It looks like you're new here. If you want to get involved, click one of these buttons!
Hi. Good news with the 31 dB attenuator. Thank you!
I don't know if this is already the plan:
Let the software adjust the attenuator in 1 dB step, to maximise the sensitivity while maintaining no clipping in the ADC. This "front-end AGC" should be dynamically with a properly selected time constant, probably in the region of one minute or so.
The front-end AGC should be selectable AUTO/Fixed of course in the attenuator settings. Maybe with the possibility to choose high or low threshold, or selectable 1, 3, 6, or 10 dB in headroom before clipping. This for to optimising the headroom according to taste or antenna situation.
This will vastly improve the sensitivity during daytime when using large or long antennas which needs attenuation on MF/LF during nights.
The change in attenuator shall be addet to RSSI-system so the S-meter reading will not be affected, preferably.
Please consider this if not doing already! :)
Regards, Per, SM7YES
Per, I split your comment out from the Kiwi-2 "expression of interest" thread so it wouldn't get lost. It is an important topic that deserves further discussion. I believe this idea has been mentioned previously someplace (maybe by Martin?)
Yes, it was discussed in this previous thread, along with some other relevant information further along.
My feeling remains that in most cases, a simple passive amplitude / frequency equaliser ahead of the KiWi would solve a lot of the problems experienced by many KiWi owners.
My KiWi (and the others) at the shared Weston site, use a 10dB amplitude equaliser, with the slope knee at around 5MHz. They also use a triplexer that I built, in order to combine a Wellbrook loop for the LF bands, a dipole with a special balun that I designed for the mid-frequencies, and a CB 5/8 antenna for the frequencies above about 22MHz.
My other KiWi's at Wessex (where I have more control over the site), use an antenna that has been chosen to naturally roll off the gain at the lower frequency end of the spectrum, and as a result no equaliser is required.
http://wessex.zapto.org:8073 V1 KiWi
http://wessex.zapto.org:8074 modified prototype V2 KiWi
http://wessex.hopto.org:8075 RasPi KiWi running Flydog software
However, both this site and Weston do have HF broadcast band notch filters ahead of the KiWi's.
Both sites perform pretty well despite this, and Wessex tends to be well up in the top ten of the Linkfanel SNR list for a lot of the time, Weston a bit less so.
Yes an equilizer is often good to attenuate MF/LF. Especially with longer antennas in the km-range where for example I designed a simple LCR-filter giving -15dB below 1600 kHz and -1 dB @ 5 MHz and 0 dB from 8MHz and upwards. This is good for this very long beverage-antenna, but there are other situations where you dont want this exact eq-profile. Dynamic Front-end does not exlude the possibility to do this also. It is a very good add-on and gives better flexibility, and improves what the equilizer has missed. 14 bit is still a limitation with large antennas, some times if you want to have maximum sensitivity. Whit extra gain you always get better sensitivity, which is always wanted at the highest frequencies 20-30MHz :)
The sensitivity is the sum of all gains and noise figures of all the stages plus the noise in the antenna, as we know. If we can maintain extra gain in the front, we get better sensitivity overall, when it is allowed by the dynamic front-end AGC algorithm. This must not be foreseen. At higher frequencies an extra 10 dB gain in the front (10 dB less attenuation) may give lets say 3 dB:s better total sensitivity, if you understand what i mean, but the extra 10 dB will also spoil 10 dB of the dynamic range, so thats why it will be so good with the dynamic front-end AGC. -you can get some extra dB sensitivity when its allowed!. There are many cases where it is useful, but of course you have situations when its not. Dynamic front-end AGC only improves. It never destroys if you for example use RF-equilizer.
One thing doesn't exclude the other! The best thing is to have both! :)
Thanks for making this thread!
Just a thought:
I don't know how you can get access to the RF-ADC total input level, for the front-end AGC algorithm. Maybe something has to be done in the FPGA code to measure and report the ADC level for future software, if it will not be done from start? Also an indicator/monitor of how much the input level is using, preferably in percent of the ADC-range, can be good to see during adjustment of the input 31 dB attenuator, both when adjusting manually or via front-end AGC algorithm, to see how much headroom is left.
I am not a programmer, just a radio system engineer :) So you know better of how to implement this of course, and if possible :)
The key aspect is Signal to Noise Ratio (SNR).
If you have >10dB noise floor above the receiver noise floor with the antenna connected, you are throwing away Dynamic Range.
The ITU noise curves are in the region of 20 to 30dB higher at 1MHz than at 30MHz. So, even allowing for poorer antennas at the LF end of the spectrum, there is usually still far too much gain.
Amateur transceivers allow you to define the pre-amp and attenuator settings per band, making it easy to automatically arrange for them to be turned off on the LF bands.
However, with the kiwi it provides 0-30MHz continuous coverage, which is shared between several users, so any changes in gain affects everyone.
The standard KiWi noise figure is just about adequate at 30MHz. But as soon as you introduce extra attenuation, this is degraded.
Fortunately, most ADC overloads tend to occur at night when there is little propagation above 15MHz, so this is not as problematic as it could be, although there are local 27MHz CB and 10m Amateur stations that may still be heard.
This is why initially I prefer using amplitude / frequency equalisation, as opposed to simply adding extra attenuation.
