How really do chip amps work?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2015
Paid Member
We all use these chip amps, TEA 2025, 8002A is what I have used upto now, knowingly, that is.I cannot quite figure out how it works. A transistor works by the fact that a small bass current will cause large swings in the current across the emitter and collector, I get that.

Sorry for asking, but the explanations I see are simply too complex.

Gainclone - Wikipedia

I am hoping the kind folks here on this forum will give an amplified explanation.
 
Administrator
Joined 2007
Paid Member
Chip amps are very like an opamp that can deliver lots of current. If the 741 opamp had more powerful transistors inside it, and if it was in a package that could be bolted to a heatsink then it could be thought of as a chip amp and used as such.

Operational amplifier - Wikipedia

Going from how a basic transistor works to a full amplifier is massive step :) If you read up on normal discrete circuit design and discrete amplifiers and can understand how they work, then it is easier to visualise what is inside the chip of a chip amp. The chip has all the semiconductors and resistors needed, but anything like big capacitors would be connected to the various pins as needed.
 
Moody is right. It's hard.

Yet there is an explanation that's not too difficult
You need three concepts.

Current gain (hfe) where icollector = hfe * ibase
Base-emitter voltage drop (VBE)
And (E=IR ... I=E/R ... R=E/I) ohms law.

Working with the idea that a transistor is a current multiplier, we use ohms law to turn that amplified current into a voltage gain. Sometimes. Some of the time. By putting a resistor (most simply, tho rarely in practice) in series with the collector and the power rail. E=IR ... volts equals amplified amps times that resistor.

The VBE is a "hill" to overcome. In practice, we usually bias the transistor so that it is partially conducting. The input signal is imposed on the biased base thus varying the conduction of the transistor in (hopefully) it's fairly linear region. Small changes of input control larger changes to collector current. E=IR turns the current to voltage gain!

After these concepts... the "tail that wags the dog" (final output) requires again using current gain, but more directly: instead of amplifying current to turn into voltage amplification , the last stage does not try to change voltage. Just amplify current.

It's a bit tricky...
But it works.

GoatGuy
 

PRR

Member
Joined 2003
Paid Member
...TEA2025.....I cannot quite figure out how it works. A transistor works by the fact that a small bass current will cause large swings in the current across the emitter and collector, I get that....

What is the problem?

"Amplify a teeny signal into a BIG signal."

What is "teeny" and "Big"? The TEA2025's input is about 0.3V into 30K, so about 1 microAmp. Its output is near 1 Amp. So one part of the answer must involve a current gain of a million to 1.

Let's look at our tool. One transistor gives a current gain nearer 100. We can't do it with one transistor.

We can find higher numbers, to 900+. However some of that gain is lost in bias and coupling. And a million is not happening. (Even with FETs.)

We can multiply current gains with multiple stages. A cursory study of transistor amps shows that multi-stage designs are VERY common, so we may be on a right track.

So a bare minimum, we want three stages. 100*100*100 makes a million.

"Amplify a teeny signal into a BIG signal."

Well, we need to break that down into sub-problems which may be solved with a single transistor (or occasionally 2-Q compounds).

1st stage: "input": accept a teeny signal and amplify considerably
2nd stage: "VAS": take that bigger signal and make it even bigger
3rd stage: "POWER": take the biggened signal and pump the big voltage and current swings to slap a loudspeaker around.

POWER amp design generally starts with the "tough" problem. Until recent decades there was no large choice of POWER transistors. We had to work them the best we could. Their fairly low gain determines the requirements on the 2nd stage. The Power stage's efficiency is very important for heatsinking and power supply cost.

The majority of transformerless designs use the power transistors as unity-gain followers. So the stage before needs significant current and LARGE voltage swing. To the rails both ways so the output uses all the precious power supply we bought. Douglas Self calls this the "VAS", Voltage Amplifier Stage, because its primary goal is voltage amplification and swing. A common emitter transistor is a fine choice.

First illustration in the attached shows VAS coupled to Output, two common ways. Next illustration shows a simple output stage, two emitter-followers with (optional) 2-D bias.

The CE VAS typically has a way-low input impedance, significant input current. To reduce the load on your preamp we add an "input" stage for more current gain, and to couple a typically center-referenced input to the rail-referenced VAS.

