T/R Switching

A quick post from the day’s experiments: the transmit/receive power switching arrangement for my next project. The scheme is very much like that in the KN-Q7; I wouldn’t have stumbled across this (very simple) setup without an excellent write-up by Andrew Woodfield, Zl2PD.

Here’s the simple schematic:

The power switching arrangement for a new project.

The circuit itself is straightforward – when the “key” pad is left floating, current can flow through the 2k2 resistor attached to the 3904, providing a small base current and driving the 3904 in conduction and powering anything connected to the +12R pad. At the same time, there is nowhere for base current to flow in the 3906, so no current is provided to the +12T pad.

When the key pad is grounded (by a morse key, or other TR switching method), the base of the 3904 is pulled to ground through the small signal diode, and very little current will flow through the 3904 and into the +12R section, effectively killing receive functions. At the same time, a small current will flow through the 2k2 resistor attached to the 3906, allowing it to conduct and powering anything attached to the +12T circuit.

In short: when the key line is floating/disconnected, the circuit is powering receiver functions. When it is grounded, the circuit powers transmitter functions.

I added the little 7808 regulator to this power board, which will be powering some NE602 mixers on receive only. I put both circuits together on a little piece of copper-clad, in something like Island-Pads-meets-manhattan-construction:


I like how the layouts ends up showing off the inherent symmetry between PNP and NPN transistors. The whole thing looks quite nifty on the bench. I left the input and output pads deliberately large, to accommodate however many connections end up being made.

I did begin construction and testing on the next project, but was stymied by my SLA-battery power source running down (it was down to 7.4 out of 12V by the time I thought to put a meter on it). Rather than switch to the bench power supply (an old computer power supply with the 12V and 5V rails brought out), I took it as a sign to call it a night.

Hear you over the sizzle of solder!


Si5331-Driven Transmitter – Beginnings

In Spring, a young man’s fancy turns to thoughts of transmitters. Having recently completed a receiver, and with the weather starting to warm a bit, I’ve got an itch to actually get on the air and talk back to the stations whose code I can now (slowly, painfully) decode.

Re-purposing some hardware and code from my DDS VFO project, I’ve been working on on a digitally controlled CW transmitter based around an Si5351. This is by no means an original thought, and my designs are largely based on Qrp-Labs’ Ultimate3s Kit. You can check out that original design over on the QRP-Labs site, under the PCB assembly instructions.

Essentially, this transmitter uses an Si5351 DDS clock chip to directly synthesize the desired output frequency, at up to 200 Mhz. This frequency is then amplified by a simple FET amplifier to approximately a 1W output level, then passed through a low-pass filter and out to an antenna. The Si5351 is controlled over i2c by an Arduino Uno, which has an attached LCD, a rotary encoder, and a couple buttons for frequency and band control. The updated code for this project is on Github.

Here’s a block diagram of the transmitter:


The nice thing about this design is that the main frequency-dependent component is the low-pass filter; the Si5351 should be stable enough for CW contacts up to at least 50MHz. (Without an ovenized environment for the reference clock or some GPS disciplining or similar, there’s still a little drift and inaccuracy, but I don’t think it will be noticeable.) Above HF, I’d expect to see diminishing returns from the FET amplifier. But switching HF bands should just mean switching LPF filters and pressing a button on the VFO.

Here’s the circuit as constructed, up to where the LPF would go:

CW Transmitter sCHEMATIC
The simple Si5351-based transmitter, with a 3-FET amplifier.
Let’s walk through the circuit starting from the signal generator and working out toward the antenna. The output of CLK0 on the Si5351 is coupled into the amplifier with a 100nF cap. This drives the gates of three J110 FETs, and is biased upward by a voltage divider formed by a 10K pot and a 4.7kOhm resistor between 5V and ground. The power end of this voltage divider is bypassed to ground with a 100nF cap.

Power for the amplifier is fed from a nominal 12V (or lower) through an RF choke, in this case 25T on an FT37-43, into the FET drains. A 100nF cap here helps further bypass RF to ground at this point. Output is taken off the drains through a final 100nF cap. The FET sources are grounded.

While an actual RF transistor like a BS170 would likely be ideal, I had a bunch of J110 FETs in my bin after my last trip to California, so that’s what I used. They’re only rated for about 300mW dissipated power, so I’ll need to be careful with my heat sinks and duty cycle until I can replace them with something a little more sturdy.

Three J110 FETs with their heat sinks on a piece of copper clad. In the back you can see the PA power coming in through the black alligator clip via the small RFC. The bias pot is on the left, signal comes in on the red wire on the bottom, RF out via the BNC connector on the right. The big white box down in the bottom is a 51 Ohm, 5W resistor I was using as a basic dummy load for testing
Preliminary results are encouraging – I hooked the output of the above circuit directly to a dummy load with no low-pass filter and ran the clock generator at 7.050MHz. I assessed power by reading off the peak voltage on an oscilloscope.I started with just 1 J110 and a 5V PA supply instead of 12V. This yielded about 9V peak-to-peak, or 200mW into 50 ohms. Installing the other two J110s bumped the output up to 10V P-P, or 250mW. Finally, after installing heat sinks on the FETs for safety and taking the supply voltage directly from a 12V SLA battery (~13.2V), the output hit 22V P-P, or 1.2W into 50 Ohms.   This last reading was verified with the power meter on an MFJ versa tuner.

