BJT Amplifiers

[Cogman]

This MAY be considered non-highly technical. But, either way, I felt this is the best place for this.

We are supposed to construct a 30Db (+- 3 Db) power amplifier. The amplifier has a 50 ohm input and a 8 ohm output impedance (IE, Line in from some source and speaker output).

So the thinking was, us a base amplifier for the first stage, a emitter amplifier for the second stage, and a collector amplifier for the second stage. I don't want to use the common versions of these amplifiers because I want to have some fine tuned control over the amplification.

Our teacher hasn't been very good at explaining how BJTs work. But here is my best shot and where I'm at, (any other information would be helpful).

First off, I wanted each stage to have a 10Db amplification, this should prevent the distortion that occurs when the transistor hits saturation. For the emitter (and here is where I'm stuck) I was thinking something like putting a 50 ohm in parallel with a capacitor and the voltage source. This, I believe, gives a ~ 50 ohm input impedance. The next problem I see is that I want to swamp out the r_pi so that I don't have a thermal dependency (or that I minimize this.) How do I go about doing this? Is it just attaching a resistor onto the base?

And finally, do I just use the resistor on the collector to determine the amplification? Is there any way to compensate for the varying betas, or do I need to just pick resistors based on the betas of the transistors I get?

Thanks for the help, and links or resources would be helpful (Though, the common base, emitter, and collector stuff is not so helpful. I get that, I just don't get swamping and when/how to correctly apply it)

 

[uclabachelor]

This MAY be considered non-highly technical. But, either way, I felt this is the best place for this.

We are supposed to construct a 30Db (+- 3 Db) power amplifier. The amplifier has a 50 ohm input and a 8 ohm output impedance (IE, Line in from some source and speaker output).

So the thinking was, us a base amplifier for the first stage, a emitter amplifier for the second stage, and a collector amplifier for the second stage. I don't want to use the common versions of these amplifiers because I want to have some fine tuned control over the amplification.

Our teacher hasn't been very good at explaining how BJTs work. But here is my best shot and where I'm at, (any other information would be helpful).

First off, I wanted each stage to have a 10Db amplification, this should prevent the distortion that occurs when the transistor hits saturation. For the emitter (and here is where I'm stuck) I was thinking something like putting a 50 ohm in parallel with a capacitor and the voltage source. This, I believe, gives a ~ 50 ohm input impedance. The next problem I see is that I want to swamp out the r_pi so that I don't have a thermal dependency (or that I minimize this.) How do I go about doing this? Is it just attaching a resistor onto the base?

And finally, do I just use the resistor on the collector to determine the amplification? Is there any way to compensate for the varying betas, or do I need to just pick resistors based on the betas of the transistors I get?

Thanks for the help, and links or resources would be helpful (Though, the common base, emitter, and collector stuff is not so helpful. I get that, I just don't get swamping and when/how to correctly apply it)

You will have to do DC analysis on each stage of your amplifier to ensure that the transistors are biased and operating in the middle of the active region to allow the maximum voltage swing.

Once your DC analysis is complete, then you will need to use the hybrid-pi or the h-model to do the AC analysis. You will most likely have to bounce around DC / AC analysis a few times before arriving at a good solution.

Since it is a three-stage amplifier, the first stage is usually a buffer (gain of -1 or close to it), 2nd stage provides amplification (30dB in this case), and last stage is the output driver in push-pull format, biased so as to stay just above the cutoff region to prevent crossover distortion.

Some thoughts to help you out:

1) The 2nd stage almost has to be a common-collector circuit due to the high input impedance.

2) A resistor in the emitter of the 1st stage reduces a lot of the transistor's properties and its effect on the circuit due to the reduction of gm. The gain would be approximately -Rc/Re, which is unity if both are selected to be the same.

3) The input impedance due to #2 is going to be high (beta + 1) * Re, so a simple RC low pass filter will satisfy the 50ohm requirement.