The third pic "Three Transistor AB AudioAmp" is about the simplest "workable" loudspeaker amp of good efficiency. In the form shown it has a real low input impedance and will suck hard on whatever tries to drive it. While we could beef the preamp, in context of a power amp it does not cost much to add another small transistor, avoiding unhappy users.

There is no basic difference between Discrete and Chip audio amp design. Chipwork limits your choice of transistors, and encourages an all-purpose part that can be sold in the millions to cover development costs. But it's just an extremity of the general audio power amp problem.

In context of chip-making, added small transistors are quite trivial, though power transistors are quirky and trickery may be helpful. On right I show the TEA2025 internal circuit. There are 18 transistors per channel, and a few more shared between channels. We figured at-least 3 stages, but chip transistors can be lowish-gain, but are nearly free, so they use 5 stages plus helper transistors all down the line. Individually, each transistor does a simple thing. Working out all the implications and interactions is one of the harder problems in audio design.
 

Attachments

  • PowerAmp-guts.gif
    PowerAmp-guts.gif
    66.2 KB · Views: 387
Last edited:
Its all black magic.
But get it wrong and the black magic smoke comes out and it no longer works.

Its all about voltage and current gain.
As we want an output of many volts and a few amps.
Some of the chip amp internals are for increasing voltage and some for increasing current.

Care has to be taken with pcb layout with chip amps. Get it wrong and it can become unstable, oscillate and blow itself up. I learned this from a TDA7294 project.
The feedback resistor path needs to be as short as possible to keep inductance of tracks down. It also needs gain of at least 22 or it becomes unstable.
 
Last edited:
Member
Joined 2015
Paid Member
There is no basic difference between Discrete and Chip audio amp design. Chipwork limits your choice of transistors, and encourages an all-purpose part that can be sold in the millions to cover development costs. But it's just an extremity of the general audio power amp problem.
PRR, thanks but first of all these chip amps are not Digital amplifiers?
Are there any examples of digital amplifiers on chips? Class-D amplifier - Wikipedia

Edit: XH-M531 Yamaha 2*20W Digital HIFI Audio Stereo Class D Amplifier Board YDA138-E

So again, these chips consist of multi-stage transistor amplifier circuits, then:

We can multiply current gains with multiple stages. A cursory study of transistor amps shows that multi-stage designs are VERY common, so we may be on a right track.

So a bare minimum, we want three stages. 100*100*100 makes a million.

"Amplify a teeny signal into a BIG signal."
Well, we need to break that down into sub-problems which may be solved with a single transistor (or occasionally 2-Q compounds).

1st stage: "input": accept a teeny signal and amplify considerably
2nd stage: "VAS": take that bigger signal and make it even bigger
3rd stage: "POWER": take the biggened signal and pump the big voltage and current swings to slap a loudspeaker around.

Several questions then: stage 1, 2 and 3 will require different transistors since the input signal is different (higher voltage swings in each case) and output signal is also different. Or is there some clever trick to use the same type of transistor in all three stages? The TEA 2025 internal circuit diagram is very useful.

Also, since discrete electronics use capacitors, how is it possible make capacitors on such a small scale, or is there some sort of work around?

Are these the same as OP-Amps or Gainclones, because as I understand it, OP-Amps were not initially used in audio amplification circuits nor were intended to for this purpose.
 
Several questions then: stage 1, 2 and 3 will require different transistors since the input signal is different (higher voltage swings in each case) and output signal is also different.

In IC design, there's almost limitless flexibility as far as the transistors you can use. In the CMOS processes that I've designed in I can scale the dimensions of each transistor from itty-bitty (and fast) sub-micron things to great big huge 1mm+ gate lengths, pushing tens or hundreds of milliamps around. I can choose different oxide thicknesses (= voltage), different threshholds, you name it.

Every single transistor can be optimised for the job it's doing, within the constraints of the process. The parasitics are practically non-existant compared to off-chip stuff, and transistors are completely free, so if I need a hundred to lower noise, well then I wire up a hundred.

Most of those constraints are actually around what sort of devices they offer - for example many bipolar processes don't have complementary parts, so you've got to do the whole lot with only NPN parts. The processes I used had limited resistors (using a nitride layer) so you avoided using resistors if at all possible, and often just used more transistors to do that. Ditto with capacitors. In monolithis processes capacitors bigger than a few pF are totally unheard of, so you find other ways to do that.
 
