Transistor Power Amp Design

[kurtlives]

Can anyone give some help or starting point on power amp design using BJTs.

I have a +5V/-5V supply voltage and a preamp to drive the power amp. I have 2N3055 NPN power transistor and it's PNP equivalent MJ2955. I am looking to get about 6W (yah ) out of this. Driving a 4R load.

I am familiar with Class A, B and A/B configurations but do not know about biasing them here.

Thanks!

[R.G.]

There is a literally huge amount of information on line about designing BJT power amps, so much that anything that can get typed in here extemporaneously is a serious understatement. So step 1 is to go do some reading. Read these, then ask questions.
http://www.electronics-tutorials.ws/amplifier/amp_6.html
http://www.st-andrews.ac.uk/~jcgl/Scots_Guide/audio/part2/page2.html
http://www.ibiblio.org/kuphaldt/socratic/output/bjtamp_b.pdf
http://www.eee.bham.ac.uk/collinst/ee3b1/slides/PowerAmps5.ppt
http://www.hobbyprojects.com/junction_transistors/push_pull_amplifier.html

How much power you get out will be determined by the power supply and load. If you had *perfect* transistors that could provide all of your +/-5V to the load, the peak output voltage would be 5V. In reality, the transistor can't do that, so the voltage out from a 5V supply will be less than 5V. But if it were perfect, the peak current would then be 5V. The RMS value of a sine wave with a 5V peak is 5V/1.414 = 3.53Vrms. The power into a 4 ohm load from 3.53Vrms is V²/r = 3.1W.

For real transistors, you're going to be pushing very hard to get them to saturate to less than 1V, so you'll really get out +/-4V peak, and 2.8Vrms, for 2.0W.

You can't get out more power from a fixed power supply voltage by changing transistors however good they are. You can only change the load to get more. And then your power supply has to be able to supply the current.

You bias the output transistors by introducing a small voltage between the bases. Generally this is close to one diode drop for every driver/output transistor being biased. So for a 3055/2955 pair, you'd introduce two diodes between the bases, ensuring that your driver pulls DC current through the diodes to force them to conduct. The references will have some leads on this topic.

Current is also an issue. Peak current with 4V into a 4R load is 1A. The gain of a 3055 is about 20 at 1A, so you have to get 1A/20 = 50ma into the base to make it pull the current. 50ma is small, but non-trivial to provide. Just any old transistor won't do, necessarily.

As I said, read the references, then ask.

[teemuk]

http://www.sentex.ca/~mec1995/tutorial/xtor/xtor4/xtor4.html


Your Ohm's law math is flawed, though. You'll never get a 6W output with such low supply voltages. Circa 2W is a more reasonable estimation.

[DavenPaget]

Can anyone give some help or starting point on power amp design using BJTs.

I have a +5V/-5V supply voltage and a preamp to drive the power amp. I have 2N3055 NPN power transistor and it's PNP equivalent MJ2955. I am looking to get about 6W (yah ) out of this. Driving a 4R load.

I am familiar with Class A, B and A/B configurations but do not know about biasing them here.

Thanks!

You might as well use Class A if you are REALLY planning to use those TO-3 transistors .
No really , your logic is very flawed if you want to get 6W from a +-5V supply .
2R + 2R BTL operation possibly , but they won't be able to do 2R per side .

[PRR]

Audio Power Amplifier Design Handbook, Douglas Self
http://www.amazon.com/Audio-Amplifier-Design-Handbook-Fourth/dp/0750680725
(Other editions also)

This starts somewhat above your level and goes way-far beyond what you need.

Even so, I think you should get it.

> +5V/-5V supply voltage ...get about 6W... 4R load.

As others say: parts-available and result-desired does not add-up.

I can easily get 12 ohm 12 Watt tail-lamps. I wish to run them in my brother's 1942 Plymouth which is 6 volts. Do math. Don't work out.

You need math. But sine-audio computations can be tricky. Always cross-check your math with someone else's. One handy source is audio power-amp chips.
http://www.national.com/ds/LM/LM383.pdf
page 3  top-center chart

The LM383, perhaps a more-efficient power amp than your first attempt (there are benefits to integration), will make a bit over 2 Watts in 4 ohms with 10V total supply (equivalent to your +/-5V). To get "6W in 4r" takes 15 to 17V (+/-8V).

In math: 6W in 4r is 4.9V. We measure in Sine but the amplifier has to cover peak-to-peak, so >13.85V total supply. We also have voltage loss in transistors (and drive schemes), and resistor loss in bias resistors. You surely need 20% more, and 40% more is not uncommon. 16V-19V required.

"+5V/-5V" sounds like a regulated supply. A POWER amp generally should not "need" a regulated supply. Levels are high, and regulation wastes precious power.

A 12V AC CT transformer into FWB will give +/-7.5VDC. For 8r loads you want over 1,000uFd total filter; two 2,200uFd 10V would do. For 4r load you double this: two 4000uFd 10V caps. For 4r load the 12VACCT transformer should be rated near 1 Amp.

