Busy... but...

I haven't done any work on any of my projects since the last post and probably will not be able to until Sunday night at the least.  I didn't even hit the wine that hard...

However, I did go off and order one of these:

...off of our favorite spot to buy stuff that "normal people" wouldn't buy*...  e-bay... For $140 shipped, how bad can it be?  60W iron, and a hot air reflow tool (which I've been lusting after for quite some time).  Coming from Omaha it's practically in my back yard to start off with.. :O)

I suppose I need to slow down on the toys.. but hopefully this will be a long term investment in the hobby.  My soldering irons have all been surplus up until now, and even though the Edsyn is a good second iron-- its UGLY... I mean UGLY... RE-F-UGLY..  So I'm hoping the iron on this is at least as good as the Edsyn is...

I've been deep in the insides of Microsoft Windows at work trying to put a square peg into a round hole in a triangular box that needs to be encased in a parallelogram.  So I've been doing late nights this week.  Maybe I'll start another blog about what I've learned about Windows Messaging in the future.. but probably not.  I usually reserve the Software Engineering for stuff I have to do for a living.  I rarely code for fun anymore.

I also managed to get control over my website, so when I actually complete some of these projects I'll do formal write-ups there.  I had not touched it in over five years... shame on me.. I guess that is what long hours at work, earning a MBA with Honors while working 40-60 hrs/week, having four young kids and having a gastric bypass and losing 150 lbs will do to ya...

Hopefully early next week I'll be blowing IC's off of old boards.  One can only hope.  If anything else it will be a good stress relief!

* I define "normal people" as people who can go into a shopping mall and be excited by all of the shoe stores there (and little else.. maybe an "Orange Julius")... I, on the other hand wonder why "normal people" need so many damned shoes....Or, as an alternate definition, "Normal people" are people who like watching sports exclusively as a hobby rather than doing something!  (I'd say something about athlete's and support... but I might offend...)

Ok.. on a more serious note...

I feel like high tech, 1977.  WOW!  Chip Caps.  "2001: A Space Odyssey!"  "Star Wars"  Not Episode III: A New Hope, just "Star Wars".

I forgot how nice solder mask is, and computer generated traces, etc.  I did get the AN-762 built in the last two evenings.  Not at all tested yet.  Tomorrow evening I'll do the bias regulator tests and then all of my stuff will have to be shelved until after next weekend as I have family commitments until then and I'm getting behind on the sleep again.

But except for somehow nicking one of the bias resistors it turned out picture perfect.  The nick on the resistor won't effect how it works, either.  The Toroidal transformer is very tight where the center screw goes through the 2N5190 and it's difficult to get it all to fit there without scratching something.  No biggie.

There are two layers of Kapton tape under T1 and T3, the input and output broadband transformers as required in the instructions.  The original design used coax braid and leads and because of that the typical copper tube broadband transformers can't be soldered at all in the case of the input transformer, and only the split single turn side on the Output transformer.  I mounted the input transformer upside down on the Kapton tape so I had something to solder the input leads to (there wasn't much if I flipped it the "normal" way) and to help keep the PCB ends from shorting to the board.  It looks weird but it should be good.

Note the 2N6488 pass transistor for the bias circuit mounted upside down to make the pin out the same as the obsolete 2N5989 ... the 2N5190 is on the bottom with the screw started through it.  The transformer (the toroidal one) is extremely tight there.. so that is why I nicked up the resistor.. probably when cleaning up the solder joint on the top of the one turn wire.   Be careful with this.. but you can't get the two transformers and the silver mica in much farther back than I have them so some mechanical slop will be needed here.  The 2N5190 needs to face down and thermally connect with the heat sink-- with silicon heat sink grease and be screwed down to the heat sink, so the screw here is unfortunately necessary.

Like I said, a test of the LM723H "tin can" regulator is next.  The "build hints" that come printed out with the kit (I have a scan of them if someone needs it.. e-mail me) state that the regulator can be pre-tested under a small load, say a 15 to 100 Ohm resistor 1/4W ... so that's tomorrow night probably.

Two more views... notice how tight the screw on the 2N5190/Toroidal transformer is.  I doubt if there is a good way around it.  Overall all looks pretty good though...

Yes there are two chip 470 pF caps in parallel with the 820 pF Silvered Mica cap.. it's pretty tight in there which makes having clearance for the 2N5190 temperature compensation diode (B-E junction of the transistor) pretty tough going...

After testing the bias circuit with the temporary resistor, I need to mount the MRF-454's and adjust quiescent bias as in the app note... Then I have to decide if I want to rig a couple of connectors on it and test the unit into a dummy load.  I could hook up the directional coupler and the 141T to it and see if IMD is in line before I build the filter bank too..  Same deal about power out.  I'd also like to get 5W or so out of the SDR-Cube to drive this to 100W+ so I need to build my single Class A RD16HHF driver too.. it is a big project making a 100W HF rig from scratch...

I also decided to try and bypass the 0.1 uF capacitors from the sr63ng with 10 uF tantalums to see if that brought up the audio level.  It did.  Not as much as I thought it would, but the unexpected pleasant side benefit is that the low audio response is remarkably better and it makes SSB and AM (especially SWL) sound even better than it did.  It's sounding great now.  Even though I probably only gained 6 dB more in the passband totally.. I can tell the response is flatter.  I'll leave it that way for now.

---

I got started late tonight as it appears that my XM radio subscription (which was last a three year term starting 3/8/2008, had them for 7+ years) was terminated 6 weeks early.. so I spent a couple of unsuccessful hours trying to get my radio reactivated via an online hit until it's up (and I won't renew) 3/8/11.. didn't do it even thought their computer showed it should work and the hits were getting sent... I don't want to call them either, except to cancel...  It seems that many of the horror stories about XM under the new "Howard" regime (Sirius) may be true.  Damn shame too as five years ago before the merger (thanks for protecting us Uncle Sam!) it was awesome and well worth the money... not now... monopolies suck--especially "entertainment" ones!  I suspect that if I call I'll have to re up or I'll remain dead.  I'll remain dead.. hopefully they won't try to reconnect and bill me and send it to collections like I'm reading is happening to many people online.  What a bunch of thieves....sigh...

'Till Next Time...

In one of those moods...

I'd like to work on the project tonight.. but instead I think I might drink...

Maybe not..  I swear the ARRL said this was a "mighty fine wine..."

I suppose to be mighty fine it needs to at least have a cork?   Back to the projects...

AN 762 Amplifier project-- mechanical beginnings...

Well.  Since the microcontroller is getting larger than planned, the TR relay is bigger than expected, and everything else is getting bigger and since I was frustrated with the 'Cube build trying to shove so much stuff into so little space...  I think the surplus 7"x9" frame will go back into the closet.  I'm going to spring for the similar BUD frame, but in another standard size-- 12" x 7" which happens to be the same size as the heat sink.   I think that the connectors will come off the short side as the back, and the controls and microcontroller display will be off the short side in the front of the amplifier.  What I'll do is form an "L" out of Aluminum for the connectors and controls and nibble out the frame when I get it to fit over the "L" that the "stuff" is on.  So the frame will be a bottom cover for the box and the whole thing can be built and tested on the heat sink.