We all use these chip amps, TEA 2025, 8002A is what I have used upto now, knowingly, that is.I cannot quite figure out how it works. A transistor works by the fact that a small bass current will cause large swings in the current across the emitter and collector, I get that.

Sorry for asking, but the explanations I see are simply too complex.
As a chip amp has a dozen or more transistors in it, it is a lot more complex than a single Class A transistor amplifier stage. There's an awful lot of things (voltages, currents, biases, etc.) going on at the same time, and that's before you get to negative feedback.

I suggest reading through a textbook, specifically "The Art of Electronics."

Gainclone - Wikipedia

Ya gotta love Wikipedia:
"The Gaincard shook the audiophile community[citation needed] ..."
 

PRR

Member
Joined 2003
Paid Member
There may be a problem of word-meaning here, discrete, chip, digital.

I was not talking about digital. Your first message does not mention any digital amps.

Yes, in practice we may "need" transistors from 3mW to 300mW. In practice 300mW is the smallest transistor we can buy. It is not difficult to design a many-Watt amplifier with only two part-types (PNP and NPN) from input to driver, and then a couple 30W for the output stage.
 
Member
Joined 2015
Paid Member
As a chip amp has a dozen or more transistors in it, it is a lot more complex than a single Class A transistor amplifier stage. There's an awful lot of things (voltages, currents, biases, etc.) going on at the same time, and that's before you get to negative feedback.

I suggest reading through a textbook, specifically "The Art of Electronics."

Gainclone - Wikipedia

Ya gotta love Wikipedia:
"The Gaincard shook the audiophile community[citation needed] ..."
I wasn't really in the audiophile community then, were you "all shook up?" How about this one?
Practical Electronics for Inventors, Fourth Edition 4th Edition
by Paul Scherz (Author),‎ Simon Monk (Author)
 
The "problem" with this link (another thing, the .gif image links don't work, but maybe they're on archive.org) and the other links given (like Youtube - the EEVBlog guy is good, if a bit too long-winded, but then I have a bias against videos vs. reading) is ...

... They tell how to use op-amps (at least for simple gain stages), but they don't tell you what you're asking, which is what's going on inside them and how the transistors inside work together to make up an op-amp.

An op-amp is NOT a series of transistor amplifier stages put together, even thought it might be "reasonable" to conclude that.. It's more complex than that, and it's more than can be explained in a thread like this.
 
…An op-amp is NOT a series of transistor amplifier stages put together, even thought it might be "reasonable" to conclude that.. It's more complex than that, and it's more than can be explained in a thread like this.

You're right of course. And operational amplifier is a series of non-obvious sophisticated stages and elements composited in such a way as to achieve useful specifications to the engineers and designers that'll go on to utilize the thing once fabricated.

There are devices on board that are trivial to produce in silicon wafer (such as dual-emitter or more-emitter transistors, or intrinsically-thermally coupled current mirrors) that simply cannot be produced with discrete off-the-shelf elements. Is this a bug or a feature? Well… it turns out to be critical for the stable operation of op amps.

Part of the extraordinary requirement for most 4th generation and beyond op amps is that they work nearly “rail-to-rail” linearly. That they tolerate reactive loads (that can “back up” considerable energy into the op-amp). That they nominally can achieve “stable operating point” at 0 dB gain in either positive-gain or negative-gain (inverting) setups. These are quite important. Add in the requirement for nearly-identical input impedance between the + and - channels (or these days, “special transistor” input, also matched, with JFETS), and you have a heck of a design tasklist.

Then you have “do this without capacitors” (because capacitors are really, really hard to integrate onto a silicon wafer … in silicon … of any appreciable capacitance), or inductors, or anything beyond modest-value resistors. (Silicon resistors, while they can be made in a wide range of values, are hard to make in large values. And run-to-run, they vary by a considerable amount.)

So the designer(s) of operational amplifiers also have to accommodate idiosyncrasies of silicon fabrication. Idiosyncrasies that aren't perfectly controlled in practice.

As posters have already pointed out, this discussion is really too complex to contemplate diving into in any serious level of accuracy and detail on this forum. It is. I could, I'm tempted to, but … being as old as I am, I know better. LOL!!!

GoatGuy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.