Driving a speaker with a single transistor

[phaeton]

I have finally gotten a little time* to sit down and play around with the breadboard, so I looked over some datasheets, made some calculations asbestos I could and devised this.  Basically an LPB-1 gain stage driving a TIP41,  which drives a speaker.

It twerks!  But, it's pretty weak.  It can drive a 16-ohm 2.5" speaker I had, along with a 16-ohm 6" speaker, but for some reason I got nothing out of a 4-ohm 6x9. I put my Rangemaster clone in front of it thinking it needed more pre-amp stages.  That helps, but it also overdrives the 2n3904.  increasing the voltage supply to the circuit helped, but not proportionally.  It gets to max volume level around 34 volts, and isn't any louder on the way up to 48V.  I also reduced the value of the collector resistor on the TIP41, thinking more current would mean more oomph.  4.7K seems to be the sweet spot.  Going lower than that, the volume decreases.  I get a big jump by bypassing the emitter resistor, but then the bias is off.  It's about speaking volume at best.

Is there something I need to change to get more performance out of this circuit? Or is this pretty much the reason why single-transistor Class-A designs aren't talked about or used much, but Class AB designs are? 

Thanks for any help or suggestions.


* I've been working 50-60hr weeks back to back since the middle of May.  All work and no play makes Jurd a very dull boy.

[PRR]

> I got nothing out of a 4-ohm 6x9

There should be "something". I can believe it may be softer than the naked strings speaking.

Removing emitter bypass cap should NOT throw "bias off".

I could teach about appropriate resistor ratios and such, but I just came off a hot roof.

At very best (you are not close) a resistor-coupled audio 'power amp' is 8% efficient, compared with >70% for push-pull topologies. That's why the last time it was seriously used to drive a loudspeaker was 1959, a 1-transistor megaphone with DC flowing in the voice coil.

[phaeton]

> I got nothing out of a 4-ohm 6x9

There should be "something". I can believe it may be softer than the naked strings speaking.

Likely true.

Removing emitter bypass cap should NOT throw "bias off".

To be fair, there was no capacitor.  When I say 'bypass' I meant removing the emitter resistor and replacing it with a jumper wire.

At very best (you are not close) a resistor-coupled audio 'power amp' is 8% efficient, compared with >70% for push-pull topologies.

'8% efficient' sounds like a Class A transistor amp is a lost cause.  Kinda what I expected.

That's why the last time it was seriously used to drive a loudspeaker was 1959, a 1-transistor megaphone with DC flowing in the voice coil.

How do you know these things?

[Rob Strand]

DC flowing in the voice coil.

The DC through the voice coil is evil on single transistor ckts.
The speaker actually sits pushed out.  If you try to bias the amp for full o/p swing the DC through the speaker is often larger than the power rating of the speaker!

When you use a single transistor with an o/p transformer the transformer deals with DC and the speaker doesn't see any of it.

[amptramp]

For a time, you could get 45 ohm speakers designed specifically for use in single-ended transistor amplifiers.  They were not quality or audiophile components by any means but you could connect them between the output transistor collector and Vcc and they did work, most often used in intercoms.  Today it would be much easier to just add transistors or go to an IC.

[Danich_ivanov]

I'm thinking that one transistor power amp is tricky to pull off. One thing I would try with this circuit is to play around with feedback biasing within the last transistor, sometimes it helps with getting more juice out of it, otherwise all that's left is to add another transistor.

[amptramp]

One transistor power amps were common in the 1950's with car radios that used 12 volt tubes for the radio portion of the signal and a 12K5 as a driver for the transistor, often a 2N176.  I have a radio from a 1957 Buick that uses this design.

[phaeton]

Thanks for all the help and recommendations, folks.  Learning above from PRR that the best-case efficiency would be 8% indicates there's not much more to wring from this design.  Since I'm good at throwing stupid effort after bad, I tried a few other things anyway:

1) Bypassed the emitter of the TIP41 with a 10uF cap.  This almost doubled the volume of the circuit.  Doing the same for the 2n3904 only caused oscillation.

2) Played around with small amounts of positive feedback.  This also resulted in oscillation before any benefit could come of it. 