4) There are three amplifier types - common collector, common base, and common emitter. Each of those have advantages and disadvantages, some of which I have already mentioned.

If you want to get fancy, then try a differential input stage.

 

[DanDaManJC]

I'd guess this is the final project in your bjt class?

What most people in my class did, along with what I did, was a 3 stager.

1st stage: Differential input
2nd stage: common collector -- as the other guy said, has a high Rin so it wont load down your 1st stage
3rd stage: class a/b push pull output

with the differential output, you can run a negative feedback line from your output to the negative input of your diff amp. this actually works to minimize the error / distortion caused by the a/b output stage. remember, that diff amp is a very basic op amp, and op amps (ideally) try to make their two input terminals equal.

other stuff to consider... look at your equations for the BJTs. Off hand I forget --- but if you daisy chain the BJTs just right, often times you'll see exponential increases in Rin, Rout or the bias currents... and often times your overall gain will be some function of Rin or Rout. So use things like darlington configs for more gain.

Another tip, remember... the input current to the "gate" (sorry, forgot the right term for the bjt) is very small compared to the collector/emitter currents. use that to your advantage (that idea ties in with the darlington config stuff)

also generally... since this amp will be a VCVS... you basically want Rin->infinity and Rout->0 for your amp. That is... you want the Rin very high so you don't have a voltage divider between your input signal and your amp --- you want all of that input signal to be processed and not lost due to voltage division at the input. Likewise, you want all of your output signal to drop across the load and NOT across some internal amp impedance. This concept will very likely be on a test too... (what kind of loading conditions would you want for a VCCS? etc)

 

[TuxDave]

1st stage) Common emitter
2nd stage) Common collector

That's probably your basic two stage amp that would make it work. Make sure you put in negative feedback to control your gain and do your frequency analysis to make sure that it's stable.

 

[Cogman]

1st stage) Common emitter
2nd stage) Common collector

That's probably your basic two stage amp that would make it work. Make sure you put in negative feedback to control your gain and do your frequency analysis to make sure that it's stable.

Wouldn't a Common Emitter have a high input impedance? For maximum power transfer we would want to match the 50 ohm input impedance. given in the specs. I agree that a common collector for the final stage is pretty much a given.

 

[Cogman]

I'd guess this is the final project in your bjt class?

What most people in my class did, along with what I did, was a 3 stager.

1st stage: Differential input
2nd stage: common collector -- as the other guy said, has a high Rin so it wont load down your 1st stage
3rd stage: class a/b push pull output

with the differential output, you can run a negative feedback line from your output to the negative input of your diff amp. this actually works to minimize the error / distortion caused by the a/b output stage. remember, that diff amp is a very basic op amp, and op amps (ideally) try to make their two input terminals equal.

other stuff to consider... look at your equations for the BJTs. Off hand I forget --- but if you daisy chain the BJTs just right, often times you'll see exponential increases in Rin, Rout or the bias currents... and often times your overall gain will be some function of Rin or Rout. So use things like darlington configs for more gain.

Another tip, remember... the input current to the "gate" (sorry, forgot the right term for the bjt) is very small compared to the collector/emitter currents. use that to your advantage (that idea ties in with the darlington config stuff)

also generally... since this amp will be a VCVS... you basically want Rin->infinity and Rout->0 for your amp. That is... you want the Rin very high so you don't have a voltage divider between your input signal and your amp --- you want all of that input signal to be processed and not lost due to voltage division at the input. Likewise, you want all of your output signal to drop across the load and NOT across some internal amp impedance. This concept will very likely be on a test too... (what kind of loading conditions would you want for a VCCS? etc)

not a final project. We haven't actually covered any of the stuff you've just mentioned. We have only hit common base, common emitter, and common collector amplifiers.

But like I said, the teacher has been really lax on the equation. We have pretty much only covered the pi model (which, from what I've google is really the small signal model that we've been taught). So I'll see what else I can dig up. All in all, you guys have given me more google fodder.

