Thursday, January 18, 2018

2018 ~ The Year of SSB Transceivers

The IC Transceiver Design (Based on K1BQT)

2/15/2018 ~ Thinking out of the Circuit.

One thing I now look at when I am designing a circuit is how can the design be used to better advantage. At times I have been criticized that something I designed was no good for those who like to strap on a back pack, climb up some hill in a god forsaken place and run a QRP marathon. That is never one of my design goals. But of late my goals have taken a twist --how to build things so that circuits could do double duty.
This is not out of laziness but time constraints. I have so many projects lined up that unless I look to building less in a project it will be impossible to build more projects. So going back to our 40 (or 20) Meter Driver stage my thoughts drifted toward how this could be used on both transmit as well as receive. Let us start with the basic circuit shown below.

This board is way too large for a compact type back pack rig but was purpose built as I knew I would be adjusting, tinkering and changing things. I often build two versions as the first serves as a prototype and the second becomes the production unit. Often these prototypes have found their way into yet other rigs. In 2016/2017 I built several "Junk Box" rigs out of such leftovers. So initially the thought was this board would be put in the junk box.
I jokingly say that I do my best thinking while I am asleep and often my brain awakens me with some sort of an amazing idea --unfortunately it often is at 3:00 AM with my never returning back to sleep.
Well it happened again several nights ago specifically with this board. So here is how my brain sent me a message. Suppose we were to introduce some relays on the board so that the signal path could be changed in that the J310's instead of being used as the pre-driver to the 2N2219 was instead "steered" so that is now was the Receiver RF Amp stage, and the Band Pass Filter that followed the J310s' was now steered ahead of the J3110s and what was the output of the BPF that normally fed the 2N2219 was now connected to the antenna circuit. On Transmit the signal path would be restored by the relays to the original design.
What we have below is a block diagram of how that would happen. The R & T designations show how the relays would be connected so that the circuit elements could function on either transmit or receive. Now all that extra board space can be fitted with SPDT relays -- four of them. Assuming that an additional RF linear amp stage is added to the 2N2219 (say an IRF510) then you would need to have the 8 element W3NQN filter following the IRF510. So the receive side would have the benefit of both the LPF and the BPF. If one chooses to use just the 2N2219 (at 300 MW) then replace the 5 element LPF with the 8 element and then the LPF could be switched between the BPF and the output of the 2N2219.
My next task would be to add the relays --All Electronics sells these neat SPDT relays for 60 cents a piece. As it turns out the dual filter test board could now be mated with this prototype board and we now will have another transceiver on line. I already know the 2N2219 stage will drive my outboard amp to 30 Watts so I may just skip the IRF510 stage. For those with the backpack stations --lets see you make some contacts at 300 MW.

BTW this also will prove the suitability of buying those filter boards from Israel for $34 delivered to your door. In one of the videos I had of the test setup with the boards spread out on the bench, I received a "negative" comment that it was such a junky mess. Well that may be true but it does work!

Pete N6QW

2/08/2018 ~ Time to Drool! The ASUS Tinker Board is Alive and working with Quisk and a Softrock.

For those radio illuminati who lurk the reflectors such as EMRFD and uBitx --let's see yours! The ASUS Tinker Board works with Quisk and is a better alternative to the RPi3.

Pete N6QW

2/5/2018 ~ Some important Updates: Stay Tuned!

Here are the final schematic for the 40 Meter Driver Stage and the plot of the output. Note the 5 element Low Pass Filter in the final schematic. If you were to use this as a standalone output stage I would suggest using the 8 element W3NQN filters to give you confidence that nothing is slipping through. In my designs following the IRF510 stage I have the 8 element filter as the final filter

News Flash ...

Over the weekend my newly purchased Asus TinkerBoard arrived. This board sports a 1.8 GHz Quad Core Processor and  2 GB of onboard RAM. It is better than the RPi3! --  You tube videos and Netflix are amazing. Here is something more ...

