This is a continuation of an earlier post, which can be found here.
Work has been continuing on my CW rig design, and now I will share another subsystem of the rig. I will get to the audio, but in this post I will be looking at the mixing function.
Many rig designs use a SA612 mixer oscillator IC. This device is a double balanced mixer and oscillator. It can actually be used with either balanced or unbalanced RF inputs, and a balanced or unbalanced output can be taken. Regarding the oscillator, essentially only the tank circuit is external. The SA612 can also be driven by an external LO input.
Why not use it? Good question. My early concepts made use of it, but the limitation of the SA612 is its dynamic range. The third order Intercept (IP3) is at about -13dBm with a -45dBm signal. The 1dB compression point is -20dB, but given the IP3 at -13dB, a signal at this level of power would be full of inter-modulation distortion. I wanted something that would give much better inter-modulation distortion performance. In looking around, I found nothing that was quite like the SA612, which is why so many designs make use of it.
You’ll be able to see what I have done in the circuit below. This is a screen shot, with the RX signal coming in from the top from the TX blocking circuit discussed in previous posts. Note you can click on the picture for a zoomed in view:
The minicircuits ADE-1L
Minicircuits have a large variety of mixers available. Most service microwave requirements, but some work down to HF. I had a good look at a number, but the ADE-1L took my interest. This device has as 1db compression at 0dBm, which is 20dB better than the SA612, along with a IP3 at 17dBm, 30dB better than the SA612. Dynamic range is therefore effectively 20dB better, and inter-modulation distortion performance should be 30dB better. Using these devices should make for a much better rig. Unfortunatley, there will need to be more work to use one (or two) of these than SA612 ICs. Lets get into dealing with the issues.
50 ohm input and output
The ADE-1L is designed for 50 ohms. The SA612 has a high input impedance, 2k is ok to feed it, while it has an output impedance of 1700 ohms. 50 ohms sounds better, but it is harder to use. What is coming in might be expected to be 50 ohms at the antenna, but by the time it gets through the TX blocking circuit, it is around 10 ohms. Essentially both devices need matching circuits – which is band specific. I discussed these networks more on a previous post.
The first mixer
The the first ADE-1L mixer, M1, takes the RF input and a LO. The RF is 7 to to 7.3MHz or 14 to 14.3MHz depending on the band. If 7 is mixed with 11, 4 is output. If 14 is mixed with 10, then 4 is output. This LO needs to change with tuning, needs to be around 3dBm, and these things can be done by the AD9834. I will look at this device more when I look at the microprocessor control, but it can be used to generate the carrier at 7 or 14 (and up the band) during TX, and generate the around 10 or 11 MHz during RX for mixing.
With these two signals mixed, I have output at 4MHz. I also have an image frequency at 18MHz for 40 metres, and 24MHz for 20 metres. There might be other products coming out of the mixer as well, but these will be well down on the 4MHz and the image frequency. I should not have much RF and LO coming out, given the isolation performance of the mixer.
All I have done with the RX coming in so far is to impedance match it to the output finals (which are like leaky closed switches during RX) pass through the TX block and then match it to 50 ohms for the mixer. The mixer loses 5.5dB mixing, and there would be expected to be 1 or 2 dB across all of the matching networks. There is a little over 1dB across the TX block. All of this is not such a big deal because we are dealing with 20 and 40 metres, and we don’t need to worry about noise performance so far. The SA612 actually amplifies the output as well as mixing it. There is around 5dB of loss in mixing, and then 22dB of gain, leaving a net 15dB gain with over 5dB noise figure (because of the mixer).
I have put a 4401 device to do this amplification, because the ADE-1L is only a mixer. The 50ohm output of the mixer is impedance matched to 700 ohms. This allows a lower current biasing network on the transistor. A BJT is used to keep things linear as they do this job better than MOS devices, especially at these low current levels. The 4401 is designed to give 23dB gain. I use it to do a small impedance transformation, back down to 400 ohms, a suitable level for the crystal filter about to come.
The crystal filter
I have an intermediate frequency of 4MHz with this design. Now using a series of 4MHz crystals, I can have a narrow pass band filter. The great thing about crystals is that they can be used with some series and shunt capacitors to give various types of high performance filters, such as Butterworth or Tchebycheff filters. Butterworth are a little lower performance, but have no ripple. I have designed this for a small amount of ripple (1dB) with steeper skirts, making it a Tchebycheff. The 330pF and 390pF capacitor radio decides the ripple. The amount of capacitance overall decides the pass bandwidth. I have designed for 500Hz. It is a reasonable compromise between selectivity and usability. A SSB filter width is too much for a CW rig. I think 300Hz is too tight, except in contests, but I am not really designing it for contests, more for SOTA activations.
The number of crystals forms the number of “poles”. 4 poles seems to be a good compromise for these filters, and many CW rigs have settled on this number. I will too. After the 4 crystals, I have a matching network to bring the impedance back down to 50 ohms for the second mixer.
The second local oscillator
The second mixer needs another local oscillator, this time pulled just off 4MHz, so that I get audio out. If I was using a SA612, I just need a tank circuit, but here I need something more. I need an oscillator, plus I need to get the LO up to 3dBm for the ADE-1L.
I spent quite a bit of time on this circuit. One of the problems is that the crystal can make the input to the first transistor quite free of harmonics, but the output is not much so. I found that it worked better with quite a high impedance biasing network. I also have gone with a Clapp oscillator fed from the emitter of the active device. The output has a collector resistor, but no inductor. This allows a moderate impedance path for the harmonics to go. The desired output goes through a series resonant circuit, to pass the fundamental, and then a parallel resonant circuit to shunt any remaining harmonics to ground. Most of the harmonics leave through the collector. The approach works quite well. I then use a second active device to bring the oscillation up to 3dB and impedance match to 50 ohms. Again, I have no inductor on the collector, so any harmonics (there is still a small amount) are shunted to AC ground.
At the end of all that, I have a near 4MHz local oscillator, controllable through a varicap on the crystal, mixing with a 4MHz intermediate frequency. This will yeild audio frequency output.
There will be, of course another 5.5db loss across this second mixer, so I have a net -5.5 + 22 – 5.5 for 11dB gain across all the mixers. S9 has gone from about -70dBm to 59dBm.
Next up is the audio circuit, which includes a automatic gain control. I’ll look into that for the next post.
Regards, 73, Wayne VK3WAM