Preamp can probably be powered from two 150 ohm resistors to two 1,000uFd filter caps. Most audio circuits only need "smooth", not "regulated". It may come out +/-5V or +/-7V, but with the filtering it won't wobble enough to get into the audio.

BYW: "+/- supply" implies direct-coupled output to speaker. This has some advantages. But it has a major disadvantage for learning: half your your first dozen mistakes will put non-zero DC on the output, and significant DC burns speakers. PDF top of page 4 shows the classic single-supply capacitor-coupled connection. Any DC off the amplifier is blocked from the speaker by the big cap.

[PRR]

Meyer's "Lil Tiger" is a lovely SIMPLE transistor amplifier.

http://www.swtpc.com/mholley/PopularElectronics/Dec1967/PE_Dec1967.htm

The article has theory, and a graph of Power versus Voltage and Load. This graph will be near-right for any decent OTL transistor amp.

2N3055-class output devices are way more than it needs, but will work fine. The small parts can be 2N390x series.

[DavenPaget]

Meyer's "Lil Tiger" is a lovely SIMPLE transistor amplifier.

http://www.swtpc.com/mholley/PopularElectronics/Dec1967/PE_Dec1967.htm

The article has theory, and a graph of Power versus Voltage and Load. This graph will be near-right for any decent OTL transistor amp.

2N3055-class output devices are way more than it needs, but will work fine. The small parts can be 2N390x series.

Oddly simple , but sure is a nice little animal 

[R.G.]

I built four copies of the Plastic Tiger. Worked first time, no troubles, lasted until I retired them for reasons that seem trivial and silly now.

I'd go for the Plastic Tiger.

[PRR]

> the Plastic Tiger. Worked first time, no troubles

Yeah, and it uses the 3055/2955 pair that Chris has.

At +/-30V it is aimed rather higher than Chris' +/-5V or the +/-10V which would be rational for 6W in 4r.

I built the Universal Tiger. Twice. One would NOT run right, I now know the design is marginally stable and my sample was oscillating. The other worked fine and I built it into a film-sound speaker case as a guitar amp (in a day when tubes were going out of style). Sweet thing. Strong. Wish it was still around. But my impression is that Meyer was pushing the envelope a bit hard. And the "marginal stability" was with original part-numbers, mostly not available today. "Equivalents" may be just enough hotter to ensure instability.

Anyway.... you bias a 3055/2955 pair with diodes. There are other ways, some "better", and dozens of patented "improvements". For a first-amp, diodes work 99% as well as any other, and are 99% less trouble than fancy-tricks.

[R.G.]

I built the Universal Tiger. Twice. One would NOT run right, I now know the design is marginally stable and my sample was oscillating. The other worked fine and I built it into a film-sound speaker case as a guitar amp (in a day when tubes were going out of style). Sweet thing. Strong. Wish it was still around. But my impression is that Meyer was pushing the envelope a bit hard. And the "marginal stability" was with original part-numbers, mostly not available today. "Equivalents" may be just enough hotter to ensure instability.

The output stage with gain that Meyer used was a big part of the marginal stability. Ditching that would have helped. Makes the thing less dependent on output device parameters. I only built one, it went up in metaphoric flames. I later found out that I had broken the mica insulators.

Anyway.... you bias a 3055/2955 pair with diodes. There are other ways, some "better", and dozens of patented "improvements". For a first-amp, diodes work 99% as well as any other, and are 99% less trouble than fancy-tricks.

The Plastic Tigers were where I did my first Vbe multipliers. Glued trimmers to TO-220  packages, bolted the TO-220s to the heat sinks with the outputs.

I used TO-220s, not the piddly packages.

[PRR]

> output stage with gain that Meyer used was a big part of the marginal stability

In heavy hind-sight, I suspect that was only part of the issue (though maybe the more random issue).

The gain-stages (longtail pair and common-emitter) make a LOT of gain and there is no major response-defining network. Compare with LM101/'741: these have the pole-splitting compensation cap across the CE stage (what Doug Self calls VAS stage). Internal on '741, external on 101/748. Gain-bandwidth can be forced to tolerable phase-shift all the way to unity loop-gain. Meyer (and many others of the day) came from lower-gain systems where the lead-cap across the NFB series leg was good-enough.

Which is one reason I pointed-out the Lil Tiger. With just one stage of voltage gain it is fairly unlikely to get into trouble. (Flipside is low input impedance and "high" THD numbers.) Also the cap-coupled output is tolerant of mis-wires. The lack of protection or even 0.2r emitter resistors is frightening, but with 3055-size devices it probably won't smoke easy.

Huh, I just noticed it calls for two different types of bias-diodes, none of them popular today. I'd try 1N4007 for all three. That 1N645 may have been a double-diode, but I'd flame-test with three diodes. It will play loud just fine.