After 4 iterations of my Wadley loop receiver, all my hard work was a disaster. So, ripped it all apart and learned from my mistakes. It now has more in common with the original Pogson Deltahet than my interpretation of it, hence the new topic.
Here is a link to the very first signals received, here on 9 megs, some weird god station, always good for strong signals during a bad time of day, but no miracles were performed despite the theological BS.
It was still a battle to get this far, and birdies remain a problem, especially at the 1 megacycle boundaries. This is not suprising as the harmonic comb generator is as yet unsheilded. Another issue was getting valve mixers to have enough conversion gain at these frequencies, I had long forgotten just how hard it can be.
Ideally I need an extra amplifying stage in the 37.5 meg IF, I found that using a hi slope pentode with grid leak injection worked fine and saved another current hungry valve, the mains transformer is working close to its limits. ... we have 15 valves!! I tried cathode injection on the Wadley loop input, here the comb feeds into the cathode, the VFO into the grid, this too works very well. Another 300mA saving in heater current was using a ECF80 , the triode portion as the VHF oscillator, the pentode section as the third mixer. The first mixer remains problematic, this one converts the RF from the preselector to the 40 meg first IF. Currently this is using a 12AT7 twin triode, but efficiency seems low. The RF amp is a 6BA6, as are the 2 455kc IFs. I stuck with the ECH81 for the tuneable IF, as I am very familiar with these. The detector is an anode bend for AM, and will see the BFO injected into its cathode for SSB/CW. There are still problems with the 40 meg bandpass filter, and am struggling to find a good coupling to prevent double humping, and signal loss.
There is no preselection yet, other than a random tuned circuit temporarily hooked up to the first mixer.
AGC seems to work fine, this uses a voltage doubler using germanium diodes, giving a negative 10 volts on s9 signsls. The voltages to the BFO, VHF oscillator and associated mixers are regulated by an OA2 neon., giving a stabilised 110 volts.
The s meter uses a bridge circuit, one side is the cathode of the third IF valve, the other side derived from a voltage doubler from the heater supply. This needs improving, it drifts all over the place as the set heats up.
Another two quick videos graphically show the workings of the Wadley loop.
The next few days I will tidy up what I have so far and try to pinpoint the weaknesses. At least with sound, it helps to track these issues.
Have spent even more time experimenting with this radio. It seems it has reached the limit of its performance insofar as sensitivity and stability is concerned. The biggest ongoing issue is drift in the tuneable 2.455 to 3.455 mhz oscillator in the final IF section. As the radio warms up, it drifts out of the narrow 455kc xtal IF filter passband, only stabilising after 10/15 minutes. The 2mm alloy chassis has lots of thermal inertia, ie, it heats slowly. I am wondering if a series of holes drilled through the chassis would create a thermal "island" in the ECH81 3rd mixer area to help isolate the external heating, especially from the mains transformer. I have now settled on 6BE6s pentagrids for the first and 2nd signal mixers, with low screen voltages and high value cathode resistors, the mixer noise has reduced considerably, conversion gain still remains high, the signal and injection grid isolation is a welcome bonus. A quick 12 dB SINAD test reveals -107 dBm AM at 30 percent mod, thats around 1 microvolt across 50 ohms, this is quite good, I think up from the -92dBm earlier. This sensitivity holds from 3 mHz through 30mHz... more than adequate for HF.. AGC is now applied to the RF, ist IF, and second IF amplifiers all using 6BA6 remote cutoffs, the RF amplifier now using a 6BZ6, these seem quieter in operation. Cosmic noise is now audible over all bands. Another slight issue is BCFM breakthrough in some higher bands, this would be eliminated with a LPF after the RF amplifier. Another annoying issue, I cannot quite get the new valve VFO to exactly coincide with the existing FRG7 tuning drum markings, so have opted for a digital display. This is due to arrive from China in a few days. Unfortunately it uses a six digit 7 segment display, so may be a tight fit in the front panel. These modules have a +- programmable IF offset so setting this to -3.455 should allow it to read 0 to 100000 as the VFO tunes 3.455 to 2.455 mHz, ie, the Wadley loop 2nd IF runs backwards in frequency as it tunes the RX frequency higher. Its a cheats way of doing it, I have spent far too much time on this project already.
With my current fascination with Wadley loop receivers, and before I forget how they work, have picked up a realistic DX300. I am wondering why the reviews were so bad, and just how bad are they, can it be improved etc. I may do a write up if anyone is interested.
And a trap for us young players.....
