Saturday 27 August 2016

Summer SMT Project

Over the past week, I've been playing with my 630m transverter summer project.

It's not the normal type of transverter that would produce a signal on 630m (~475kHz) but rather, one that will produce a signal on twice this frequency. My630m amplifier, using switching FETs, was designed and built with a 'divide-by-two' input circuit, to allow for greater frequency resolution when first constructed as a 2200m system. I really don't want to modify the transmitter and prefer just to be able to unplug my DDS and plug-in the transverter, allowing me to run WSPR, JT9 and several other digital modes when needed.

So far I've not been able to get the transverter to produce the desired low voltage 950kHz squarewave output, although I'm closer than when I began the troubleshooting.

My circuit is basically a G3XBM 630m transverter, without the FET final amplifier. Instead, the driver feeds a  full-wave rectifier frequency doubler for 630m, designed and published a few years ago by Jay, W1VD. Although I can get a squarewave end result, I think harmonics creeping into the process earlier down the mixing chain are causing non-symmetry in the output.

I'm using a 3.200MHz crystal oscillator with low level RF drive from my IC756PROIII transceiver at 3.675MHz feeding an ADE-1 double-balanced mixer in order to utilize the 'difference' frequency, 475kHz.

3.2MHz signal out of oscillator buffer stage

475 kHz signal out of mixer stage

Signal out of 'squarer' stage

Signal out of full-wave bridge

Signal out of LPF

Asymmetrical squarewave from 4046 output

I've tried a 3-section LPF with a cutoff of ~600kHz between the mixer output and the next stage, trying to knock down the 'sum' frequency of 6.875MHz and any crystal harmonics but have since replaced this with a 630m diplexer at the mixer output to pass anything below 1MHz to the squarer and anything above 1.5MHz to ground. I'll keep playing with it ... until it works as it should or I have no hair left to pull out. I may also just rebuild the circuit, Manhattan-style, with normal sized parts, making it much easier to change or modify as I need. The scope grabs shown above are with the diplexer in place at the mixer output. It seems like something in the full-wave frequency doubler is amiss. I've looked at T1, the bifilar transformer several times but don't see anything odd but that's the stage where things seem to go south. Maybe it's an impedance matching-thing, something I'm yet uncomfortable with when it comes to properly matching various stages.

I also now realize that the section of transverter I have chosen will require post-mixer filtering, not shown in the schematic. Not using the last stage (FET PA) and its associated filtering, has no doubt caused more harmonics to appear in the output than anticipated. If I redesign and rebuild, I will add an LPF after the crystal oscillator as well as at the RF input port and pay more attention to post-mixer filtering.

Another route to try, rather than the W1VD doubler, may be this 4069 IC doubler or a similar 4011 doubler.

Although the project outcomes have not yet been achieved, the process has been a positive one in a couple of respects and well worth the time spent so far.

I have learned that working with the 1206-sized SMT parts is much easier than suspected and in fact, populating the board with these tiny parts was a lot of fun.  The 1206-sized parts are at the 'large' end of the SMT size-spectrum ... rest assured I won't be going any smaller, at least for now.

My previous taste of SMT-construction was many years ago, when I built a small 40m QRP transceiver using seventy-two SMT parts ... it was an inexpensive kit from one of the QRP clubs. Happily, it worked with no problems but I found the process tiring and laborious and only soldered a few parts on the board each day until it was completed. The present experience was the exact opposite as I found it a very enjoyable experience ... I'm really looking forward to doing more SMT construction in the future.

The other positive outcome was knowing that my PCB design process can work well with the narrower lines, pads and much smaller IC footprints. I still design my boards using MS Paint and have always been happy with the ease at which it works. I know many might scoff at MS Paint but it really is a very versatile little program, with a short learning curve, unlike many of the more sophisticated PCB design programs.

Although I will work with SMT again and, similar to this project, I'll not be concerned about shrinking the board too small, although I'll work towards that goal. My main reason for trying a homebrewed SMT board was to see if I (my eyes) could still handle it and since so many parts are no longer available in the 'normal' size we have been used to for the past several decades ... particularly obvious when trying to source IC's in the good old DIP package. I am seeing a lot of 'obsoletes' pop up for some of my wanted chips, as the larger DIP packages are no longer in as much demand.

And one more positive outcome. I used the project as an excuse to purchase a new soldering station ... one that I had been lusting over for a few years as it is very SMT-friendly compared to my old Weller clunker. It also has an electrically isolated tip, unlike my Weller. The Hakko FX-888D turned out to be everything I had hoped and was no doubt part of the reason that I found the SMT process so enjoyable.

Hakko FX-888D
If you haven't tried building with these smaller components, don't be afraid to test the waters. I have no idea how long our 'normal-sized' leaded parts will be readily available but I now know that their eventual demise won't slow me down when it comes to homebrewing.


Anonymous said...

Hi Steve

Potentially lots of problems with that schematic. A diode doubler better needs a higher output impedance -- usually an RFC to ground, then we'll often run an BJT amp with collector tuning to remove any extra frequencies. Also , after a diode ring, we would normally place a band-pass filter [although a low-pass can work depending on the frequency scheme ] , then an amp to drive the diode doubler.

1 strategy is to mix your signal to 4X F, , then after the mixer, ply a band-pass filter and then drive a simple transistor switch set up for CMOS drive levels. Then drive a D-flip flop like the HC7474 to get a perfect 50 50 duty cycle square wave at half the input frequency. A Gilbert cell mixer seems perfect for what you are doing -- gives some conversion gain and cheap.

My 2 cents

Steve McDonald said...

Tnx for your input. I do agree with your comments and the need for more filtering. I will study your suggestions and add them to my growing list of notes regarding 'ways' to improve the outcome!