Above is the latest Quisk SDR Software which has a built in test signal --look close and you can see the test blip on the screen. This is the year of transceivers at N6QW -- so one more in development.
You have to stay tuned for this --it may be a bust; but so far this may be one of the first SDR transceiver implementations using the TinkerBoard. BTW -- I was able to take most of the files that I used when I installed Quisk on a PI2/3 and simply loaded them onto the TinkerBoard. Some of the files required that you use a later version as explained in the latest Quisk documents -- but that went on without a hitch. I guess Linux is Linux.
Check my website and there is a link to the documents used with the Pi2/3. Once again -- N6QWmay be ahead of the curve. See if you can find any info using the TinkerBoard on any of the radio illuminati reflectors? I hope to update the documentation for the TinkerBoard as there are some unique changes that are required. But for now just drool!
BTW -- the TinkerBoard is very easy to interface. I plugged in my 5 inch  HDMI Touch Screen and boom it worked. With the Pi3 --there is a lot of hand-hockeying to get it to work. Music playback is very crisp.
Pete N6QW

2/2/2018 ~ 40 Meter Driver Board is Finalized!

This morning I soldered up all of the components that would go on the 40 Meter Driver Board including the Low Pass Filter. Success! We are getting close to 10.5 Volts Peak To Peak through the Low Pass Filter measured across a 50 Ohm resistive load. With the driver stage potted I will now start on the main transceiver board. With a bit of effort two Driver stages can be built on a single 4X6 board which likely will be a PC board versus the single sided copper vector board.

More info ... This afternoon I connected the output of this driver stage to my commercial amp which is a Communication Concepts Inc. AN762 --good for almost 150 watts output. With just the drive coming out of the 2N2219 --I was hitting over 30 Watts PEP. So this combo is now on par with the amp that was originally designed for the K1BQT project. I could even envision this small driver stage with one of the $15 amp kits being sold on eBay that are good for about 45 watts.

A new day has dawned and for very little $$$$ invested you can have a rig that is good for making lots of contacts.
Pete N6QW

1/30/2018 ~ 20 Meter Driver Final Stage Data.

Yesterday I posted the 40 Meter results and this morning using LT Spice I simulated what would be needed for the 20 Meter version. There are three circuit changes including the transformer on the output. Since we are bit higher in frequency the FT-37-43 had a few hiccups and burps to get it to peak on 20 Meters. That is when I went to the FT-50-61 core and with 16 Bifilar Turns put it smack on 20 Meters. I also changed the in/out coupling caps to 10 NF (0.01 uf). Having the Band Pass Filter ahead of this will insure just the 20 Meter energy is getting to the stage. This would then be followed by the 20 Meter Low Pass Filter. This could be a breakthrough module used in many driver applications.

Pete N6QW

1/29/2018 ~ A Breakthrough -Driver Finalized!

I attempted to use the BS170 as the final stage -- and while it looked good on paper it flunked the smell test. So I shifted gears and went to a conventional BJT to do the heavy lifting and we were successful! Boom this stage is potted.

While the sketch shows 12 Turns on the BFT --use 13 turns as that value produced the following plot from the LT Spice simulation

Notes on the schematic the center wiper on the 200 Ohm trim pot is connected to the 100 NF cap and the settings are proximate. Look at the output on a scope and find the setting for the best linearity. The Resistor R7 is a 200 Ohm trim pot connected as a variable resistor is in parallel with a 100 ohm resistor. These two trim pots give a coarse and fine setting of the output. Tune for max smoke and best linearity.

Finally we have a video. Even I am impressed at what we now have for the IC Transceiver. With the 20 Meter Driver we will need to put less turns on the FT-3743 Core. I will determine that value and post it on the blog.

I plan on building the LPF and will rerun the tests --but for now we have moved the needle.

Pete N6QW

1/28/2018 ~ Adding the Band Pass Filter.

I built the filter and it wouldn't tune up and the pattern looked like crap. I thought maybe a bad capacitor --but that wasn't the case. The problem was on boot up the Arduino was set for 20 Meters and the screen was not facing me. Then I had an Ahh Ha moment --the filter is on 40 Meters. A quick change of frequency in shirt order showed it was a solid design -- And I get out about 3 Volts Peak to Peak. That should be a good place to start when I finish off the BS170 stage.
Pete N6QW


1/27/2018 ~ Pete's Build of the Driver Stage.

Today I started work building the 40M version of the driver stage. This is a prototype build and thus quite large real estate wise. Ultimately this circuit will be built on the single sided copper vector board. I often build two or three prototypes before we "freeze" the design. Of these prototypes end up in new transceivers.

