Saturday, 4 July 2015

Motley Tales of Woe and Rapture

Notes on a few experiments for my Jupiter receiver.

I departed from my 5 normal month radio hiatus to blog because it’s super-hot outdoors and our basement feels oh so cool. Sometimes, too:  if I wait too long to post stuff, I lose passion or worse yet — forget details and/or lose data.

A few readers shared that they look forward to the outcome of my Jupiter direct conversion receiver experiments – so do I!  At this point, I haven’t nailed it down and go/went/reach in all directions like a drunken sailor.  I’ll show a few of my primitive style radio design experiments and often feel stuck in a groove with my mostly 1980’s-90s circuitry.  This blog flaunts neither Arduino stuff, nor C++ objects coded to perfection; just really ugly analog circuits and some measures. Analog radio plus binary thinking still works for me.


I love the works of Anton Chekhov. Physician, scientist, and — a staggeringly good short story author — who penned crisp, slice-of-live narratives that both entertain + prod you to think deeply. Chekhov inspires my aude sapere — “my dare to know” self and supplies the tone for my blog: a thin slice of 1 particular QRP workbench.  

This post arose because I wanted to design my own Jupiter receiver, sought fresh circuitry — and in the end, wasted a ton of time and parts just to learn a few new things. But that’s what I do.

Every builder, I suppose, learns in their own way — I believe that making and measuring real circuit’s yields dividends par none.  Hats off to the SPICE kings — simulation might be the next best thing for those who lack the instruments needed to measure data and make comparisons — and for getting starting values for an experiment.  To me, however — real versus simulated life including electronics just feels better. 

Above  — Jupiter Block Diagram

20.1 MHz VFO

With some oscillator details gleaned from Jim Sky, my target receive frequency = 20.1 MHz.  Jim shared that we’ll want to avoid WWV 20 MHz and also need to occasionally steer around any nearby carriers. 

So I set out to make a frequency agile oscillator and struggled, sputtered and got beat up in the process. Old Murphy, stupid mistakes, bad parts and the-like tangled up nearly every LO experiment.  As a result I didn't build the best possible LO due to fatigue and frustration — but at least I felt reinvigorated to explore mixers.

My LO strategy involved mixing a 16.93 MHz xtal oscillator ( xtal Q = ~100K ) with a 3.121 to 3.216 MHz Hartley L-C VFO. I planned to use the EMRFD Figure 4.24 method to extract low noise + distortion from the xtal oscillator and mix it with the VFO signal in a Gilbert cell mixer like the NE612.

A reader sent me 4 NE612s last year — it turns out all of them were fried. Sadly, I didn't suspect these mixers until much in-situ debugging ---- wasting parts + time. I didn't want to wait until fresh NE612 mixers arrived and set out to make a homebrew mixer on the bench.

First I'll show the L-C VFO:

Above  — VFO schematic with a buffer giving a 50 Ω output impedance The output power measured -8.7 to -10 dBm across its tuning range. With practice, it's fairly easy to make a temperature stable VFO in the 1-3.5 MHz range. I employed light resonator coupling and perhaps overkill DC filtering + output buffering. While testing, I could not pull the VFO frequency with downstream manipulation despite trying hard to do so.

Above  — Another super buffer schematic with measures. In Winter 2015 I explored ways to cascade common base amps to develop "super buffers" possessing really high reverse isolation. L-network matching tends to work best @ 1 frequency, however, with experimentation, they can work over a limited range with respect to VFO output amplitude flatness.

Above  — VFO schematic built on single sided Cu board. The resonator T50-6 toroid is both zap strapped and epoxy glued to a piece of copper-clad board. A red colored wire provides the 1 link output to the VFO buffer.

 Mixer Experiments

 I built a few active mixers and will refer to just 2:

Time to add some balance:

Above  — A mixer with differential RF and IF ports. Some tuning on the output serves to further suppress the  LO and RF at the IF port. I applied discrete, matched transistors, but sometimes build them on a CA3046 BJT array for even better transistor matching.

Above  — Analysis of the BJT mixer output. The LO and RF are roughly down the same power from the desired sum IF @ ~20 MHz. (LO and RF = 36-38 dB down). From my real-world balanced mixer experiments, you've achieved reasonable balance when the LO and RF are >= 30 dB down from the LO & RF sum + difference frequencies.

Although usable, I realized I wouldn't have enough room to build all the needed filtration plus amplification circuitry and thus went to a tiny MCL TUF-1 diode ring mixer. 

To run this diode ring mixer, I changed the buffer on the L-C VFO to that already shown and adapted the ever-evolving 16.91 MHz xtal oscillator circuit to chop the TUF-1 LO port with a solid 50 Ω termination @ ~ 7dBm.