This means I'll need to Digi-Key $30 more for the case and shipping but it will give me 1" more thickness to work with and the entire heat sink to mount stuff.  It will also be massively larger than it needs to be for even 100% duty cycle use.. but bigger is better here like in many things.. :O)

But because of this I decided to start mounting the stuff from the center of the heat sink out..

Here is a small green Sharpie with markings of the AN 729-140 Amp, the K5OOR HF Projects amplifier board and a VSWR bridge that I've built.  this leaves all of the aluminum on the right for the microcontroller.  The little bit on the left will have to house the 20 mW to 5W Class-A amp and maybe the TR relay.. but I could mount the TR relay on a bracket also. 

Even at 12"x7" it will be a little bit tight if I stay with one layer of boards.  I'd like to also put an Aluminum shield in between the amp and the microcontroller, too... that'll be on the right too.

Look good enough.  Lets drill and tap.  The only way to get this all accurate enough is to use a drill press, the #43 drill and actually drill through the holes on the board and transistors.  The "dots" from the Sharpie are not precise enough.  Take your time, and do one corner first.. drill and tap and screw down.  Do the opposite ("kitty") corner.. drill, tap and screw.. then do the remainder of the holes.  This goes for both the AN-729 board and the K5OOR board (which is really tight tolerance!).  For the RF transistors.. I partially drilled the holes with the MRF-454 in place and then finished them off with the transistor removed.  Be very careful when doing this.. but it's the only way to get precise enough holes with home equipment.   This is probably the most delicate and most important part of the build and is probably the most important part to get right for the amp to work.

There are five holes-- the standoff holes and the D1 hole to drill for the PC board.  Then there are the four screws to secure the MRF-454's to the board.  All need to be precisely drilled and mounted.  Take your time and use plenty of tap oil when tapping Aluminum.

The K5OOR 100W LPF board has 4 mounting holes also that should be drilled/tapped.  These are extremely tight tolerance (the 4-40 screws basically take a little of the solder off in the plated through hole they are that tight) so get these right too.

Here is what is should look like when done.  I'll save anyone building this about $20 at the hardware store to show you what is needed for the perfect fit needed for both the D1 bias (thremal runaway preventer) transistor (used as a diode, B-E junction is used).. and the proper 1/8" spacing for the MRF-454's.  I could only get nylon for spacers.. and they didn't fit well.  I bought a whole box of 4-40 Stainless Steel flat washers.. and low and behold.. 4 of the under the board spaces the board perfectly  for both D1 and the MRF-454's.  (D1 mounts in the middle of the board.. it's just sitting at the edge to show the board spacing fit!)  Of course, the 4 washers under each screw are impossible to mount underneath unless you bind them together somehow.  A drop of Krazy Glue brand super glue (cyanoacrylate ester) carefully applied when four of them are on a 4-40 screw works well enough to hold them together for placement under the board.  This means that a total of twenty-one stainless steel flat washers-- including the top washers- are needed for the AN-729 board alone (buy a box of 100 from Ace Hardware if you are getting them locally--it will be less than 25 at piece rate).   There are 25 of them in this picture on both boards.   The D1 2N5190 transistor is maybe 1/64" shy of 1/8" but I figure that you could either mount it flat and let the board bend inwards a little (what I'll do) or put a washer under and over and have the board bow out a little.  D1 needs to have silicon heat sink compound on it and have the metal tab side down towards the heat sink.. but securely wedge with the center screw... drill and tap that hole as well.  

The Pass transistor Q3 on the bias regulator circuit (pass transistor for the LM723G or MC1723G regulator IC) does NOT attach to the heat sink.  DO NOT drill a hole for it.  The original 2N5989, 2N5990 or 2N5991 transistors that were suggested for this design are all obsolete (and are available, but from $7-13 each!).  There is no newer transistor that has the odd-ball pinout that these transistor had, so a 2N6488 TO-220 cased transistor is substituted.  The transistor needs to be mounted upside down and hardware is supplied for that but nothing else.  I wish the 1/8" spacer supplied for that could have also been supplied for the four standoff corners, but CCI clearly doesn't sell these as kits, but boards and parts.

The screws needed are quantity 4 - 1/4" long Stainless cap screws (cap screws are needed here) for the MRF-454.  The board screws are quantity 5 - 3/8" long Stainless cap screws ...  washers on top.

On the K5OOR 100W LPF.. the mounting is 1/4" aluminum spacers (Ace Hardware) with 1/2" long 4-40 SS Cap Screws and a 4-40 washer on each.  I had to go to two different hardware stores to get enough SS cap screws to do this job.. so some are from Ace, some are from True-Value.  True-Value is cheaper.. Ace had more selection.  :O)

Here are a couple more views of the test fit.  It's vitally important to get this all correct before mounting parts on to the boards.  You should spend an afternoon doing this alone....  maybe two...

I also fixed that old power supply.. and added +3.3V/0.8A LM1117-33T to it this weekend see below:

I also built a TR relay based on the one in the Communications Concepts EB-63A amplifier.  I probably will only use this particular circuit for the initial use.  Eventually I'll have PTT in and Microcontroller based RF "VOX".. but for initial build and testing a good RF-sensed relay will be handy.. it was only about 1 hour of work.  Two more Teflon RF cables need to be soldered on.. but some future evening for that... 2N3904's were substituted for the 2N4401's and 1N914's were used instead of 1N4148's.  I didn't have a 1.5K resistor handy in non-SMT, so I used two 3.3K in parallel.  It's all not that critical.  I actually think the SSB delay could use a bigger delay.. maybe a 33 uF cap instead of the 22 uF electrolytic?  We will have to see in actual use.  There are 0.1" headers and jumpers for the SSB delay and from the diodes (so that the RF sensing could be bypassed and a +5V keying could be used instead...)  Break is at D3 and C11..

So that's it for this weekend.  It didn't seem like I did a lot until I sat down here tonight and documented it all...

Chinese #$ap!

Ok.. I'll take much of what I said about Microchip back.  I hooked my PIC32MX protoboard up though the same cable that hooks to the ICSP ZIF board.. and noticed that all of the voltages were about 1/2 of what they were when hooked directly.

I thought.. naw, 4 inches of cable can't drop that much.  So I bugged the cable.  The green pin (which I used as ground in my tests) was not properly crimped on one end of the cable!  So.. that explains the issues I had with the PIC16F871.

I crimped the pin back down with needle nose pliter and flowed solder into it on both sides (since I didn't know which one wasn't connected and obviously the insulation on one side wasn't stripped off before crimping.. so melting it and flowing with solder was the only way I could insure a good electrical and mechanical connection since I didn't have replacement pins) and now the cable is well.  I still see more voltage drop than I'd like in the charge pump (so it still sucks IMHO) but it seems that most devices will now work even with a little supply sag.  I now have a working PICKIT 3.

