|
Home | Articles | Projects | Products | Bookmarks | About |
Posted at - 25th June 2017
Edited at - 15th August 2017
This is an old project started somewhere in March 2013.
Being a VHF lover at that time, someday I was wondering what if I'd give to HF spectrum a try. I built a direct conversion receiver to listen on 80m band. I have enjoyed listening to that band, many contests from other countries, late night QSOs, etc. I remember about a late night contest where I collected all the received callsigns then I have searched them on QRZ.com to see where they were from. I was able to hear stations from all over the Europe with just a piece of UTP cable as an antenna. After that extraordinary experience I wanted to further experiment radio propagation on that reliable HF band. Also, in our country the local QTC is transmited on 80m. I searched on the internet for a super simple SSB transceiver and I've found the BITX20 project by Ashhar Farhan, it looked very promising to me.
An old friend of mine - YO2MHO - confirmed that the bitx does a good job on the air if is correctly tuned. Therefore, together with another good friend - YO3IMR - we've started to build our bitx80 projects. First, we have chosen the critical components in the following manner:
For its construction the following schematic and PCB design were used, of course, with BAT41 diodes on mixer and product detector, 9.8304 MHz replacing those 10.000 MHz crystals, etc.
My PCB had some problems, but hopefully they were easily fixed.
I remember about a small issue of the original PCB which had a flaw with a pin of the LM386 amp, thus leading to no audio on the speaker.
After all components were in place and LM386 problem solved, I have adjusted the VCO between approx. 5.8MHz - 6.3MHz with a PIC frequency counter borrowed from my friend, YO3IMR. On transmitting mode, the potentiometer from the product detector should balance the detector/mixer by cancelling out the BFO oscillation, but in my case, the oscillation cancelling did not work. I verified almost all components if they were erroneously placed on the board, all traces on the PCB, I found nothing wrong. On receiving mode seemed to work OK. I started to worry about this issue, all things seemed to be OK, but it does not work as it should on transmitting mode... that continuous wave on 9.8 MHz was always there.
I have verified oscillator signals with my scope, I modified biasing resistors in order to get bigger and cleaner signals, also the oscillators were screened with copper plates, no improvement at all. Sometimes I felt like I would smash this piece of junk with the hammer... it drived me crazy many times, everything seemed to be alright but it does not work as it should... Finally, I thrown this into my junk box, the project was abandoned.
But later in 2016, I tried to resurrect again this project, armed with more patience and after several days of debugging I have found the root cause of all problems! When I measured between two components on the board which they were on the same PCB trace, I've found the resistance between them was about 4Kohms, I did not believe my eyes, how could be possible to have 4K instead 0 Ohms between two components on the same trace??? Simple! cold solder joint. That was between product detector potentiometer and 22 Ohms resistor, affecting both the emission and the reception. Once this problem fixed all worked good, 9.8 MHz from BFO was supressed successfully, plenty of sound volume on receiving mode, the I.F. signals on the scope were more bigger, etc.
Then I made countless modifications including the following schematic.
The DDS is a must nowadays and it will replace the local oscillator for intermediate frequency. Due class A polarisation of the RF power amplifier the chassis will get hot during long QSOs and the base frequency becomes uncontrollable. So, when I designed this DDS I was thinking about a very cheap DDS chip which can generate frequencies around 6MHz. Studying the AD DDS products, I've found AD9837 suitable for this purpose. Using a 16MHz reference clock it can generates signals of max 8MHz. The sine wave above 1MHz becomes noisy and should be filtered. I used two amp stages with three tuned LC tanks to obtain a clean 6MHz signal for the I.F.
The controller I used was Atmega32 with 16MHz clock together with a 2x16 HD44780 display and a rotary encoder. The software was written in standalone C, no arduino nor bootloaders or something like that. The code is in pre-alpha state at this moment. Unfortunately, I didn't make the DDS schematic at that time, but the controller hardware schematic and software can be found on my GitHub repo.
First, I was live without DDS. The frequency was very unstable but despite that I made my first contact in autumn of 2016, YO8SJR answered to my first call from 240Km away. I was astonished! My antenna was a 2x20m dipole at 2m above the ground, no antenna tuner involved and 3W output at that time.
