Hello to all, Here is an extract of the answer I did in the Multipsk group plus some other results.
>> I have also ran into at least 8 different stations running MultiPSK >> 4.7 or newer that I can't decode UNLESS I run MultiPSK 4.7 or later. I tried here generating Olivia 32-1000 towards an old Multipsk version and Mixw...no problem. This is logical as there was no modification on Olivia for some versions (however a regression error was always possible but there was no such error). >> I decided to do a side-by-side receiving/decoding test with DM780 >> software (1806 beta) and MultiPSK 4.8 and found that DM780 >> consistently could decode very weak signals 80-90 pct. copy that >> MultiPSK would only get 10pct or less of. I was surprised - but did I tried here comparing Mixw, Multipsk and OliviaAid with a very noisy transmission: the decodings are more or less equivalent (Multipsk being slightly better and OliviaAid slightly worse with a relatively important decoding delay). The sound card speed calibration was the same for every softs. Note: OliviaAid is worth for DM780 and Fldigi. The results are the following (with a signal at -13 dB of S/N Gaussian which is the limit for Olivia 32-1000 decoding): * OliviaAid: 120 characters decoded, * Mixw: 164 characters decoded, * Multipsk: 208 characters decoded. However, I tested on Gaussian noise, when real conditions could be very different from a Gaussian noise (QSB was not simulated for example) and so the results could be different. I think the decoding's problem of Gary is linked to the sound card speed calibration (button "Sampling freq." and follow the instructions). If it keeps on failing, Gary must check that the button "Freq. search" is pushed (to have an equivalent to Mixw or OliviaAid) . If it keeps on failing, let's Gary contact me or better the Multipsk group. A good thing would be to send me a record file (at 11025 samples/sec, 8 bits, mono) recording a phase when Multipsk decodes badly. I will check why. I tested previously thanks to the Multipsk group the 48 KHz sampling speed (adding a provisional measure of 48 KHz sampling speed). It is in general very good (my poor on-board sound card becomes suddenly excellent), but it is, unfortunatly, not a general case (I hoped that it was a general case to progressively switch to 48 KHz). > special soundcard alignment that I haven't already done - too many As indicated above, it is true that with a 48 KHz sampling frequency the probability of an immediate good result is perhaps of 80 or 90 % but unfortunatly it is not 100 % so, in all cases, it is necessary to calibrate his sound card. With Multipsk, it is automatic, it takes 2x3 mn, so I don't think it could really be a big problem. And it is one time (you can control from time to time, in case of). So in any case, it is important to calibrate the sound card, moreover in Olivia. This because it is a MFSK mode with a lot of possible carriers (32 in the standard mode), so you must not to mix one carrier with another carrier or simply have a shift which makes that you assess your carriers (through a FFT or an inter-correlation) at a bad location. The other issue is the synchronization. You must synchronize on the right moment. In general you have a capacity of auto-synchronization (about +/- 1% in Multipsk (and also about the same for Mixw as far as I see) through a digital PLL) but for the best performance it is of course better not to have to modify the standard speed and to be just at the right speed. So if for a PSK or a RTTY transmission, you could have a certain tolerance with a slow degradation of the performance around the real sound card sampling speed, in Olivia you have a quick degradation of the performance. This problem of soundcard alignment is extreme with MT63 which works with 64 carriers... 73 Patrick ----- Original Message ----- From: Jose A. Amador To: [EMAIL PROTECTED] ; digitalradio@yahoogroups.com Sent: Tuesday, May 13, 2008 4:56 AM Subject: [digitalradio] Re: [olivia] Re: MultiPSK 4.8 doesn't decode all that great on Olivia garylinnrobinson wrote: > I didn't do my comparison's on MixW and Olivia Aid - I did them with > DM780 and just recently FLDigi on a separate computer but same sound > feed from transceiver since FLDigi is on Linux. Same results. > > You can say it's just Gary but I don't believe it. And it is doesn't > apply to PSK or RTTY they work abt the same on all the progs. > > If I have to own a special computer or special soundcard or do some > special soundcard alignment that I haven't already done - too many > hoops for the regular user let alone a guy who has worked in the > computer industry as a tech and programmer as I have. There are several factors to consider in order to achieve a fair relative evaluation, and I am sure you know with your claimed background. First, with the data you have at hand could you achieve a quantitative evaluation? As Lord Kelvin stated a long time ago, in science and engineering you actually need numbers to avoid fuzziness. To dissipate doubts, it could be useful if you could also provide your data sets for independent evaluation. Second, are your two computers identical? Same sound card, processor, speed, memory, you certainly wouldn't need to be told all the factors to weigh. Third, as I understand, the "AC97 timing syndrome" only happens on Windows. On Linux and Unix derivatives, queues, semaphores, etc, have different priorities, and so far, Linux fares better with "run of the mill" soundcards and associated delays, even when that does not make differences insignificant, for many reasons, not related exclusively to timing. Signal levels, distortion, noise, A/D and D/A converters linearity, Hamming distances of different modulation formats, FEC, data interleaving are also important factors and certainly have an influence on received BER. Something that would be quite peculiar, if proven true, is that all modes show exactly the same problems. It seems important to sort out this particular allegged behavior with valid data to substantiate it. Linux certainly could give an edge to FLDigi, which is, in fact, also a good performer. It might be interesting to evaluate also GMFSK or other available programs, for sake of completeness. I feel that the last paragraph of your posting above is particularly unfair. In many aspects of life, there exist well known price/performance tradeoffs, be clothing, cars, CPU's, soundcards, just to mention a few well known and some relevant ones. When the multimode boxes were predominant, there were designs and brands that were undoubtedly superior to others. I believe that it is a formidable feat to achieve a similar perforance between dedicated boxes with single tasking processors and computers with multiple running tasks on a multitasking or task switching environment like Windows, at the cost it gets achieved. I have not made any well documented comparisons myself previously, and I am using an average card for receiving, an Audigy 2, which is not a Delta, an EMU, or a higher cost cousin, but neither an AC97. So far, I have not found substantial differences between MultiPSK and MixW, before I began using MultiPSK almost exclusively when versions 4.xx appeared. My soundcard does not require a noticeably different setting from its default. Nevertheless, hardware differences may be so many among users, and behaviors under different OS versions that an independent developer cannot evaluate all possible influences without the beta testers and users feedback. Other programs I also use corroborate such a situation. I believe that all users could certainly gain with a fair evaluation that unveils problems that a developer alone cannot certainly find. 73, Jose, CO2JA
