0
\$\begingroup\$

Below shows the phase shifting network I designed for a FM detector. The goal is to create a 90° phase shift between the input and OUT1 at the centre frequency 10.7MHz. This circuit drives the bottom differential pair of a Gilbert cell mixer to eventually demodulate the signal. OUT1 connects to one of the transistors of a differential amplifier while OUT2 goes to the other transistor.

Schematic

How the circuit was designed

R8, R9 and R10 is a voltage divider with voltages of 3.3V, 2.75V, and 1.1V. It is used to bias the transistors. L1 allows both transistors to have the same bias level but since it forms a high-pass filter with C3 I chose 39uH to not attenuate the input signal. C3 was chosen to be small to not interfere with the resonant network formed by C2 and L2 which give the 90° phase shift at 10.7MHz. R13 is there to decrease the Q of the circuit. C6 and C7 are there to block DC. C9, C8 and C5 are needed to create a stiff voltage for biasing.

How does the circuit perform

Below shows the magnitude and phase of OUT1 simulated in LTspice. At 10.7MHz, it sits at 0dB with a phase shift of -270° (which is equivalent to 90° phase shift). When I built this circuit on a prototype PCB, the input signal was at around 1.2 Vpp, but the signal at OUT1 was at 200 mVpp. I can't figure out why it is so attenuated. I'd really appreciate getting suggestions on how to improve this circuit.

Simulation

\$\endgroup\$
6
  • 1
    \$\begingroup\$ Did you use an oscilloscope? And a 10X probe, for example? \$\endgroup\$ Commented May 24 at 9:53
  • \$\begingroup\$ Yes, but I used a 1X probe. If it would help, I can attach photos of the measurements. \$\endgroup\$ Commented May 24 at 10:00
  • \$\begingroup\$ You forget then ... adding a 1Meg // ~30 pF ... at ouput. \$\endgroup\$ Commented May 24 at 10:02
  • \$\begingroup\$ It's a very rare situation where a 1X probe is actually useful. 10X probes load the circuit under test much less. You also can't get a bandwidth on a 1X probe much more than maybe 10 MHz, while 10X probes go up to 500 MHz easily and a few manufacturers offer higher-bandwidth ones. \$\endgroup\$ Commented May 24 at 16:17
  • \$\begingroup\$ Why so many unnecessary components? Replace L1 with C2 || L2 || R13. The resistive divider + bypass could be simplified too. \$\endgroup\$ Commented May 24 at 17:21

2 Answers 2

Reset to default
2
\$\begingroup\$

As @periblepsis allready pointed it, you forgot the probe "capacitance" which "shift" the resonance peak ...

Here is a simulation with microcap v12 which explains that.

enter image description here

Red curve is for cc = 5 pF, other's are for 15, 25 and 35 pF.

If you make a measurement at 10.7 MHz, you see that the voltage shuts down from 901 mV to 278 mV.

\$\endgroup\$
2
  • \$\begingroup\$ I tried using a 10X probe and I measured the input signal at 2 Vpp and OUT1 at 600 mVpp (so less attenuation). Just want to check my understanding, so what you and @periblepsis is saying is because of the capacitance of the probe and input to the oscilloscope, it will shift the resonant peak of the circuit which will cause me to see an attenuated signal on my scope? Those this mean that the circuit is performing as expected (like the simulation)? \$\endgroup\$ Commented May 24 at 10:54
  • 1
    \$\begingroup\$ Yes. Probing ... shift the resonance peak, so measuring a lower voltage. The circuit is performing like the simulation (when all "devices" as a probe is wired). This can be "corrected" when a low capacitance probe (just some pF) is used. \$\endgroup\$ Commented May 24 at 11:19
0
\$\begingroup\$

The phase shift network can be simplified for feeding a multiplying-type discriminator like a Gilbert cell:

schematic

simulate this circuit – Schematic created using CircuitLab

For the F.M. broadcast band, the 10.7MHz resonator of L1, C2 might be chosen with a loaded Q of about 66 (or less) for +/- 75kHz deviation. The resonator loading isn't shown in the crude block above...one might roughly assume that the multiplier has high-impedance inputs, and that the voltage source V1 has some low-impedance output. V1 need not be sinusoidal - the I.F. output from limiter stages are often square(ish).

The tradeoff here is that high-Q gives more audio amplitude, but with higher distortion. With a Q of 66, distortion for +/- 75 kHz deviation is in the ballpark of 3%. A more complex double-tuned coupled network can give lower distortion, but this simple network might be used for non-audiophile applications.

Many old FM I.F. chips (MC1357, ULN2111) have 10 MHz I.F. amplifier, along with a Gilbert cell multiplier. Some follow with an audio buffer/amplifier.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.