacher asked that we take two weeks to figure out what we were going to build, how we were going to test it and assess its performance (measurement criteria). We had to present this information verbally in front of the class. Simple projects were OK, but we would be judged based on performance to our own criteria, pending project acceptance by the teacher AND the class vote.

I stated that I was going to build a car stereo system, and my only criteria was that the teacher's head would hit the roof of my car when I cranked it. The teacher asked for more "technical" specs, which I agreed to provide, but the class demanded the original test. The project was approved. There would be 4 channels of 20 watts RMS at 1% or less 20 Hz to 20 KHz, and 1 channel of 50 W RMS at 3% or less 20 Hz to 1 KHz to feed a 15 inch sub (required to get the teacher moving).

The audio amps were 4 X Plastic Tigers and 1 X Universal Tiger. All of the audio stuff was already built in an amp with 8 X Plastic Tigers and 1 X UT that I built for my van during the Motorola audio club days. It wound up in the closet after the van got parted out. The power supply and some of the Plastic Tigers were gone, so I built a new SMPS using an SG3525 chip, 4 Motorola mosfets and a home made transformer.

I put all this in my Dodge Omni (far smaller than the van) with a 15 inch sub mounted in my daughter's toy box, a pair of 6 X 9 car speakers, and a pair of 5 1/4 inch car speakers. It was LOUD, and to my surprise sounded far better than I expected. I drove around for while and found jut the right music for the test.....Mylon LeFevre's Crack the Sky....https://www.youtube.com/watch?v=D5Ov8SyfEH4 It's the first track on the CD, there is 3 seconds of dead silence, 4 bars of very percussive drums followed by a screaming lead guitar.....just right.

On test day I was the last to present my measured data, which far exceeded my meager specs. Then the entire class and the teacher convened in the parking lot for the demo. I had left the volume control at full crank when I got out of the car.

With the teacher in the passenger seat I start the car, and without touching anything, pop in the CD, without warning the drums hit, and he was out of the car before the guitars even started.....I got my "A."

I had that system in my car for several years. One day I had it cranked to full tilt and there was a familiar stink followed by silence. Autopsy revealed an electrolytic had shorted in the power supply. The others had all vented, or were bulging. Those were the only dead parts. I replaced them and all was well. Several months later it blows up again, same failure. The common element in both failures.....ZZ Top, Sharp Dressed Man, both times at full crank. I fixed the amp, and remembered NOT to play that CD at full crank. No other failures. I eventually retired the amp when I traded the Omni for a Shadow convertible, which had no room for a big sub or more than two door speakers.

The remains lingered in my junk collection until I sold some of the parts and scrapped the rest. See this thread for a picture and discussion:

http://www.diyaudio.com/forums/solid-state/41926-universal-tiger-17.html post #169 through #181
 
#30
tubelab.com said:
You just reminded me of an event that took place in the early 90's.
I stated that I was going to build a car stereo system, and my only criteria was that the teacher's head would hit the roof of my car when I cranked it.
The audio amps were 4 X Plastic Tigers . . . I built a new SMPS using an SG3525 chip, 4 Motorola mosfets and a home made transformer.

I had that system in my car for several years. One day I had it cranked to full tilt and there was a familiar stink followed by silence. Autopsy revealed an electrolytic had shorted in the power supply. The others had all vented, or were bulging. Those were the only dead parts. I replaced them and all was well. Several months later it blows up again, same failure. The common element in both failures.....ZZ Top, Sharp Dressed Man, both times at full crank. I fixed the amp, and remembered NOT to play that CD at full crank. No other failures.
Click to expand...
ROFL
So we are agreed here, ZZ Top Sharp Dressed Man is a great amp, power supply, and speaker test.
Coincidently, today I received from digikey my first two UC2526AN Pulse width controller IC, to either repair 1987 organ switcher supplies that are toasted all the way back to the IC, or in my dreams build my first 12 v- to 27 switcher supply using all these dead printer 24v toroids laying around here from the trash pile at work. the goal to avoid buying a patented $155 lithium battery for my Dell laptop computer, and power it with a 12 v gelcell in my pocket through the charger inlet. Remember the 1950's TV show, Mr. Wizard sets his lab coat on fire!
 