So there you are.  Typical Chinese crap that isn't tested before it's sent off on a month long journey.  Grade: C- at most on this for Sure Electronics.  The fact I ordered 12/23 and they didn't ship until 1/7 (according to the HK Post) and then it had a defective cable with it would definitely make me think twice about buying from them again.

I bought most of my 0805 chip caps from them in an assortment previously.  They are fine.. all I've measured were accurate.. but it took over three weeks to get those too.  Most Chinese sources on E-bay are much faster.  having a defective test cable really sucks though.  I can't get back those three hours of my life!

So there you are.  Now that I have it working it's probably better than the real thing (since the ZIF socket thing isn't an option from Microchip and the assortment of cables is nice too.. when they work...).. but they obviously don't test the stuff before they ship, so you are taking a risk you'd wouldn't for $20 more with the real thing.

Oh, SURE!

I've been pretty busy this week.  I made up for the lost day at work and then some.

I did play a little on Tuesday night and again tonight.

I received the following toy on Tuesday:

This is a clone (exact, it takes real Microchip firmware) of a PICKIT 3 programmer.  It's from Sure Electronics in Hong Kong and.. well it's available on everyone's favorite site, e-bay.  Overall it's an exact clone.  Having said that, it's a typical Microchip programmer.. could have been great if thought through a little better, but it is not quite there.  Most of the complaints you'll hear about the PICKIT 3 online are true.  But I needed something to program PIC32MX's as they are one of my favorite choices for Microcontrollers now as they are very powerful, not very power hungry, have lots of I/O and are as cheap or less than the 8-bit ATMEL's* (like the one I built below.)  The 3rd generation programmers like the PICKIT3 and the ICD3 are the only ones that do the newer PIC32MX's.  They also program the DSPIC33's like the SDR-Cube uses, so since my older serial-port and parallel port based programmers couldn't do these newer devices... I decided to go for it.

The Chinese clone was nice as it's a good $25 cheaper shipped... and for a couple of bux more you can get that nifty ZIF socket board for the older 5V DIP 8-bit PICs.  But unfortunately, do to the fact that the charge pumps suck in the PICKIT 3 design... not all will work while internally powered.  So you'll see in a future post me redoing a simple 7812/7912/7805 based supply that I built for small microcontroller work in the late 1980's to also add a TO-220 based +3.3V supply to it for use with the newer microcontrollers.  What I am saying is that the charge pump is really only good for the 20 mA that Microchip claims.. and some of the older devices are so close to that hairy edge in normal draw that they don't work in self-powered mode.  Life will be better when I stop trying to power the devices with te charge pump on the PICKit3 and instead use external power.  It's supposed to be an "in circuit programmer" so one (probably a marketing guy from Microchip) could say that the internal power is a "bonus."  It's really a "bogus" in actual use though.

I tried a PIC 16F877A in the socket with the PICKIT 3 self powered and it works fine.  A PIC 16F871, basically the same series.. draws too much current.. the supply sags to 3.5V and the device ID's okay, programs OK, Erases OK but can't read back!  If I hack a cable to provide external 5V (and 3.3V) to that ZIF socket board, I assume all will program fine.

There are also several older devices I use (like 12C and 16C devices) that the PICKIT2/3 can't program either..  so I'm disappointed that I can't chuck my older 2001 era serial port based programmers.

Anyway.. I read the SDR-CUBE's data in with it OK (which is externally powered, of course), a CUI32 board's (PIC32MX) data being powered from the PICKIT 3, the 16F877A was programmed and read back.  And tonight I got my e-bay special board ($1.65) from Thailand (etteam DSPic/PIC32 protocard) and a sample PIC32MX360F512L (which has the same pinout as the DSPic33 on the SDR-Cube, but unfortunately the wrong pitch at 0.4mm instead of 0.5mm) to program.

This bugger gave me fits.  The trick here is that ALL Vcc and Vss lines need to be attached-- including AVcc and AVss-- and most should be bypassed with 0.1 uF.  But in addition.. on the DSPic's and PIC32MX's the cores run at 1.8V on these 3.3V devices.  Microchip put on an internal regulator.. which is the squirrliest one I've ever used.  Follow Microchip's advice on this.  10 uF 10V or greater Tantalum capacitor.  Don't try an electrolytic.. even a low ESR one.. and don't try only a 1 uF Tant.. it won't work.  It looks like that regulator rings on startup with all but the correct 10 uF Tant...  It was muy muy bogus figuring that out.  I suppose if I had 50 10 uF's instead of 1 uF chip tantalums in the surplus section of the junk box I would have not figured this out.  I'll have to get me some from e-bay China :O)  I'm using a leaded tant tack soldered for now on this board.

I'm starting to wonder why I use DIP IC's for anything now.  There are easily 2x the IO lines on that board than the ATMEL board I'm using for the PA project.  This sucker has 2x the memory and 20 times the processing power, consumes less current.. has many 5V tolerant lines on it... I almost wonder if I should use it instead of the ATMEGA.  Not necessary as there need not be much processing power.. but now you see that you can use the most modern processors in your design.  They are faster, cheaper and better than the older processors.

I'd still like to port the SDR-Cube code over to this processor.  I have some TI CODEC's that I sampled around here.. and could probably make a "lite" version that's controlled over RS-232 instead of the very fine GUI for proof-of-concept in the port.  I'm sure this will take me awhile.  But I think the performance will be better than 2x the throughput of the DSPic33.  

I'm very tempted to sample off a MX7xx series part, a and a PHY, and try to get Ethernet running on my protoboards.. but I'll worry about that in a few months time.  The Lite MX SDR-Cube streaming UDP over Ethernet seems like an interesting concept, tho.

Case nibbling revisited --- More AVR fun

George Heron sent me a private e-mail today that I think is worth mentioning.  I don't think, in my case, it would have made a difference because of the big variety of 1/8" stereo connectors I use.   I still think opening the slot up in the back of the case was the appropriate thing to do for me and the future diverse uses of the SDR-Cube.

HOWEVER, you might want to try George's suggestion first.  That is moving the I/O card PCB farther back in the case by sliding it back as far as it goes (I did that actually and it wasn't enough) and possibly by elongating the four mounting holes (I didn't try that) on the subchassis to make it go farther back into the case.  If you get the metal rims of the phono jacks a little past the outside of the case, then the connectors will not interfere.  But you will no longer have the +12V coaxial power jack flush with the case either.

I think, personally, I like the connectors slightly recessed which means nibbling the case.  But the other option is valid if you have issues once you assemble the cube.

-----------------------

I had a "snow day" today.. we got about 4 in of snow out here in the great white North of Eastern Iowa.. I drive a front wheel drive car (my 4WD 1998 Explorer has a coolant problem and I've stalled for about a year on deciding to get it fixed or not.  I really want a full size 4WD pickup, used, to replace it.. and putting in $2K into a vehicle worth 2K?!?  I dunno...).  Anyway, had that thing still be running I would have had no issues.. but it was just bad enough today with the VW TDI that I thought "Vacation Day."

