[linrad] Re: Linrad Network
Hello Joe, I've been experimenting with the Linrad Network. All seems well using options 4, 5, 6, or 7, but when I turn on option 8 (Send timf2) or option 9 (send fft2 transforms) the AFC graph and S meter graphs become very slow and jerky and the main spectrum and high resolution graphs seem to slow down some. If I try to X out, the screen freezes and I have to close the window and restart Linrad. CPU utilization is low, so I'm wondering if something is causing a conflict in the network processing thread for options 8 and 9? I'm using WinXP on a P4 with 1.5G RAM and Linrad 2.35. OK. I tested for this and I can reproduce the error. Consistent results on my laptop as well. I haven't checked how it behaves under Linux yet, but I have a hunch it's the same. No, I can not see this problem under Linux. The network is complicated because each time a big FFT has been completed a lot of data is released in a big chunk. Linrad has to send the data in small packets that are evenly distributed in time because the network does not like to be heavily loaded some times and idle in between. I have forgotten the exact details, but packet losses was the problem. Maybe it was the switches in my network that required this complication. I will look for the error. In the meantime you might run a first instance of Linrad that sends raw data to the network as well as timf2. You can then use a second instance of Linrad in which you hopefully will be able to click on signals and see normal behaviour of S-meter and AFC. 73 Leif / SM5BSZ # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: MAP65 and Linrad Network
Hi Joe, Yes, it should be possible to run everything on a single fast computer with plenty of memory. I have not yet established that the necessary CPU-sharing during time-critical parts of each program is handled adequately for this to work without glitches. Neither Linrad nor Windows is a very good real-time O/S, and one must work around their limitations in this area. I do not think there would be any problems:-) Moreover, one will almost certainly want separate screens for Linrad and MAP65 -- both of which generally use most or all of a normal-size screen. Yes, I agree - but those who have only one standard computer should know it is possible - if it really is;-) For testing MAP65 when real signals are not available, and anyway so that I can get 100% repeatable results, I solved this problem in a slightly different way. I saved some data by using the Linrad S command during the ARRL EME contest last November. With a slightly modified Linrad I converted the data to 16-bit TIMF2 format, and wrote it to a file. A simple program that I call plrs (for pseudo-Linrad send) can read this file and multicast it in the same way that Linrad would do, so that MAP65 can receive it. The data file, 11 minutes of original data, amounts to 507 MB. After compression with bzip2 it is 228 MB. I will make the data file available soon, together with plrs and MAP65, to anyone who wants to participate in testing MAP65. Hmmm, I would have liked a file that allows me to play with Linrad on the data that are sent to MAP65. Presumably you have a couple of spurs and perhaps other interesting interference that MAP65 (or at least the waterfall graph) would benefit from not getting. Surely, for MAP65 testing and development I agree that your strategy is more convenient - but from my point of view the original file is far more interesting 73 Leif / SM5BSZ # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: MAP65 and Linrad Network
Hi Leif, Moreover, one will almost certainly want separate screens for Linrad and MAP65 -- both of which generally use most or all of a normal-size screen. Yes, I agree - but those who have only one standard computer should know it is possible - if it really is;-) When MAP65 is available, I will be sure to document how it can best be used, and what the hardware requirements will be. Hmmm, I would have liked a file that allows me to play with Linrad on the data that are sent to MAP65. Presumably you have a couple of spurs and perhaps other interesting interference that MAP65 (or at least the waterfall graph) would benefit from not getting. This will not be a problem. I have the original raw data file and have placed it at http://physics.princeton.edu/pulsar/K1JT/06_0744.raw.bz2 It is 263 MB in size. One of the nuisances of using the original data file is that Linrad's raw data are not time-stamped. As you know, JT65 transmissions must start at the top of a UTC minute. To use the raw data file effectively with MAP65 you will need to do something to ensure that the multicast packets contain times that are close to being correct with respect to the original time (modulo 60 seconds). Surely, for MAP65 testing and development I agree that your strategy is more convenient - but from my point of view the original file is far more interesting Yes. It probably will be far more interesting, for you especially. You will find plenty of birdies and other junk in the test file!! See http://physics.princeton.edu/pulsar/K1JT/MAP65_1.JPG . -- Joe, K1JT # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: MAP65 and Linrad Network
Hi Joe,
I have the original raw data file and have placed it at
http://physics.princeton.edu/pulsar/K1JT/06_0744.raw.bz2
Excellent:-)
One of the nuisances of using the original data file is that
Linrad's raw data are not time-stamped.
