OSHW redesign of the Pixie-2

[rhb]

I thought I'd see if anyone here is interested in discussing this.  Out of curiosity I bought a kit for one of these:

https://www.ebay.com/itm/Assembled-9-13-8V-S-PIXIE-CW-QRP-Shortwave-Radio-Transceiver-7-023Mhz-Case-Cover/193561205516

It is basically the Pixie-2 designed by WA6BOY in 1995 which copied the Foxx designed by GM3OXX in 1982 but substituted an LM386 for the transistor AF amp of the Foxx.  So far as is known, the Foxx was the first instance of the PA also serving as the mixer.

Mine came from a US seller with 10 xtals.  It's a *very* minimal radio, 2 transistors and one IC.  The PA is also used as the mixer for a DC receiver.  It puts out about 500 mW.

The output RF filter doesn't comply with current FCC regulations, though that is easily corrected by changing the filter parts.

Here's what the xmit spectrum looks like on an HP 8560A:



Additional issues with these are:

no keying shaping

no selectivity

minimal gain

no useful VXO, just RIT of 700-900 Hz

I want to design some  low power drain QRP transceivers.  So as a warm up exercise, I'm redesigning the radio from scratch under the constraint of 2 transistors and one IC.

Goals are:

~ 5 kHz VXO range using 2 xtals in parallel with separate RIT

1 uV detectable signal

1-2 watt RF output at maximum input voltage

50 Hz wide audio filter with 80-90 dB of adjustable AF gain

operation over a wide input voltage range

FCC spurious emissions compliance

freedom from key clicks

minimal current drain both Rx and Tx

GPL'd board files with good RF layout

The PDF schematic for the kit I bought and the other is what I used to finish the kit when I couldn't read the values.  I've also attached the schematic for an AF filter with potential for 80-90 dB of gain and a Q of 14-20.

The Elsie calculated responses for the filter in my kit and an alternative Cauer design are attached as well.

Have Fun!
Reg

[rhb]

I used TIna for the first time today and designed an AF filter and amp using a TL084 quad op amp.  Haven't built one yet.  I need to find my tube of TL082 parts.

The active filter design came from Don Lancaster's "Active Filter Cookbook, Fig 7-9.  I added the AC coupled amp at the input to allow varying the gain, though the schematic doesn't show the pot.  I set the input level at 1 uV and ran the AC analysis from 100 Hz to 10 kHz.

If the simulation is accurate it should make a reasonably selective DC transceiver with pretty good sensitivity.

If anyone sees a mistake I'd appreciate an explanation and guidance.  Never having used a Spice package before it was a bit confusing.  I used TIna in hopes that it would accurately model the TL084.

Have Fun!
Reg

[rhb]

I got parts and built the new filter.  It's a significant improvement over the kit filter.

I transferred the parts from the kit to a small PCB and measured the response:



This is a prototype of the new filter on a similar PCB:



These are the test filters:



This is the key down output with the new filter:



And this is key down with the kit filter:



Have Fun!
Reg

[fourfathom]

That new filter is nice!  Actually much better than required by the FCC, but hey, why not?

Have you looked at "Class-E" design for the RF output stage?  You can get significantly better efficiency than with the Pi network (90% vs 50%).  Even if you don't need the lower supply current, reducing the temperature on your output transistor is usually a good thing.  I can dig up some Class-E pointers and links if you're interested, but the essence is using a series-tuned LC at the driver, which goes inductive at the harmonics.  This lets the transistor operate in more like switching mode.  You tweak the values to let the transistor switch close to the zero-current portions of the collector/drain signal.

[rhb]

I'm generally familiar with class E operation.  This started out as a redesign to maximize performance with 2 transistors and 1 IC.  But the the combined PA/mixer concept produces more LO radiation than I'm willing to accept and I don't see a cure for that using 2 transistors for the RF section.  If you know of a way to pull that off please post it.  I was deep in EMRFD  and other references this morning and I can't see a way to do it.

So I'm looking at a 3-4 transistor design with a separate mixer and diode switching to prevent LO radiation.

In any case, it's been an interesting and quite educational exercise.  Though I still don't understand how to specify the parameters for a Cauer filter to Elsie to get what I want.

BTW I've not measured efficiency yet, but class C is usually good to better than 70%.

Have Fun!
Reg

[iMo]

I built/modded the Pixie as well as Pixie-like trxes in past:
 Flea aka Pulga https://sites.google.com/site/ea3fxf/flea
 Mosquito  https://sites.google.com/site/ea3ghs2/mosquito
and there are also many modifications of the above (g...le).
It works basically, do not expect miracles, however.

[rhb]

I was looking at the mixer/PA this morning.  The current arrangement puts out about 10 mW in receive mode which I don't consider acceptable.  However, I think that if one switched the LO feed to the collector and the RF feed to the base when receiving that might solve the problem.  It might even provide some RF gain in receive mode.  I plan to model it with Spice and prototype the LO and PA/mixer on separate boards so I can test the two configurations easily.

The change to a 3 element Cauer filter suppressed the 2nd & 3rd harmonics by over 60 dB which is a huge improvement for 1 additional part.  I *really* like that.  I still haven't corrected the keying waveform, but that will be the next change to the kit.  And the end of the line for the kit.  There's not much room on the PCB for changes.

The Chinese kit I bought is basically the WA6BOY Pixie 2 with RIT added.  With no audio selectivity and no true VXO operation it's really pretty limited.  But if swapping the mixer inputs works I think it can be developed into a fairly decent radio within the 2 transistors and an IC constraint by providing adjustable AF  Q, gain and center frequency with  a quad op amp.  I ordered a bunch of 2SC5706 transistors which should let me bump up to 3 W out.  But even a 2N2219A or 2N2222A with a heatsink should do over a watt in class C.

If I can resolve the LO radiation problem I'll look into trying class E operation as battery efficiency is a prime goal.

If I can get it to work reliably in NVI mode into a dipole during the day over a 100-200 mile range with 5-10 kHz VXO range and a BoM under $15 I'll be pretty happy.  It would be a significant step up from the current kit.

Have Fun!
Reg

Edit:  Making the filter PCBs for testing has had an interesting side effect.  I now have a couple of 7 MHz RF filter modules.  I plan to stick electrical tape over the caps and chokes and then cover it all with copper foil tape with a label on the foil side of the PCB and the copper tape soldered to the PCB.  That will provide good isolation and some excellent modules for my RF tinker toy set.