Discussion Forums

Community Meeting 2012-09-18

1 reply [Last post]
sigblips
sigblips's picture
Offline
Joined: 2010-04-20
Posts: 733

[11:02am] sigblips: Welcome to the weekly setiQuest community IRC meeting.
[11:03am] sigblips: jrseti won't be here today.
[11:04am] sigblips: I have a couple topics that I'll go through. If you have topics or questions then please speak up.
[11:05am] eb3frn: this is Iban from Spain
[11:05am] sigblips: Hello Iban.
[11:06am] eb3frn: I wish reproduce the voyager data files here to do some tests with autodrift
[11:06am] eb3frn: but need the strip header utility
[11:06am] eb3frn: it is available?
[11:08am] sigblips: I haven't released it. No one was really interested in it. It is really simple. Let me find it and post it today.
[11:08am] eb3frn: then I'm very interested :)
[11:08am] eb3frn: ok, tnx
[11:08am] eb3frn: I'm trying to detect voyager using my radiotelescope, and it is a nice example to do some tests
[11:09am] sigblips: Are you interested in reproducing what I did here http://baudline.blogspot.com/2011/03/setiquest-voyager.html ? What is your "radioscope"?
[11:10am] eb3frn: yes exactly
[11:10am] eb3frn: I' have a 8ft dish with feedhorns for 8.4ghz and 1.4ghz
[11:11am] eb3frn: currentlly I'm receiving a lot of spaceprobes like MRO, MEX, VEX, STEREO A/B without problems
[11:11am] eb3frn: but voyager is a challenge
[11:12am] sigblips: Ohh, so you have your own radio telescope. Cool. Unfortunately you won't be able to detect Voyager since it the signal is too weak. They can't detect it at the ATA anymore and they have 42 * 20 foot dishes.
[11:12am] sigblips: What sample rate do you collect your data at?
[11:13am] eb3frn: depends of the sdr used, mainly I'm using rfspace sdr14, or now the rtlsdr's
[11:13am] eb3frn: the samplerate for DSN should be over 250ksps
[11:13am] eb3frn: now I'm doing tests with a 125+125msps sdr for pulsars and seti at 1420
[11:14am] sigblips: 125 mega samples per second? That is fast. That ADC must of been expensive.
[11:14am] eb3frn: yep, not expensive
[11:14am] eb3frn: over 800
[11:14am] eb3frn: is the phi sdr
[11:15am] eb3frn: check this: http://www.eb3frn.net/phi-bb.jpg
[11:15am] eb3frn: with baudline in a i7 I'm getting over 7 / 10 fps at 250msps
[11:16am] sigblips: Wow, I'm impressed that an i7 can keep up at that data rate.
[11:17am] eb3frn: yep, it is a pci-e card
[11:17am] eb3frn: the screenshot is using the real mode
[11:17am] sigblips: What is real mode?
[11:17am] eb3frn: here anoter one using complex: http://www.eb3frn.net/phi-rf-108Mhz.jpg
[11:18am] eb3frn: one channel of data without imaginary part
[11:18am] eb3frn: the card works in baseband using real data, and using simple converion rx with complex data
[11:19am] eb3frn: yep, really amazing how is working baudline with this big amount of data
[11:20am] sigblips: No I channel? It looks like got I/Q data to me, slight negative frequency offset.
[11:20am] eb3frn: yep, only one channel
[11:21am] eb3frn: the card us sending one short to send 1 sample in real, and two shorts to send i and q un complex mode
[11:21am] sigblips: Hmm, you're using the -quadrature flag, if the channels were duplicates (real only) then the spectrum would be a mirror image around 0 Hz.
[11:22am] eb3frn: the card have two adc's here http://www.eb3frn.net/phi-rf-108Mhz.jpg I'm in complex mode then the card is using a adc for I and the second adc for q
[11:23am] sigblips: So you are using the -quadrature flag and sending I/Q (2 channels)?
[11:23am] eb3frn: yes
[11:23am] eb3frn: here http://www.eb3frn.net/phi-bb.jpg , forget the params that ara shows in the top right xterm
[11:24am] eb3frn: they are used before to take the complex screenshot