I had a pulsing effect on my secondary HT rail, this one feeds the tuneable IF section and demodulators with regulation performed by an OB2 regulator neon. These stike at around 150 volts and then shunt regulate at around 105 volts. The rail was pulsing between the above figures, cycling at around 2 cycles/second. Turns out, I had mistakenly put a 50 mFd cap across the neon to decouple that rail to ground. This turned the neon into a relaxation oscillator, it would strike at 150v, then pull the rail down where it would cease conduction, the rail then returning to 150v, then striking once again......the time constant being the series HT feed resistor, and capacitor. We rememeber something every day.....
A short video of the Deltahet receiving American CB on 27 megs, over 8000 miles from here.
A careful alignment has netted an MDS of -110 dBm, AM, 30 percent mod, and -92 dBm for a SINAD of 12 dB. These figures are not dramatic, but perfectly adequate for good HF.
Common mixer/oscillators here were the 6U8 family and its follow ons, 12AT7's, the pentagrids 6BE6, 6BA7, 1R5, etc. and then the beam deflection types 7360, etc
I believe my 75A-4 uses 6BA7 mixers, but it's been a while since I have been in it. The 51J series may have had the mil equivalent of 6BE6's (5750?) in it - it's been a couple of decades since I have owned one.
Win W5JAG
After listening to the Deltahet over the last weeks, I decided to address some annoying issues. Firstly, tuning the megacycle selector was much too critical, the bandwidth was too sharp. This too affected the second mixer, the CIO injection level was too low This was why that mixer originally used the "leaky grid" method, where the 37.5 and 40 megahertz were applied to the same grid of the 6AU6, These two signals are so close together in frequency that they interfere with each other, not a good design. .... tuning the 40 meg BPF would pull the 37.5 meg tuning. A 6BE6 was tried here, having separate control grids, but the signal levels were far from optimum to achieve good mixing.
So, another amplification stage was in order here at the 37.5, allowing steeper selectivity skirts, and a final amplitude increase from 350 mV to around 1 volt. The coils could now be stagger tuned to increase the effective bandwidth, the final tuned circuit was tapped down to allow cathode injection to the second mixer, thereby further isolating the RF and injection ports.
This has made tuning easier and the circuit more robust. There are now 3 double tuned circuits at 37.5 megs, with a 4th single tuned at the output, the 40 meg signal BPF retains the 8 top coupled tuned circuits, the coupling reduced from 2.2 to 1 pF per section to give the 800 KC bandwidth with minimal ripple
. The existing second mixer was repurposed as the third amplifier stage, the ECC81 twin triode first audio was replaced with a 6BL8 /ECF80, the triode section remaining as the audio, the pentode now becoming the new second mixer.... there was limited room to use another valve, the 7 pin miniatures have only limited possibilities
The second annoyance was the many different valves used, I figured it could be improved here as well. The EF80s were replaced with 6AU6 RF pentodes, with only a slight decrease in conversion gain. The radio now uses 6AU6s in all the mixer and loop amplifier circuits, and the remote cutoff 6BA6s in the signal circuits using AGC control. The 4th mixer and product detector still uses the ECH81s.... these operate at higher signal levels. The ECF80 is now out of place, but needs must. Fortunately with valved equipment, many mods can be made without too much major surgery.
The performance of the radio is very good, as far as sensitivity and stability is concerned. The off signal noise level is high, this is mainly mixer noise, this reduces quickly on signal. It is also exaggerated by the very wide first IF bandwidth.
The wider the window opens, the more dirt blows in......
I have managed to receive the American truckers at 27.025 for long periods, just as stable as say, WWV on 5 megs, thats a distance of 8000 miles or so...... sensitivity is in the 1 microvolt region right now....
Once I am happy with the radio, I will measure the actual performance, drift, sensitivity, dynamic range etc, and come up with a few figures.... I am especially interested in the dynamic range and image rejection figures.
Today had a MAJOR breakthrough. I had suspicions with the first mixer being inefficient. The radio was sensitive but struggling even with fairly strong signals, and appeared to overload easily despite the AGC working its little heart out. While messing with the mixer and even changing it to a different design, I turned my attention to the RF amplifier. Turns out, the geriatric moron that assembled this thing fitted a 47k cathode resistor instead of a 470 ohm..... yes, I am a bloody moron. Now this radio is really alive. Later when the sun goes down, I will do a decent band scan on video. The ssb detector is now functioning correctly, the s meter is behaving itself, and ALL the birdies have vanished. The good thing I guess is I have optimised the weak signal handling of the first mixer......... always look at the positives.
All for the sake of ONE bloody resistor.
I temporarily placed a small fan under the chassis to increase ventilation, this works a treat. The long term oscillator drift in the second IF was driving me nuts,, this slowly increased as the chassis heated up, not much, but enough to be a nuisance. I dont know whether lifting the power transformer slightly off the chassis will help here, this will break the thermal transfer to the chassis, and then to the coil cans. Oddly, the 40 to 70 meg oscillator sits within 100kc of its setting, it drifts very little considering the higher frequency, and of course matters not, as its drift is cancelled anyway. The tuneable IF drifts a few 100s of cycles , on;y a pain on SSB.. The BFO uses a ceramic filter, this is relatively stable, despite not liking to work into high impedance valves, they are low impedance devices.
