I just received QMX kit so I took T30-6 and did a very quick test of the LPF. My order for the proper components did not arrive yet, so I just used some cheap 1608 ceramic chips. I only had those in E6 so I had to combine values to approximate my design... so components have errors. I just built it quick, didn't tweak anything, and I left the leads long so that I can actually use them as-is.

The filter deviated from theoretical prediction mostly in the passband ripples and the stopband ripples, but no huge major change in the overall characteristics. That's as expected.

The filter looks a bit aggressive into the 10m (I only do CW so I didn't care about slight loss beyond the CW subband) but at the same time I got pretty decent attenuation at 36MHz, so I might relax the cutoff frequency a bit upward, depending on how the actual emission looks like.

So, I think 60m to 10m implementation of QMX is one step closer.

开云体育

Hi Nick,

Yes, my traces look incredibly low tech. I still use those numerical design tables at the back of filter theory books and my HP calculator to design filters, so NanoVNA V2 Plus 4 is a bit new to me haha. Yes I know NanoVNA Saver, but I don't have a computer right at my bench, but maybe I should look into it. I also use TinySA Ultra (x2) and another more powerful signal generator and getting readable traces can be a bit of work...

I didn't capture the trace, but I also looked at Z1/S11 while output port terminated at 50. Those ripples correspond to where the input impedance goes a bit lower and slightly capacitive. So, it is entirely possible that the "ripple" can be partially offset by greater power pulled from the transistor, resulting in slightly lower eta but not much change in the output power at the BNC. We'll have to see that aspect of it in the actual board.

My filter is 5th order elliptic, one more zero in the transfer function than the topology in QMX's stock filters. It's barely visible in the photo but both inductors have parallel caps. A part of the lousy return loss and passband ripple is component values being slightly off, but this is actually not bad compared to a lot of wideband amplifier matching networks. The filter is also better than some used in other amateur transceiver kits. So, knowing the actual filter will probably be slightly better than this with values from E24 sequence that are closest to the theoretical calculations and also with higher quality Murata/Kyocera C0G/NP0 caps, I'm not too worried. I also implement those caps on the back of the board with SMD components so if I see a better design later, they are easy to change.

In terms of the design challenge, I think the hard part is that we have no good specification requirement as to the minimum attenuation of the second harmonic. If it's 15 dB, I could design a filter with higher cutoff and less passband ripple. It's a trade-off but I don't know what I'm trading off. So, the design shown here probably turns out wayy too conservative in terms of meeting FCC emission limits and too aggressive on the passband, but I'd rather err that way on my first version.

Incidentally, in relation to another thread of a mismatched load causing high voltage at the transistor's drains, I also looked at Z1/S11 while the output is open or shorted. Of course, the contour of Z(f) on the Smith chart looks more like a very long transmission line, basically marching right on the unit (outer) circle where the phase moves rapidly with the frequency. So, a new, easier way to find a point that maximally stresses the transistor is to short or open the antenna connector, and sweep the transmit frequency while observing the drain voltage. Why didn't I think of that while spending time on that thread.

Thanks for the good discussion, though.

I also looked into the receiver BPF. It's a bit of a puzzle, but I currently postulate the fact that two inductors wound on one core with a tap is complicating things due to the mutual inductance or the transformer effect (they are not two separate inductors). I'll be winding two separate inductors on T37-10 (just because I have a bunch from other job) and see if that moves the notches high enough.

So I just got my Mouser package arrived so I assembled QMX for real. I skipped the 80m LPF but installed my filter in place. I also skipped BS170 but installed LDMOS AFM907NT1 x 2 in place. I just uploaded the firmware, looking at the spurious emissions, etc.

First of all, there does not appear to be any parasitic oscillation or instability in using this UHF power transistor in an HF rig without any additional component to tame the excessive gain and super excessive high frequency response. I did take good care in wiring the transistors in dead bug style to minimize unwanted coupling/feedback while maximize the heat dissipation.

I am driving QMX with a 7.5V supply and measuring 34 to 36 dBm on 80/60/40/30/20m (bands that are activated by default).

Second harmonic seems about -27dBc in my setup. (I measured it at 3.5MHz while running through my custom LPF of 30MHz cutoff, so the harmonics are intact.)

I'll figure out how to configure the rig properly and report more later. For those who don't want to bother incremental reports but want to read one story, I intend to write up something once everything is done and tested in the field.

If anyone else is experimenting many-band QMX modifications, I recommend putting the experimental filter in L510/L513 slot rather than what I did. I put my filter in the 80m slot, since I wanted to leave stock filters for 60/40/30/20m and didn't want to shuffle too many things on the board. However, the 80m filter slot (L511/L506) is a lot more crowded than where 20m stock filter is located. I ended up having to solder SMD capacitors sideways on the back of the board to implement my fifth order elliptic filter (needs one more cap than the stock filter topology). I think I never soldered SMD capacitors sideways before.

Also, if anyone is experimenting the receiver BPF using two separate cores, I would use T30-4 or smaller. I used two T37-10 (one of my stock components) and while they could fit, it looks awkward. The filter Q is so low there that the difference in QU (unloaded Q) is not going to matter.

Using my filter and transmitting at 18.068MHz, the second harmonic is about -55dBc. This is about 10dB better than the second harmonics of 60m and 30m through the stock filters (they are at about -46dBc) and it is more than 10dB better-than-required harmonic suppression. The price is that the 10m output is a bit lower due to aggressive cutoff frequency. I think the next refinement may be to relax the filter to minimize the attenuation at 10m.

In short, QMX can have one TX LPF for 17m-12m, as mine is working fully, and the same filter can very likely be tweaked to function on 10m fully.

Now on RX bandpass filter.

My design worked largely as intended, although the filter peaks may be a bit off on some bands from 60 to 15m. For some reason, the BPF for 12/10m is behaving very strangely. I used all four channels of the multiplexer for switching two inductors and four capacitors. At leas 60/40m, 30/20m, 17/15m filters seem working fine without strange behavior previously reported.

The receiver sideband rejection, image rejection all look good. I did not change anything in the receiver other than the BPF bank.