Another tremendous excitement I felt was during Christmas time in 2016, when I used same 2x20m dipole with its center at a much higher height, about 7-8m above the ground, fed with a flat 2x0.75 speaker cable, tuned with a balanced transformer at its base, the power about 4W and then I was able to hear my CW and SSB signals in UK on 160m.net websdr, about 2200 Km away. Later that night I heard a foreign station calling and I have answered to that call to see what happens, I was surprised to hear that station answering back to me, it was S55L from Slovenia, about 1000 Km away. Those days I made contacts all over my country and neighbors like Bosnia and Greece.
The frequency drift was annoying during QSOs and next days that period, actually in the new year night, I read the AD9837 datasheet in order to code a driver for it. So after a couple of weeks after that the DDS was installed making a huge difference.
In the Easter time 2017, using same antenna but with a better matching system, I was able to work the entire country with 5W output. Everybody heard me effortlessly. In general, I've got S6 to S9+15 signal reports on the air accompanied by a very clear and penetrant modulation.
Audio Recording - S6 signal strength over S4 noise - WebSDR Server Bucharest Romania
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
 |
After two years of continuous service, the RxTx relay died mainly because I made a design mistake in the past: I used a big capacitor ~1000uF for Tx supply rail, so every time when the relay closed its contacts, big sparks slowly corroded them until unfortunately they glued together.
Therefore I made a MOSFET switch which will replace the old relay, now the big advantage is that it will not wear out so easely with or without that big cap. on the Tx line.
Based upon the resistive divider R1,R3, while the PTT is open, Q1 MOSFET-P stays open and Q3 MOSFET-N is closed down to GND, bringing the GND to Tx line which in turn closing Q2 and opening Q4. When the PTT is closed all MOSFETs operation is reversed. Simply to put, initially Rx line is positive, Tx negative, when PTT is pressed Tx becomes positive and Rx negative. Bringing the GND to unused line will help to discard possible charged capacitors, thus preventing from unwanted autooscillations.
This circuit works great, but it was not tested enough. A high precision digital scope is needed to analyze the opening/closing (raise-fall) time of the MOSFETs. For example, for short periods of time, miliseconds or maybe nanoseconds, both Q1/Q3 or Q2/Q4 can conduct at the same time when switching RxTx and TxRx.
The C1,C2,D1,D2 protect against inductive loads, for example, other relays on the tx line, etc.
The power MOSFETs IRF5305 have low RDSon, therefore voltage drop across them is negligible, about few milivolts for 1.5A flowing through.
Because I just opened the "can" again fixing the broken relay I was somehow forced to put several transistors together and make an Smeter indicator, and if I made an Smeter maybe I can add another few transistors together and make an AGC circuit... At first, I said "OK, I have not enough time to build this one" but shortly I realised that's very annoying to listen to very strong singals without AGC, so I had to do it. Also I have no time to draw everything in CAD, but here are the pen&paper prototypes:
In the first image is depicted an 11V voltage regulator used for Smeter positive rail supply and in the lower part of the pic, a darlington pair with its Ib and Rb ecuations, the stage exhibit high input impedance and high current gain in order to be processed further. So, for 330 Ohm load resistor a 43-44MOhm base resistor is needed in this case. In the last pic from above is the Smeter prototype. The signal from IF - taken from emitor of Q3A in bitx schematic - is amplified by high impedance darlington stage. Here I used a 4N7 capacitor because 100n will charge very slowly due high impedance resistor (44MOhms) thus lagging a bit the emitor voltage over RL. The amplified current will pass through the primary winding of the transformer (3:45 turn ratio), transforming high current into voltage. The diodes are 2x1N4148 together with a potentiometer are used to set the voltage at the "knee" of conduction for 2n3904 transistor, thus this will act as a polarised diode, working in class B, allowing only positive half cycle of the RF wave to pass (like an AM detector). Also the 2n3904 is acting as a impedance converter, using RL = 1000 ohms in emitor, at the base will have a high input impedance and a high current gain. Then the current is sampled and amplified further by the PNP transistor which is a 2n3906 type. I had to design this circuit because I wanted high gain with small amount of components, since the IF voltage was about 250mVp at maximum it needed about +20dB gain to reach 5Vp, at first I built a stage with 2n3904 and had only 8dB gain, therefore about three stages were needed in order to reach the 5Vp, then other two or three stages of detection, too many components... Even though, after I built the transformer variant I realised that there could be an even much simpler way to measure the amplitude without a transformer with the same idea (transistor as a biased diode - class B):
The RSSI circuitry in this case dosn't have to be linear because other stages in transceiver are not linear neither, therefore the linearity and calibration will be implemented later in software. Also, the transistor biased in class B here working into a high impedant load has an unstable conduction threshold due ambiental temperature changing, currently I have no idea how to fix this issue without using a long tailed differential pair.
