Very Simple Transistor Amplifier Design Question

[SonsOfThunder]

A speaker is inductive...Not counting the low-frequency resonant "hump"

Quite an over-simplification, but for Ian's purposes, good enough. FYI, the resonant hump is the point where the speaker is going to be most efficient, acoustically speaking. :grin:

maybe an R-C conjugate load across the speaker

Commonly called a "Zobel filter", though I have struggled for years to determine whether or not "Mr. Otto Julias" actually had anything to do with this.

You probably wouldn't believe me

Yes, I would. Been there, done that, passed on the T-shirt. :thumb: Sometimes I'd like to go back to college, just to be able to have fun with the prof's mnd on stuff like this. Also, I'd like to have the opportunity to "remind" a few profs exactly who is working for whom. (I think I used that correctly) Too many profs bully their students about, when the student really should ask for more of the prof. At least that was the case where I went. I actually had a prof admit to me (10 yrs later) that they were under explicit instruction to flunk a certain percentage of students. Mind you, that this was senior level EE classes.

 

[PRR]

> our hi-freq (4k) gain is higher, something like 2.2-2.3, while everything below that (400~1k) has a gain of 2.0.

1.2dB (+/-0.6dB) variation over a wide band is nothing to sneeze at for a First Lab project.

However it is only 400-4,000Hz, and you "should" be able to get a lot flatter.

Are you now testing on a resistor or a speaker? Will the Prof test be on speaker or resistor?

The rise from 1KHz to 4KHz very well could be speaker inductance. (But I would expect another rise in the 70Hz to 400Hz range depending on cone and box stiffness.)

> we made Ce waaay big (~2000uF)

2,200uFd at 400Hz is around 0.2Ω reactive, right? And you hope for around 4Ω in the emitter to give gain of 2 into 8Ω. Hmmm. That smells like a 5% error and should decrease above 400, while you see 15% error flat 400Hz-1,000Hz?

DO you understand that the 100uFd input cap reflects into the emitter as about 100uFd*β? But that should give an exceedingly small error.

Note that the design spec calls for a Voltage gain, and you have made a Current source. If the load were a resistor, gain is fixed. If the load is whatever lumpy-Z speaker the prof brings in, you fail.

It is no big secret that most loudspeaker amps are voltage-source, not current-source. That is, they present a near-constant output voltage for any load impedance (within reason). And that means Voltage Feedback. You need to sense the voltage at the load and force it to be the desired value (2*Vin).

It is actually difficult, in real-life, to do good voltage feedback and voltage gain in a 1-transistor amp. But the prof's specs are so generous that it can be done. The obvious thing is to get the absolute maximum voltage gain you can, then tie a resistor from the output back to the input, and refine. You also need DC bias control, but when asked for 0.75V p-p from a 5V rail, you can be very wasteful.

Transistor must be a big 500mA part like 2N4401(?). Why? Or can a skiiny part work OK?

Caps are semi-hi-fi, few-dB down around 20Hz. Proper values for 10% error at 400Hhz is left to the student.

Estimate (with your lab-chum) the bias current, AC gain, input Z, output power.

All of these are somewhat β dependent. You should first rough it out, use that rough answer to refine, then check for extreme β. Figure out what things make big differences, and which are minor corrections. There is probably an equation to write, but if you know your poop then you can eyeball it faster than deriving the right equation.

Why did I use nearly the maximum current allowed and a $0.20 transistor, instead of the "rational" plan of reducing energy consumption and a $0.15 transistor? Maybe I'm lazy? Energy-hog? Transistor salesman?

How can a 120Ω input resistor (maybe) work with a 150Ω input spec and a "virtual earth summing node" (or is it?)? Or am I leaning on what -I- know about portable headphone players?

What happens if the prof throws you curve-balls? A 2-way with severe impedance rise/dip at crossover? A 16Ω speaker? A 50Ω speaker? A choke in parallel with a 50Ω speaker? A 1,000uFd cap and a speaker with one leg permanently tied to the lab bench ground? (Early car speakers had grounded frames to save one wire.) Same grounded-frame speaker with no cap?

