Designing a 20/40 band CW rig – Part 2

Hi all,

This is a continuation of Designing a 20/40 band CW rig – Part 1.

Shown below is a schematic of the TX and initial RX parts of the rig. I have been successful in simulating all of these parts, and have a reasonable amount of flexibility if the real world performance of the components does not match the simulated performance.

Schematic of the TX and initial receive sections of the rig

TX Driving

This being a CW orientated rig, all that is needed is a oscillation on the desired TX frequency. CW merely turns it on and off. I did mention that I wanted to retain cabability of PSK and FSK modes. PSK needs the phase of the oscillation to be changeable, while FSK modes need the frequency to be changeable. An Analog Device AD9834 fits the bill. Steve KD1JV is using this in his Mountain Topper. This device is not a PLL and VCO combination, rather it generates the output digitally, and then feeds a DA convertor to generate the waveform. The output is needed for two things – it is the Local Oscillator for a mixer, and can be used directly for a TX frequency. The AD9834 has a number of capabilities that I plan to use, but this will be for another post.

Unlike Steve, I have decided to take the analog output of the AD9834. This output will drive the initial mixer, and the device selected has a 50ohm load. I’ll talk about this device on another post, but I will be using a different approach to the other designs I have seen because I want more dynamic range and better inter-modulation distortion performance.

The input to my driver is going to be taken off this output. The AD9834 can provide a balanced or unbalanced output. I need unbalanced. This signal is shown as OSC. It then feeds the base of a MMBR941, Q1. This is a RF BJT device in common emittor configuration. One of the nice aspects of BJT devices in common emitter is there is a fair degree of flexibility in setting the input impedance and the output impedance of the device. I decided on a 150ohm input impedance. I am also driving the device fairly hard, but not too close to device limits. The gain on the device is 26dB.

To save power when not TXing, I use Q2 as a switch. This 4401 device is being used as a current sink, switched on during TX, and off when not. It deprives Q1 of its DC ground, so no DC current will flow through the device. The bias network is shut down, but is still kept alive on AC so that the load presented to the AD9834 is not significantly changed. The effect of Q2 kills nearly all of the output power from the collector of Q1.

TX Power

My objective is to deliver 5W into a 50ohm load (antenna). There are quite a lot of choices to go about this, but I want to keep things as simple and as cheap as possible. I considered initially using another BJT in emitter follower mode, by using Q1 to deliver the desired peak to peak voltage AC signal and then the emittor follower would supply the current. I used an inductor on the emitter to improve the efficiency, but at the end of the day, it is still being used in Class A mode. This is a CW rig, so Class C amplifiers await!

The schematic shows three BS170s that are driven by the output of Q1 on their gates. These are N-channel MOSFETs that can dissipate about 800mW of power. If they are run 66% efficient, then that means that each one can deliver 1.6 watts. The efficiency is a function of the peak to peak voltage supplied by the MMR941, and what level the bias is set at. The bias mid point needs to be below the pinch off voltage on the BS170’s for Class C operation. The further away, the better the efficiency, but it cannot be set too far away, as there are limits to the peak to peak output I can get from the driving device. After a fair amount of experimentation, I set the bias level using a voltage divider resistor network. R9 and R10 (the R10 going to ground) form this network. (The other R10 nearby on the base of Q3 has a new identity as R18). This level ends up being around 1.2V, around a volt below the BS170 cutoff to ensure Class C operation.

The job of Q3 is to act as a switch, on during TX, so that R9 and R10 do their job as described. Off during RX, so R9 is taken out of the circuit, and R10 pulls the base of the BS170’s to ground so that they are cut off from doing anything. The BS170s then act as open circuits.

Presenting the output to the load

Class C waveforms have less than half of the initial sine wave present. The BS170s also present a changing load to Q1, so what it produces is not a great looking sine wave either. We actually have quite an ugly looking waveform on the output of the BS170s. This needs to be cleaned up. Also, the load is not 50 ohms. It can’t be if this circuit is going to operate off 12V. The best that can be hoped for, in terms of an output wave form is 24V peak to peak, due to the action of the inductor L2. I used the complex model of L2 in my simulators, rather than using a perfect inductor. The complex model output was practically indistinguishable from the ideal model. Coilcraft make some good inductors, and I plan to save builders of this rig from having to wind their own inductors, by using these Coilcraft chip inductors.

24V peak to peak really only allows for 12 ohms load impedance directly on the finals. This needs to be transformed to 50 ohms at the antenna connection. All rigs need to do this transformation. Steve’s Mountain Topper Radio provides a network for each band. The FT-817 provides one for each of its bands that it can TX on, and they are switched by relays. The FT-817 set of finals are operated in a push-pull configuration, but off a 8V rail. This means that the load impedance that these are driving will be lower than 12 ohms, perhaps about 9. I should analyse the inductor/capacitor networks for a given band that Yaesu have put in there. If you were to look at the circuit diagram, these matching networks take up most of the power board module schematic.

