Amp Chip -> BJT AB Stage -> more oomph?

[KarenColumbo]

For years my only means of testing that which I threw together on the breadboard was a H&K "Trilogy" 100 W tube head in a Mashall 4x12". not ideal, to say the least, especially in a populated area like an apartment house.
So I bought some TDA7267, took one and build what the datasheet suggested and powered it with a 18V wall wart. Now I got (almost) clean sound into an Celestion 10" speaker that's JUST about the loudness I can cope with when testing out some stuff. It's very stompbox-friendly, too.

https://www.mouser.at/datasheet/2/389/stmicroelectronics_cd00000202-1204084.pdf

But it's just a tad too small (it puts out approx. 1.2 Watts).

My question:

Can I add a BJT AB stage at the chip's output to crank the signal just a bit more? I'm thinking of something like this:


This is just a question whether I should follw this line of thought or if it's technically a stupid idea because it's not possible in the physical realm.

[Rob Strand]

Hmmm, the data sheet says you should get 2W into 8ohm with 12V, admittedly that's at 10% distortion.   I expect something like 1.2W into 8 ohms is when the supply is *12V*.   (If you want, cross-check with the LM380 datasheet at 12V).  You have 18V so I wonder why you are only getting an output corresponding to 12V.  Is your 18V supply unregulated?  and is it 18V when it is unloaded?     That would mean it's probably 12V and the supply is dropping to 12V under load.   Alternatively you might be overloading the supply and it is shutting down (if regulated) ?

A bigger question is what power do you really want out of it?

The circuit you suggest can work but it has some problems which need fixing.  More the problem is it won't give you much more power.

[KarenColumbo]

i wouldn't mind if it's not much more power, to be honest. I just have a picture in my mind that it adds a bit more "amp feeling" - it's a push/pull thing after all, isn't it? But I guess mind pix and reality differ in many ways ... but if you could point me in some direction I'd be willing to dig into this

[Ice-9]

Why not build a small chip amp using TDA7030, this will give you more output without annoying the neibours.

[KarenColumbo]

Why not build a small chip amp using TDA7030, this will give you more output without annoying the neibours.

I've been considering this (and i have a couple of 3886 in the box, too, so I could go pretty loud), but where would the fun be in that? I'd like to learn something. I will dive into this AB-amping thing - got some books about amps, I will have a reading session.

[amz-fx]

Look at the LM384. It is a cousin of the LM380 and still being made. It will give you more power than the 380.

regards, Jack

[R.G.]

I've been considering this (and i have a couple of 3886 in the box, too, so I could go pretty loud), but where would the fun be in that? I'd like to learn something. I will dive into this AB-amping thing - got some books about amps, I will have a reading session.

I just got first article PCBs back for a 30-50W amp using the LM3886. It's indeed a workhorse.

Good for you for digging into discrete AB amps. Your output stage is only about four low-power trasistors away from a basic class AB amp. I suggest that you find a copy of Douglas Self's book on audio amplifier design. A local library may have a copy or be able to get one on loan. There is a huge amount of finesse expressed there, but the basics will be:
1) a differential bipolar input stage
2) a bipolar voltage amplifier stage, with either resistive or constant current collector loading
3) enough drivers to push the current into the output devices.

Surprisingly, much of the power you'll get out is determinied by your power supply. ignoring for the moment that real speakers have varying impedances, the AC power you can put into an 8-ohm resistor is dependent on the size of the AC voltage. You know, that power equals v-squared-over-r thing.

So if you have an 18V total DC power supply, you can't drive more than 18V peak to peak into a load resistor without fancy stuff like stepping up the voltage. In fact, you can't get that much, because your output devices won't saturate to 0V, so with 18V power supply. you can only get maybe 15V peak to peak AC for an output.

That gets you to 7.5V peak, and the AC rms of a sine wave with a 7.5V peak is Vrms= 7.5/1.414 = 5.3Vrms. The power is then 5.3*5.3/8 = 3.5W.

You can only do more power by raising the DC power supply voltage or lowering the speaker impedance. And this is all before you say anything much about the power amp itself.

If you want to learn classic AB amp design, go for it!   If you find you want to get this amp on the bench faster while you wait for the learning, and want to do a chip, I recommend the LM1875 or TDA2030 in a TO-220 package. These will get you up to 12-15-20 watts if you like. The actual power does very much depend on the power supply you feed it. The LM3886 can in fact be used - it will work with power supplies of a >minimum< of +/-10V, and that's pretty close to the 18V total we were talking about.

Internally, most of the chip amps are just versions of the diffamp-voltageamplifier-driver-output design of the classical AB amplifier, adapted to the limitations of IC design.

[PRR]

> Can I add a BJT AB stage at the chip's output to crank the signal

It already HAS a BJT AB stage!

Your add-on will reduce power a hair (due to Vbe drops and poor base drive).

If your "18V wart" were really holding even 12V at full wail, you'd have nearly 2 Watts.

