Hi guys,
I'm working on a power amp project and need to decide how man // devices the output stage needs.
Ian Hegglun helped me with a nice thermal model for the output devices and I can now graph the junction or case temperature versus dissipation and time. Next step of course is to related that to the data sheet SOA curves.
But. SOA is specified in terms of duration of the disspation; there's curves for 1ms, 10ms, 100ms etc. But how do I find the curve I need to use? I mean, suppose a trumpet has a fundamental of 3kHz (just for discussion) I don't think I should use the 300us SOA because the trumpet blow may last 200ms. So, do I use the 300us or the 200ms curve to determine whether that 3kHz tone stays inside the SOA?
And then there's the question of how often a particular 'burst' repeats in typical music...
Jan
I'm working on a power amp project and need to decide how man // devices the output stage needs.
Ian Hegglun helped me with a nice thermal model for the output devices and I can now graph the junction or case temperature versus dissipation and time. Next step of course is to related that to the data sheet SOA curves.
But. SOA is specified in terms of duration of the disspation; there's curves for 1ms, 10ms, 100ms etc. But how do I find the curve I need to use? I mean, suppose a trumpet has a fundamental of 3kHz (just for discussion) I don't think I should use the 300us SOA because the trumpet blow may last 200ms. So, do I use the 300us or the 200ms curve to determine whether that 3kHz tone stays inside the SOA?
And then there's the question of how often a particular 'burst' repeats in typical music...
Jan
Your a serious pro. However, you need to study music. put on a record and look at the scope on the output.
Classical music, like that using a trumpet, has 55 db signal variation, with 99% of the time being in the 10-20 db above noise area. Jazz is a little less variation, more mono volume. I find I can use a single NTE60 (MJ15003 equiv?) pair with peaks of 40 v on 1.5 Vpp average music. I've got a 80 v power supply on those two dinky soa transistors. That amp is perfect for classical music in the home. when I tried that amp in PA service in a church, about 10 watts, I melted the solder on the wiring after a 3.5 hour rehearsal because the heat sinks didn't lose the heat enough.
Techno and house music, <10 db variation loud to soft. The current flows all the time. You need waaay more output transistors to reproduce this "music". Four and five pairs MJ15024/25 are the survivors in this service. AKA peavey CS800s/PV2000, Yamaha, QSC, Crown, the big boys of PA amps. See what the repair men say. PA service also needs fans for class AB.
Guitar amps, when they turn the overdrive on, the flat top waves really flow the current. Again survivors have a lot of SOA on the output.
For the modeling, the software in microprocessor protections is hidden. But the soa model of the cs800x is right there on the schematic diagram, network around the protection op amps. I follow the pros, since I don't sell my work.
Classical music, like that using a trumpet, has 55 db signal variation, with 99% of the time being in the 10-20 db above noise area. Jazz is a little less variation, more mono volume. I find I can use a single NTE60 (MJ15003 equiv?) pair with peaks of 40 v on 1.5 Vpp average music. I've got a 80 v power supply on those two dinky soa transistors. That amp is perfect for classical music in the home. when I tried that amp in PA service in a church, about 10 watts, I melted the solder on the wiring after a 3.5 hour rehearsal because the heat sinks didn't lose the heat enough.
Techno and house music, <10 db variation loud to soft. The current flows all the time. You need waaay more output transistors to reproduce this "music". Four and five pairs MJ15024/25 are the survivors in this service. AKA peavey CS800s/PV2000, Yamaha, QSC, Crown, the big boys of PA amps. See what the repair men say. PA service also needs fans for class AB.
Guitar amps, when they turn the overdrive on, the flat top waves really flow the current. Again survivors have a lot of SOA on the output.
For the modeling, the software in microprocessor protections is hidden. But the soa model of the cs800x is right there on the schematic diagram, network around the protection op amps. I follow the pros, since I don't sell my work.
Last edited:
As Jo says, different music styles have different average to peak values. Well mastered pop, rock and jazz will have an RMS to peak value of 18dB. Good classical and big band jazz, about 22 dB. More modern pop, R&B and vocal genres 14dB. Metal and dance, 10dB or less.
As to the duration of the peaks, I can't say - but may be able to figure it out somehow.
As to the duration of the peaks, I can't say - but may be able to figure it out somehow.
You might have a look at this, done by NAD I think. Finding that peaks can last as long as 1/2 second
https://secure.aes.org/forum/pubs/conventions/?elib=4898
https://secure.aes.org/forum/pubs/conventions/?elib=4898
I like Tomchr's analysis here LM3886 chip power amplifier thermal design. as a good primer/reminder of things.
It's very easy to get dynamic range, crest factor and SOA duty cycle peak demands confused. At least for me. The worst recordings in my collection have about 3dB compression added so RMS is -9ishdB with peak always around 0dBFS.
Good classical has RMS down around -23dB
And of course you wont necessarily listen to compressed studio music 10dB louder than you would classical, which confuses more.
If you assume a 33% duty cycle at worst case load you should be safe. Based on the link that is 6dB crest factor.
And I've confused myself again
It's very easy to get dynamic range, crest factor and SOA duty cycle peak demands confused. At least for me. The worst recordings in my collection have about 3dB compression added so RMS is -9ishdB with peak always around 0dBFS.
Good classical has RMS down around -23dB
And of course you wont necessarily listen to compressed studio music 10dB louder than you would classical, which confuses more.
