Tuesday, April 11, 2017

A New Line of Transceivers - Difx (On Steroids)

4-13-2017 ~ A small Radio story.

Last evening (4/12) my XYL and I went to our favorite Chinese restaurant. Great food and really reasonable prices. As we walked in the door we saw a large party of about 25 people who were celebrating a birthday. I noted an older gentleman seated at the head of the table and our booth must have been only about 5 feet away. I leaned over to another gentleman sitting closely to me and inquired if it was a birthday celebration. He responded back that it was his father-in-law who turned 97 that very day.
 
I wished the celebrant a Happy Birthday and then commented to the other gentleman that his father-in-law experienced the evolution of radio broadcasting, the stock market crash, the end of prohibition, the New Deal, completion of Hoover Dam and the SF Bay Bridge, WWII, TV, space travel and the computer revolution. With that there was a bit of a buzz at the table as the other well wishers suddenly grasped what this man had seen.
 
Then the Birthday Boy spoke up and said I built my first crystal set when I was age 9! Boom that was amazing. I then asked if he wound the coil on an Quaker Oats box and used a piece of aluminum as a slider tuner -- and of course finding the hot spot on the galena crystal. He then said yes and asked how I knew that. I explained that about 18 years after him I built my 1st crystal set. What an experience. Just think anyone in their 90's has seen a dramatic change in our world. Of the 16M who served in WWII only 600,000 are still around. Not many left.
 
73's
Pete N6QW

 

DifX A Dual Conversion Transceiver!


In 2017, VU2ESE announced his uBitx (Micro-Bitx) transceiver which uses an up-conversion technique to a 45 MHz 1st IF and then a lower 12 MHz second IF, which handles the normal transceiver functions. This is a well known Gain and Selectivity approach to minimize "birdies" and to better process a lot of crud showing up on our beloved ham bands.

In theory the higher IF provides the gain and the lower IF the selectivity. But there may be more subtleties with actually having the higher IF to also provide a degree of selectivity much like a "roofing filter". So the first IF must be designed as such as to have  gain and bandwidth parameters in keeping with up conversion and crud prevention.

The magic key in a successful transceiver design is a reasonable gain distribution with attendant selectivity over the entire rig topology. Super high gain front end amplifiers that overload a receiver can and should be eliminated and the gain made up in later stages. When you hop up the signal gain at the front end --you are doing the same to the noise coming into the receiver chain! So what have you gained (that was a pun)?

A dual conversion has other desirable attributes such as the second filter selection. My KWM-4 has a 455 KHz "primo" Collins  mechanical filter embedded as the second filter; but the choice is mostly open to the many of the popular, 8, 9 ,10, 11 and 12 MHz filters. I say mostly as you really have to perform a frequency analysis to determine unwanted mixing products and ones where harmonics of BFO's or LO's end up in the middle of some conversion process. Later you will see such an analysis for this DifX rig. Don't overlook the 4.9152 MHz IF frequency as used in the Elecraft K2.

In fact there is some very strong support for keeping crystal filters in the range of 4 to 10 MHz and avoid those above and below that threshold. The reasons are many especially with the higher frequency filters where  stability is a very major factor. Yes stability -- usually crystal stability ratings are in PPM (parts per million). The inexpensive units (C^3 = Cheapo Chinese Culls) may be rated at 50 PPM or worse. So lets run the numbers at 4 MHz and 50 PPM that is as much as 200 Hz and at 12 MHz (like in the Bitx) = 600 Hz. The better units are 30 PPM (real crystals and more expensive) so our 4 MHz example is now 120 Hz and the 12 MHz versions are 360 Hz. When the 40M SDR Police report you for being 30 Hz low, you can see where this is headed. By the way the lower end filters like say 1600 kHz units (that notably were used in the hallicrafters SR-150) present some issues with regard to image rejection.

Later mention is made of commercial monolithic crystal filters -- the superior characteristics of these filters comes from the use of a common base crystal structure and tight control of the manufacturing processes. That is hard to do with 6 or 8 discreet crystals tack soldered to a PCB board that may be varying in frequency all over the place.
 
Speaking of crud, as I earlier mentioned, I am suspicious that someone in my neighborhood (Southern California) is growing "pot" in their garage, as I hear what appears to be the cycling of "grow lights" especially on 40 Meters. Thus some receiver architectures are better than others in handling such man made interferences. My only hope is that with the recent legalization of pot, my neighborhood grower will soon be out of business. I walk in my neighborhood on a daily basis. Up to now I have been looking for the culprit from among my neighbors who seem very happy (all of the time)  and sporting beer bellies from consuming too many munchies (Overly happy YL's and one's with the onset of mid-drift bulge are included). But I need to be more get scientific and use my compact 20M Transceiver (DifX II) with a battery pack and do some serious direction detection. You all saw that rig on the cover of a 2015 QRP Quarterly --right?