Last edited:
#32
Hi Guys

The safest approach for designing an audio output stage is to use the DC curve of the SOA, as Bimo stated very early in this thread. If you really want reliability, you have to be conservative.

The pulse curves of the SOA graph is for switching applications. Taken to an extreme, look at the SOA for a "super-power" mosfet - the ones with Trench or Polar or Mesh in their name. These things have crazy power ratings that require liquid cooling even in the pulse circuits they are designed for. When you look at the SOA there is NO DC curve. If you extrapolate one, the 1kW+ device is now 125W !

If you are expecting to play all kinds of music, compressed, dynamic and otherwise, you have to build a big output stage. You could take it to the ESL-capable extreme, where the phase angle can be 90-degrees, meaning current can be maximum when output voltage is minimum - full V across the devices at full current - short circuit current, really. I am in full agreement with Tubelab with respect to staying well away from the secondary-breakdown portion of the SOA.

Using the short-time (pulse) SOA curves is folly when designing a linear amplifier. Your amp will only be reliable when YOU use it - others will break it easily. and not be too impressed with how 'clever" you might be with math.

Have fun
 
Last edited:
#33
I use the formula:

Maximum reliable output power is ~ Total device Pmax / Derating Factor (DF).

That formula and the devices it is modelling is affected by secondary breakdown and as a result differs between mosFETs and BJTs.
I came up with DF= 4 for mosFETs for typical domestic duty into severe reactance domestic style 2way speakers.
For a BJT output stage DF= 5 to 6

These DF values came from using Bensen's spreadsheet (based on the David Eather paper) which allowed modelling of reactive load vs mosFET device SOAR.

Some years later R.Cordell's interviews came up with a similar device rating vs maximum output power, but stated in a quite different manner. The answers were about the same and gave me confidence that I was using the Bensen/Eather analysis in a sensible manner.

An example:

for a 100W power amplifier, use 500W of BJT output devices. DF = 5
One pair of MJ15003/4 have a 500W rating and would give a reliable 100W of maximum output power.
That could be 100W into 8ohms using a 35+35Vac transformer, or 100W into 4ohms, using a 28+28Vac transformer.
 
#34
Struth said:
If you are expecting to play all kinds of music, compressed, dynamic and otherwise, you have to build a big output stage. You could take it to the ESL-capable extreme, where the phase angle can be 90-degrees, meaning current can be maximum when output voltage is minimum - full V across the devices at full current - short circuit current, really.
Have fun
Click to expand...

Actually the application is a direct-drive ESL amp... 😎
 
#38
in my dreams build my first 12 v- to 27 switcher supply
Click to expand...

I had to design a 12V to 28V boost converter for a mobile LTE two way radio at work about 10 years ago. I used the National (now TI) LM3478 chip. I would still go that way today if you don't need input to output isolation, and can solder an 8 pin SOIC. It is good to about 150 watts at those voltages. The inductor is a simple 10 uH coil, and the circuit comes from the Web Bench simulator. Use just enough mosfet to do the job and run at 200 to 500 KHz. Too much gate capacitance kills the efficiency and makes mosfets fry.

For higher power out there is a two phase boost converter from Linear Tech I thin. I don't remember the number. Tiny little SMD package too.
 
#39
Thanks Andrew for a serious moment.
I suspect based on recent experience, on classical music with low incidence and period of large peaks, one can cheat a bit on your formula, but only on peak power, not in average power. Certainly not on any popular music.
Thanks tubelab. I realize now a 180 vdc (rectified mains) to +-12v +5 power switcher is a BUCK supply and an 11v to 27v is a BOOST supply. further baby steps will move to the Power supply forum. At least a Nat Semi LM3478 datasheet is in english.
 
Status
Not open for further replies.
Top Bottom