Above  — Low noise, low distortion, well buffered 16.93 MHz crystal oscillator. 

20.1 MHz mixer filter and buffer amplifier

I still needed to filter and amplify the TUF-1 mixer output to make a usable 20.1 MHz local oscillator for my Jupiter receiver. I didn't have a lot of board space left and had to compromise.

Above  — Post mixer filter amplifier sandwich. I've never tried this before in a post-mixer filter scheme and ran T68-6 toroids [ should have run size 50 or less 'roids to save space ]. At some point I must have shorted and fried the MPSH81 and had to clip it out of the crammed board and replace it. This oscillator project fought me from start to finish. 

Frequency domain analysis of the entire 20.1 MHz local oscillator output. All non-20.1 MHz tones are down by at least 54 dB. Not great for me, but OK. Prior to placing the board in its metal box, the LO and RF tones were down closer to 60 dB, but the metal box de-tuned, and/or de-Q'ed my T68-6 filter resonators.

Lacking both time and patience to spend more time wrestling with this oscillator. I just bolted the lid on and sighed in relief when it was done. 

The 16.93 MHz LO tone is especially tough to filter since it's close to the IF. A 16.93 MHz notch filter might be employed downstream from the oscillator box. The output power = -6.79 dBm.

Above  — The final box. A lot of work and parts for a simple 20.1 MHz VFO! It might be better to make a synthesizer with a Si5351A or a VXO affair, but I went my own long way.

Triple-Tuned Filter Preamplifier

I'll show the current favorite Jupiter receiver preamplifier circuit. I'm not sure if I'll keep it, but its transfer function and gain look fabulous.

 Above  — The schematic of the input filter with PNP common gate amplifier.

Above  — S21 in TG plus SA sweep. High Q tank parts, careful tuning, and proper port + part matching may yield wonderful results. Admittedly, it's hard to get outcomes this good without some serious test equipment.

Above  — Filter-amp breadboard photograph. 

 Mixer Experiments

I'm still experimenting to find a good mixer (product detector) for my Jupiter receiver.
Here's a couple of candidates I've tested so far:

Above  — A JFET bridge variant like this might be seen in high IIP3 professional grade receivers.

Above  — SA plot with the RF @ 10 MHz and LO @ 25 MHz. As you turn the 10K balance pots, you'll see the RF and LO tones change in power. I tweaked the 2 pots to get the best balance (the lowest LO and RF power). Spectrum analyzers = serious fun.  After getting the best possible balance, I then performed the mixer's 2 tone testing.

 Above  — Highest mixer IIP3 I've ever measured on my bench.

Above  — Breadboard photo. Double sided copper board with via ground wires. Another replica only yielded an IIP3 of 23.4 dBm. Lots to learn!

I plan to investigate the Trask (N7ZWY) KISS mixer. Here's my first experiment.

Above  — I lacked many of the specified parts such as the Fairchild FSA3157 and those expensive MCL transformers, but made do with a dual FET bus switch, the CBT3306.
The LO transistor switch + D flip-flop come right out of EMRFD.

I recently ordered some parts to asses more KISS mixer variants including the FSA3157 and also stocked up on some obsolete, but very fast 74F series logic including the 74F04 and 74F74 that will clock up to 125 MHz or so.

Above  — The FET bus switch evaluated as a mixer. The LO and RF are ~ 35 dB down @ the IF port. You can see tones at ~ 12 and 14 MHz  — I've got to get used to working with digital circuitry in my RF mixers.

Above  — IIP3 measure with the tones 40.5 KHz apart. All three of my 2 tone test signal generators feature very low phase noise.

Above  3 — KISS mixer breadboard photo. Again, to provide proper RF ground for all output frequencies, I employ double-sided copper board with copper via wires around and especially adjacent to every pin that requires RF ground. I found this may improve mixer balance and even IIP3 depending on frequency.


Above  — For those who have trouble visualizing mixer math, this frequency domain sweep tells the story of mixing 2 frequencies in a diode ring.

This spring/summer I watched yet another YouTube video stating that mixing the RF + LO results in 2 frequencies. I wish it were true!  In general, we can describe mixer output frequencies with the equation:

IF (output freq) = N x LO port freq +/- M x RF port freq -- where N and M = integers such as 1, 2, 3 ....

Above  — 2 tones from the output port of my 6 dB hybrid combiner into my DSO

Above  — Vishay MLCC HF bypass caps. Summer is the time to buy parts for Fall-Winter building.

Above  — Great, ten NE612 ICs came yesterday.

That's all.  Thanks.