3) Using small amounts of negative feedback along with an emitter bypass cap on the 2n3904, hoping to find a balance of more gain/no oscillation.  No dice.

4) Doubling up transistors.  Plugging a second TIP41 into the board with the pins of both transistors jumped together: emitter->emitter, base->base, collector->collector.  I have seen pairs of transistors on each side in some AB designs and (of course) pairs of tubes in various output sections.  Was feelin' lucky, but it was just a feeling.  Now I feel... sleepy.

A few things still strike me as odd, though.  For starters, the power transistor never got hot.  Not even warm.  At idle it was only pulling 17mA, even with a 34V supply and a 4.7K collector resistor.  I would have expected it to require a heatsink at all times.

DC flowing in the voice coil.

That sounds like a terrible idea.  This is why I used a capacitor.  I'd have used a transformer but I don't have any other than tiny 1:1 types.

One transistor power amps were common in the 1950's with car radios that used 12 volt tubes for the radio portion of the signal and a 12K5 as a driver for the transistor, often a 2N176.  I have a radio from a 1957 Buick that uses this design.

Back when transistors cost more than tubes, I bet.  Then they got cheaper, and I bet now a NOS 2N176 would cost more than a 12AX7.  I've ridden in some cars from the 50s and had some all-solid-state radios in my cars from the 60s.  I don't remember any of them being super loud, but definitely louder than what I've got here.  Of course, those radios were made by real engineers, not bedroom hacks like me.

[Rob Strand]

One transistor power amps were common in the 1950's with car radios that used 12 volt tubes for the radio portion of the signal and a 12K5 as a driver for the transistor, often a 2N176.  I have a radio from a 1957 Buick that uses this design.

I remember some of the radios had one 2N301 power transistor.  I'm pretty sure those designs used a transformer.  This looks like the thing,
http://archive.siliconchip.com.au/cms/A_30859/article.html
It's more of an auto-transformer/parallel inductor.  The DC across the transformer and speaker is small.  I suppose it reduces the size of the transformer since no isolation is required.

Many TVs in the 70s had single transistor power amps and they used a transformer.

[PRR]

> One transistor power amps were common in the 1950's with car radios

With output transformers (chokes).

> I suppose it reduces the size of the transformer since no isolation is required.

The core still has to be sized for the DC current. The DCR will be a touch lower since you share most of one winding for both source and load.

---------------------------------------
> 45 ohm speakers designed specifically for use in single-ended transistor amplifiers.

The usual excuse was for intercoms. You want 1000 feet of wire between stations. Working at 4 Ohms, you have huge loss (or way-huge wire cost). At 45 Ohms you can use 1% as much copper. They were also worked as the "microphone" in intercoms, and the higher impedance gave more voltage to overcome 1st stage hiss. (These were also made with-horn for factory intercom systems.)

> the DC through the speaker is often larger than the power rating of the speaker!

Those megaphones ran on battery. Battery cost meant super low power. You might typically have 1/2W DC heat in the voice coil, and the cheapest horn speaker was rated 4W. Yes the diaphragm sat off-center but was super-stiff and moderately tolerant of this.

[phaeton]

What a neat article.


The title of the other article in that issue "Are electric brains taking over?" probably spawned at least one story arc in a scifi comic book at the time.

[Rob Strand]

The core still has to be sized for the DC current. The DCR will be a touch lower since you share most of one winding for both source and load.

Agreed,  DC is going to be determining factor.

Yes the diaphragm sat off-center but was super-stiff and moderately tolerant of this.

A Hi-Fi purist's dream. The fact the button is only on for short periods probably helps the cause as well.

[Rob Strand]

The title of the other article in that issue "Are electric brains taking over?" probably spawned at least one story arc in a scifi comic book at the time.

The thing I find hilarious is they actually have.    Nowadays a large proportion of the population spend most of their lives being shown an image on their computer or phone, then they click a button in response.   It's like one of those rat experiments where a  reward is given after clicking the right button.

[stringsthings]

The title of the other article in that issue "Are electric brains taking over?" probably spawned at least one story arc in a scifi comic book at the time.

The thing I find hilarious is they actually have.    Nowadays a large proportion of the population spend most of their lives being shown an image on their computer or phone, then they click a button in response.   It's like one of those rat experiments where a  reward is given after clicking the right button.

