|
I did extensive tests today after Gerald did. Gerald in Texas and me
in N.J just to confirm what Gerald found. We came up with exactly the same results. NOT ONE PEEP, not a single spike from a second order response. Taking a large output from an 8640B putting it inside the QSD bandwidth. Take a second 8640B, put it inside the QSD bandwidth but 11025 kHz away from the first and passing them through a decent combiner, I looked. I tuned around, up, down, every which way. I can see that the image rejection does indeed need to be automated with the pulse generator but the image was always within 10 dB of my noise floor (at -140 dBm) and at any place I stopped, I could calibrate the image to the noise floor. I SAW NO OTHER NONLINEAR RESPONSES outside the of the 3rd order ones (2*f1 - f2, 2*f2-f1) of the type we have used to measure the IP3 and IMD-DR. Lest it not be clear to people, we are talking about testing the modified QSD with appropriate load termination for the front end filtering as well as replacement of the active device on the RFE for much higher IIP3. This is what the ARRL measured in their labs and it is the result of the trip I took to Texas to work with Gerald on this some months ago (last spring). Rod, WN8R, received the first modified receiver before Dayton after I completed the mods on his radio, for the first time. The ARRL got theirs shortly thereafter in late June. We all still owe Rod a debt of gratitude for allowing me to make him a guinea pig! The ARRL tested with the M-Audio Delta 44. Again, these sound cards make a large difference in the performance of the radio. ALL stages participate in the production of mixing products, especially close in large signals where the sound card's performance really has an impact. You cannot get on the order of 100 dB of IMD-DR at 2 kHz without having an excellent A/D pair and having the image calibrated. Any dynamics testing done WITHOUT these modifications or without this caliber sound card has nothing to do with the current hardware and I am not one bit interested in hearing about it. The front end is, without DDS issues, bullet proof. If you have a laptop with a poor sound card, that is not a problem of the software or the hardware. It is a thing we have desperately been trying to remedy without much success for less than a thousand dollars. Hopefully the USB widget that is coming from Phil and others will fill the bill for the laptop owners. Another complaint bit the dust today when it was learned that complaints we have been subjected to these complaints here in this forum are based on a misunderstand of the noise blanker. This gating things on and off is how the QSD mixer actually works! If you have large impulsive repetitive noise, expect to have a large 30, 40 over 9 signal inside the QSD passband to be mixed into the filter when you engage the noise blanker (NB). Sorry, that is the nature of the beast. It happened all the time with my several thousand dollar high end radios. It will happen with this one as well in its current form. That said, we will not have to endure this forever. You can mitigate this current situation somewhat by changing the threshold. The Noise Blanker 2 (NB2) is a more linear mitigation but it is a simplistic approximation of the Asbrink removal and you cannot afford to set the threshold where all pulses are removed because the probability of false alarm is too high with this simplistic algorithm. It makes an "educated guess" based on the neighborhood, of what the sample should be. This is not the same caliber impulse noise removal that is available in LINRAD but it does demonstrate its power. From the IK3MAC recording, you can hear what these things can do for you under favorable conditions (listening for weak signals in an impulsive noise environment where you are absent extremely strong signals). The correct strategy is to remove the worst pulses with the NB, and remove the weaker impulse noise spikes with NB2. This is done by setting the threshold higher for NB than NB2 and having them both on. It is the reason for the default settings. How can we do better? We can in the future activate the impulse generator on the RFE board. With this we will be able to accurately estimate the impulse response of the front end. 1) With this impulse response, we can easily eliminate image problems in the receiver in software without calibration by the user. We have already demonstrated this in a Matlab^tm demonstration. It is one of those things that I should have been working on rather than the keyer,which cost months and wore me completely out, and for which I owe various parts of my body as replacement parts to Frank for doing the keyer. 2) With this impulse response, we can match the incoming signal against this pattern and determine impulsive events and subtract them. This is a linear noise pulse removal algorithm. It will not have the mixing products of the noise "blanker", which is as decidedly nonlinear as it can be. Not related to the impulse generator: 3) On the transmitter, we have most everyone using the Delta 44 and we need to encourage that where possible. As such, if you have the Delta 44, we are going to give you monitor out at audio just as you currently have and we will mix the "to be transmitted signal" up to the IF. ALL of the spurs visible in the ARRL lab picture in the review article, VANISH when we do this. The total composite noise goes down. Since the outgoing signal will have been terrifically filtered, there will be nothing much to mix with the junk near DC on the sound card and the signal on the air will improve dramatically, "the echo" will disappear, and image rejection will work as well as, if not better than, it does for RX. 4) We can avoid the DDS spurs almost completely by being much more aggressive than we have been in using the entire 40+ kHz we have inside the QSD. The worst spurs you can imagine are near 50 MHz (50 .011025 MHz in SSB, 50.000 in SPEC) from clock leakage. You will notice nothing of them at 50 MHz in SSB and large signals do not mix with them to any great extent. We can move even further away from them with a new strategy. These are the worst spurs. We can mitigate their impact almost completely. 5) We can enable all sorts of keying strategies for the radio, enabling it to work with all CW generating comport programs, with probably a couple of hours work and then extensive testing. When will we do better? Do not expect this soon from me. I am very busy at this time. Gerald has been gently twisting my arm (to the breaking point) and I have put that arm into a bullet-proof cast to which I am now applying setting compound with this note. I have business, personal, AMSAT, and ARRL commitments (in that order) that come before this new work will start. I need to *use* the SDR for my purposes. I put two years into this to get here and I am doing the consumption of the technology for my purposes now in both professional and hobby settings. My professional requirements, which should have been finished by the end of fiscal 2005, will be done by the end of calendar 2005 instead to everyone's dismay. If someone else wants to do some of this work in 1-5, I will be glad to give you verbal advice. This will not involve my writing a single line of code or writing a single line of mathematical equations. I will not get sucked into working on this before I have time to complete it. The only thing to expect from me in a timely fashion (Bob time, i.e. as available) is the new agc(s) and the new RX and TX equalizer. More testing and debugging is needed for both but they are progressing very nicely. Expect them soon after I get the PySDR and PySDX stuff done with Frank and Bob Cowdery. The AMSAT meeting is in 4 weeks. The symposium is canceled, but we are having a BORED meeting (sic) and an AMSAT Eagle design meeting. I will demonstrate the SDX at that meeting (Pittsburgh, second weekend in October). The ARRL work must be completed soon to have a large impact on some regulatory issues they are about to pursue. I am hopeful we can modify their direction. The new chairman of the FCC, in a very public address in Virginia to industry and amateurs at AMRAD, told the ARRL to pursue SDR and to pursue wideband digital comms using SDR on HF and VHF for ENCOMM and other purposes. It appears they are not really listening carefully to what this well placed amateur(!!) is saying. That certainly beats the political general's son we had before. I hope to convince the ARRL board with a single BER test of a system that they can be the sponsors of a revolution in communications using SDR and be a leader in the communications world again as we were in the SSB days. I am completely justified in my mind doing this AMSAT (SDR and spacecraft OS work) and ARRL (OFDM HF, VHF, and SHF work) with the SDR I have worked on for so long with my friends. There are very good people working on all of this with me helping. I am not asleep or not working. I have had to change my focus based on the current exigencies (the spacecraft is to be launched before the end of 2006 and the League proposals come out at the end of 2005). I know there is impatience but my bullet proof vest has a Naquadria powered personal shield based on my read of the correct priorities. Please, stand by. 73's Bob N4HY |
