A Not Bad test result from active mixer using BFR93 transistor

After get rid of the parasitic oscillation, test can be resume. A 27MHz -20dbm and a 28MHz -20dbm signal is feed to the RF and LO port of the mixer, resulting an output of 1MHz low side output signal, the amplitude is about -20dbm, which is 10dbm larger than using PN2222 transistor. The output is quite clean, not much spurious product. The two strong peak on the right side are the RF and LO signal, and IF output on the left side.

A clean looking downconversion using BFR93 transistor mixer.

Get ride of 1.2GHz parasitic oscillation

Don't exactly understand why, finally I could get rid of the 1.2GHz parasitic oscillation from the active mixer.

It is done by adding a 3.3pF capacitor across the collector loading resistor, as shown on screen, there are no high frequency spike above the mixer product signal frequency.

Output spectrum from low frequency to 1.8GHz

Parasitic oscillation using high Ft transistor

The active mixer does not work well above a few MHz of RF/LO, the output level can drops from -14dbm at 5MHz to -30dbm at 27MHz, so I replaced all transistor, these PN2222 general purpose transistor, to RF transistor BFR93A (Ft = 5GHz), surprisingly, even the inputs are all terminated in 50 ohm, there are a ~1.2GHz -20dbm parasitic oscillation at the output! If I could not get rid of this oscillation I won't be able to continue with any other test! I tried to added some smaller value supply decoupling capacitor but it doesn't help. The problem may come from the board layout. When I put finger on the board, it's not easy to disturb the oscillation, it just shift down the frequency a bit when finger tap on the upper side transistors. May be I should change to a lower Ft transistor.

Discrete RF circuit, is it useless?

In this modern time , most of the radio goes system-on-chip, i wonder what is the value of doing discrete RF circuit like using a transistor to build an LNA or mixer, is there any application of doing RF using discrete compoent ?

Gilbert cell active mixer fix

Discovered a mistake in the active mixer board, one of the inverting input was not AC grounded! It was left opened. That should be a major cause of the weird result of last test.  By shunting a 10nF capacitor across the inverting input to ground, the mixer works much better. As the screen showing a 1MHz tone mixed with a 1.1MHz tone, the 100KHz product signal is clearly seen and it's much stronger and spurious products is reduced, compare to the result from last test.

RF:1MHz -20dbm

LO:1.1MHz -20dbm

Troubleshooting the active mixer by circuit simulator

Trying to find out the cause of the poor performance of the active mixer, I switched from soldering desk to circuit simulator and it looks like a little good hints is found after some tweaking with the resistors values. Assuming all transistors are identical, if the base current of the lower side transistor and that of the upper side transistor are having a ratio of 2:1, the output would make a lower distortion and higher output amplitude. However, how to verify the base current of the transistors on the actual mixer board? I don't know. May be I should apply the bias resistor value from simulator to the actual board.

Gilbert cell active mixer test

Spent 2 hours to solder the Gilbert cell active mixer board, bias current is selected in 1.8mA at each of the output  loading resistors, and -20dbm for both RF and LO input, the result was disappointing. There are so many spurious signal product. The spurious signal product would be fewer for -30dbm input, but obvious the dynamic range is not enough.

-30dbm input power for RF and LO

-20dbm input power for RF and LO

RF/LO on the left side and output on the right side