Few months ago I had interesting conversation with a colleague who got a kit from me and built it. He was using it on 27 Mhz and according to him, the breakthrough from the Local Oscillator (LO) could be heard two miles down the road via another receiver. This was interesting problem to investigate, i could have worked on a fix immediately as i have a Spectrum Analyzer, but it would have been foolish to ‘play’ with the RX mixer and pre-amp without having proper baseline measurements for sensitivity. And this was not possible as the Tracking generator of my Spectrum Analyzer is stuck to specific output value, i think the attenuator circuit is cooked, maybe the previous owner was careless and transmitted into it.
So after some searching on Ebay, watching videos from Dave (EEVblog), i have narrowed it down – i desperately needed Marconi Aeroflex 2024 RF signal generator. I was lucky to get reconditioned unit in almost new condition, ex-MOD. It even had Option 4 installed – the oven oscillator! The most incredible thing about this signal generator is the built-in precise attenuator, with sub dB steps, that can go as low as -140 dBm. It has also great protection against accidental TX into it’s output port of upto 50W! And it can switch units on screen – either dBm or uV.
The Test Setup
I was ready to to test the sensitivity and LO leakage on the antenna port. I have chosen to test MDS (Minimal Discernible Signal) while using simply my headphones. This method is highly subjective, but it was ok for my comparative testing of improvements and mods. It is possible to do a fully scientific measurements, but this is beyond my current needs. I have decided to use my Audio Technica ATH-M50 studio headphones, plugged directly to the mcHF headphone jack.
Despite the protection of RF signal generator, the first thing is to disable TX from the Advanced menu of the mcHF. Being extra careful never hurts! I am using v 0.5 RF + UI board for this test.
The results from the sensitivity test(MDS, 3.6kHz filter, RF gain at 50):
- 80m, -118 dBm
- 60m, -109 dBm
- 40m, -118 dBm
- 30m, -130 dBm
- 20m, -130 dBm
- 17m, -130 dBm
- 15m, -129 dBm
- 12m, -133 dBm
- 10m, -130 dBm
The basic obvious thing from this test is that at lower frequencies the RX pre-amp is not as sensitive, and the 60m band dip is caused by using the 80m bandpass filter, instead of it’s own, dedicated one.
Local Oscillator Leakage
Next on the list was checking the LO leak to the antenna port, using my Spectrum Analyzer, results as follows:
- 80m, -83 dBm
- 60m, -73 dBm
- 40m, -72 dBm
- 30m, -66 dBm
- 20m, -63 dBm
- 17m, -62 dBm
- 15m, -58 dBm
- 12m, -61 dBm
- 10m, -61 dBm
Again, some quick conclusions – there is more than 20 dB difference between lowest and highest band, so most probably the leak is capacitive breakthrough, instead of design related. Also the worst possible value of -58 dBm converts to 1.6 nW of output power at the antenna socket, so even if we leave it like this, it is not something to loose sleep over, specially on our noisy HF bands where noise level is usually between S5-S9.
Still, i wanted to try few things. One suggestion was to completely rework the RX mixer into double balanced mode, but as no schematics with detailed practical measurements was available, only some pdf with lots of theory and simulations, i have decided to skip this option as it would be too much work, with not guaranteed result.
Another idea was to use low noise MMIC monolithic amp, there are some nice chips similar to the AD one i have selected for the v 0.5 TX mod, but with much lower noise value and again with built in bias circuit. The main criteria would be noise figure and reverse insulation.
But even before starting of the selection process, i needed some extra measurements. For example, the RX mixer is sensitive enough, but there was a need for buffering to reduce LO leak and to compensate the losses in the BPFs and LPFs, some amplification was needed.
Without going into detail, after insulating some stages, i was able to measure the loss in the LPF/BPF/antenna switch chain per band. It averages at 3.88 dB.
RF Pre-amp gain
Measuring the gain of the BFR93A gain in the v 0.5 configuration is not exactly an easy task. I was too lazy to build it on small proto PCB, so what i did is remove R41 and R42, then feed pin 1 and 3 of the transformer T1 to the Analyzer, then RF signal generator was fed to the input via bypass of R4 in the BPF. Again i did measure gain per band, but as we need average value, it comes to 9.21 dB. I did not insulate the center tap of T1, so we can safely assume that the uniform gain value is about 10 dB.
