Gain

That depends on the frequency of operation, the CE capacitance, the beta, and the Early voltage. The first and last are made much easier if you use a cascode, but that's two transistors already.

Plus your input swing is pretty small--that 20 mV swing will make I_C change by a factor exp(20 mV/26 mV) = 2.2 peak-to-peak at room temperature.

With an output swing of 8 V, the quiescent bias will need to drop about

11 V from a supply of at least 15 or 16 V.

The other issue is bias stability. If you care only about the AC, you can use nV_BE feedback to stabilize the bias. Unfortunately, that puts an even greater premium on high beta, and multiplies the V_BE drift by the same factor: with a voltage gain of 400, your output will drift by

400 * 2.1 mV/K = 0.84 V/K, so you're probably stuck with an emitter degeneration resistor paralleled with some huge capacitor.

(If you care about the DC, you're pretty well screwed with either approach.)

For sufficiently slow signals, something like a 2SD2704K cascoded with a

2N3904 would probably work, if you can figure out the bias issue.

Alternatively you could try an integrated Darlington such as an MPSA14, though that would reduce the transconductance by another factor of 2 because there are two V_BE's in series instead of only one.

How about a nice LM358A?

Cheers

Phil Hobbs

CB

Realistically, one really needs two stages with this sort of spec.

Well.....

beta really has only a minor effect on the actual flatband gain, it's a second order effect.

The Early voltage will only matter if the biasing load is a current source. Its usually swamped by the collector load resistor.

The inherent basic gain, with a collector resistor, is given by

Av = 40 x Vdcr

Where Vdcr is the voltage across the collector load resistor

This is explained here:

If there is an ideal current source, the limit of gain is

Av = Ve/25mv = 40 x Ve

This is a large input signal if not degenerated by an emitter resistor, which will lower the gain.

The notionally distortion of a basic stage is:

%nd distortion = Vi in mv

so, a 20 mv signal is heading to around 100% 2nd harmonic

This is explained here:

One only wants to drive a raw transistor stage with < 5mV

For a diff pair, the even distortion cancels, thus one might get around 50 mV signal handling

Yep. beta is a tad of a problem in setting up a bias.

There are ways around this, but it all adds up to more transistors.

Yep......

Why anyone would use a discrete transistor stage today is indeed a mystery...

-- Kevin Aylward

SuperSpice

You need to get out more. ;)

Cheers

Phil Hobbs

HaHa! That would be too easy, Phil. :) I omitted some key paramters in the original posting as I wanted to just keep it general. The input signal will be 50Hz to 10Khz. I don't know exactly the source impedance but it will be around 600-800 ohms. Distortion is (unusally) not a major consideration here. The first attempt I made from hand calcs was very current frugal (battery power in practice so need to keep it down) and when finished only measured 0.01% THD. But it produced an output of ~1V p-p and wasn't readily adjustable to a more useful level up or down without screwing up the quiescent current draw. Nothing's ever as simple as it seems at first, clearly. :-/

Yeah, it seems like the relationship between re and the Rc||Rl combination dominates in this configuration so just about any small signal BJT will do the job.

Less than 5mV? Really?

Oh, I could use an IC; I have dozens of them lying around, but I just felt like I could use a bit of refamiliarisation with discrete components having last looked into this topid over 20 years ago.

<snip>

To get an 8 V output swing at a gain of 400, your poor transistor's Early voltage would have to be, like, 3200 V, not counting the actual collector resistor.

A cascode might work, apart from the Class-A current drain problem mentioned upthread.

Cheers

Phil Hobbs

Some of the RF people are ashamed to admit they are selling transistors. They label the pins RF IN and RF OUT and GROUND.

Sometimes a transistor is just what you need.

Seriously?? <boggle>

Yup, particularly in front ends.

Cheers

Phil Hobbs

That is possible to find. ZXTN2018F Va = 4600 tho' beta is low, 100@10mA Ref: AoE3, pg. 501

400 is possible but difficult. You can get far more if determined. How about a millioon?

But to get a high enough gain, even with typical values, the collector load would have to be very large--about triple.

1/(1/3200 -1/4600) =~ 10500.

A cascode fixes this very handily.

Cheers

Phil Hobbs

Hi CD, If it were me, I'd grab a National Semiconducor Linear Applications handbook, look at AN-222. While its main topic is using the LM394, the FIGURE 4 is fairly relevant to your project. For the low-noise NPN, the AoE has table with many to choose from. cheers, RS

I will; many thanks.

More details (for those who need it) AN222 may be tricky for some people to find.

1994_National_Linear_Applications_Handbook.pdf page 435 (p. 460th in pdf) my point is, it's a "2" transistor low-noise pre-amplifier. The LM394 could just as well be any other low-noise NPN (or multiple matched units in parallel).

in lieu of LM394 , many choices are given in Horowitz&Hill Art of Electronics, in table of "low noise BJT transistors"

cheers, RS

FETs are better in that role IMHO.

I have several thousands of all sorts of transistors 'in stock' here and have been rummaging through them all today trying to find a matched pair of complimentary BJTs for the eventual power stage. Guess how many pairs I found? None! A few dozen matched pairs, but *all* totally unsuitable for one reason or another. I was totally certain I'd have several to choose from at the end of my hunt but it was not to be. What I *did* come across during the rummage, however, was a decent selection of NOS TDA series chips. I'm getting evil thoughts about cheating now. Sigh... :-/

For what in particular? ;)

There are a very few JFETs that are competitive with or even superior to BJTs in wideband, low-level front ends. (People have often used even noisy JFETs such as J309s in receiver front ends, but below VHF, the atmosphere is so noisy that you don't care very much, and the reduced IMD is a help.)

I do a fair few mostly-discrete front ends for various things. ICs are great at what they do well, and can often be repurposed for things their designers never intended, but even in 2022 there are a lot, a lot of things you can do with discrete front ends that no IC can touch.

The one we've been discussing has an InGaAs pHEMT, two tiny GaN FETs, and three 60-GHz Si BJTs. Good luck getting all that on a chip. (And for Kevin's benefit: no, you couldn't integrate the equivalent function in plain silicon.)

Cheers

Phil Hobbs