Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
QDX: Anyone else having a problem with image rejection?
#qdx
|
I'm seeing signals from 27 KHz below 14.074 MHZ... strong Morse code signals actually (when checked on another receiver).
I understand why it's 27 KHz, and that image rejection depends on the balance between I and Q magnitudes (IC7A and IC7B outputs). I thought I'd fixed it earlier by removing the solder bridge on IC7, the problem is better but still there. Before I start poking around with a scope: Has anyone tweaked their image rejection? 73 Andy WM8Q |
|
No, I haven't tweaked it but I measured it. I found the unwanted sideband response 24kHz below dial frequency on 40m and it was about -50..-55dB below the wanted signal. Quite good in my opinion, considering that the rig has no means to adjust I/Q gain/phase balance in the hardware. Gain balance can be adjusted by trimming either R26 or R27. Phase balance is affected by the tolerances of C44 and C45 but I haven't paid much attention to how much it can contribute. However, R26/C44 and R27/C45 puts a -3dB lowpass corner at about 1.6kHz (1 / (2*pi*R*C)). This sounds a bit low to me since the IF is picked at 12kHz. The IC7 gain at 12kHz IF is about 17 or 18 dB down compared to the gain at zero beat/DC. Maybe this is a design tradeoff, to prevent the ADC from seeing to much strong signals. Anyway, the QDX RX still works great, much better than I ever expected :)
toggle quoted message
Show quoted text
73 SM6LKM AndyT wrote: Before I start poking around with a scope: Has anyone tweaked their image rejection? |
|
Hello Andy,
I am not aware of an adjustment for image rejection, as that is in the software unlike in the QCX line of transceivers where it is analog.?? What is the Audio Sweep showing as far as image rejection?? When I do my scans with a dummy load connected I am seeing better than 60db rejection. I would also make sure you are on the latest version of firmware 1_02. One last point, I remember one of the changes of version 1 to version 2 of the hardware was adding bias resistors to the BPF switch to help with strong signal rejection.? Here is the message link: /g/QRPLabs/message/74675 73 Evan AC9TU |
|
Hi Evan- Thanks for the reply.... right after building it, my scan looked great. Today, not so. I also noticed that the image rejection looked like it went to more normal after I unplugged the USB and re-connected it.... but no power cycle. No explanation?for that (yet). 73 Andy WM8Q On Sun, Dec 19, 2021 at 2:12 PM Evan Hand <elhandjr@...> wrote: Hello Andy, |
|
Hi all There are two different things that generate unwanted signals on RECEIVE (not on transmit): 1) Unwanted sideband cancellation. Where the amplitude of the I and Q channels mismatch, or the phase shift is not 90-degrees, the unwanted sideband (normally lower sideband) is imperfectly canceled. This is shown?on the AF sweep function in the QDX terminal applications.? 2) Image: QDX digitally implements a SUPERHETERODYNE receiver having a 12kHz IF, and then a phasing-method direct conversion receiver to get SSB audio. There will be an image response at 12kHz the other side of IF, i.e. 24kHz from the reception frequency. There is no tool in the QDX terminal applications to view the image rejection.? Both of these imperfections can be quite substantially reduced by software in the QDX SDR. I did not implement any kind of corrections to compensate for these errors. My reasons for not implementing these were:
I'll certainly implement the corrections, when I have some time. It's the same code on QSX too and I did it, I just hadn't fully tested it and it's tricky.? 73 Hans G0UPL On Mon, Dec 20, 2021 at 1:16 AM AndyT <andy@...> wrote: Johan- |
|
Hi Colin Firstly, are you correctly referring to image rejection (the superhet image 24kHz away from the wanted signal) and if so, how are you measuring that? Or are you referring to unwanted sideband rejection as shown in the AF sweep screen and if so, a screenshot would be useful.? Secondly is this a Rev 3a board and if so, have you cleared the infamous via short at C41?? Thirdly are you sure you got all the T2 wires in the right holes and without shorts, with proper soldering etc.?? Finally if you get through all that... There is no way currently to adjust amplitude balance or phase between the I & Q paths. I didn't bother with that because in all the QDX radios I built and tested it was already so good I didn't deem it urgent. Things just reached the stage where the effort required to get from 99% perfect to 99.9% perfect was growing exponentially and I really just wanted to get thing out the door and on the street. But it was something I did (do) plan to come back to later since there are good ways in software, of correcting these imperfections and indeed, doing so automatically.? 73 Hans G0UPL http://qrp-labs.com https://www.buymeacoffee.com/g0upl -------- Original message -------- From: Colin Kaminski <colinskaminski@...> Date: Wed, Aug 10, 2022, 7:46 PM To: [email protected] Subject: Re: [QRPLabs] QDX: Anyone else having a problem with image rejection? #qdx I was just digging into this. I am only 20 db down on 20M and 80M and 10 db down on 30M and 40M. Is there a way in hardware to adjust the phase of the IQ signal? |
|
Thank you Hans,
It took me a few days to check everything. I am only using the built in tools. They have changed since the manual and I didn't notice the scale changes till today. I am using a Rev 3 board. I ran through T2 with a fine tooth comb. It had one extra winding at the beginning so I removed it. The solders have continuity. I am attaching 10 screenshots of the built in diagnostic tools. I played with the 40M RF filter for some time. The peak is still too high at 20 turns on both inductors. I am wanting to fine tune everything as much as possible before I close up the case. This is an exercise in understanding as much as anything. I am interested in what you see? -- Colin Kaminski - K6JTH https://www.qsl.net/k6jth/ |
|
On 10/08/2022 20:18, Hans Summers wrote:
are you correctly referring to image rejection (the superhet image 24kHz away from the wanted signal) and if so, how are you measuring that? Or are you referring to unwanted sideband rejection as shown in the AF sweep screenHans, I came across your comment and wondered what it meant. I was under the impression IQ balance was responsible for reducing, if not eliminating image responses. The unwanted sideband is an image. If fed with the basic "audio" IF an SDR program will show say 48KHz bandwidth, the LO in the centre with all signals duplicated each side of centre. A USB signal will appear opposite the centre as an image, a LSB signal. If the QDX is set to receive the USB signal the image is a LSB signal 24KHz below it. Feed the SDR program with IQ signals then images are cancelled. IQ balancing differentiates between negative and positive frequencies, signals below and above the LO. The unwanted sideband is an image and is cancelled along with any real signals on that frequency. I have no idea about how the software works but that is the way I see soundcard SDR working. There is almost always an offset. It is recommended to maintain a fixed offset so IQ cancellation is optimised there. IQ balancing is needed because of small variations in the hardware. I do not remember anyone suggesting image and unwanted sideband are different. Presumably any software frequency conversion would have perfect IQ balance if that came into it? 73 Alan G4ZFQ |
|
Hello Alan It may be a mix-up of terminology and perhaps I am using incorrect terms or explaining poorly, my apologies if so.? It's best understood by analogy to a conventional superheterodyne?receiver. Imagine we have an 80m SSB receiver, using an old common 455kHz Intermediate Frequency for which filters were available. We wish to listen to a station naturally, at 3.560 MHz (the QRP frequency). Let's say we have to tune the VFO to 3.560?+ 0.455 MHz = 4.015 MHz, so that after the first mixer we'll have the wanted signal at the 455kHz IF frequency, and pass it through the 455kHz IF filter. The IF filter might be a ceramic type and have an SSB bandwidth such as 2.8kHz. It's job is to provide selectivity, including rejection of the unwanted (upper) sideband. The BFO will be placed just past the edge of the filter such that the second mixer will then take that filtered IF signal and mix it down to audio.? There are TWO unwanted signal possibilities?here.? First is the superhet image. A signal on 4.015 (the VFO frequency)?+ 455kHz (the IF frequency) = 4.470 MHz, will also mix with the VFO to produce the 455kHz IF frequency. To attenuate this unwanted image frequency, we have to use front-end RF filtering. A bandpass or low pass RF filter such that 4.470MHz is greatly attenuated relative to the wanted 3.560 MHz signal.? Secondly, there's the unwanted (upper) sideband. To attenuate this, we are using the 2.8kHz IF filter.? So in a superhet receiver there are two different categories of unwanted signal, which are each dealt with differently, so that we end up only with the nice clean wanted signal. The front end RF filter can help with the superhet image but not the unwanted sideband. And the IF filter can help with the unwanted sideband (and other good selectivity) but cannot help with the superhet image at 4.470 MHz.? In a direct conversion receiver, also sometimes called a zero-IF receiver, there is no "superhet image", only the unwanted sideband.? An SDR can be anything, it does not necessarily implement a direct conversion receiver. The problem often around 0Hz is that it can be difficult to eliminate power line noise and other hum sources. For this reason, in QDX I implemented a superhet with a 12kHz IF. The VFO is set to 12kHz below the dial frequency. It has an image response that is 24kHz above the dial frequency. Additionally it has the unwanted sideband. Two different effects.? Software conversion and processing is not generally perfect; the degree of perfection is increased the more CPU power you are prepared to throw at it - but typically it is much more perfect than analog.? In QDX both the unwanted sideband elimination and the superhet image response 24kHz away, are affected by the I-Q balance and phase accuracy of the analog circuits preceding the digital processing. It's also possible to compensate for any imperfections in the analog circuits, within the digital processing. Even to do so automatically. I did not, when it came to QDX release, implement such coding because it appeared to me that already without any error compensation, QDX has an extraordinarily high performance. And because firmware updates are so easy every owner can do it free in moments without any? additional hardware or software, and on any OS. And because, you know how it goes with the creeping feature creature. So at some point you have to draw the line and postpone enhancements into the future, and kick the thing outta the door.? 73 Hans G0UPL On Tue, Aug 16, 2022 at 1:23 PM Alan G4ZFQ <alan4alan@...> wrote: On 10/08/2022 20:18, Hans Summers wrote: |
|
The receiver bandwidth is given by
1) Input and low pass filters. This bandwidth is quite high. 2) Filtering by the Tayloe detector. The filter bandwidh is given by the resistance of the switched mixer and by the capacity of the sampling capacitors. 3) Low pass of the IQ amplifier feeding the AD coneverter. This filter is not very sharp. 4) And finally, the digital filter inside the AD chip. 4) is important here. The AD is a sigma-delta converter, sampling with a low resolution converter at very high sampling frequency and then low pass filtering the result. The AD converter parameters are well documented in the datasheet. I believe Hans configures the AD converter to 48 samples per second, using the single rate filter, which has a frequency response documented on page 13 of Figure 19 shows the behavior around the transition frequency, which is +-24kHz from the VFO. These graphs are not very detailed, but I eyeballed that the signals at +-24+3=+-27kHz from the VFO will be suppressed by -50dB. This AD converter is quite good. I remember studying the datasheet of the chip in mcHF and it was a lot worse. Vojtech OK1IAK |
Colin Kaminski
to navigate to use esc to dismiss