 

[TuxDave]

Wouldn't a Common Emitter have a high input impedance? For maximum power transfer we would want to match the 50 ohm input impedance. given in the specs. I agree that a common collector for the final stage is pretty much a given.

Ok, now rereading your specs for the 3rd time, I totally glazed over your input impedance requirements. Yeah, CB, CE, CC is the obvious choice. I'm not sure if you want to bother with any gain in the last stage. Build your 30dB from the first two stages using a feedback loop.

 

[PsiStar]

My favorite xsistor pair is a Darlington ... been a long time since i built an amplifier tho.

 

[esun]

From the I/O impedance requirements, you need a CB input stage and CC output stage. No other reasonable way to achieve that. 8 ohms output is going to require a lot of current in your output stage so I hope you have some beefy transistors (or have a bunch you can put in parallel).

There's no way you're going to get 10 dB in gain per stage. In order to get 30 dB of gain you will need a second gain stage (aside from the CB input stage) which will likely be a CE stage (easiest to bias in that topology). So your overall topology sounds correct (CB-CE-CC). Even if you don't ground the base/emitter/collector, we still call these "common-base/emitter/collector" amplifiers.

The general design procedure goes like this:

Input resistance of a common base is 1/g_m, where g_m = I_C/V_T, meaning you need V_T/I_C = 50 Ohms, or I_C = 0.52 mA (recall V_T = 26 mV). So you need to bias your common-base stage to achieve that collector current. You'll use something like a current source bias in the emitter of the CB amplifier and capacitively couple in your input signal (you could also use a resistor but will have to tune it to maintain the proper R_in based on the resistor value and resulting I_C).

Picking R_C for the CB stage depends on how you want to connect to your CE stage. If you have separate bias circuitry and capacitively couple, then you can basically pick any R_C you want as long as your keep the CB stage in forward active. To maximize swing you'll want the output biased around V_CC/2, so you could shoot for that. That will determine your gain, though. If you couple directly, then you'll need to pick R_C carefully to set V_B for the next stage. You'll almost certainly need emitter degeneration to bias that stage reasonably.

CE stage will not load your CB stage too much due to its relatively large R_in. If it does, degenerate it with a resistor and tweak that value. Again you'll have to design the R_C for the CE stage based on necessary biasing of the final stage and gain requirements.

For the CC stage you'll only be losing gain. Generally to get the most out of a CC stage you need to pump tons of current through it. If you have extra transistors, put then in parallel with your CC stage and you should be able to get closer to unity gain out of it.

Don't bother with differential. If you don't have a CMRR spec then there's absolutely no reason to go with differential. Matching can be a pain and you'll need twice as many parts (which I know can be sparse in an undergraduate lab). Biasing is going to be the hardest part generally speaking so think carefully about that.

BTW, 10 dB isn't a lot of gain for a stage. You don't have to worry too much about distortion with a gain of 10. For example, if you expect your input signal to be of say a 1 mV amplitude, then your output would be around 10 mV amplitude, perfectly reasonable. After 30 dB it'll be up to 1 V, so that's an issue for large input swings (although this depends on your V_CC which I'm guessing for discretes is at least 5 V).

Ok I just saw your comment about biasing the first stage (with a voltage source, resistor, and capacitor in the emitter of the CB amplifier). So my first question to you is: What is the output resistance of the structure you describe (ignore the cap for now, just consider a voltage source in parallel with a resistor)? If it isn't obvious, ask yourself what the resistor is doing (if anything). Or you can draw the small signal model and see. Let me just say that is not what you want in your emitter for a 50 ohm input impedance.

 

[Howard]

Dunno if you really care about powering a speaker from a line level source, but you want a higher input impedance and lower output impedance. Most speakers are designed for voltage drive, or high damping factor, which is the ratio between the speaker's nominal impedance to the amp's output impedance. When damping factor drops, strange things happen.