I did a couple of hours worth of work on the AVR board.  I decided that to band switch the amplifier I would use a 74HC138 3-8 demux (I acutally used an LS part).. then I discovered for the positive drive I needed for the 2N3904 transistor switches for the +12V coil relays.. that the '138 had the wrong output logic polarity (active Low)... So I figured, OK, I wanted LED's for a band indicator anyway.. I added a 74HC04 (actually another LS since I had it in the junk box) to the design.  I'm doing point-to-point wiring on that protoboard with scrap Teflon wire.  It makes you appreciate the invention of printed circuit boards.

Anyway.. I did get the whole thing working while taking only 3 of the lines from Port C (after 30 minutes of fiddling I remembered to set the fuse to disable JTAG on the part so there WAS a PORTC!) of the ATMega644.. since we only want one band selected at a time.. that is a hardware way of doing it.  Also if 12V gets to the logic it'll blow a 74LS04, a thirty cent part rather than the $8 AVR.

The board that is (temporarily) under the LCD is the LED band indicator (six LEDs).  Blue, of course!  :O)

The missing ground wire on the connector mounted to the LED board will go on a second connector for a couple of pushbutton switches.  I've not quite decided on how many.. 2 or 3.. I need to think through the GUI for this.

Only attempt this sort of point-to-point wiring if you can get a hold of Teflon wire.  PVC would have been a smelly melted mess:

Now I don't have room on the protoboard for the MAX232 driver for a serial port.. I'll have to figure that one out in a few days when I get to work on it again.  This is part of the reason I really like SMT on finished circuits (not protos like this one).. there isn't that much circuitry there for the size of the components which are quite literally huge for what they do.

'Till next time... it may be awhile since I'll need to make up some time at work now.  No snow in the forecast for the rest of the week, but highs in the single digits Fahrenheit.  I don't think human beings are meant to live at this Latitude...

ATMega board.. piece of cake...

I said it would be trivial to get the board going for eventual inclusion into the amplifier.  Only a little more work than using a bargraph display or a rotary switch and it has a cool blue display!

Had to change my "stock test code" to a different port on the ATMega644 and also burn that funky fuse that needs to be burned (lfuse = F7 on the '644) to use the XTAL osc.  I normally use 18.432 MHz but I'm  out.. had several 14.7456 MHz crystals, so I also changed the clock rate to that.. plenty fast for this application.

Anyway.. here is proof that it's pretty easy to do.  Code is in "C" I'll document and put it somewhere safe later.  The 16x2 module is a Chinese clone HD44780 (made by "Tinwell" TC1602a off of e-bay... more expensive but reliable American source).  The AVR programmer I use is a usbtiny, from that more reliable American source.  The board is a Sanguino.cc board.  I bought several blank boards, which seem still to be available (at a higher cost than I paid a year ago, of course!).  Now there also appears to be a complete kit for $25.  I got ATMega644's from Mouser (who had some in stock at the time.. see my comments about ATMEL and AVR's in a previous post) and later I got some ATMega644P's from Digi-Key.  All the rest of the parts are junk box.  The 644P's have two UARTs on them.. if I determine a need I can replace the 644 with the 644P.  These are "bigger, more powerful" Arduino clones, and even though I prefer tighter code written in straight "C" with WinAVR, Arduino is easy if you are not a programmer.  40-pins, simple as sin to program, lots of memory, better peripherals (and more memory-- did I say more memory?) than a PIC in the same class.  Not much not to love about them.  Big 40-pin DIP, easy for prototypes or one offs, and the code is compatible with chips in more modern PLCC packages for production.

SDR-Cube.. done, but just the beginning.

As I type.. I'm listening to Arnie Coro CO2KK on Radio Havana 6.050 MHz.. blasting in... like usual.  The SDR-Cube sounds really good on SWL stations...

I spent about 3 hours today verifying that the 20/30m TXPA is functioning.  It is.. at 13.0v it put out about 1050 mW, about what was expected.  I suppose I should have taken a picture of the Wattmeter, but I did it out in the shack in my detached garage, the 20m band was starting to die with sunset, so it wasn't worth heating it and attempting QSOs.  IT WAS COLD out there.  Oh well.. Next weekend.

When I tried setting up everything outside the first time I couldn't hook my other set of amplified speakers up, nor my Bencher because the slot was, again, too small to put certain 1/8" stereo connectors through to the connectors on the PCB... even when it was mounted flush with the proper screws (which again were BLACK ones I added.. :O) ).

I, quite frankly, am tired of that.  I want to be able to use any molded 1/8" stereo connector and the very fine gold connector on my Bencher paddles.  So I decided to do it.  I got out the metal nibbler again, a file and a Sharpie.  It turned out okay.. and now I have no issues with any of the connectors I use.

I widened up the slot one nibble on each side, but the side that meets with the PCB is the one that needs to be wider.  I'm recommending that everyone do this or you will have to be very picky about the type of 1/8" stereo jack you use with the SDR-Cube.

I then verified the rig worked on TX, listened around on 20m on a good antenna.  It works well, but I still think there could be more gain on the line output of the CODEC (i.e. an Op-Amp following it, since I have to turn up all of my amplified speakers up pretty high) and in my low noise environment, I think more AF gain into the CODEC also.  I think maybe another Op-Amp stage with maybe 10-20 dB of gain, but configured as another 0-8 KHz LPF would work.  I experimented with more RF gain but if I do that I get significantly more "blow by" of  SWL stations.

Anyway.. I will only briefly comment on the mounting of the X-LPF.  You all know by now what I think of things that are not screwed down, and the X-LPF attaches with solder.  Also the instructions say to solder it to the connector and then use the connector tab to solder to a landing on the PCB.  This only works, however, with the supplied connector if you cut off the Teflon spacer (cut it nearly flush with the connector edge).  Then it will reach the landing.  You have to be careful not to short the body of the BNC against one of the filter caps, though when you do it.

Also, on my Cube (which probably has a slightly longer stand off set on the sr63ng than came with George's kits on the "real" Softrock 6.3's),  I had to flip around the heat sink on the plug in to be able to get it on and off of the sr63ng because of the awkward placement of the X-LPF-- which is in the way.

Here is a picture of it.  It's ready for a 20m QRP CW QSO right now.  I'm going to try to also wire up a MIC and a PSK-31 connection to the Cube in the next few evenings.

It all seems to work nominally now.  Not horrible for a project of this scope.

Overall.. I like the rig.  If I had it to do again I probably would have only bought the boards and not the case, and neither cable kit (I bought both).  But the case is cute, and it will make for a GREAT demo at the local ham club meeting..

