Why always differential stages?

[ashade]

why always differential stages?

Hello, my name is Alex and I'm studying electrical engineering - 4th year. Last 2 weeks, I was trying to make my first audio amp. Iwas looking at some high power amp designs and they always begin with differential amplifiers - I mean, the first amplification state is always the differential pair. Why the others configs (ie. common emitter, common base, etc.) are not used?

PS: is there a good book for engineers covering audio amp design (not analysis only, like Microeletric Circuits) that you recomend me??

 

[EC8010]

The standard book is Douglas Self's "Audio Power Amplifier Design Handbook (3rd Ed.)". It deals with the optimisation of the most popular circuit (originated by H C Lin), and yes, it uses a differential pair as the input stage. A good counterpoint to this book is Ben Duncan's "High Performance Audio Power Amplifiers". Randy Slone's "High-Power Audio Amplifier Construction Manual" is remarkably similar to Self's book.

Oh, and welcome to the forum.

 

[Jennice]

Hi ashade, and welcome to the club


Why use a differential pair...?

The differential pair is there for a reason - especially in (high) power amplifiers.

It is difficult enough to make a high quality small signal amplifier without overall feedback, as you want to operate all active components (transistors) within their most linear range to minimize distortion.
This makes many of these no-global-feedback designs inefficient in terms of power. Therefore you'll only find very few (expensive) power amplifiers without feedback for error-correction.

this brings me to the point: The differential pair is a well-proven, cheap, flexible and good way of creating a global-feedback path.

You'll find some amplifiers which have local or nested feedback loops, where the feedback only works on a given stage of the amplifier, but the the majority rely on the differential pair for the mentioned reasons.

Regards,
Jennice

 

[Mr Evil]

Differential pairs offer:

• Differential input (duh!)
• Symmetrical input
• High input impedance
• Symmetrical output
• Low distortion
• Low DC offset
• Thermal stability.

and probably a whole load more advantages I can't think of right now. They're hard to beat not only for audio, but for all sorts of applications, which is why they also feature prominently in op-amps.

 

[ilimzn]

You should also expand your search, there are other input topologies used. Most commonly, either single ended or complementary current feedback, the first is used mostly in simpler amps, the second is very popular with complementary FET inputs (and is usually cascoded). You might also find that most of these alternative topologies use some sort of DC error correction, or are AC coupled so a small but not well controlled DC error is not a problem - this should be one clue as to why differential inputs are so popular.

 

[TonyTecson]

there was an article published in the AUDIO magazine about the evolution of the amplifier circuit topology! i couldn't now recall which year and month? a very good read if it can be found.

 

[AndrewT]

Hi,
the late John Linsley Hood wrote "the art of linear electronics" publisher Newnes.
Quite a bit more than just audio amps but well written & thoroughly recommended. A good complement to Self.
He also deals with one & two transistor single ended amps for the front end & then quickly moves on to LTP etc.
BTW. single ended are reputed to be significantly quieter than LTP.
I wouldn't bother with Horowitz & Hill nor ARRL handbook.

 

[tlf9999]

I have been a big fan of SE input stage (aka JLH) for its speed, simplicity and ease to compensate.

However, it does have poor control of zero voltage so using a capacitor on the output is mandatory.

 

[jcarr]

Ashade:

It is possible to close a global NFB loop through the emittor/source of a single input device, so ease of global feedback shouldn't be the obvious answer for the popularity of LTPs.

As Evil and ilimzn mentioned, LTPs have some significant advantages, and IMO issues relating to DC stability are the most notable.

And as AndrewT mentioned, LTPs are less efficient at achieving low noise levels than single (or complementary single) input stages, and for amplifying low-level signals like those from microphones or MC cartridges, may not be the optimal choice.

Regarding reading material, the other posters have made some good recommendations in English, but if you care to expand the choice of language outside of English, I am sure that there must be additional recommendations as well. I could offer some good recommendations in Japanese, but I fear that they may not be of much immediate use to you .

hth, jonathan carr

 

[ashade]

you know, my great problem is due to low experience. For example, I was trying to make a simple amplifier for a mic and a speaker, so I started gettin some tech info about it. After some google searches, I found out that for the mic I have, I had to make an amp able to amplify the signal about 800 times to hear something on the speaker. So I started this way:

1) I mixed the input stage and the high gain stage into one, using a 150k resistor with a common emiter config. Testing with spice, with a resistive load of 100k I got about 1000 times of amplification and a very high impedance input.

