thanks for the reply...although I really have no idea what you said ...lol (I'm sure alota people here do.....and appreciated the details) What I did understand tho...I haven't had ANY of those problems :) _And dont get me wrong Bob...everyone appreciates your work._ We know that our hardware revolves around volunteers donating their time to program the software that runs our investment:) so in the words of Gomer Pyle...Thank Ya - Thank Ya - Thank Ya!!!!! If you weren't already richer then God...id send you a donation :)
And I realize us SDR1000 owners are Not a revenue stream for the company any longer...with that said I was looking for a very broad timeline of when the new algorithms for the NR/ANF you eluded to in a previous email "might" be finished? Thats all.. / (((I expect we will be able to greatly improve the ANF and NR algorithms since we will be able to afford to do a complex LMS in regular widrow form and not block form)))/ But I guess my answer is........... a long long time from now...and dont hold your breath I think I get it now kw5tx Robert McGwier wrote: > Steve: > > That is a little difficult to say because the debugging process can be > painful. In this case, it is debugging SIMD assembly language > programs that need to work across multiple OS's and development > platforms. > > On September 4, I get to return to full time work on GPL software for > software radio for my work. I will be working on DttSP, GnuRadio for > work. Both of these need the speed up of this library though GnuRadio > already makes some use of SSE and AltaVec as does DttSP. We can do > more and we need to. > > The algorithms that need to fully implemented ASAP are: > > 1) The full scale assault on image rejection on both transmit and > receive using a derivation I did after seeing a hint on the Rocky SDR > page. > > 2) The full scale assault on impulse noise using methods associated > primarily with linrad. > > 3) The full scale assault on adjacent channel interference (sidebands > that leak into your passband) > > 4) The full scale assault on TX IMD through use of predistortion and > pre-modulation filtering. > > 5) The full scale assault on the use of better tools for graphics > display. I would rather use OpenGL since DirectX is MS only but I do > not carry all of the vote. It really is beginning to be impossible to > buy hardware that does not do OpenGL and this is cross platform. > > Contemporaneously with this set of battles, we need to get ready for > serious full duplex operation (satellite, etc.) and the second > hardware receiver, due out by the end of the year. We need the lower > CPU usage to make both of these available to a much larger range of > computer owners. Thus, the library work will go on in parallel. > > Recently, Eric has really carried almost 100% of the load. The other > person doing lots of work is Bill Tracey on making Janus/Ozy really > usable. Eric had to concentrate on the management of the internals of > the Flex 5000 and its manufacturing and testing. I can definitely > tell you it has swung away from this and back towards features and bug > fixes and renewed interest in doing better and newer algorithms, etc. > It is these kinds of things 1-5 that will make the Flex5000 really > unique and distinguishable from others I think. > > 73's and thanks for your support. The Flex people, the guys you never > hear from at the factory sitting behind Eric, Gerald, and John are > doing the hardest work of all and doing it well. Bringing out a thing > as complex and 'avant garde' as the 5000 is not easy and they are > doing it well. > > Bob > > > Steve Kirk (KW5TX) wrote: >> THANK YOU BOB! >> I really appreciate that update and the insight into workflow at FLEX >> programming >> ... and I understand about time constraints...we all have them...but >> any idea when? 3 months? 6 months? >> >> rock on >> steve >> kw5tx > > --- snip --- >>> >>> I am writing an sse/sse2/3dnow library that will help DttSP, >>> GnuRadio, Flex, etc. do lots of different things faster. It is >>> about 1/2 done for just the sse part and it makes no sense to >>> release it in this state. After this library is done, there will >>> be a collection of routines that call it to do things like sums of >>> squares, iir filters, etc. This will be a huge speed up in all >>> sorts of areas for us. With the release of the Flex 5000 and all of >>> the algorithmic work I need to do on new features, I just don't >>> have time to finish it right now. But when it is finished, I expect >>> we will be able to greatly improve the ANF and NR algorithms since >>> we will be able to afford to do a complex LMS in regular widrow form >>> and not block form. The agc percentage in the profiler will >>> plummet, etc. We will get to this. >>> >>> If we can just get our graphics impact on the CPU to plummet, we >>> would have wiped out most of the problems of running this code on >>> any modern computer. >>> >>> Bob >>> >>> >> >> _______________________________________________ >> FlexRadio mailing list >> FlexRadio@flex-radio.biz >> http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz >> Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ >> FlexRadio Knowledge Base: http://kb.flex-radio.com/ >> FlexRadio Homepage: http://www.flex-radio.com/ >> >> > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.flex-radio.biz/pipermail/flexradio_flex-radio.biz/attachments/20070810/30a218df/attachment.html _______________________________________________ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ FlexRadio Knowledge Base: http://kb.flex-radio.com/ FlexRadio Homepage: http://www.flex-radio.com/