But it really doesn't fit together that well.  For a Ten-Tec case, I am rather disappointed by the lack of a "D" punch on the BNC and the hack job on the DB-15.  Greenlee makes punches for both that could have been done by hand if anything else.  The plug-in philosophy of the Softrock 6.3 has been discontinued by Tony Parks, and for good reason.  The cabling and the plug ins will require maintenance because they will come loose.  I gave some suggestions on how I partially mitigated this.. but I can't fix everything that is floating loose in this case.

Putting the boards out flat would have been better.  I am not the slightest bit surprised that the first working cube was a board only kit because many hours of fiddling, mistakes, rework, swearing would have been avoided by me if I'd not have built it into the cute little case.

If you want a little CW and PSK QRP rig that is the size of an oversized Rubik's Cube.. then it's worth all of the effort to put it into the cube case.  This, admittedly is Midnight Design Solutions normal customer base, and I can not fault him for "cubifying" the rig.  The QRP crowd loves this sort of thing, and for their little portable uses it is an excellent fit.

But for those of us who want to use it for a building block for something bigger the mechanics of the cube enclosure will get in the way.

I think the CODEC needed an op-amp on both Input and output.. I don't like almost anything about the I/O board (including no op-amps for the CODEC!).. I may redesign this card in the future.

The software is excellent.. the latency is controlled better than any SDR I've ever used, and the audio is super clean.  You'll be proud to put this on the air.  CW is pleasant with the rig, and will get better with some software tweaks (like being able to reverse the paddles they are set up for a lefty by default!)

AGC would make the rig unstoppable but that may happen with software in the future.

I'd like to potentially try undersampling a HF RX with a filter on a IF.  More gain could be used and the blow by issues will go away.  I think the Softrock IF is a weakness for this rig, but still it works remarkably well overall.  With only an 8 KHz IF bandwidth, a "roofing filter" at an IF and then undersampling probably would work much better with this design.

 I think I know what I am going to do to multiband this rig while leaving it into the Cube package.. which is different than my initial ideas.

I'm going to build a 20 mW output broadband (not filtered) PA plug in.  That will go out the DB-15.  On my PA will be a 5W Class-A RD16HHF amp running into the AN762 with filters controlled by a microcontroller.  Then I'll try to come up with an all band RX-AMP at the same time in the 'Cube and then the Cube will be a 20 mW (unfiltered) receiver/exciter.  The filter and band switching on TX will happen in the other box.  On RX, a multiband RX-AMP arrangement will do the band filtering.

The problem with putting 5-10W in that little 16 sq. in box is dissipating 10-30W of heat in it.  Not worth messing with IMHO.  20-50 mW out driving a 5W driver makes much more sense.

As for the Microcontroller.. I did a little preliminary work on the ATMega644 (or 644P I've not decided if I need the second UART or not yet) board that will go into the 100W amp:

The green board already has the contrast pot (it works) on it and the power to the LCD.  I need to use a different pin out than on past AVR projects because I want to leave the analog inputs alone instead of using them for the LCD.  So I haven't wired up the data lines or ported my LCD driver code over to that processor board yet.  Should be fairly trivial to do.

"Build this Solid-State Titan..."

So states the June (pp. 27-31) and July 1977 (pp.27-29) issues of QST.

Well.. I received my Communications Concepts AN762 kit.. I also decided to buy another heat sink, too so I can still complete the WA2EUJ amplifier from the 40m amplifier homebrew competition using the vastly inferior (and I'm not kidding for the price) RF Parts heat sink.  That amplifier uses four Mitsubishi RD16HHF RF MOSFETs in push-pull and parallel.  I have that 2/3 built but really I wanted a full 100W for the Flex-1500, the SDR-Cube and all of my other little QRP rigs I have lying around now.

One of my main goals on the SDR-Cube project is to have my own completely home built (rebuilt in the case of my SB-200 amp.  I'm counting it anyway) station.  I've wanted to have home brew HF be a primary (if not the primary)  station since day one of being licensed.

Anyway, the 33.5 year old AN762 (K7ES/OH2ZE) design is still as elegant as the day it was designed and really, it's the grandaddy of all of the newer designs including Push-Pull FET designs.  I've wanted to build one since I saw it in the first ARRL Handbook that I owned after getting licensed in 1986 (1986 Handbook).  I didn't have the money, or skills back then and I also didn't have any thing to drive it with...

Fast forward 25 years, and I have all of that now...

I bought the AN762-140 version, which uses the less expensive, and slightly less power (nominally 140W though I consider the amp to be a 100W design with these finals) rather than the AN762-180 version with the $50 more expensive MRF421's.  I also decided on the MRF454's because of the slightly better gain out of those devices which should better match the QRP rigs I have and the stages I plan on designing for the SDR-Cube's Softrock.  (With little actual noticeable on the air difference, 1 dB isn't worth $50 more for the MRF421's either.)

Anyway.. the Picture of the AN-762 kit is below:

I also bought CCI's larger heatsink, recommended for this amplifier, but probably big enough for the 300W plus amplifiers.  This thing is big.  Granberg suggested that 7.9" of an equivalent (AAvid Engineering 60140) is enough for the 180W version to run 100% duty cycle, CW at 40 deg C ambient.  This thing is the same size with the same fins but is 12" long.  I have a "frame" or an aluminum box that I have surplus that's roughly the same size but only 9" long.  That's still 1" longer than the original Granberg article suggested, so I'll probably cut the 3" off of this beast!

Needless to say this heat sink is bigger than any 100W amp I have on a rig is.  Here is an opportunity to build up something better than I can commercially buy so I am excited.  Usually one can only do that with antennas any more.  (The Heatsink is a BARGAIN for the price.)

The frame I'm going to use is below.  $2 from a surplus store a long time ago.  I'm going to cut a hole in the top where the AN729 board mounts so the MRF454's contact the heatsink directly. 

The dots are a test for mounting.  Those holes, plus four more for the MRF454's themselves need to be drilled and tapped (I think 4-40).  The thicker line is the 8" that Granberg suggested as the minimum for the 100% duty cycle (with no fan at 40C at 180W also) the thinner line is the 9" that the surplus frame is long.  I suspect I'll cut it off and save the other 3 inches.. otherwise I'll need to have 1.5" hanging off of either side.. which would be OK but ugly.  I'll have to decide soon.  With heat sink, bigger is better, tho.

Oh.. all of the toroids on the CCI amp are professionally machine wound.  They look beautiful and I don't have to wind them!  Yay!

I'm also going to mount the very fine HF Superpacker PRO LPF board that I received today into the same box to probably by controlled by an 8-bit AVR that I picked up several boards of ("Sanguino.cc" and the ATMega644 and ATMega644P 8-bit devices.. but I don't typically use them as an Ardiuno preferring to do real "C" instead.. since this is all simple slow control.. the 8-bit devices should be A-OK for this).  I've also homebrewed a VSWR bridge which is pretty simple to do for a power meter and VSWR protection... which will feed into analog inputs on the AVR.