[11:24am] sigblips: What are the specs of your computer that is running this? What type of i7 do you have?
[11:24am] eb3frn: for the real mode dont using quadrature, and only using 1ch
[11:24am] eb3frn: it is a i7 3770k
[11:24am] eb3frn: only 1chip, have 4 cores and 8threads
[11:25am] eb3frn: if need some test, will be a pleasure do all that you need
[11:26am] sigblips: Just looked it up. 3.5 GHz with a turbo up to 3.9 GHz. 8 MB cache.
[11:26am] eb3frn: yep
[11:26am] eb3frn: but dont using the turbo mode
[11:27am] sigblips: Yeah, that's only for when the thermals are low like when single threading or for short CPU bursts.
[11:27am] eb3frn: to receive pulsars with the card I'm looking to convert the data to psrfits or filterbank, to use presto and sigproc
[11:27am] eb3frn: but to use it in seti, dont sure yet how manage all the data
[11:28am] eb3frn: because not seeing fft packages to do lerge fft's
[11:28am] sigblips: What you're doing is really cool. You should write it up, blog about it, or something.
[11:28am] eb3frn: yep, when have a bit of time, I will do
[11:29am] eb3frn: I hope encourage some people, to do RA and seti
[11:29am] eb3frn: I'm looking some fft package with cuda, but dont see nothing
[11:30am] sigblips: I have been working on the next version of baudline that does larger FFTs (up to 16M) and I've done some major optimizations to the -stdin data pipeline. What you're doing would be perfect for a beta test. Let me build a test version for you later today.
[11:30am] eb3frn: wow! very niece to hear it!
[11:30am] eb3frn: of course, will be a pleasure test it
[11:31am] sigblips: I'm still surprised to see that a 4-core 3,5 GHz i7 can sustain 250 MSPS into baudline running a 65536 point FFT.
[11:31am] eb3frn: yes, runs very smooth
[11:32am] eb3frn: only crash if using overlap
[11:32am] eb3frn: when do something with the scroll control, the program stops and frizes
[11:33am] sigblips: With an overlap of 100% that is 3846 FFT/sec.
[11:33am] eb3frn: not, bad, hehh
[11:33am] sigblips: Let's see if the new -stdin pipeline I mentioned helps.
[11:33am] eb3frn: ok, tnx
[11:34am] eb3frn: get up to 16M of fft will be great for seti
[11:34am] eb3frn: you have doing something for cuda?
[11:35am] eb3frn: the "big boys" of the pulsars are using it to dedisperse signals of more than 100Mhz of bw
[11:35am] sigblips: Problem with 16M point FFTs are that they are slow. A new high-end Intel i# core CPU can do about 3 per second per core.
[11:36am] eb3frn: with a 2gb cuda grafic card can do 64M fft
[11:36am] eb3frn: not bad
[11:36am] sigblips: Do you know how many a second it can do?
[11:36am] eb3frn: then the issue will be check how much bw can use to do 16M point
[11:37am] eb3frn: who? dont understand the question
[11:37am] sigblips: How many 64M point FFT/second can the CUDA graphics do?
[11:38am] eb3frn: dont know exactly, but sure that a lot
[11:38am] sigblips: I'd guess that a fast Intel Xeon would do less and 1 per second per core.
[11:39am] sigblips: At that point it is a memory bandwidth problem. Not sure AVX will help much.
[11:40am] eb3frn: of course I'm talking using the last ones like gtx680 that have over 1500 cores and peaks of more of 2teaflops
[11:41am] sigblips: It would be interesting to know those benchmarks.
[11:41am] eb3frn: this cards are using ddr5 of 6Gbps with mem bandwidth of 192GB/s
[11:41am] eb3frn: I will find
[11:43am] eb3frn: the trouble is that most of the benchmarks are a bit old, using the firts cards, over 4 or 5 years
[11:43am] sigblips: From your screenshots I see that you are getting some font errors with baudline. Check out the two font question/answers here http://baudline.com/faq.html#Fedora13_fonts . They might help.
[11:44am] eb3frn: ok, I will check
[11:44am] eb3frn: tnx!