Makes hardly any difference which chip you use, or if you go discrete. R.G. gives the math for any ideal amplifier, and notes there are are always losses. The commercial chips have low losses. This TDA7267 may have lower losses than the old classic LM380 or LM384 (because we can put fat PNP parts on chips now). The LM3886 actually has higher losses (because aimed at much higher rails). But really no different to the ear. (Except the LM3886 may not even stay on at 12V.)

If it is dropping 18V idle to 12V slammed, that's a LOT of drop for a hard-pushed audio amp.

The current approaches 0.4 Amps at full wail. Is the wart rated that much? Is the rating real?

[KarenColumbo]

Yeah, I see - it's a bit like building an engine on top of another, I get that now. Well - I'll have some of those chip mentioned with a higher power rating ordered Douglas Self "Audio power Amplifier Design" bought online. Can't wait to learn what goes on inside.

[Rob Strand]

Yeah, I see - it's a bit like building an engine on top of another, I get that now. Well - I'll have some of those chip mentioned with a higher power rating ordered

You should just build it, experience the problems, and try to fix them.   Doug Self's book is one of the better books but some of the points made could pass by beginners.

The problems would be crossover distortion and thermal run-away. 

A third problem is outright distortion.  When you design an amplifier the output stage is inside the feedback loop and the feedback reduces the distortion.   Your proposed "buffer" is outside the feedback loop so you loose that correction mechanism

Here's the general picture:
- If you underbias the output stage you usually don't get thermal runaway but you get crossover distortion.
- To remove crossover distortion you need to to bias the transistors slightly on.
- When you bias the transistors on they get warm.   
- When a transistor gets warm the Vbe drops (roughly 2.2mV/deg C).   
  The diode bias is fixed, so when Vbe drops the collector current increases.
  That causes the transistors to get hotter and they keep getting hotter (runaway) until something breaks.
  This process is called thermal runaway.  It will be in Self's book.
- A partial solution to this is to put the diodes in thermal contact with the transistors.  So when the
  transistor heats up  the diode heats up.  The diode's bias voltage drops at roughly the same mV/degC
  as the transistor.  That makes it track Vbe and it reduces the runaway problem.
- As an added measure you can/must add emitter resistors to the output stage.  When the bias current
  increases the voltage across the emitter resistors increases and reaches a point of equilibrium.
- What you will not find in the text books is how to choose the emitter resistors.  You will see small values
   from 0.1 ohms upto 20 ohms.      The emitter resistor actually depends on the size of the heatsink on
   the output transistors.

You can read a bit more here,
http://educypedia.karadimov.info/library/PushPullAmp.pdf

The designers of chips do all the hard work for you!

[PRR]

> Douglas Self "Audio power Amplifier Design" bought online.

I believe this (and Rob's TLDR) are over your head, at your present point on the path.

Long term, you will learn stuff.

Short-term, IMHO you should have put the $20-$80 into a beefy power-pack, probably your main limit. (In fact 3 minutes with a meter would confirm/deny wart-sag for free.) 18V audio amps are cheap from China and many of them work (unlike high power LOW-price offers with chips worth faking).

Self's older Small Signal Audio is also a strong if over-kill resource. As R.G. says, power amps are just baby-amps with bigger output stages, many of the concerns are the same.

[PRR]

> how to choose the emitter resistors.  You will see small values from 0.1 ohms upto 20 ohms.

In Caveman Design, it is obvious that they must be "small" relative to load. Yet non-Zero. Say we accept 5% loss. For 8r load this leads to 0.4r emitter resistors. For 600r line, 20-40r.

And to the Caveman Designer, obviously the heatsink depends on the power-out and the tolerable temperature.

> The emitter resistor actually depends on the size of the heatsink on the output transistors.

Yes, naive Caveman Design leads to the Discovery Of Fire, or at least epoxy-toast, until the thermal NFB loop is understood or plagiarized (complete). The hidden goal is to fiddle Re until the thermals are stable with the heatsink needed for power.

It's not 1976 any more. The Chip Designers know some more than the Caveman, and you get their extended thinking for sub-pennies on the dollar.

[amptramp]

One way of getting current amplification is to follow the design of the LH0002 which is a buffer used for video op amps:

http://elcodis.com/parts/1342966/LH0002CH.html#datasheet

This gives you the equivalent of a Darlington output for current gain.  Take DC feedback from its output and since the gain is unity, you can take the AC feedback from the amplifier output / buffer input.

[Rob Strand]

For 8r load this leads to 0.4r emitter resistors. For 600r line, 20-40r.

The hidden goal is to fiddle Re until the thermals are stable with the heatsink needed for power.

Some up front guidance and then post fiddling and testing is probably the norm.

[KarenColumbo]

Thank you for this information-rich answers! This sounds like it has all the makings of great tinkering - smoke included Never thought amping is so complex ...  sounds like interesting times ahead

[PRR]

> Never thought amping is so complex ...

https://www.diyaudio.com/forums/solid-state/

Even leaving out the cork-sniffing, there is a LOT to "amping".