Anyway here is a picture of the LPF board from K5OOR:

It is pretty nice looking with filters designed for the AN762 power level (The Superpacker Pro used basically the same RF deck design as this amp).  If I were to design a filter board, it would look just like this one (but I might have done the 12m/10m bank as normally closed with all the rest normally open in case the microcontroller failed for some reason.. but that's the only thing I'd do differently... This board was designed so a rotary switch could be used instead, so that wasn't an option.)  The filter kits that come from CCI are 300W rated, $14 a band and really too physically large.  This was about $20 cheaper, appropriately sized and has the switching relays for me.  Like I said it is perfect for this project.

Did I say, Yay! earlier about not having to wind Toroids on the AN729?  Well this will make up for it.  In the case of a filter bank like this at power, there is really no choice but to use hand wound toroids.  Ideally (and since I can because I can get access to test equipment) tuning and flattening the pass bands of the filters is a good idea.  So it's hand built and hand tuned anyway.  Toroids with "non-standard" values and hand tuning is needed in this design, so my complaints about being Toroid-ed to death by ham projects do not apply here.  K5OOR has specific instructions for winding that differ a bit from my standard way of winding and has L/C meter and other checks in here that should help, too.

I'm thinking I'm going to build this entire project "backwards" and build and tune the LPF's first.  This way I won't be tempted to run a poor IMD amp on the air even for "testing" purposes.  :O)

Then I will do all of the mechanical work on the amp.  The amp electrical build itself is very simple.  The mechanical work is the critical part (as with any power electronics) and will take the longest time.

I'm a little busier at work and at home now (as expected) so I expect progress on this and the SDR-Cube to slow.  I need to verify TX on the SDR-Cube tomorrow and maybe (if TX works) try it at QRP level on the air.

But this is like my 6 year olds at Christmas... I don't know which toy to play with first.  I did want to inventory and document all the stuff for the project.. but in addition I built the VSWR bridge I'm going to use with the AVR for power output indication and shutdown of the amp (VSWR protection).. just because I found all of the parts in my junk box for it.  (It's a simple "Stockton" design from the W1FB notebook, but I used SMT caps and a SMT Schottkey diode rather than a 1N34.)

I suppose once I test it (tomorrow?) I'll take a picture of it.. but it's a pretty standard design.  I made it on a smaller hand etched (with a Dremel and diamond bit) board that should fit near the amp deck and the LPF.

I do need to wrap up the SDR-Cube (for now to the "version 1.0 state") first though... so many toys.. so little time...

TTFN...

RX issues solved more?

Well.. after talking with George Heron by e-mail, he was thinking that I probably still had issues with the SoftRock.

Some good diagnosing information was measured by George.  I did turn out that I was -11 dB from his Softrock 6.3 RX TX with my sr63ng.  I went through and verified that I had strong LO from the 74AC74 divider and then.. well got fed up and reflowed all of the RX stage components.

It seemed to bring up the gain to within a dB or two of George's unit.  I need to verify with at least one more test that I have it A-OK quantitatively (measure with the signal generator and the oscilloscope again).. but the initial test George proposed (50 uV "S9" or -73 dBm @ 50 Ohms) should just peak the bandscope with the RX Atten setting at 2.0 dB.  It does do that now.  It seemed about 12 dB off before from the units' own RX Atten scale.  Measuring 11 dB on lab equipment (8640B and a Tek 2235 Scope) when I guessed 12 dB from the rig and George's test description is a testament to the care taken in the design of the gain distribution on the rig.

I should also say that the late evenings I've been working with the radio.. have had pretty bad band conditions..  so I've done some SWL with it as well.  Of *any* radio I've used LSB/USB to listen to AM on this one sounds the best.  This includes (IMHO) the Flex-1500 and Flex-5000.  There has been a lot of thought in the DSP in the SDR-Cube, and it shows.  It is REALLY clean.  Once it gets some AGC, it will be unbeatable.

I also ordered off the CCI AN-762 amp I posted about previously.  So the beginnings of an 100W+ amp that I've always wanted to build for all of my QRP rigs is coming in 3-4 days.

I'm going to design a replacement plug in card for the TX PA (PAF) that is broadband and more like 3-5W out to the amp.. I'll get there eventually.. I don't think I could handle a kitting operation like George but I will make the design open.

Ultimately I'd like to design a sr63ng type rig, but with a PA and filters on one board like the MOBO.. but without all the plug in's.. maybe I'll get to in in 2011.   Maybe not..

I'll try to tackle the "Electronic Origami" that assembling the Cube is and test the 20m/30m PA I built in the next several days.

I did put a right angle connector on the I2C connection on the DSP board.  I'll upload a picture of it later when I screw all of the fun stuff back together and start packing up the SDR-Cube.

Until next time..

RX issues solved.... Rig Receive is ALIVE!

Well.. After studying the schematics for everything, I decided that I really wanted as much gain as I could get out of the RXAMP.  If you recall, there are three options for gain.  I initially went with the "middle option" which was 18 dB of gain, 22.0 Ohm bias resistor.  I decided to instead replace it with the 10 Ohm 24 dB gain option.  What I got was a lot more than 6 dB of gain.  The radio came alive.  I mean from nothing to "Hey there is RTTY Contest going on!  WOW!"

I wonder if I didn't fix something else with it when I did that.  I'm almost wondering is the 22.0 Ohm resistor was bad or there was a bad solder joint, or whatever.  The rig needs more gain in RX, probably still!  So I highly recommend that all builders use the 10 Ohm resistor as is in the schematic as the original value and do not substitute gain down/resistance up based on the values in the table.  The front end attenuators will fix any overload situation.  The higher gain is desirable.  I think most computer sound cards have more gain (like an amplifying buffer) in front of their CODEC's at audio than the DSP Board in the SDR-Cube does.  Thinking about this even slightly, it make more sense to distribute the gain at RF (until the point of saturation, though) than at audio since it's so bloody easy to instead pick up 60 Hz (or 50 Hz for you Europeans) noise instead with big gain at audio.

This is actually good that the Cube designers did that.  What this means for external SDR front ends is that either more RF or AF gain from the rig will be needed though.  It's a simple hack to hack in the RXAMP to the Ensemble II RXTX for example.. I'll have to document and post that at a later date.

Subsequently after I it "came alive",  I've tested the RF/Antenna path all the way back to and through the PAF-3 (TXPA) and it seem good.  I also calibrated the si570 oscillator (after the video was taken) against WWV @ 10 MHz.  It was about 7.5 KHz off... which is about the worst si570 I've ever used.  But it's all good now.  I suppose getting forced to become intimate with the sr63ng wasn't a bad idea in the grand scheme of the project...

The audio on the video is pretty low.  I would have still liked more audio gain (and will probably add it in my cube) but it is not nearly as bad as it looks in this video. I don't understand what was up with the camera.. it was loud enough to hear in the next room.

I also took a narrow band CW video.. but the audio is even lower.  I'll do that again when I button up the project in a few days.

One thing that takes getting used to but is not nearly as bad as you would think is not having AGC.... You'll find that you think the rig has less gain than it does.. because most rigs will crank the gain way up with no signals.  Once you start using this rig.. it's not that big of a problem.