[11:45am] eb3frn: dont see nothing, when have a bit of time, will try to do some benchmarks
[11:45am] eb3frn: this is the package
[11:45am] eb3frn: http://gamma.cs.unc.edu/GPUFFTW/
[11:45am] sigblips: Also enabling backing store might improve your graphics performance a bit. http://baudline.com/faq.html#turn_on_backing_store
[11:47am] sigblips: GPUFFTW is for CUDA. I also saw that there is a CUDA version of Intel's IPP library.
[11:47am] eb3frn: will try but the -sysinfo says that not is usefulll
[11:48am] eb3frn: yes
[11:48am] eb3frn: what fft package you are using?
[11:49am] sigblips: My own that I wrote many years ago. It is scalar. It's impressive that it is as fast as it is. I'm looking at integrating IPP for the next version of baudline.
[11:50am] eb3frn: wow, then congratulations
[11:50am] eb3frn: it is rally fast
[11:51am] sigblips: With baudline -sysinfo having backing store = NotUseful means that backing store isn't enabled.
[11:52am] sigblips: I don't know why X11 calls it NotUseful when it really means it is Off.
[11:52am] eb3frn: yep
[11:53am] sigblips: Do you have any other baudline questions?
[11:54am] eb3frn: not for today :)
[11:54am] eb3frn: thanks a lot for your help
[11:54am] sigblips: You're welcome.
[11:54am] sigblips: OK let me quick mention a couple setiQuest news items.
[11:55am] eb3frn: of course
[11:55am] sigblips: The ATA is still off-line for an air conditioner upgrade. It is scheduled to be back on-line Sept 24th. http://setiquest.info/
[11:56am] sigblips: Uwingu extended their funding by 10 more days. Currently they are at 79% funded, they've raised $59,230 with 7 days left. http://www.uwingu.com/do-you-uwingu/
[11:57am] sigblips: The SETI Institute will get half of everything they get that is above $75K. So Uwingu fuding could help SETI.
[11:58am] sigblips: Here http://www.csicop.org/specialarticles/show/pointing_the_telescopes is a great interview with Jill Tarter. She talks about a number of topics including setiQuest.
[11:59am] sigblips: I also heard that NASA is building a couple new 34m dishes in Australia to expand its Deep Space Network (DSN). http://www.cdscc.nasa.gov/Pages/antennas.html
[12:01pm] sigblips: That's it for the setiQuest news updates.
[12:02pm] sigblips: Ohh one more important item.
[12:03pm] sigblips: There has been some discussion about canceling these weekly community IRC meetings. What are peoples thoughts on this?
[12:04pm] eb3frn: it is a pain, will it is my firts one
[12:04pm] eb3frn: I hope that it not be my last tuesday here
[12:05pm] sigblips: Nothing has been decided yet.
[12:06pm] sigblips: Some people only communicate with setiQuest via IRC so cutting off their only communication channel would be a bad thing.
[12:07pm] eb3frn: yes
[12:08pm] eb3frn: in my case I'm trying to connect here scince a lot of weeks, but I was connecting when the meeting was closed
[12:09pm] sigblips: Next week jrseti may elaborate on the SETI Institute's position on this. My guess is that they feel that they aren't getting enough value from it.
[12:10pm] eb3frn: a last comment from here, about voyager it is very important check the sfos to see if the xband status
[12:11pm] eb3frn: http://voyager.jpl.nasa.gov/mission/soe-sfos/tracking_schedule.html
[12:11pm] sigblips: sfos?
[12:11pm] eb3frn: it are the operations schedules
[12:12pm] eb3frn: they are very usefull to dont try to detect the probe in blind mode
[12:12pm] sigblips: Wow, I didn't even know that NASA had a page like that. Thanks.
[12:12pm] eb3frn: yes, it is very usefull
[12:12pm] eb3frn: to check the x-band perofrmance of the ata, can be nice the detection of some probe, like stereo-a
[12:13pm] eb3frn: it is a strong signal and his schedule is public,