@G8JNJ It all depends of what antennas we are using. I know many fighting with extra attenuators, and loosing sensitivity on higher frequencies, and who have not the possibilities to build equilizers. Dynamic front-end AGC must be very slow to loook like manual adjustment. So it will look like fixed and adjust for the current antenna situation, local transmitter, nigh or day etc. As i said "One thing doesn't exclude the other! The best thing is to have both!"
if you dont want the dynamic agc, please tell me why, and how it eventually would interfere with equalizers.
But obviously you don't like to have the 31 dB attenuator in the Kiwi at all... So what are we discussing?
@G8JNJ "The key aspect is Signal to Noise Ratio (SNR).
If you have >10dB noise floor above the receiver noise floor with the antenna connected, you are throwing away Dynamic Range."
Yes. Please read again what i wrote :)
We are agreeing with each other, I was just trying to emphasise the point.
At Weston, we actually have a fairly sophisticated MQTT controlled external automatic attenuation and notch filter system.
This monitors the total RF power in a few separate filtered bands and applied RF notches and overall attenuation as required using MQTT
However, most of the time this was not required because the 5MHz equaliser did the job without degrading things too much.
As you can see, with the equaliser on, it is still just about possible to maintain more than 6dB noise floor delta even at 30MHz (almost).
Gain distribution graphs for signal chain with distribution amplifier and successive passive splitters.
Ok thanks, Looks good. Nice effort you have put into with filtering and notches. And yes if you have 6 dB increase in noise at higher frequencies, the antenna must be quite good in effeciency att higher frequencies. I dont see that with long wire antennas on my remote noise free locations. But a single 1 km wire for 0-30 Mhz is a different story from your multiantenna setup. With a single long wire we want the maximum sensitivity on higher bands. I think this situation is not so uncommon.
And this is what I am trying to emphasise on, that antennas and locations looks veruy different, and also many have no possibilities or skills to do such a nice installation with notches etc. Some antennas have less output levels especially on higher frequencies, some QTH have less noise etc.
It would be very nice to have an automatic adjustment of the 31 dB attenuator (slow like almost we do it manually) So it is optimising for every situation to have best sensitivity with no ADC-clipping. And also nice to be able tu put in a preamp, like a low gain 4x2sk125 or similar without "destroying" the reception.
There is definitely a good reason to have an automatic attenuator "front-end AGC", although it should be switchable to manual if preferred of course. And if we can monitor the attenuator settings and ADC percentage loading it will be a lot of help when setting up and administering a site.
I agree, allowing the KiWi V2 attenuator to provide automatic broadband attenuation when required is a desirable enhancement.
I just wish many of the KiWi's had a decent enough antenna to make it occasionally operate.
Yes and in addition to the above; For Kiwi owners with large antennas, who are not so experienced in the "dynamic thinking" (which is probably a large percentage of Kiwi owners) it will make everything better and easier. But even for us more experienced it will make it easier, for example when adding preamplifiers and passive splitters that give a little extra gain, will no longer be a headache. As long as the preamp itself is not overloaded, the 31dB attenuator setting will take care of that easily. IP3 on the attenuator also does not seem to be a problem to cope with fairly high levels. This is going to be great!
So now it's up to the software developer if it can be done. If possible, it will be a good improvement to the overall system.
I may buy one and do beta testing, if this software feature will be introduced. I have access to both signal generators and long wire antennas which usually overload the current Kiwi, unless I use an antenna equalizer.
Perhaps the initial suggested algorithm in this thread is a good start, along with the monitor readout of the ADC's total input level loading on a 0 to 100 percent scale. Also if the current selected RF attenuation is shown, that would be nice to see.
This is not shown for the clients using the Kiwi of course. Only for the admin-login, what i think.
We can always wish, as I usually say :)
All the best!
Another way to see this is that the RF front-end AGC acts as a protection. If ADC is not clipping then fine, and that is always good to strive for, with rf eq, careful gain adjustments and so on as Martin stated, but as soon something climbs up too high in level, the protection climbs in :)
It would of course be nice with all the suggested UI-functionality with indicators for ADC-level and Attenuator settlings, but if the UI consumes too much of programming work it could be active invisible in the background instead and we at least have the possibility to turn RF-AGC on and off, maybe with two levels of thresholds.
Just my thougths for now, thank you.
@G8JNJ Btw, i forgot to mention, thanks for all the info, Martin. I listen on your Kiwis and they are sounding really good. I agree that it must be very well aligned EQ with still about 6 dB in noise floor från antenna at higher frequencies. Its a very serious work you have done there.
Good source also with the SNR-list Kiwis.
As a follow-up to the subject or Amplitude / Frequency equalisers.
After a bit more research, I came across this document published in 1977, relating to the use of an Amplitude / Frequency equaliser at HF listening stations operated by GCHQ. One of the UK's main intelligence agencies.
UDC 621.372.55: 621.391.822.4: 621.396.62.029.55
Noise equalization in h.f. receiving systems
D. C. BUNDAY
Journal page 195 - 206 / PDF page 3 - 14
The paper analyses a concept of noise equalization that can be applied in order to reduce the noise contribution due to intermodulation effects without affecting the basic signal to noise performance. Some practical applications of the noise equalization
Technique are outlined and the resulting operational benefits demonstrated.
The article describes pretty much what I had previously said, but in much more detail and with the associated theory and supporting maths.
It also includes a typical circuit, which is also very similar to the one I had described.
I hope this is of general interest.
Happy New Year (in advance).