I believe the assignment calls for a volume control, and suggests an input pot. What is a good value? (We may be in trouble here.) Is there another way to reduce gain?

A question you must not think about: how loud is 0.0088 watts in a typical speaker? Is this thing even worth building??? Is it possible to develop a "better product" within the general assignment?

 

[Ian MacGregor]

Are you now testing on a resistor or a speaker? Will the Prof test be on speaker or resistor?

We are testing with a speaker. I am quite sure that the professor does not supply a speaker, we are tested with the one we have sourced.

DO you understand that the 100uFd input cap reflects into the emitter as about 100uFd*β?

This I am a little confused about. Doesnt the input cap set our low frequency rollof point by making a HPF with the Zin of the amplifier?

And that means Voltage Feedback.

Ahh... interesting. This seems like the best way to do it. The wierd thing is that we haven't been taught anything about feedack in BJT circuits at all. Might give the prof a little suprise if I pull that out!! haha...

I kind of have the feeling that this project was assigned the way it is in order to make us bang our heads against the wall a little. This class is the junior EE's first REAL design class of any kind. I think they are trying to make us realize how much thought/troubleshooting goes into design and that what makes sense on paper does not necessarily work in real life.

Again, thanks for all the help. I feel like I've learned a lot more than I expected from this project due to the help from the group. Our project checkout is tomorrow evening. (I wish I had posted here a hell of a lot sooner!!) I'll let you guys know how it goes.

Ian

 

[PRR]

> make us bang our heads against the wall a little

I'm sure many students will "fail", at least on some detail of the spec. I think it is a goal; the real lesson is in the journey even if you don't get to the 19th hole.

Of course that means you have to travel the path, and also document where you've been, for your clarity and so the Prof sees the work.

It is certainly possible to meet and exceed the spec even with the unstated 1-transistor limit. But figuring out why it almost meets spec is good learning.

I'm harping on cap values because the Prof gave you the hint: "the value of DC blocking capacitors". You Were Warned! Blocking, bypass, it's all how you look at it. Caps are made in a 1,000,000,000:1 range (few-F to few-pF), for some reason, so you can't just pick "4.7uFd" out of thin air and assume it blocks or bypasses.

The spec does not say one-transistor. Me, I'd spend Labs 1-10 on single transistor, then 11-20 on 2-transistor Compound Pairs, the workhorses and building blocks of discrete design. But peeking ahead, I see that Lab 2 has a tighter spec that in past years was solved with a 8-transistor push-pull design (actually a fairly deluxe small hi-fi amp). So he's rushing you right along. Too late now, but a 3 or 4 transistor design for Lab 1 would cream the competition for loudness and efficiency without breaking any stated rule. And not really be harder to design. And be more like real-world solutions. (Thus you could do more plagiarXXXX..., er, Research into how similar problems have been solved before.)

> we haven't been taught anything about feedack in BJT circuits at all

"feedack in BJT circuits"? So you were taught feedback somewhere else? Dynamos and Steam Engines 101? The idea is the same, just the place to stuff the wires changes. If you understand the gear, the dynamo feedback can be applied to a BJT amp. If you don't make leaps like that, you will be playing catch-up your whole career.

Interesting that you have not been taught voltage feedback, but Lab 2 will expect you to prove a thermal stability criteria (I can't do that rigorously). It would be insane to meet that spec without global feedback, "simple" with feedback and common parts. (Wonder what the limits are for "zero volts on the output".) Though considering it only asks 6V p-p on +/-5V rails, 0.265W, and suggest push-pull Class AB, I wonder why he gives information for mica washers. You could solve this with TO92 transistors and quite sloppy thermal coupling. Of course it really begs for LM375(?), but that would be letting National do your homework for you.

 

[PRR]

> the professor does not supply a speaker, we are tested with the one we have sourced.

Ah! Loophole! Knock the magnet off the speaker, fill the voice coil with pennies. You will get a non-resonant low-inductance resistance that will probably be +/-5% across 400-4KHz or DC-10KHz. It won't make a sound, but sound was not in the spec! You have Solved the assignment! If they wanted groovy noises, it should have been in the spec!