I still have to do I/Q balance on RX and TX and do all of the lab measurements I've been threatening to do so there is still several more weeks of entertainment on the 'blog for this project.

The Cube also does OK listening to SWL broadcasts beating against the carrier in USB/LSB.. I'd still like to figure out a way to do real AM though.. that will be difficult in the current 8 KHz bandwidth.
I'm thinking that PIC32MX touch screen platform with 16 KHz or better 22.1 KHz could fulfill the promise of a complete HF IF.  This is close and will be very nice for CW ops and OK for SSB..
That could be a next phase for this "small embedded no-PC" SDR concept.  Of course, an Android tablet has crossed my mind for a platform as well :O)

Integration. Always the killer...

Well.. The cube survived me throwing it against the wall repeatedly (ok I only pictured this in my mind.. but you get it... didn't do this literally even though I wanted to at several points...)

I first built the 20/30 TXPA (or in original Softrock parlance, PAF-3).. and except for those #$#%ing toroids it was easy.  All parts were there, etc.

If someone from the Softrock crowd is reading this, could you please do us a favor, and where possible switch to commercial RF transformers?  Coilcraft, Wurth (like in the RF-AMP), Mini-Circuits, etc. all make suitable transformers.  Except where they are in a tuned circuit they should work better.  Where they are in a tuned circuit design that part so that a broadband transformer would work instead!  YES WE WILL GLADLY PAY THE $2 EXTRA!

If you insist on forcing 11,000 hand windings of Toroids, could you at least put the six holes on 0.1" spacing so those of us who would first like to solder it on to a header (like a commercial transformer) could?  (Also on the audio jumpers on the Ensemble RXTX spacing at 0.1" instead of wire jumpers would have easily allowed I/Q flips with a short header or cable rather than  desoldering.. geeze...)

Anyway.. as it turned out that was only the start of the fun.  It will be awhile before I worry about TX.  Now it did draw the right amount of current on TX.. so it's probably going to be okay smokey.

But...

The sr63ng/cube/PAF-3/RX-AMP combo didn't receive anything.  So I decided I needed to test the CODEC by the internal tests that are on the AUX serial port menu.  No RS-232.  Well.. it seems that the AUX connector needs to have the cable completely snug when using it or it won't work.  Unless the I/O is screwed down flush, the slot in the case is too small to fish connectors through.  Two hours of messing around I got fed up and took off the subassembly bracket.  Plugged the cable in.. switched RX/TX back to correct.. and voila.. came right up!

Now I'm scared S-less that I'll short something (okay.. I've done that already:  Correction, short something and take something out!) but it's easier barely possible to debug that way.  The cable two posts back is correct and does work for RS-232.

Anyway.. I'm still afraid that the Radio-Shack type 1/8" stereo connector that I used for the AUX might still not fit when the thing is properly boxed up.  I may need to nibble the connector slot open more.  I am not happy about that at all.  Plus the cabling in the cube is like a rats-nest.  I am starting to think that I would have been happier and better off if I went with the original setup, that is the Cube in a case and the Softrock/IF radio external.  But now I am thinking that building it into a bigger box and laying out the boards flat might still have been better.  I suppose that is the price for having the "cute little case".  I would like to demo this radio... and the MBA in me values the Marketing that case provides as much as the Engineer has learned to hate the cube physical design.

I also have to figure out some sort of grommet to keep the ribbon cables from rubbing in the slot.  This isn't a good situation either.

As you can see, the inside of the case is getting quite "busy".  I'm frankly more than a little uncomfortable with it.  Also I am going to have to put a right angle connector on the I2C connector of the DSP card.  This is bad, bad, bad the way it is.  Luckily I dug some up so I can fix this major issue--

The bigger issue is that it sure seems that the SDR-Cube is "deaf".  Compared to a sound card IF, either a lot of CODEC gain needs to be dialed in (which I think isn't good) or the preamp in the RX-AMP card must be used for this thing to act like a radio (I *think* as mine isn't working).  I'm wondering if there shouldn't have been an additional op-amp stage on the DSP card prior to the CODEC.  That may have killed dynamic range.. but I expect "hiss" out of my radios when I turn them on unsquelched!

I think my RX-AMP might not be working.  After several hours of playing with the cube, measuring voltages, resistors, caps, on the SR63ng... verifying with the frequency counter the si570 is working and tuning (it does), verifying the 74AC74 /4 divider works (it does) and several different things, I took the RX-AMP off and hooked a long piece of wire to one of the connectors on the BPF plugin on the sr63ng.  The signal generator nearby instantly became audible to the cube.  But it's still significantly more deaf than the Kenwood R-2000 I also had monitoring the signal farther away with a shorter wire antenna.  I have to figure something out.

If it's the RF-AMP board, and the SDR-Cube needs the pre-amplification I can understand what is going on.  It's giving me fits now though.  Maybe I do have something that's affecting the CODEC input.. but hooked to the PC, the audio levels seem reasonable, so I don't know.

One thing that is really neat-o when testing is to make a cable to hook your computer up the line input of the Cube and go to the softrock40 groups file downloads and look for the DL6IAK Tone Generators and I/Q generator download and get the nifty I/Q generator.  I knew the cube was working when I tried than and then assumed it had to be something wrong with the sr63ng.  I now think the sr63ng (at least LO and receive) is perfectly fine.  I think the RF-AMP isn't working.  That's next in the debug.

And... yes.. one or both of those frequencies can be NEGATIVE and when you do that they end up on the LSB side of the display of the SDR-Cube.  ;O)  Cool toy.

I'll get there eventually.. probably soon actually.. the only thing I can say it's extremely difficult to do debugging with the cube in the way.  But I'm afraid of shorting everything and blowing it up to kingdom come if I started to disassemble things.

I am debating building another RX-AMP up (since I have two more) or building a classic filter on perfboard as from the original Softrock design.  But another option would be to build a broadband transformer on a board for testing.  I need to do something to continue.

Filter coefficients revisited...

I mentioned in a previous post that "if you don't like the filters that Juha has chosen for the rig" or "if you want to add more of your own" that you can simply design them in software and change or add coefficients.

I also stated that there was a program I liked better than the one I posted but I forgot what it is.  Well I remembered and also noticed that it's the program that Juha himself used.  If you look at the Microphone IIR source code header (miciir.h) he reveals his source of coefficient generator.

// Microphone test LPF filter
// fs = 8000 Hz
// f1 = 2100 Hz
// pass band riple 0,3dB
// designed with Toshio Iwata DFALZ1 program
/*
a0 = 15723
a1 = 31447
a2 = 15723
b1 = 7747
b2 = 22379a0 = 6398
a1 = 12796
a2 = 6398
b1 = -12827
b2 = 5651
*/

The dfalz1 program is available on http://www.digitalfilter.com/ (link to Digital Filter Analyzer page) (direct link to download) and applies to the FIR filters as well. So you can change characteristics of the filters in this rig quite simply if you would like. I think Juha's choice of filters was quite good and most people shouldn't mess with it.. but if you want to mess with it to learn about these filters.. you can always easily change it back.

