# Advice for an AM detector using a BJT

I'm designing an AM detector that uses a 2N3904 transistor in common collector configuration. Pardon me if my design's silly because I'm still quite a novice with using BJT transistors as amplifiers, but after a bit of calculations and a lot more trial and error, this is what I've been able to come up with:

In my simulation, I made the input signal (V1) a 1 MHz sine wave with an amplitude of 15 mV modulated at 1 kHz (imported from a CSV file), as shown in the plot below:

I measured the output signal and while it does seem to have the right frequency of 1 kHz and be pretty free of the carrier frequency of 1 MHz, the signal only has an amplitude of about 0.8 mV. That's less than a tenth of the magnitude of the input signal!

Would anyone be able to recommend any modifications that I should make to my circuit so that the output signal has a larger magnitude? (While still being "smooth" and having the right frequency, of course.)

• Are you only concerned with 1kHz? or perhaps audio range up to 20kHz? R2C2 product is chosen to reproduce the highest frequency expected, while suppressing the 1MHz carrier. Your R2*C2 product is too high, even for 1 kHz. However, it does suppress the 1MHz carrier very well. May 13, 2023 at 15:15
• @glen_geek Yes all the audio frequencies (around 20 Hz - 20 kHz) should be reproduced May 13, 2023 at 15:20
• You have the transistor biased for a collector current of about $2 \mu \mathrm A$ -- the 2N3904 isn't really designed for collector currents below 1mA. It'll go down that far, but the gain and bandwidth will suffer. Why the low current? Are you trying for a super-high impedance input? May 13, 2023 at 15:23
• "1 kHz sine wave with an amplitude of 15 mV modulated at 1 MHz". It's the other way around: 1 MHz sine wave with an amplitude of 15 mV modulated at 1 kHz. May 13, 2023 at 16:37
• @Davide Andrea Thanks. I stand corrected. May 14, 2023 at 0:48

## 2 Answers

Would anyone be able to recommend any modifications that I should make to my circuit so that the output signal has a larger magnitude? (While still being "smooth" and having the right frequency, of course.)

Expectations are high: for such tiny AM source signal (15mV), a silicon transistor junction is not an efficient detector no matter how you bias it. It is operating in its square-law region for such small signals...if signal amplitude is cut by half, the detector's output reduces to one-quarter.

For more efficient AM detection that is also more linear, first boost the amplitude of the signal presented to the detector. Shown below is a demo AM detector simulation as input signal levels are varied. Perhaps not optimized, but it demonstrates how square-law hurts you...

 AM 1MHz in    detector_out

7.5mV        2mV
15mV         7.8mV
30mV         26.6mV
60mV         75.6mV
120mV        185mV


Doubling input from 7.5mV to 15mV causes output to nearly quadruple from 2mV to 7.8mV. This is square-law in action.
At the higher end, doubling input from 60mV to 120mV causes output to increase by a factor of 2.45. This is somewhere between a square-law detector and a linear detector...a linear detector would increase by a factor of two.

TLDR: A proper AM detector requires large input amplitude. One also must consider what load impedance must be driven at detector output: ideally, the resistance that the detector drives is high.

• Thanks. I'll try putting a common emitter amplifier before the detector to bump up the input signal to an amplitude of about 100 mV. May 18, 2023 at 13:11

Try biasing Q1 near to cutoff .An easy way on the sim is to greatly reduce R3 like starting with 68K and optimise this .In real life I use a diode connected BJT to give better thermal stability.R2 could also come down or C2 could come down if slope distortion is an issue.

• So you would personally use a diode connected NPN transistor in place of R3? May 14, 2023 at 14:03
• And sorry, what's "slope distortion"? May 14, 2023 at 14:04