In an RF context, a mixer is a device which “mixes” two frequencies, producing the sum and difference of said frequencies at its output. In the time domain it works as a multiplier, multiplying both input signals.
There are several circuits capable of behaving as mixers, this article will cover the design and behavior of a mixer using a Gilbert Cell encapsulated in a MC1496 IC. This post assumes previous knowledge of the MC1496 and its applications. To familiarize yourself with the chip you can read this post about it.
This handmade board is the first step to creating a proper evaluation board for this chip. I don’t trust myself so ordering a batch of PC boards with no prototype seemed dangerous. It is also a great excuse to play around with SMD components (this was my first time soldering them).
Building the board
Board layout is very important when designing high speed or high frequency circuits. Some basic guidelines are:
- Traces need to be kept as short as possible to minimize the effects associated with transmission lines.
- Component leads should be kept short, or else they will act as inductors. Their parasitic properties need to be taken into account above certain speeds/frequencies. Tolerances also play an important role.
- Signals need a plane to couple to for effective transmission and good signal integrity, this usually involves a ground plane below the signal traces.
Therefore, I decided to use a one-sided copper board and copper tape. The copper in the board will act as the ground plane, and the copper tape on top will ideally behave as a microstrip. The board has a 1.6 mm thickness, which means a 3.2 mm wide strip of copper tape should behave as a 50 ohm transmission line (approx.).
The KiCad design files this board is based on can be found on this repository. Basically most if not of it consists on the DC biasing network for the Gilbert Cell. The target voltages are:
Pin (s) | Function | Voltage (V) |
---|---|---|
5 | Bias | 1.25 |
1, 4 | Signal | 5 |
8, 10 | Carrier | 8.2 |
6, 12 | Output | 11.1 |
Said voltages are obtained with voltage divider networks, any valid resistor combination will do, however the data sheet recommends the networks to sink 1mA.
I used 0603 SMD components, a DC barrel jack, and a potentiometer to make the DC bias of one of the signal pins variables. For the few ground connections needed, I made poor man vias by drilling a hole and sticking leads cut from THT components.
Images
Here are some images which document the build process, a schematic and components values can be visually obtained from these pictures. All capacitors were 10nF.
Results
The test signals were generated with a GW Instek AFG-2012 function generator and the oscilloscope used for the measurements was a Tektronix TBS 1072B. The function generator had a 50 Ohm output impedance and the output was measured at a 50 Ohm SMA dummy load. No impedance matching was made, so results should be horrendous.
DC Bias Point
Pin (s) | Function | Voltage (V) |
---|---|---|
NA | Vin | 15.13 |
1 | Signal | 5.08 |
2, 3 | Gain | 4.37 |
4 | Signal | 5.08 |
5 | Bias | 1.194 |
6 | Output | 11.5 |
8 | Carrier | 8.15 |
10 | Carrier | 8.15 |
12 | Output | 11.44 |
Mixing two signals
Conclusions
The circuit worked at some degree, which is already good. The DC biasing network worked almost perfectly, with the biggest discrepancies being found at the output. Both outputs were about 400mV higher than expected.
This means that the biasing current was not 1mA. According to the calculations it was 930.76uA, an error of around 6.923%. Considering that no biasing rules were broken, and that part of it depends on the temperature and characteristics of the internal diode, it seems good enough.
The output however is definitely horrendous, as expected. The mixer circuit presented in the MC1496’s application note is stated to have a 13dB conversion gain, in this case the mixer had a negative conversion gain, since the output was smaller than the RF input.
This is probably caused by the lack of an impedance matching network. Ideally this mixer will be used for IF down conversion, bringing a 153 MHz signal down to 21.4 or 48 MHz (to be determined), that’ll be the real test for this circuit, and a matching network will definitely be used there. In this case it did not make much sense to make a matching network for a quick test.