Signal for AGC is sampled from RSSI output circuit, a potentiometer is used to adjust the closing threshold for MOSFET 2N7000. When weak signals are being received the mosfet stays opened and the capacitor is charging through A + B resitors keeping the BJT conducting at full rail - 10V regulated. This voltage is then used to supply Q1 in bitx schematic. When a strong signal is received, the RSSI voltage will close the mosfet, rapidly discarching the capacitor, A + B becomes a resistive divider lowering the voltage given by BJT emitor follower, therefore the gain of Q1 decreases. After the strong signal goes away the mosfet stops conducting, C is then slowy charging, and Q1 amplification is slowly increased, being a pleasure to listen now :) The component values (if I remember correctly): pot = 1M mosfet = 2N7000 C = 22uF A = 6k8 B = 3k3 BJT = BC547B During very strong signals the AGC voltage will decrease reaching about 2Vcc. Unfortunately, I used AGC voltage to supply only one transistor stage in Rx chain, therefore it worked ok but not so effective. If I could supply more Rx stages with ACG voltage, then the AGC efficiency would have been increased by multiplying the gain of those stages.
Design Errors
I tried to speed up things and the design I made was not very good. I made an RSSI circuit which work by incereasing the voltage, then an AGC circuit which is working by decreasing voltage while it could be done in a much even simpler way: the RSSI and AGC working in the same stage by decreasing the voltage. But I could admit that I have enough electronic components so this doesn't have to be a problem. But imagine if this kind of waste of components, space, weight, electric current will eventually reach a production plant, then hit the market, there will be a big loss for the company. Few components costs about few dollars, not so expensive right? but multiplied by a million of produced items, it turns into few millions of dollars lost!Other Modifications
Badly biased stages: I don't know what type of transistors the designer used, but in the schematic were specified 2N3904 or BC547, therefore I used BC547B for all bitx transistors and they were erroneousely biased by default. Over the collector resistor should be dropped around Vcc/2 in order for that stage to have enough voltage swing, so by default they're not! Therefore I modified almost all the biasing resistors. I have looked with the scope while injecting signals and it proved that I was right. Now the received and transmitted signals greatly improved their fidelity. A substantial difference can be heard by listening the following records: Before: After: Cooler: Since my final stage is a class A, after long transmitting periods it gets hot and amplification factor decreases. So a small fan was mounted in order to cool down the transistor. The fan is a 12V - 1W in series with 100 Ohm 3W resistor in order to decrease its rpm and noise. The fan is active only during transmision. Imperfect Contact: Again I had throuble with this kind of problems. After transmitting for long periods, the case of the transceiver gets hot and sometimes I got no output signal frustrating me over and over again because I wasn't able to locate the problem quickly. Anyway, it was found: a coaxial cable between final and prefinal stages got shorted between central conductor and its braid through dielectric, and that happen only when it gets warmer... jeezzz! It was replaced by 10cm of RG316. Output Transformer: The matching between output of final transistor to 50 ohms impedance was made with an L-matching circuit at first, using an unknown ferrite core which gets warmer after long Tx sessions. The L-match was replaced by a transformer wound on a 43 ferrite material (I don't remember the turn ratio). Tx Command: In order to trigger power amplifiers, etc. a connector was mounted over the case connected to the Tx rail through a 2k2 resistor.Input Impedance Analysis
Full CSV file here.
The Latest Inside View
|
Home | Articles | Projects | Products | Bookmarks | About |