You want the input impedance to be high so as to minimize power transfer.

EDIT: You can get a surprising amount of sage advice from diyAudio's solid state forum.

 

[Modelworks]

50ohm input impedance seems odd for audio, usually it is much higher 10K or so , what frequencies will it be tested at ? the whole audio range or just a narrow set ?

Are you limited in supply voltages ? Split supplies allowed ?


If the input is 50ohm you can adapt it to any circuit easily using a transformer for the input.

 

[Cogman]

50ohm input impedance seems odd for audio, usually it is much higher 10K or so , what frequencies will it be tested at ? the whole audio range or just a narrow set ?

Are you limited in supply voltages ? Split supplies allowed ?


If the input is 50ohm you can adapt it to any circuit easily using a transformer for the input.

Unfortunately, none of this information is available. Here, I'll post the actual assignment so you can see what I'm working with.

Amplifier Lab
Purpose:
1. Design a multi-stage amplifier with a specified power gain.
2. Determine an appropriate amplifier configuration (CB, CE, CC) for each stage.

Procedure:
Design a multi-stage audio amplifier with an overall gain of 30 dB 3 dB. (That is a very wide tolerance!)
Review of decibels: http://www.allaboutcircuits.com/vol_3/chpt_1/5.html
The DC supply voltage, the number of stages, the type of BJT (NPN, PNP), and the amplifier type used for each stage (CB, CE, CC) are all your decision.

Design Advice:
1. Each stage should have a stable power gain (Ap) between 10 dB and 13 dB. This can be accomplished with emitter degeneration (swamping). Keeping the gain of each stage modest will make the amplified signal more linear, have less distortion, and be less affected by temperature variations.
2. The input source is an audio-frequency sine-wave generator (20 Hz to 20 kHz) with
Zout = 50 Ω. Your first stage should have Zin approximately equal to this for maximum power transfer.
3. The load is an 8 Ω speaker. Your final stage should have Zout approximately equal to this for maximum power transfer. A speaker requires power, not just voltage to operate properly.

 

[Howard]

Damn it, why is somebody who doesn't know audio trying to teach audio?

 

[Colt45]

Who says they are teaching audio. I'm assuming they are teaching BJT amplifiers, and naturally you start with AF, not RF. It doesn't matter if it is mellow or has clarity or whatever else audiophool words you want to use. It's to learn the transistors.


I suppose they haven't taught you about diff amps and push-pull output stages yet, eh? Emitter followers for output are painfully inefficient, but if you're just starting that's probably what they'll want you to use.

 

[Howard]

Who says they are teaching audio. I'm assuming they are teaching BJT amplifiers, and naturally you start with AF, not RF. It doesn't matter if it is mellow or has clarity or whatever else audiophool words you want to use. It's to learn the transistors.


I suppose they haven't taught you about diff amps and push-pull output stages yet, eh? Emitter followers for output are painfully inefficient, but if you're just starting that's probably what they'll want you to use.

Doesn't matter... why would they throw in that nonsense about maximum power transfer?

 

[William Gaatjes]

Doesn't matter... why would they throw in that nonsense about maximum power transfer?

Both of you have a point. When reading about impedance matching, it seems that cogman will sooner or later have to design some rf circuit or at least learn about the basics.

But i agree with you that impedance matching should be a separate design challenge. 50 Ohm impedance matching for maximum power transfer is irrelevant when designing an audio amplifier of this kind. Modelworks is right. 10KΩ input impedance at least. 50kΩ is pretty standard though...

It would be better if the teacher would to keep the subjects apart for explanation and then combine both subjects, impedance matching and amplifying in 1 design.

 

[Colt45]

Doesn't matter... why would they throw in that nonsense about maximum power transfer?

Because it's correct?

http://en.wikipedia.org/wiki/Maximum_power_theorem

 

[WhoBeDaPlaya]

Watch the number of stages. Compensation gets tricky with the additional poles and zeros.