This IS very much part of the engineer's training. Know what is and isn't specified. If you have seniority in a solid company, let them write the specs right and you just do what they say. In the real-world, you need to object to bad specs. No consumer cares how many volts comes out! You want an SPL number, that might be met with a lame speaker and big amp, or hot speaker and micro-amp, depending on price and battery life and size (all of which need to be sketched in the spec). Of course in a big shop, the Senior Engineer knows that 1 watt in a 10" or 10 watts in a 3" make about the same racket, and knows the size and costs of each, so you the Junior Kid will just get an electrical spec. It would be unproductive for you to question the total acoustic package, not your job nor your specialty. Unless you are sure the boss is headed for disaster that will bring you down too.

If you have time, dummy-up a speaker so it IS 8 ohms flat. Glue in the gap will kill the resonance, though you really need to get the iron out to kill the inductance that is causing your top-rise. Even air-core has inductance: a shorted turn (pennies inside the coil) will kill most of that. (Quicker: put a $2 speaker on a 100W guitar amp, whang a few chords, then solder a pure 8 ohm resistor across the now-open voice coil's terminals.)

Of course when someone objects, you will have to bring out a working speaker. And the overnight cure for the rising top is the time-honored C or R-C across the speaker, trimmed to taste or spec. (Look in any good kitchen tube radio: 0.005uFd across the output plate winding to tame the top-rise. In AM radio, we want an eventual fall-off so just a cap. If your inductance is rising at 1KHz, a cap to catch it by 2KHz may be dropping at 4KHz. You may need a series resistor around 10 ohms if you have to work several octaves into the inductive region.)

 

[NewYorkDave]

It won't make a sound, but sound was not in the spec! You have Solved the assignment! If they wanted groovy noises, it should have been in the spec!

You've worked as a government contractor, I take it? :wink:

the overnight cure for the rising top is the time-honored C or R-C across the speaker, trimmed to taste or spec.

Yep, the "conjugate load", as I call it, or "Zobel" as Thunder calls it. RCA likes to call it a "corrective filter":

http://www.kbapps.com/audio/tubemanual/images/030.gif
(last two paragraphs)

 

[SonsOfThunder]

[quote author="PRR"]Knock the magnet off the speaker, fill the voice coil with pennies[/quote]I laughed until I cried on this one. For the first time, I have to say, I would not recommend following PRR's suggestion, though this is the sort of thing I would like to do, were I to go back to school.

In the real-world, you need to object to bad specs.

You gotta be careful on how you go about this though. Mess this up and you can lose business for the company you are working for. I am dealing now with a situation where there are tests that are supposed to be performed on a product and data gathered...but no performance goals or limits...! The guy who wrote the spec (at one of the big 3) can't tell me what I am supposed to be looking for at all. He apparently just inherited this as a "legacy" test. What's more is that "they" don't even have a test instrument to verify or deny that we meet the spec. Kind of on the honor system here, I guess. Lots of funny stories that I'll be able to tell in about ten years.

Peace!
Charlie

 

[Guest]

[quote author="NewYorkDave"]

[/quote]
If you ommit that automatic bias {emmitter R//C}, you
can get bias partially from secondary {via simple diode},
and you will have A-class amp with dynamic bias.

I remember, that on big AM transmitter there is
controlled carrier via modulation. In 2 MW transmitter it
can do many spare.
But in electric power station are not so glad

xvlk

 

[Ian MacGregor]

Wel... we had our lab checkout yesterday. We had to hook the circuit up to a scope/gen and verify the requirements with a TA standing by. Our circuit passed every requirement! We were pretty stoked. We used the same circuit diagram that I posted earlier, but the values were different:

Cin = 100uF
Cee = 3300uF
Rb1 = 500
Rb2 = 1k
RC = 8 ohm speaker
ReAC = 3.3
ReDC = 51

Thanks again for everyone's help, my partner and I learned a lot. We've got project number 2 coming up, I might post about that one as well. I see PRR has got a head start on us... :grin:

Ian