I just finished the first part of the build which is the substitute circuit for the 40673. I am very encouraged! The power out just from this stage was 10 Milliwatts. Follow on work will involve the Band Pass Filter. the BS170 RF Amp stage and then the 40M Low Pass Filter. Watch the video.
This is starting to get exciting.
Pete, N6QW

1/26/2018 ~ IC Transceiver Builders Report.

Nick, G8INE has done several things to take on this project. First he purchased a table top CNC Engraving/Milling Machine. Unlike me it did not cost him $250K. He is using one of the machines that can be had in the $200 to $300 range. Next he taught himself CAD and designed a SMD board for the IC Transceiver Driver stage. Finally he cut a board and wired it up (should say no wiring involved). The board will shortly undergo testing to verify the performance as indicated in the LT Spice Simulations. Here is Nick's handiwork.
Kudos to Nick and his board. We will report on his test results. Nick did comment to me that once he had a board that it was a lot easier to build the module. Now the icing on the cake --when he goes to build a second board he simply chucks up a piece of PCB stock, calls up the program and punches the start button. That easy!
I was a bit waylaid on building my board as my time was spent on the SBE-33 whose transmit signal sounded like crap. This prompted replacing all of the electrolytic caps in the HV power supply and the microphone amplifier stages. I suspected the rather unusual voltage regulation circuit may need a rebuild too. I did make one on the air contact and there was a report of FM'ing on voice peaks. That is a well known problem traced to impacts of voltage regulation on VFO circuits. Now if there was a Si5351 in there we would not be talking about that subject.
Thanks again Nick for sharing your efforts.
Pete N6QW

1/24/2018 ~ More Trips back in Time. The SBE-33.

Final fixes included opening the TR relay case and added a drop of De-Oxit to each of the contacts. Tribal Knowledge use the end of a wooden toothpick to apply a small drop to each contact. I also (using the low setting) applied a bit of De-Oxit spray into the Volume Control and Microphone Control. SBE recommends a dynamic microphone or a powered crystal microphone but avoid a pure crystal microphone. Note an Electret Microphone will not work with this directly --no bias at the microphone jack. I hope the OBTE have taken notes with their latest Iphone.
I hope to check into the Vintage Sideband Net tomorrow evening.
Pete N6QW

1/23/2018 ~ Inconclusive Results. Trial By Test.

I tried several simulations and the results were not what I reasonably thought they should be. First I changed the resistor from Gate #1 to Ground to 10K Ohms just as in the original design without changing the turns ratio on the Input matching transformer. Boom the output jumped another 5dB to something in excess of 30 dB for the 40 Meter version. Most of the heavy lifting is in the final stage with something close to 25 dB just in that stage. Without changing the turns ratio that 10K would look like 225 Ohms at the primary. So that certainly left a big question mark.
Several other configurations were evaluated that step down the 10K impedance using various simulations. A simple 10 Bifilar turn FT-37-43 was connected "step down" was tried. Using the bifilar approach results in a 4:1 --so I could make 10K look like 2.5K which then would mean changing the Gate resistor to a 2.5K Ohms and this would replace the matching transformer. That configuration resulted in less gain. So we are back now to testing configurations with real hardware and evaluate which one works best.
One option would be for the final configuration would be to "build per print" with the 10K resistor on Gate #1 as T1 does give a 4:1 so that the third winding results in a 10K value.
However to evaluate the amplifier block with actual hardware and a BS170, I will use the 2.2K Ohm with the matching transformer first specified as I will be driving this stage during the test and evaluation from an ADE-1 which has a Zout of 50 Ohms. I have the board cut so now I need to start adding parts. The evaluation essentially has two components --1) does the circuit work as an amplifier and the gain developed and 2) what works the best when connected to U7.
Pete N6QW

1/22/2018 ~ Not so Fast Batman --The Batmobile may have a problem!

For many years now it has been obvious that while asleep my brain keeps churning away thinking about problems/issues/concerns with designs I am working on at that time. It is both a blessing and a curse. A blessing because I may find a problem before it raises its head as a problem. The curse part is that it is usually around 3:00 AM and I am awakened --never to get back to sleep. It happened last night.
Sometimes when we chunk a project into modules --we miss something at the interface. This might be the case with the K1BQT original design. Below are two parts of the original schematics. Take a minute to look these over. Read on.