I modified the AGC slightly to give a fast attack and slow decay, and added a fast/slow, this will use one of the missing front panel switches.
Also added a triode audio preamp stage to give a tad more audio grunt, had an unused spare..
Its pretty much finished. The Kilohertz tuning is slightly non linear and is not quite accurate on the Yaesy dial drum, that is annoying and needs looking into.
The ECH81 product detector works well, but having the BFO centered in the 3rd IF passband iis not ideal, the audio image makes copy difficult at times. This needs improving too.
I will now do a final video of it running......
I have looked at them with my Red Pitaya and I agree, the Si5351 output is filthy compared to a conventional L/C oscillator.
My first iteration of my double conversion superhet used dual gate MOSFET's for the mixers, and between all of the filth and the cross tallk, it was completely unacceptable.
However, I find that double balanced diode mixers, like the commercial mini circuits types, are pretty much immune to all of this, and work well with nothing more than a coupling cap between the Si5351 and the DBM.
The true DDS modules looked at on my mediocre test equipment have a strong harmonic content, but are otherwise pretty clean. In my homebrew SSB transceiver, I use a DDS module and, on the transmit side, I pre filter the 23 MHz LO with a SAW filter to get rid of the harmonics before they get into the mixer. The transmitter has a very clean output - it easily meets FCC requirements at its power level with only a five pole low pass filter at its output.
I think if you wanted to use a DBM with a Si5351, you would probably be okay for a BFO / product detector.
73,
Win W5JAG
Wow.
That is absolutely incredible, Dayle. Fantastic work.
If solid state can be tolerated, QRP Labs make a very compact fixed frequency Si5351 synthesizer/controller for $20 ish USD. Programs with DIP switches. I have one and it works great - the project it was to be used in is abandoned however.
I agree low frequency crystals are hard to source, and no guarantee they are on frequency after this time.
73,
Win W5JAG
I am slowly making progress between fixing cars and cutting down trees that are growing so fast they are blocking my solar array....
The Deltahet is working very well, several more things to do, especially adding SSB capability.
The mechanicals are largely completed, wondering if there is enough panel space for my EM81 magic eye tube, and just where to put it.
A few pics.....
Here we are locked and tuned to the 7 meg band. Managed to add an FRG7 type lock indicator, here I cheated and used two germanium diodes and a BC547 transistor, Im running short of room, and another valve seemed wasteful. Also had major problems with way too much harmonic generation, so ditched the vacuum state approach and settled for a small transistor harmonic generator module. This outputs a few millivolts only and cured the 1 meg boundary breakthrough. This I will revisit at some stage.
There is still a front end preselector coil to wind, 1.6 to 4 megs, this will be next. The vfo coil is as yet without its screening can. There is room behind the speaker to add the BFO and product detector, will probably use another ECH81 here. I need 455kc crystals, these are VERY hard to source. My experiments using ceramic resonators are unreliable, they have HUGE frequency tolerances, and dont like the heat, like 14 valves a glowing, at 1.5 watts of heat each, plus the hard working power transformer in a confined space.....
Its looking a tad messy and busy underneath, mainly due to having to isolate and bypass every stage to prevent interaction. I may have gone overboard. The bent tin sheild is sheilding the 40 meg bandpass filter, made from a scrap of tinplate from a can of CRC. I managed to switch the band indicator LEDs with the same switch contacts that switch the serial coil primaries, even rotary switches seem to becoming rarer by the day.
I wasted 4 hours replacing the three gang kilohertz tuning capacitor.... note to self. When drilling holes in an aluminium chassis, cover the tuning gangs, else they will become scratchy forever after......
I know its easy to over praise ones performance on homebuilt designs, but this thing receives radio Pyongyang 20 minutes before either my NRD93, or my NRD515, both very sensitive receivers, have any listenable audio.......
I have picked up yet another FRG7 Yaesu, rusty and unloved...... Ill need to , ummm, upgrade that one too I guess
Sorry to not add to the technical aspect of your article. I don't have any experience in this area of tubes and this design. Great stuff, interesting to learn about all this. I really appreciate the contributions made here on this site, like dayleedwards here. Thanks
Here is another video of a 9 meg band scan after the debugging process of removing spurs and birdies. Performance is now smoother, the birdie on the one meg boundaries remains, but WWV at 10 megs can still be received, so its down in the microvolt region Im guessing.
https://www.youtube.com/watch?v=aP5ppbreLlM
Wow, very impressive results, Dayle!
73,
Win W5JAG