How all fits together
So we have an idea what to look for: MMIC chip of min frequency at DC, gain at about 10 dB and as much as possible reverse insulation figure. Also interesting conclusion – out of the total RF pre-amp gain we have only 6 dB for amplification, while the rest is to compensate for filter losses.
After using the above criteria I have selected three chips that were possible match for what was needed and were available for immediate order:
- TRF37D73IDSGT, Gain 10 dB, Noise 3.25 dB, Reverse Isolation N/A, package WSON-8
- HMC476SC70E, Gain 10 dB, Noise 3.9 dB, Reverse Isolation 18 dB, package SC-70
- BGA416E6327HTSA1, Gain 14 dB, Noise 1.6 dB, Reverse Isolation 60 dB, package SOT-143
So all those chips looked nice on paper and used similar connection method, more or less the way the current BFR93A is wired (on purpose).
Lots of soldering
There was no other way than removing the RF pre-amp components, installing a small socket and building a replacement board for each amp. It was a lot of work, but it had to be done to have a comprehensive result from the test.
It is worth mentioning that you need to be be brave to do this prototyping work with WSON-8 package. It is not easy to imagine how small it is, it basically fits onto standard 2.54mm PCB pad and have 9 terminals to solder!
My plan was to post detailed results per band for sensitivity and LO leakage for each chip. But unfortunately non of the chips scored any better than the current amp with BPF93A. It was a nice reminder why i have chosen it few years ago. Overall the HMS476 had reverse isolation of 15 dB worse than the BFR93A and the rest of the chips, I was not able to see 60 dB isolation for the BGA416 chip, and it’s gain was very low, probably around 0 dB. The only chip that was impressive was the TRF37D73 – it had nice uniform gain across all bands, average sensitivity resulted into -130 dBm and reverse isolation was not worse than the BFR93A configuration. Obviosly the results were bit skewed due to using extra PCBs, sockets, i would believe this chip would perform even better if soldered on the main board with careful layout.
Off course there is always a but, and this time was the package. This thing make the SC-70 package we used before for the Schmitt Triggers look massive! I am sure the pick and place machine in the factory will handle it without problem, but there are many OMs who choose to solder their RF board by hand.
Time for decision
It might sound cool to use modern chip, but i didn’t see a reason to switch from the current design. Instead i have decided to come back to my previous conclusion about capacitive breakthrough causing the LO leak. So i decided to take another look at the schematics and PCB layout and search for a simple solution. Few things caught my eye:
- R36 connected to the collector of Q1 might be good as negative DC feedback, but can contribute for poor reverse insulation
- All MMIC chips datasheets were insisting on 1 uF tantalum in the Vcc bias circuit, we don’t have it here
- C75 blocking capacitor needs to be close to the center tap of T1, unfortunately i have placed it by mistake far away, next to the mixer bias divider
So the mod is as simple as cutting the R36 trace that leads to the Q1 collector, wire R36 to the R35/C65 junction, install small 10 uF cap on top of C65 and move C75 next to the T1 center point.
And here the modified schematics:
The final result
Now the important question, is this mod any good ? Here the results of my test:
- 80m, MDS -133 dBm, Lo Leak -87 dBm
- 60m, MDS -125 dBm, Lo Leak -88 dBm
- 40m, MDS -132 dBm, Lo Leak -86 dBm
- 30m, MDS -140 dBm, Lo Leak -76 dBm
- 20m, MDS -134 dBm, Lo Leak -68 dBm
- 17m, MDS -135 dBm, Lo Leak -71 dBm
- 15m, MDS -130 dBm, Lo Leak -66 dBm
- 12m, MDS -130 dBm, Lo Leak -67 dBm
- 10m, MDS -127 dBm, Lo Leak -66 dBm
There is an obvious 5-10 dB improvement in the LO leakage values, with existing v 0.5 PCB and simple HW modification. The increased sensitivity must be attributed to removal of the negative feedback, so some changes in the base biasing might be needed to reduce gain (R36/R37 values).
Again as all other mods, feedback, comments and ideas are welcome. And this mod is obviously completely optional.