 

[Modelworks]

I think the teacher is looking at his function generators output, which is usually 50ohm because of the BNC connectors and coax used. Not really meant to be used as a test source for an audio amplifier without something to match it in between. Some of these can output up to 10 volts , I hope his output isn't planning to go that high, it doesn't say what the range is. Big difference between an amp designed to gain 10db at 1.1V max and one at 5V max.

If you want linear then look at class A amplifiers, it doesn't get anymore linear than that and you can add a stage at a time to make it as high a gain as needed. They are also very simple in the layout.


Two such amps with good explanations on why they are designed the way they are:
http://sound.westhost.com/jll_hood.htm
http://sound.westhost.com/project36.htm

 

[esun]

Because it's correct?

http://en.wikipedia.org/wiki/Maximum_power_theorem

The other poster's point is that maximum power isn't usually the goal in audio, it's accurate representation of the analog voltage meaning maximizing voltage gain is more important. Max power transfer may be important in, say, a portable GPS receiver where the input signal is at -130 dBm.

 

[Howard]

Because it's correct?

http://en.wikipedia.org/wiki/Maximum_power_theorem

So? It's still nonsense for audio.

 

[Born2bwire]

So? It's still nonsense for audio.

All audio is nonsense, but this is proper engineering and that is what they should be teaching. You need to impedance match to not only ensure the maximum power transferred but also signal integrity since impedance mismatches mean reflections. This doesn't happen in audio because they don't really care and lack of standards. Just look at any audiophile cable. You can find a myriad of designs that will have varying impedances but will not even match to the terminations that they use. But for any kind of application in the RF and above it is very necessary to pay attention to impedances.

 

[Cogman]

All audio is nonsense, but this is proper engineering and that is what they should be teaching. You need to impedance match to not only ensure the maximum power transferred but also signal integrity since impedance mismatches mean reflections. This doesn't happen in audio because they don't really care and lack of standards. Just look at any audiophile cable. You can find a myriad of designs that will have varying impedances but will not even match to the terminations that they use. But for any kind of application in the RF and above it is very necessary to pay attention to impedances.

Well, we got to ask the teacher a lot of questions about this project. He basically said. "It should just be in the ball park." In other words, as long as our input impedance was something less than 500 ohms and our output impedance was less than 80 ohms, we would be good. I was going to shoot for exact impedance matching, but I guess I don't have to anymore.

Thanks for the help so far guys, in some ways this is more educational than the class has been

 

[Howard]

All audio is nonsense

So you're saying there's no rhyme or reason to audio engineering?

 

[William Gaatjes]

All audio is nonsense, but this is proper engineering and that is what they should be teaching. You need to impedance match to not only ensure the maximum power transferred but also signal integrity since impedance mismatches mean reflections. This doesn't happen in audio because they don't really care and lack of standards. Just look at any audiophile cable. You can find a myriad of designs that will have varying impedances but will not even match to the terminations that they use. But for any kind of application in the RF and above it is very necessary to pay attention to impedances.

Oh come on. You do know there is a difference between being able to hear something and measuring it ? Besides, in those low ranges of frequency and power it does not matter. The electrons really do not care about impedance in this case. (Actually they do but your hearing does not).
The only thing that is important is the frequency characteristics of the input and those are usually designed properly or at least good enough to not hear it if you do not know what you are looking for. Add the fact that most cd music is surpassing the 0dB limit and there really is no use for it for impedance matching. Add the fact that a speaker setup has more effect then a simple 10kΩ to 50kΩ impedance mismatch.

You, are full of nonsense i am afraid. I would almost think you are the kind of person who buy expensive ofc copper cables from that one brand because in your mind it is better. And someone who paints his cd collection green for better sound.

For those who want to know what i mean with cd music quality and 0dB.

A little bit info is here :

http://en.wikipedia.org/wiki/Loudness_war