If you assume a 33% duty cycle at worst case load you should be safe. Based on the link that is 6dB crest factor.
And I've confused myself again
Can you share how do you do it?I use DC SOA. I make a DC SOA table on LTSpice and then compare it with transistor dissipation.
I use DC SOA. I make a DC SOA table on LTSpice and then compare it with transistor dissipation.
The thermal model is a series of RC sections, but the values have to be matched to the thermal data in the data sheet. Ian did it for me so I am not sue how to do it, other than trial and error.
Jan
Using the transient thermal numbers is great when you have control over the pulse width, height, and rep rate, like a switchmode.
With music, you have no control. All you can do is design worst case which is 50% duty cycle at 20 hz.
You don't want to be in the position where you find that the devices are about 30 bucks (a pop).
John
Actually one vendor does control the input to the amp. DDT by Peavey is a circuit that turns down the gain if the signal has too much high freq content- ie is a square wave.
Square signals are popular these days, as is compression in pop music tracks. Watts ratings are a great marketing game. Even the old Fair Trade RMS figures watts were for a sine wave signal, and a limited duration. That was a fine rating for home listeners of classical music, but pretty light duty if you tried to use those watts to play crunch guitar on a stage more than 10 minutes.
That is where brand name achieved prominence in music performance circles. See the reports of one of the amp repairmen on here. Certain brands of guitar amps are known to produce crunch waves (square) all day without failing. Certain brands of PA amps are known that the watts rating is 24/7 for beach bars playing house music. They have BIG fans and heat sinks, not just a lot of parallel SOA output devices.
You have to know your market to rate your watts usefully for the customer. Low buck amps are rated with about a 3 second rating on sine waves.
Last edited:
I suspect Jan is not speaking about that kind of music.
And the discussion is about trying to go further out the SOA curve by utilizing the periodic nature of music with the higher pulse transient dissipation ratings of the devices due to the heat capacity of the materials in the thermal stack.
Peavey's auto-dim is not consistent with listening to great material in the home. It is consistent with how I use my amps, you know..turned up to 11.
John
amp design depends on the intended use, ie the market. For classical music in my home I'm upgrading the sound of a very loud , very light duty amp. 200 w/ch music power on drum hits, average power 1/2 watt. If I listen to a rock track with a lot of drum or crunch guitar, I turn that amp down. I don't like more than 90 db average.
For my keyboard/synth performances in churches at 10-40w average I repaired a CS800s with the DDT circuit, a fan, and a lot more idle watts drain than the home amp. Jan needs to pick his intended use to pick his SOA ratings, heat sinks, optional fan.
Last edited:
Yeah, this ST120 with single outputs is as loud as the CS800s, which is 26 db gain. Just the ST120 won't play loud for very long without overheating.
Sometimes classical music has silence, sometimes it is one flute (sine wave) like the beginning of Bolero. Next track I might play 1812 overture with cannon shots. I like the cannon shots to be about 100 db+, the solo flute to be just above the hiss of the gas burner. There is 70 db range on a good cd, I want to hear it that way. I was a live musician in high school, a real band can produce that kind of dynamic range. We had flute solos, we did a tutti hit once that echoed off the freeway retainer wall 1.2 mile away and bounced back as an echo. My speakers will handle it, they take up to 54 v music. If I get successful enough to play outdoors in the city park, I have a 52 V PV-1.3k amp to make some real noise on the same speakers I use in the living room at 1/2 watt average.
Look at organ amps, Jan lives where a famous manufacturer is based that is known for selling wimpy systems. A serious organ has 12 channels 100 W amps and speakers, with the music split out that way from the organ. Those 100 W Allen S100 amps have only one MJ802 pair, but 125 cubic inch heat sinks. Stand on the contra-trombone (a squarish wave stop) and play a 90 minute Widor organ symphony, see how hot your heat sink gets. Totally different service requirements than my wimpy keyboard playing piano samples with high peak to average voltage level. But I could play crunch guitar out of the keyboard, it is a sampler after all. Hear Stryper crunch has been done in church.
Sometimes classical music has silence, sometimes it is one flute (sine wave) like the beginning of Bolero. Next track I might play 1812 overture with cannon shots. I like the cannon shots to be about 100 db+, the solo flute to be just above the hiss of the gas burner. There is 70 db range on a good cd, I want to hear it that way. I was a live musician in high school, a real band can produce that kind of dynamic range. We had flute solos, we did a tutti hit once that echoed off the freeway retainer wall 1.2 mile away and bounced back as an echo. My speakers will handle it, they take up to 54 v music. If I get successful enough to play outdoors in the city park, I have a 52 V PV-1.3k amp to make some real noise on the same speakers I use in the living room at 1/2 watt average.
Look at organ amps, Jan lives where a famous manufacturer is based that is known for selling wimpy systems. A serious organ has 12 channels 100 W amps and speakers, with the music split out that way from the organ. Those 100 W Allen S100 amps have only one MJ802 pair, but 125 cubic inch heat sinks. Stand on the contra-trombone (a squarish wave stop) and play a 90 minute Widor organ symphony, see how hot your heat sink gets. Totally different service requirements than my wimpy keyboard playing piano samples with high peak to average voltage level. But I could play crunch guitar out of the keyboard, it is a sampler after all. Hear Stryper crunch has been done in church.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.