2) Then I satarted with the output stage and everything gone dark. I tryed lots of output configs but I don't know why, none of them was able to get a 5V peak signal as input and make a 5V peak signal to the 4ohm output without clipping or distroying the sine form of the wave.

You got my problem? it's about output stages or stages that need to handle large signals (200mV +). What you can tell me about it?

 

[sam9]

The book by Ben Duncan, mentioned above is especially interesting with respect to the historical development of topologies. For instance with regard to differential pairs -- it wasn't always so. Early SS designs often didn't use them.

 

[tlf9999]

Early SS designs often didn't use them.

that may have something to do with 1) costs and 2) difficulties in finding matching transistors (wide manufacturing tolerance).

 

[forr]

For amplifier circuits not using a differential pair at the input, have a look there :

http://www.angelfire.com/ab3/mjramp/

~~~~~ Forr

§§§

 

[bear]

you know, my great problem is due to low experience. For example, I was trying to make a simple amplifier for a mic and a speaker, so I started gettin some tech info about it. After some google searches, I found out that for the mic I have, I had to make an amp able to amplify the signal about 800 times to hear something on the speaker. So I started this way:

1) I mixed the input stage and the high gain stage into one, using a 150k resistor with a common emiter config. Testing with spice, with a resistive load of 100k I got about 1000 times of amplification and a very high impedance input.

2) Then I satarted with the output stage and everything gone dark. I tryed lots of output configs but I don't know why, none of them was able to get a 5V peak signal as input and make a 5V peak signal to the 4ohm output without clipping or distroying the sine form of the wave.

You got my problem? it's about output stages or stages that need to handle large signals (200mV +). What you can tell me about it?

Are you *sure* you're a 4th year EE student?



If you have a simple input amplifier with the requisite voltage gain, try a FOLLOWER for an output stage, that will give you current gain to match your voltage swing. Since P = I * E that works.

You will probably have to learn something about level shifiting if you want to keep it DC coupled, but start with AC coupling and that will be very simple and easy, no level shifting required.

Make sure you have biased your output follower properly - then it will swing. One transistor? It's Class A - bias it so.

One can always copy a design from a magazine or book to start.

Keep in mind that the conversion from Mic level to Line level is usually a separate amplification stage, as is the conversion from Line level (~1Vac) to "Speaker" level. They're not usually done in one circuit and certainly not in two stages of one circuit. Practical circuits are not two transistors either, not with this overall gain for sure.

_-_-bear

 

[sam9]

Are you *sure* you're a 4th year EE student?

Maybe they just do software these days.

I recall many years ago an RF engineer with whom I was aquainted had to spend a week do interviews at Stanford. He came back real grouchy. Everybody he talked to had concentrated on digital. His comment: "Any fool can count to one."

 

[PMA]

Input differential stage fits to VFB amp topology. In case you decide to use CFB amp, there is no special need for differential input stage.

 

[Danko]

Maybe they just do software these days.

I recall many years ago an RF engineer with whom I was aquainted had to spend a week do interviews at Stanford. He came back real grouchy. Everybody he talked to had concentrated on digital. His comment: "Any fool can count to one."

R0TFL R0TFL R0TFL R0TFL R0TFL R0TFL

 

[tlf9999]

Input differential stage fits to VFB amp topology. In case you decide to use CFB amp, there is no special need for differential input stage.

LTPs are also used in CFB amps too, except the inputs aren't on the bases.

 

[azira]

Lots of good reasons thrown out about why differential stages are good, but I think the most obvious reason (atleast my oppinion) why they are used wasn't really mentioned.

It's really easy, simple, and straight forward to use a differential stage to apply voltage feedback. Your input and output both (hopefully) have zero DC offset.

IMO, it's a considerably more difficult challenge to apply global feedback from a zero DC offset output to a non inverting amplifier that has a zero DC offset input signal.

A diff pair lets you do that without even thinking about it. Just hook it up.
--
Danny

 

[PMA]

LTPs are also used in CFB amps too, except the inputs aren't on the bases.

Not necessarilly. Depends how one solves the task.