28 - Diode Ring Mixers:

 Diode Ring Mixers:

One of the four building blocks in making a transceiver is a mixer. (the other three are filters, amplifiers and oscillators). Mixers have two inputs and one output. Usually the output contains the addition of the two input frequencies and the subtraction of the two frequencies. I.e if you apply f1 and f2 to a mixer you get (f1+f2) and (f1-f2) at the output. (or (f2-f1) if you prefer - if the subtraction gives a negative frequency you just pretend it is positive)

The uBitx Diode Double Balanced Mixer (DDBM)

The design of the uBitx requires three mixers, Two and a half really as the third is half a DDBM. They are each used twice, once for receive and once for transmit. In receive a low level signal of 3 to 30MHz is mixed with a Local Oscillator(LO) that can vary from 48 to 75MHz. For example an input of 14MHz is mixed with a LO of 59MHz. The two largest signals coming out of the mixer are the sum and subtraction of these two frequencies, i.e 45MHz or 73MHz. and lots of other frequencies, albeit at a lower level. The output of the mixer is passed to a 45MHz filter with a passband of 15kHz so most of the spurious signals are blocked from proceeding further in the signal chain. The other DDBM has an amplified 45MHz signal (+/ 7.5kHz) which is mixed with a second local oscillator of either 33 or 57MHz depending on whether you are receiving USB or LSB, the frequency difference frequency is 11.059MHz - again the other frequencies are mostly blocked by a 11.059 Crystal filter, this time it has a bandwidth of 2.4KHz and this is the desired SSB signal. A further amplifier passes this SSB signal to a third mixer, this time not a full DDBM and a third local oscillator, conventionally known as a BFO is mixed to convert the SSB signal into audio. This is a dual conversion superhet.

So how does a ring mixer work? The dots show the start of a winding.

Mathematically multiplying two sine waves together gives two new sine waves at the sum and difference of the input frequencies. Again, exactly what we want in a mixer.

Cos(f1) x Cos(f2) = 1/2 Cos(f1+f2) + 1/2 Cos(f1-f2) more or less...

Note this is simplified; the correct mathematical expression includes t and traditionally uses angular frequencies - in Radians/second rather than Hertz.

Below is a photograph of my diode double balanced mixer. I used 7 or 8 turns of trifilar would enamelled wire on a FT37-43 as per the uBitx instructions, I selected 4 diodes from a packet of ten that were within a millivolt or two on the diode test function of my multimeter. The transformers winders could be neater, I will make another and compare the test results.

Testing: Let us perform two tests with the mixer,

Test 1: We will apply a single tone - a frequency of 13.035MHz (one of my home made oscillators can generate this at just over 0dBm i.e 1mW or 224 mV into 50Ohms) and apply a strong 10MHz local oscillator voltage of 1.4V initially (this is 16dBm). The spectrum analyser (really my SDRPlay rsp1a sdr) will allow the following measurements.

What can we expect with two frequencies f1 and f2 entering a mixer?

In came as a shock to me when I actually tried this, there were quite a few more frequencies being generated. I showed last month how even just two frequencies can easily generate over 60 different combinations of all the harmonics, below is a real spectrum plot. At least the highest peaks are the desired f1+f2 and f1-f2.

It would be tedious to measure the height and reason for every peak. ~There is a facility in the spectrum analyser software to save a CSV file of the screen (or the entire FFT) and when I imported this into Excel, I was able to sort on the strength of each peak. There is a slight frequency mismatch if using the screen data although trawling through the FFT data showed the correct frequencies. I did not exhaustively search the FFT data as it is so large. (over a million lines long) – it would have been exhausting! I present a summary of the screen data below; I left the SA running in “average” mode which reduced the noise floor to -100dB and kept the peaks correctly, with some frequency drift of the test oscillator (tens of hertz)

The desired frequency of f1+f2 (23.035) shows a amplitude of -4.6 dB and since the incoming RF is at +2.8 dBm this gives an insertion loss of only 7.4dB which is ok - figures of 6 to 8dB usually quoted

It is interesting to note that the 3rd and 5th harmonics of my test oscillator seem to feature in the IMD products, perhaps I need a cleaner oscillator!

Whilst not shown above the response to the 10Mhz signal generator coming out the IF port is way down at -49.7dB, the 13.035MHz signal does come out the IF at only –33 to 35dB or -37dBc below the input. So one input is balanced out and the other one is nearly balanced out (-37.8dB is about 3mV of leakage. I will build another mixer and test it, I might just need some shielding. But as it stands it is fine for a receiver mixer, might need to be careful when used as a transmit mixer to make sure the carrier gets suppressed properly. The advantage of having my £92 Spectrum Analyser is that I can measure it easily.

So, the mixer has passed Test 1, insertion loss and carrier suppression is ok.

Test 2 is to see what the strong signal handling capability is like. Usually done by looking at a two tone response and using the third order intercepts, either referred to the input (IIP3) or output (OIP3). These differ by the loss, which we now know is 7.4dB.

We need three oscillators to measure IMD, a two tone oscillator with the two tones close to each other, mine are 13.0352 and 13.049, 14kHz apart, these will be applied to the RF port of the mixer, the LO is connected to a third signal generator. This time I will set it 59MHz as this is closer to what will happen in the uBitx.

The mixer should produce two new frequencies (if we concentrate on the f1-f2 terms)

i.e (59-13.035) and (59-13.049) these are 45.965 and 45.951MHz respectively

However, the third order product of (2f1-f2) and (2f1-f2) will also be present.

i.e ( 59 – (2*13.035 – 13.049) which is 45.979MHz

and (59 – (2*13.049 – 13.035) which is 45.937 MHz

The Intermodulation Distortion is defined as the distance below the main tone, and we get the intercept figure by adding half this to the power of the tone, either at the input or the output to get IIP3 or OIP3. The first time I did this the readings were low – I was overloading the mixer, I added a 20 dB attenuator and got better (expected) results.

(actually I think the IM3 is actually the height of the green bar - a mistake in the application note?)

I zoomed in and took readings of the power of one of the tones = -14.3dBm and the power of the 3rd IMD product = - 57.7 dBm. The difference between these (as a positive number) is  +43.4, half of this is 21.7, added to -14.3 gives a value of +7.4 as the OIP3. Coincidently the same number as my insertion loss.

The IIP3 is +7.4 + the insertion loss, measured as 7.4 = +14.8 dBm. My signal generator was outputting about +7 dBm. A rule of thumb is that the IIP3 should be 9dB more than the LO power so it looks like the mixer is working ok. (7+9=16!)

In summary, we want mixers with;

1. Low insertion loss (i.e a loss of 7.4 dB, most will be between 6 and 8)

2. Low LO breakthrough – mine is over 57dB below the input

3. Low RF breakthrough – mine is 33 to 37 dB below the input, need to watch this when using it to generate SSB, it affects carrier suppression, although the filter will help reduce it further

4. Good signal handling capability, whilst an IIP3 of nearly +15 is not great it is ok, a higher LO power would help it a bit, but I will use it as is in my uBitx, and see what LO power is best. I can improve it later.

One more of the building blocks done, next month I will cover the three transistor amplifier, 4 of them form the heart of the uBitx.