T4 on the output side of U7 is a trifilar wound transformer consisting of 9 Turns X 3. Thus two of the windings are connected bifilar fashion which essentially puts two of the windings in SERIES for a total of 18 Turns. The third winding, the output is stand alone with one end grounded and the other end that ends up on Gate #1 of the 40673 which has a 10K resistor to ground.
So we have an impedance transformation from the output of the MC1496 to the input of the 40673. So lets look at that transformation in light of turns ratio squared. Thus 18^2 = 324 and 9^2 = 81. We have a classic 4:1 transformation (For the OBTE 324/81 = 4). So the output impedance looks like 1/4 the value in terms of the input to the 40673.
You might want to Google Maximum Power Transfer Theory in relation to impedance matching. You get max juice when the impedances are matched. So now the Question would be what is the Output Impedance of the output of the MC1496 and what does the input impedance look like for the 40673. Not withstanding the actual values --there is a 4:1 transformation taking place. So we need to go down that bunny slope to assure there was matching taking place as that impacts what we do in the redesign.
In the Simplecever Plus Project I addressed this issue by forcing the input of the J310s' DGM to be 2.2 K Ohms and then did a match from 50 Ohms to 2.2K Ohms with a 3 Turn to 20 Turn Broad Band Transformer. Thus 3^2 = 9 and 20^2 = 400 resulting in 400/9 = 44.4. Now if you take 2200/50 = 44 --close enough. Thus 50 Ohms at any output will look like 2200 Ohms at the Input of the DGM.
Below is an example of how this matching was done on the Simpleceiver IF stage. The 1st match was 3 to 20 turns and the 2nd out of the crystal filter was 170 Ohms to 2200 Ohms thus a slightly different turns ratio but always presenting 2200 Ohms to the input
I have now modified the two driver amps to include input matching from 50 Ohms to 2200 Ohms
Some tribal knowledge here -- if you simulate these in LT Spice and add the inductor mutual inductance notes (K9 L8 L9 1) be certain that when you make that note to check the dot that says invoke Spice Command --otherwise you will get no output.
These changes added about 10 DB to the output. Now lets see if we can track down the output Z of the MC1496. Maybe K1BQT accounted for this in the Out/In impedances but given I am changing his design I want to make sure I have accounted for the matching.
You can scroll down this posting and look at the original plots. Matching is important!
When I get the output impedance of the MC1496 we can modify Transformer T4 on the schematic (if necessary) so that its output winding is 50 Ohms. Now we could be really clever and once we know the output impedance we could use a single transformer so that the primary Zout matches the Secondary Zin to 2200 Ohms. I prefer two transformers but it could be done with one.

So let us now check the MC1496 Data Sheet that can be found here. The Output Resistance for the Parallel Outputs is 40K Ohms. So if we look at K1BQT's transform it is 4:1 and thus 40K looks like 10K into the base of the 40673. Thus the design as presented is absolutely correct. K1BQT was spot on!

Now how do we match that into my 2.2K Ohms?

Well lets us start by looking at T4 so we would have a 40K to 50 Ohm match which is 800:1 and with only 18 turns on T4's Primary -- you would end up with less than a half turn on the secondary --that won't work. So now lets look at the transformer on the input of the J310s and what can be done with the primary side. A 10K to 2.2K transformation is a 4.55:1. So now if we modify my input transformer to have a 15 turn Primary and 7 turn secondary we have the following: 15^2 = 225 and 7^2 =49. 225/49 = 4.59:1 is close enough. So I will run some plots with this transformer.

Now alternatively we could change the input to 10K and see how those plots look with a source of 10K pumping RF into Gate #1 with a 10K resistor to ground.

This exercise has served the purpose of looking at impedance matching and how to adjust input / output impedance matching.

Stay tuned tomorrow for the plot data with the new input transformers.

Pete N6QW

1/20/2018 ~ Prototype board layout for the Driver.

I will start the build of the first prototype Driver stage. My process begins with using G Simple to lay out island squares and the cut it on my CNC. The layout facilitates parts changing, circuit changes and lots of measurements. Once I get a working circuit I will convert this to the single sided copper board which will be another prototype. Finally I will build two "production units" for 20 and 40 Meters. I have left space on the board to add shields should that be necessary. If one desires this could be done over a sheet of copper PCB using Rex's W1REX MePads.  For reference purposes the CNC will cut this board using a single piece of 4X6 board and pretty much fills the board.

That is the beauty of having a CNC Mill. Five minutes after I finished the design I had a board in my hand.