[12:14pm] sigblips: We might of seen Stereo A or B in the L4 dataset. I'll post something in the forum today. http://setiquest.org/forum/topic/baudline-analysis-lagrange-4
[12:16pm] eb3frn: but you have detectd it in x-band?
[12:17pm] sigblips: Not sure. The capture was at 3991 MHz which is the wrong frequency I think. What are your thoughts?
[12:19pm] eb3frn: Get the signal from st-a or st-b at 8.4ghz to check the performance of the ata at x-band
[12:20pm] eb3frn: dont sure but I feel that can be a bit deaf, becasue I was hoping a better voyager detection
[12:21pm] eb3frn: here is the schedule of the stereos
[12:21pm] eb3frn: http://stereo-ssc.nascom.nasa.gov/plans/dsn_schedule.txt
[12:22pm] eb3frn: is the schedule of the DSN mode when are sending data to the earth, the rest of time are in beacon mode
[12:22pm] sigblips: Some new ATA datasets would be great but I don't think the SETI Institute is interested. I'm still trying to convince Gerry though. I haven't given up yet.
[12:22pm] eb3frn: the difference of SNR from dsn to beacon is over 6db
[12:23pm] eb3frn: well, it is only a idea to check the performance at high freqs
[12:24pm] sigblips: Yeah, the higher frequencies are more difficult. There are only a couple setiQuest datasets in the 8 GHz range. http://setiquest.org/wiki/index.php/SetiQuest_Data and most of those have some issues.
[12:25pm] eb3frn: yes I was check it recentlly but the spectrum don shows signal of the probes that are txing all the time
[12:26pm] eb3frn: you know what bin_size is used in the spectrum showed in the list?
[12:27pm] sigblips: I think the DSN is using a 70m dish (or a 70m and 35m combined) to communicate with Voyager.
[12:28pm] sigblips: So are you saying that you've seen Voyager on your 8 foot dish?
[12:28pm] eb3frn: no, only 70mts, but using all the ata dishes sure that can be detectable
[12:28pm] eb3frn: not yet, but I hope do in few months
[12:28pm] eb3frn: I'm training and doing some tests with fake signals, for this reason wish use the ata dataset
[12:29pm] eb3frn: luis cupido was detect it using a 5mts dish, using 15m of integration
[12:29pm] eb3frn: http://www.qsl.net/ct1dmk/dsn.html
[12:29pm] sigblips: Well let's bring that up at the next IRC meeting. An ATA Voyager detection might be a fun event that SETI Institute is interested in.
[12:30pm] eb3frn: ok, then the next tuesday more
[12:30pm] eb3frn: thanks a lot for all
[12:30pm] eb3frn: see you the next tuesday!
[12:31pm] sigblips: Well if someone can detect it with a single 5m dish then 42 of the ATA's 6m dish shouldn't have any problem detecting it.
[12:31pm] sigblips: One more thing. The setiQuest forum http://setiquest.org/forum passed 5000 users this morning.
[12:31pm] eb3frn: sure
[12:32pm] eb3frn: for this reasing, I belive that exists some trouble in the x-band and can be nice check with another probe
[12:32pm] eb3frn: well, can discus it the next monday
[12:32pm] eb3frn: bye
[12:33pm] sigblips: Great meeting. Same channel, same time, next week. {for now} ): Bye.
[12:35pm] sigblips: Please voice your thoughts about the future of the weekly setiQuest community meetings in the forum.

sigblips
sigblips's picture
Offline
Joined: 2010-04-20
Posts: 733
Summary:

Some SETI news about the ATA air conditioning, Uwingu funding, Jill Tarter interview, and new NASA 34m DSN dishes can be found in the the middle of the IRC transcript at 11:54 AM.

The setiQuest forum reference link for the Voyager signal analysis topic is http://setiquest.org/forum/topic/baudline-analysis-voyager

The possible cancellation of the weekly community IRC meetings is so important that a new topic has been created. http://setiquest.org/forum/topic/future-irc-meetings Please comment about the future of the IRC meetings there.