The transmitter spread out on the bench, with the display and Arduino at the top, the amplifier visible on the copper-clad in the middle, and a bulky MFJ tuner at the bottom acting as dummy-load and power-meter.Power provided by the 7Ah, 12V SLA battery at top-right.
I’ll need to put a little elbow grease into low pass filters before putting this on the air, because even on a scope the signal looks a bit gnarly. But first the first time, I’m responsible for 1W of Homebrew RF power. Watch out!



On the Importance of Fiddling and Good Power Supplies.

Troubleshooting continues apace on the new NE602-based direct-conversion receiver. As I mentioned in the previous post about it, the receiver develops and unfortunate, LOUD squeal whenever the 10K audio gain pot at sits between the NE602 and the LM386 is advanced past about the 20% position. This is a terrible impediment to reception, so I’ve been working on eliminating this problem.

TL;DR: The receiver is a whole object tied together by its power system. Good power and a good ground are important.

My suspicions centered on the LM386 chip – after all, an audio amplifier with a couple of feedback caps seems like a prime candidate to turn itself into an audio-frequency oscillator.  This, as it turned out, was a red herring – I’d like to publicly apologize to the LM386 for ever doubting it.

The first step I took was to meter the potentiometer’s resistance at the setting just below squealing, and to replace the pot with two resistors that replicated the resistance at this setting. A little fiddling showed that the squeal could still be induced by varying these resistances slightly, which was encouraging – the problem was somewhere else in the circuit, and not a phenomenon of the pot itself.

So I took an entirely different tack to connect the NE602 and the LM386. The 0.01µF cap in series with a 10K pot was derived from a number of other designs, but it seemed worthwhile to utilize the complementary outputs from the NE602 as a means of input to the LM386’s complementary inputs. I stole this linkage directly from the EMRFD. The NE602’s pin 5 is connected to the LM386’s pin 3 through a 220nF series cap with a 10kΩ resistor to ground. Similarly, the mixer’s pin 4 was connected to the audio amp’s pin 2. Both lines were tied together with a 100nF cap.

You can see the three new blue caps that sit between the NE602 and the LM386, a well as the large white jumper bridging the 10uF cap to pin 8 of the LM386

Long story short – no help on the oscillation-front, but a little more stability in the gain, it seems?

Next, I set about fiddling with the feedback circuitry for the LM386. As the datasheet shows, there are gain configurations from 20x to 200x gain with a simple RC network between pins 1 and 8. My original configuration had a 10µF electrolytic cap in between these two pins. I tried replacing that with a 4.7µF cap plus a 2kΩ resistor, leaving those pins totally unconnected… in all configurations, it was possible to get the audio to oscillate. So no such luck there. I put the 10µF cap back in, and joined it to pin 8 with a small piece of female header, so that I could insert various resistors in that position to adjust the gain later, if need be.

Oddly, both reception gain and AM bleedthrough is increased whenever I touch the leads of the gain resistor. The AM bleedthrough sounds a lot like what happens when you touch one of the un-connected leads of the LM386, but why do I hear CW signals a lot clearer as well?

It was this oddity with the gain resistor that gave me the vital clue. “Well,” I thought, “maybe touching and fiddling with the other components will give me some clues.” I just played around on the board for 20 minutes with it powered on, tying this point to ground, putting that point high through a resistor… interesting things happened, but squealing was still very much a possibility. That is, until I happened to accidentally pull the oscillator crystal out of its holder, and the squealing stopped.

“Now that’s odd!” I thought. “I assumed this problem existed entirely in the audio half of this receiver. Why should removing the crystal, and thereby halting oscillation in the mixer, have any effect?”

A little poking around with a meter and a scope and I had my answer – the little (used!) 9V batteries I was using as a power supply for this receiver were woefully under-powered. Under load (i.e. with audio coming out of the headphones), the voltmeter read about 7.3V, and dropped by about 10 mV per second. Oscillation seems to occur when the voltage  drops too low. (I’m measuring voltage as a proxy for available power, in this case, I think.)

So, after borrowing a nice big 13.8V, 8aH SLA battery from work and bodging together a quick-connect to battery-clip connector… wait for it…

No more motorboating! (And a bunch more audio output to boot.)

Used 9V batteries just won’t cut it! The big SLA battery provided great power and more audio output.

I still need to boast the overall gain of the system, since it takes a pretty strong signal to get into the receiver, but at least it’s not likely to HOWL in my ears!

Next thoughts: to increase gain, perhaps stealing the alternate LM386 gain configuration from AA7EE’s WBR Recevier, or one of the broadband, ~18dB IF amplifier designs from the original BITX or QRPKit’s variant.