CNC Engraving/Router Mills can be had for about $200 to $300 on eBay. So within reach of many homebrewer enthusiasts. This  board is way larger than needed but since it is a prototype we will be doing lots of "peaking" "tweaking" and adjusting and thus the extra room

Stay Tuned.

Pete N6QW

1/19/2018 ~ Cleaned up Schematics for the 20 & 40 Meter Driver Assemblies.

I took some time to clean things up on the schematics so that the components were more readable and to include any coil winding data. The plots previously presented are essentially the same.

The 40 Meter version has slightly more gain but the minimum gain is 5 DB in both cases. Note that "R" on the schematic is a combination of a 10K trim Pot (Drive Control) where the top end is connected to the bottom end of the 22K resistor, the Center Wiper connected to gate #2 and the other end is connected to a 2.2K resistor which has its other end grounded.
K1BQT has a ferrite bead connected to the lead on Gate #2. Since my trim pots are board mounted I have not found this necessary. That said should you panel mount a pot then more care in lead routing and a ferrite bead may be required.
Caveat Emptor -- These circuits were simulated in LT Spice and not actually built which will be the next step. However, The J310's stage and the Band Pass Filters have been built and used in other rigs so these parts are known quantities. The substitution of the BS170 for the 2N7002 and the simple Low Pass Filters are new and thus will be evaluated. Should these two circuits work well then they will become a standard module that could be used in any rig design. I am hopeful this is the case.
I would encourage those with more sophisticated test gear than I have to build one of the circuits and run a Spectrum Analysis and share with us what you find.
One more thing: I used T-68-2 Red Cores for the 14 MHz networks. I invite your attention to the information in the chart below provided courtesy of Amidon.  Red Cores are OK at 14 MHz.  The Q's are good enough with the T-68-2 cores (Al = 57) . But if your shorts are all wadded up, then with the inductance values provided you can calculate the turns required for the type 6 (yellow) cores for the 14MHz networks. Al value for yellow T-68-6 cores = 47. 
Stay tuned -- this is the year of SSB Transceivers!!!
Pete, N6QW


We have earlier presented the reference document for this project and in looking over what will be kept pretty much intact versus design anew, the first piece to be designed is the driver stage following the IC lower level stages.
The basic LM1496/MC1350 pieces will literally go untouched. That said the driver, final RF amp and Audio Amp stages will likely get a going over. My initial thoughts were for a two band rig that will be switched by a simple DPDT toggle switch. This starting point affected what I did with the driver stage.
From K1BQT's design the driver stage looked like this:

In 1985 this was state of the art and we will build on the idea of a Dual Gate MOSFET driving a N Channel Enhancement Mode FET. Since 40673's are not like a stock item I will build one using two J310's. This is a proven approach and works perfectly! The IRFD1Z3 is like un-obtanium and so we'll substitute a BS170.  Touring Rick's design the 40673 can have an adjustable "Drive" level. We will do likewise with the J310s. The 40673 was followed by a Band Pass Filter and then to the IRFD1Z3 with a Low Pass Filter on the output. K1BQT was very careful to state that if you use bands other than 75 Meters you will need additional filtering.
So if we use just the J310's and the BS170 as a single amp we would have to be able to switch the Band Pass and Low Pass Filters which adds a lot of diode and/or relay switching.
So what if we built two complete separate driver assemblies and then it would only involve a SPDT relay at each end of an  Assembly to switch between the Assemblies and then a third SDPT relay to provide power to the Assembly being used. This approach does several things for us including making the switching a bit less complicated but more importantly each driver assembly can be optimized for the band being used. The added cost would be for a couple of devices (2X J310 and 1 X BS170) and a few resistors, caps and ferrite cores.

Today I took a crack at designing or should say simulating two separate Driver Assemblies. Should mention that the Dual Gate MOSFET Stage is a stock design that I have used in several rigs as is the Band Pass Filters. What has not been previously tested is the BS170. So after presenting this design I will build a prototype and run some tests. the simulated plot as a starting point shows great promise. 

Take a hard look at the schematic and note that the Band Pass Filter is terminated at each end with 50 Ohms. If you want a rude awakening remove the input side 50 Ohm resistor and run the plot. Do not remove this resistor.

Again these are the first pass and I will build a prototype to insure they works as deigned.

Stay tuned and start ordering the parts.

Pete N6QW