I reckon a bit of convenience is a good thing, so I am going to use a relay to switch between the two bands. This relay is K1. It is DPDT, so I can use the one relay for both ends of the matching network.

The match network has a LC tank which does most of the job of restoring a clean sine wave, and then a LC series, presenting a high impedance to remaining harmonics. This second LC series begins the process of impedance matching, where I only need two more capacitors to complete the job. This is per band of course.

Now reality bites, and there is the need to use real world components. Also, for these networks, X7R dialectic is not acceptable, but for cheap capacitors, 1nF or above tends to be X7R. So to get around this, I have paralleled up some caps to get the values I want, and to continue to use NP0 dialectic caps in 0805 SMT size. There was one cap where this approach was too long in the tooth, so I use a 3.3nF ATC cap that is 1111 size. It’s a much more expensive cap, but I’m only using one!

After these matching components, I have a sine wave output at 50 ohms supplied to the antenna. It is the job of the operator to take it from there.

Receiving

I have to receive as well, this is a transceiver after all! Now, approach 1 could take the signal straight off the antenna jack, but there is the matter of the TX output to deal with. At 50 ohms, this is not going to be a 20V peak to peak output any more. It is closer to 83V peak to peak. Of course, all of the caps need to have dialectics rated to 50V, as 83V gets 41.5V from ground, each side.

Now, I could use a relay to switch between RX and TX, to shut out the TX signal from going anywhere but out the antenna jack to the antenna. The FT-817 does. One problem – full break in on CW. It is not too good to have a relay clatter (twice) every time you send a dit or a dah. I want full break in capability on this rig, so this means using a transistor approach. A transistor is going to have to hold back this 83V, but that is too much.

Instead, I have taken a similar approach to Steve KD1JV, by putting a blocking circuit at the 12 ohm area. This cuts down the voltage that the blocking network needs to resist on TX. It also has the benefit of providing a filter through the same LC networks that dress the TX output for the antenna.

The main transistor to block is Q9, a N-channel MOSFET 2N7002. I have used a biasing network, that is assisted by BSS84 Q10 and a 4401 Q11.

Q11 is a current sink. When switched on, it takes the gates of both Q9 and Q10 to near ground. Q9 is a NMOSFET, so a ground on its gate will block any signal that is about 0V or higher. Near ground on Q10 is a PMOSFET, so it is switched on, there being nearly -12V between its gate and source. Q10 switched on takes R15 out of the circuit, leaving a voltage divider of R13 and R14, meaning Q9 is biased over 10V. The input signal is going to swing between 0V or so and 20V on the drain of Q9, with 0v at its gate. If something makes it to the source of Q9, it won’t be much below the gate, meaning Q9 should stay switched off, switching off TX signal from sensitive RX circuitry.

When wanting RX, Q11 is switched off. This then means the gates of Q9 and Q10 are pulled to 12V by R17. 12V on Q9 gate will tend to switch it on. Q10 will be switched off, meaning that the bias network is now R15+R13 against R14. The drain of Q9 will be biased at around 1V. There is 12V on the gate now, and around ground on the source. Q9 will be switched on as hard as I can make it, so get as much of the RX signal through.

Unfortunately, when TXing, quite a bit of the signal still makes it through Q9. Too much by itself to be safe. The transistor itself is ok, it is not being used anywhere near maximum limits, but the mixers awaiting later will not like what gets through. I want to block this signal, so I use another LC series resonant device to block the harmonics. This is because what does get through Q9 is horribly mangled with most of the power in the harmonics, not the fundamental. This cuts things down nicely. Also this network starts the process of impedance matching from 12 ohms back up to 50 as the mixer I want to use wants 50 ohms as a source impedance. Now this matching network is band specific, so this means a second relay. This relay, like the TX relay is only switched during band changes, so there will not be any relay clatter here. Q12 and Q13 are little helpers. They are switched on during TX to shunt away any TX signal that still makes it through this far, but because of the LC network, it ain’t much. The job of Q9 is also easier to block the TX signal when it is mostly high impedance sitting behind it.

Wrapping up

This about wraps up this post. Next up I will look at the audio processing part of the rig. The other major parts are the RX mixing and filtering and the microprocessor. I’ll look at these with later posts. My plan is to get one of these rigs built in about a month to two months as a prototype, and we’ll see how it goes from there.

Regards, 73, Wayne VK3WAM

This topic is continued at Designing a 20/40 band CW rig – Part 2.

Advertisements

One comment on “Designing a 20/40 band CW rig – Part 2

  1. […] This is a continuation of Designing a 20/40 band CW rig – Part 2. […]

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s