 
 
In 2012 I built a dual conversion SSB transceiver known as the KWM-4 which of course, was a DifX (Different than a Bitx.) Now we are moving to another DifX (on steroids) which uses that earlier topology but takes advantage of newer technology not readily available just a short five years ago. Essentially I will be strapping on a mixer conversion stage ahead of the basic single conversion DifX stages that I have been building for some time and most recently in the 60M Rig, the Big Kahuna and the DifX II.

I guess my 1st use of a DifX topology was in  the Spring of 2010, when I built the 20M MMIC Based SSB Transceiver which graced the cover of QRP Quarterly. Now that was a bit of innovation --MMIC IC's used in bilateral amplifier stages. For my friend N2CQR, that rig started with an analog VFO. But as you can see, soon morphed to the "Digi" stuff --- 7 years ago.

 
 
For my newest rig, the directed approach is to take advantage of a packaged 45 MHz, 7.5 kHz wide commercial monolithic crystal filter available from ECS (Digikey). While the uBitx employs a 45 MHz homebrew multi-pole filter, unless one is extremely lucky it will be difficult to match the performance of the ECS unit. Yes I know, there are the purists who want to homebrew everything and certainly that is commendable. But when I have an opportunity to utilize a device with known specifications at the outset, then it is a simple engineering decision.

I liken this to building homebrew double balanced mixers. Was I successful building DBM's  and did they work --yes? BUT an SBL-1 beat the pants off of the six that I homebrewed. It is hard to match components built on a manufacturing line using precision parts and tight process control while attempting to do the same with non-precision stock components, in a cold garage with an 80 Watt Radio Shack Soldering iron with a "Fat Tip"!
 
At the end of the day I want a top notch transceiver not just one that works. The packaged ECS filter is less than $17 USD and the cost differential of what would be spent building a multi-pole filter is not so great as to warrant a non-consideration of the commercial "black box". Keep in mind you might need to buy two dozen 45 MHz crystals just to find six close enough in frequency. With a matching circuit provided by the manufacturer the 45 MHz four pole filter with a 7.5 KHz bandwidth has a Z in/out of 50 Ohms which is perfect for insertion into many of the receiver topologies. The Si5351 third clock provides the 2nd mixer frequency. The matching circuit is shown below and L1 and L2 can be 11 Turns of #24 enameled wire on a T-30-6 core. (T-30-6 ~ Al = 36)

The first task I tackled was coming up with a PCB layout for the filter (really small) and the matching network. I simply took the manufacturer's pad layout which was in mm and I made a drawing at 10X in the normal inches (forget that metric BS). Once I had that done I used the design program (G Simple) to scale everything by 1/10 and then using the fact that 1 meter = 39.37007 inches (1 inch=2.54 CM or 25.4 mm) I had the design program further scale the design down to mm. I made a test cut on my CNC and was very close. I then had the design program scale everything by 1.10 (10% Increase) and it was dead nuts on. The alternative for those without a $250K CNC machine --you can just turn over the filter and wire it dead bug.

Below is a quick and dirty first cut prototype after the 10% upward adjustment; and the overall width is less than 5/8" so you can see fairly small. I have dxf files for this pad layout and the  ADE-1L DBM which I will be happy to share. I used a dull engraving bit for the prototype and when I make the final unit the lines will be crisp and sharp. It is envisioned that the conversion board will be about 2X3 inches and contain the 1st and 2nd Mixers (ADE-1L's) the matching network and the ECS filter. In discussions with the manufacturer the whole assembly will be shielded.



 
Just think the space you'll conserve by not building the VU2ESE multi-pole filter. Four poles of filtering having a 7.5 KHz bandwidth at 45 MHz is very in keeping with the gain and bandwidth  parameters. The specifications for this filter are 30 dB down at the half power points ( 3dB) and an ultimate  stop band   of 80 dB. The stability will exceed the homebrew filter!
 
 
 
.
 
 
 
Lets us start with the premise that the 1st LO (tunable) would up convert everything to 45 MHz. Thus in going from 1.8 MHz to 30 MHz the table below shows the conversion frequencies. My first thought was to use some packaged crystal filters for the 2nd IF such as those from the GQRP Club or INRAD. These filters are at 9 MHz. If you do the math the 5th harmonic of the BFO is right in the middle of the ECS filter pass band (5X9 = 45). So that choice is not a good one. If you use the uBitx approach, you avoid that problem. But this being a DifX I picked 11.5 MHz for the second filter frequency.
 