What?  ... oh, wait a minute, i've got to check my smartwatch .... ( minutes pass )

[PRR]

> an LPB-1 gain stage driving a TIP41,  which drives a speaker.

What is an LPB made to drive? Something like 50K Ohms or above; a g-amp input.

What you driving here? A few-Ohms speaker.

The amplifier MUST be proportioned to what it will drive.

The transistor has no magic power to supply current out of thin air.

Look at the plan you drew. ALL the current to the speaker (and transistor) must come through the 10K resistor.

> the power transistor never got hot.  Not even warm.

You know Ohms Law. Apply it.

18 Volts through 10K to a variable transistor. Assume the transistor turned full-on. 18V on 10K is 1.8mA. 18V times 1.8mA is 0.032 Watts of heat in the resistor. And zero in the transistor (full-on, zero voltage).

A full-on transistor is not gonna pass audio. We normally set-up for the resistor and transistor in "a fair fight", each one taking about half the supply voltage. Now we have 0.9mA, 9V across each device, 0.008 watts each. Of course a ~~1W part is stone-cold.

ALL the current to the speaker and transistor must come through the 10K resistor. Use simple numbers. Suppose you had 2 miles of hair-fine #40 pickup wire, which is about 10K Ohms, and you tried to drive a 10 Ohm speaker at the far end? Simple proportion says that to get 1 Watt at the speaker you need to lose 999 Watts in that over-resistive speaker cable! Which is why we don't use 2 miles of #40; a speaker wire is #20 and not over a few dozen feet. Less than 1 Ohm.

The 390r resistor limits pulldown, not as much as 10K, but also needs to be far smaller.

OK, with this insight let us propose changing 10K to 1 Ohm (a fair speaker wire). With many other mods this WILL drive a speaker! BUT it takes 9 Amps at 18V or 81 Watts of heat. The peak pull-up is only 8/9 of 9V or 8V peak, or 4 Watts Sine RMS. 95%(-100%) of that 81 Watts is just waste heat.

We could try "matching". Make collector resistor 8 Ohms. I just saw a Famous Designer do this. It works and does not suck. But actually you want to "match" two values: the collector resistor and the transistor minimum resistance (under 1 Ohm), and then you need to allow for DC to AC conversion.

If you go round and round: for 8 Ohm load, the optimum collector resistor is 1.414X or 11 Ohms. (Emitter resistance should be zero; you need another way to enforce bias.) The optimum collector point is 0.3 of the supply voltage, 5.4V. Supply current is 1.1 Amps. 6 Watts in transistor, 14 Watts in resistor. Sine power output is 1.8 Watts.

Note that you probably want to drive the whole thing with sources making <1V in >50K. i.e. the sources can make 0.02mA of current. At the end you want swings over 1 Amp. You need a total current gain of >50,000. You might think you could get there with two devices of hFE=224. But bias-losses make that hopeless. And TIP41 is not assured to give gain of even 50 (@1A). Unsophisticated design can get you to signal current gain near square-root of hFE. I'd pencil three high-hFE parts and your TIP. I'd combine them in feedback pairs for less loss of current-gain.

[amz-fx]

regards, Jack

[amptramp]

I exhumed the Buick radio from under a couple of tuners and the output stage was a Delco D501 transistor going into a ground-referenced autotransformer to the speaker terminals.  This is still a rather large, bulky radio.

[Rob Strand]

I exhumed the Buick radio from under a couple of tuners and the output stage was a Delco D501 transistor going into a ground-referenced autotransformer to the speaker terminals.  This is still a rather large, bulky radio.

(exhumed  ) Thanks for digging that up.   Good to know that scheme was common.

The article I posted before showed the Ferris brand.  I believe the company was owned by the Ferris brothers and they manufactured them in AU.  A popular brand in the day.

[phaeton]

Hey guys, thanks for the tips, and thanks to PRR for the logical breakdown.  Looks like I don't quite understand what I was trying to do nearly as well as I thought I did.  I worked a couple more 10hr days but I might have a chance this weekend to read up on some more theory.  Maybe then I'll be able to take a proper stab at it.

Thanks again!



p.s.: wow! A Jack Orman sighting!