Here is a screenshot of a certain "hacker" duplication the narrow filter (FIR) on the cube.. the 200-700 Hz filter. Note all of the design parameters of the filters are in the comments of the filter headers. Really cool of Juha to have the open source be that open.



OK.. the screenshot needed to be small to fit here... but the coefficients I generated with the program are an exact match (when designed with the parameters that were specified in the comments) with those that are in the existing project header file.  You might be able to tell that from the picture above!

Make sure you set the coefficents to 16-bits on a 16-bit fixed-point processor like the DSPic33!

I'm going to eventually have to post why it's much more difficult to have a project be open source (hardware and software).. but lets just say that it's wonderful that Juha and George approached the project this way.

"Aux" Serial Cable.

OK.. well.. Apparently I had Tx and Rx swapped (basically because the definitions were ambigious on the SDR-Cube schematics).. On my cable, because I had the issue (see newer posts) of the 1/8" stereo connectors not fitting in that little slot that I nibbled out more eventually.. I swapped RX and TX back and forth a couple of times.  I thought I swapped it back to the January 2011 revision drawing.. but apparently I swapped it back to the corrected diagram below. 

George Heron has a prettier diagram on his site now.. but I still was first with this.  :o)

-----------------------------------------------------------------------------------------------------------

Decided to quickly design a serial cable for the SDR-Cube.  I've not tested it yet, but as long as RX on the Cube schematic hooks to TX on the PC and vise-versa it should be correct.

This hooks to the "AUX" jack (which worries me because it could easily get confused with audio and maybe damage something.. like a microphone..) and then to a RS-232 port on the PC.  I, of course, recommend the cable be built to tie the PC handshaking lines together as shown.  Maybe someday the code will support CAT over this port and have to work with an application that expects functional handshaking lines.

The Beauty of Customization

Okay.. some corrections to the previous posts. The one pin connector that comes off of the TX Audio connector is, of course, PTT. On my sr63ng, anyway... the way the cable dimensions are specified in the instructions.. it's too short to make it to the PTT connector. I suspect it would be on a "real" Softrock 6.3 also. I extended the connection and changed the sex to match the sr63ng be making a 3" cable with pin headers on each end and then I heat shrunk the splice parts for a "permanent" splice. So it had nothing to do with eliminating the LDO regulator for the 3.3V supply for the si570 as I thought based on the missing regulator in the official Softrock 6.3 base kits. The regulator is coming for the "real" Softrock 6.3 RXTX base kits in the "service pack" as stated on the Yahoo groups site.

Okay.. secondly since I'm using a sr63ng... some of the updated parts for the SDR-Cube specific use obviously are not in the unit I already assembled. The sr63ng is almost an exact clone and most of the circuitry is exactly the same except for the fact it uses the much better to build 0805 parts (IMHO) instead of "shoalds" of bent through hole resistors and through hole capacitors.

I discovered tonight that the RX op-amp will have to have the supplied 220 pF capacitors changed to 4700 pF, just like on the Softrock "base kit". The reason for doing this at first is not obvious. The SDR-Cube samples at a rather low 8 KHz to keep processing in line with the DSPic33 processor capabilities. Because of this, a 192 KHz wide filter, like is on the Softrocks/SR clones and most other SDR's that would externally hook to the SDR cube would have aliasing issues with the excessively wide filter on the rig. George calls this "BCB blow by" but the reason it would be happening is indeed aliasing. Mix products would be pretty ugly if it were left at 192 KHz. I also made the realization of why AM and FM would be difficult to do maybe even if the PIC32MX with 2x the processing power was used, with a 16 or 22.1KHz sample rate. To properly do those modes, a Zero IF wouldn't be feasible like it is with USB/LSB and CW. I'll get into that later. AM is doable and maybe NBFM but it would be more interesting to do as there would have to probably be a conversion stage (in software) like the PC based SDRs do at low KHz IF. Of course.. if at some point the sample rate is increased in the future this filter will need to be widened up. :O)

The RC filter uses a 4.99K and a 220 pF cap for the 192 KHz analog filter. 4700 pf / 220 pf is 21.4 times difference. 192 KHz / 8 KHz is roughly a BW reduction of 24 times. Adjusting for standard values.. that's close enough.

The thing I don't like about this is that when using with an external SDR, either the SDR would have to be modified for the SDR-Cube or an additional op-amp filter is needed. I think that additional filter should have been on the DSP board on the cube to keep the Softrock or other SDR front end "standard".

I have "472" 0805's in my collection of chip caps.. so I will "customize" my sr63ng for the cube.. tho I'm not really wild about doing this.

The other "customization" that has been coming out on the Yahoo group is the TX balance adjustments. There is no good way to do this on the cube currently. One suggestion is to add adjustment pots to the DSP board. I will further investigate that. I wonder if a board with I2C or SPI EEPOTs wouldn't be a better solution. This again is an addition that probably should be done to the cube for clean image rejection on TX.

But the main purpose of this post is to show my solutions to the mechanical issues I don't like so far. I decided to tie down the regulator instead of relying on double stick tape. I left the tape on... but left the backing on the back side so it's not stuck to the aluminum. I like the actual physical connection with the nylon Ty-wrap much better. To do this a hole needs to be drilled on the other side of the regulator on the sub-chassis bracket.


Also, on my cube one of the electrolytic capacitors was rubbing on the case. It was actually already starting to pull off the DSP board from the manipulation I've done to mount boards. I initially thought that I would just use a piece of Kapton tape to try to insulate the cap. But once I noticed that it was getting physically damaged I thought I needed to do something about it.

I do testing for a living including a (very little) ESS testing. You can do whatever you like with your cube.. but wedging a PCB in with a connector and a standoff to "bend the board" violates most accepted quality standards... therefore, I respectfully disagree with the "no bracket is needed" statement by the cube designers. Here is what I did to fix both of these issues:

Notice the cutout. That bloody capacitor won't rub now. (Metal nibbler and a Sharpie.) Of course look at the bracket.. that could have fixed it by itself *if* the NUE-PSK mini-DIN connector I/O board would have cleared the ribbon cable on the DSP without the cap rubbing against the case. It didn't.

This fixes both the "floating DSP board" and the "rubbing capacitor" issue. The bracket is 1/16" thick aluminum with a 90 degree corner bent in it. About 3-7/8" long .. 1/4" on the back wall side.. 5/8" long into the DSP card. Metal nibble out the bracket so it doesn't short.. as necessary. The screws are 4-40... nylock nuts. Nothing too special.


The above picture shows the additional screws on the back.

I feel much better about the mechanical situation now. The cabling.. because I intentionally put on longer heat shrink the cables will physically wedge between the boards.. so I feel better about that as well. If they do not properly wedge in place they will be tied down.

Whether or not you choose to do any of this is up to you but I highly recommend it.