My frequency analysis shows you avoid the problem of having harmonics of the BFO end up in other parts of the rig. In checking the current price of the 11.5 MHz crystals in the large can form factor, 25 pierces can be had for 30 cents a piece form Mouser. I suggest getting the 25 pieces so that you have enough stock to find at least four that have no more than a 50 Hz spread across all four crystals. Most likely you will end up with several filters so not all bad.  If you can find five or six within that frequency constraint then you have the makings of a superior second IF crystal filter
 
 


 

The above table looks like it will avoid the problem of a tuning range sitting on the same frequency as the second mixer so we should be good.  I think we can now move forward


We now can take a peek at a first cut of a block diagram shown below. In a more detailed view all interfaces and matching are to 50 Ohms. So if you don't know --time to get smart on broad band matching and turns ratio squared! The block diagram below was based on the original thought of using 9.0 MHz filter; but is essentially the same with the 11,5 MHz homebrew filter. So where it says 9 substitute the 11.5 and the appropriate USB/LSB frequencies. Since we are using broad band amps for the bilateral amps no other frequency adjustments are necessary. CLK1 will now be 56.5 MHz and CLK2 will be either 11.4985 MHz or 11.5015 MHz. There are no changes to CLK0.

 
 
So lets examine closely what is happening with our frequency scheme and why does this and how does this arrangement work. Keep in mind ahead of this is a broad band amp stage (2N3904) that is adjustable and can provide up to 10 dB of gain. Typically I run these at about 1/2 the gain --enough to perk up the signal but not overload the downstream stages. That amp feeds a bank of relay switched band pass filters.
 
When the 42IF123 transformers were available on the market this would have been the tool of choice --so now you are stuck with buying a stock of TOKO transformers or building the BPF's from discrete components. If you are lucky enough to own a copy of the SSDRA (Solid State Design for the Radio Amateur) you can hand calculate the BPF's.
 
There may be a computer program in the EMRFD that will let you do the same; but I have never used the CD that came with my EMRFD. As you can tell, even though I have a copy, EMRFD is not my first choice for a reference document. Right now my EMRFD is a pretty expensive book end. After hand calculation, using LT Spice you can run your BPF's and fine tune them. The hand calculation process coupled with LT spice let's me really get inside my BPF's and when it is necessary to make a new one, I know how to do it!
 
So now the signal entering the block diagram will be relay selected and covers just the ham band we picked. So already there is some signal clean up in play. The 1st LO from CLK0 in the Si5351 up-converts the ham band signals using the difference frequency (59.0 MHz -14.0 MHz = 45 MHz or 59.2 MHz - 14.2 MHz = 45 MHz) The sum frequency would be (59 + 14 = 73 MHz ) way out of the pass band. Now for the first bonus -- the ECS filter has a 7.5 KHz bandwidth --so anything 3.75 KHz +/- away from the desired frequency is also filtered out! You are screening out (attenuating is a better word choice) the California KW station 10 KHz away from your rig!
 
The signal is now passed on to the second mixer stage operating at 56.5 MHz and so the down mix is at 11.5 MHz. In both of our mixing processes the mixing frequency was above the signal (most desirable) and in effect we have two sideband inversions -- so the lower BFO frequency will yield USB and the higher BFO frequency LSB. Again all interface matching is at 50 Ohms and three ADE-1L's are used as the double balanced mixing devices. The ADE-1L's are 3 dBM devices so low drive requirements -- set the drive level in the Si5351 to 2 Ma.
 
I will now outline some of the performance specification for the new DifX rig.
  • Dual Conversion ~ 1st IF @ 45 MHz and 2nd IF at 11.5 MHz
  • 10 Band Operation: 160, 80, 60, 40, 30, 20, 17, 15, 12, and 10 Meters 
  • USB/LSB Operation
  • 15 Watts Output
  • Color TFT Display
  • PIN Diode and Some Relay Switching
  • Si5351 PLL Clock Generator
Some words here about the VFO. Forget building a quality rig like the DifX with an Analog VFO as you are really limiting yourself. Time to move up the Big Boy's sandbox.  The use of the Arduino in addition to providing the digital VFO capability also provides the color display, Tune function, USB/LSB, band selection and on and on. Analog VFO's can be fun in some rigs, but not this one! This is a DifX so move on up.

When I build a transceiver I always start at the back end and as I progress forward each part of the completed build is now part of the test system that will be used to test new additions. If I build a circuit and install it into the test systems 99% of the time it is what was just installed is where the problem lies as everything up to that point works! Thank you Heathkit
 
Thus starting at the back end is the audio amplifier. The audio stage deserves some respect so use the low noise version of the NE5534 driving an LM-380. Forget that crap about building a discrete component amplifier with 2N3904's/2N3906's --yes building one using that approach is like earning a Boy Scout Merit badge. It is also like doing brain surgery with a rusty spoon; thus it is time to move on to something more robust. Another NE5534 can be used as the microphone amplifier. This is an uptown rig and so move on  past the low rent district.
 
 
This is enough to start your heart pumping and the insatiable to desire to heat up the soldering iron. Welcome to the world of the DifX.
 
73's
Pete N6QW