Oooh! But they are:-) Your file is stamped 1163231107.413487 which
was the output of the routine current_time() at the start of
the recording.
double current_time(void)
{
struct timeval t;
gettimeofday(t,NULL);
return 0.01*t.tv_usec+t.tv_sec;
}
This is the time in seconds with six decimals since Epoch
and if you run Linrad on the file with a fast waterfall
you will see the time on the waterfall starting at 07.45.08
Unfortunately I have missed to put the correct time in the
time variable of the network packets. Currently the time
is the time when data was read from the hard disk and not
when it was stored there. I will correct for the next version:-)
The 'S' file only has a single time stamp for when write was
started. With Delta 44 soundcards, the sampling rate is
pretty accurately 96 kHz (96014 Hz with the card in this
particular computer.) I think one can safely assume that
the error is well below 0.05% so in a 1 hour recording
the time error at the end should stay well below 2 seconds.
As you know, JT65
transmissions must start at the top of a UTC minute. To use
the raw data file effectively with MAP65 you will need to do
something to ensure that the multicast packets contain times
that are close to being correct with respect to the
original time (modulo 60 seconds).
OK. I hope that you have found the time to be correct when
receiving timf2 data from Linrad in real time.
Surely, for MAP65 testing and development I agree that your
strategy is more convenient - but from my point of view the
original file is far more interesting
Yes. It probably will be far more interesting, for you
especially. You will find plenty of birdies and other junk
in the test file!! See
http://physics.princeton.edu/pulsar/K1JT/MAP65_1.JPG .
Fine!
I do not know how to interpret the MAP65 screen. The polarisation
data of AA1YN seems odd to me. They change very rapidly:
0746: 90
0748: 135
0750: 135
0752: 135
0754: 0
0756: 45
Which signal on the screen that has this behaviour is not obvious to me.
The processing screen says Freq 128 and DF=0 (or very close.)
When I look at the file, I find the weak signal at the center of the
lower waterfall to start at 144.129823 and drift down to 144.129814.
That is a 1.818 Hz frequency shift as compared to the MAP65 screen???
There is some info Options in the upper waterfall, but that does not
fit. The signal at 144.12982 is right hand circular in Linrad and it
does not change polarisation during the 10 minutes.
The strong signal at -880 Hz in the lower waterfall is at 144.12894
in Linrad. This signal is elliptic, about 50% left hand. Polarisation
does not change with time on this signal either.
How do I compute the true frequency of the sync tone from the data
in the MAP65 screen?
73
Leif / SM5BSZ
#
This message is sent to you because you are subscribed to
the mailing list linrad@antennspecialisten.se.
To unsubscribe, E-mail to: [EMAIL PROTECTED]
To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED]
To switch to the INDEX mode, E-mail to [EMAIL PROTECTED]
Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: MAP65 and Linrad Network
Hi Leif, The combination works OK, and but a significant number of packets (around 400 in each minute, or slightly more than 1% of the data) are dropped during MAP65's most compute-intensive parts of each minute. Since plrs does no significant computing, I imagine that the problem may be worse when running Linrad + MAP65. It could also be the other way around. Indeed, it could. In that case I will be surprised, but it will hardly be the first time I have been similarly surprised. It may be necessary to add a call to Sleep(0) under Windows or to sched_yield() under Linux at regular intervals in all routines that might lock up the CPU for a too long time. Those calls are effected by lir_sched_yield() in the OS independent code of Linrad and I found it necessary to place about 45 such calls within Linrad. Hmmm, I may have trouble doing the equivalent. For example, MAP65 does some very long FFTs using tthe FFFTW3 library. I can't very well put calls to relinquish CPU control inside these tasks. On the other hand, there is a threads version of FFTW3, which might be what's needed. Right now such issues are not a top priority, though. -- Joe, K1JT # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] MAP65 and Linrad Network
MAP65 Status I have now established the necessary network connection between Linrad and MAP65. So far, I have found nothing to prevent a happy marriage! MAP65 receives a full 90 kHz of xpol signal from Linrad, finds all the JT65 signals in that bandwidth, matches the linear polarization angle of each one, decodes the messages, and provides the operator with a band map showing callsigns, operating frequencies, polarization angles, and message contents over the past 15 minutes or so. The program provides a full-width waterfall display and another one zoomed in on the frequency of the station currently being worked. Questions about Linrad's TIMF2 Multicast Data - Header information accompanying Linrad's multicast TIMF2 data includes two single-byte parameters, userx_no and passband_direction, about which I have questions. I understand that userx_no should correspond to the number of RF channels, with negative sign indicating floating format. I would have expected the value -4 for my system using xpol antennas and the WSE converters (I and Q for each of two polarizations), but in fact I see -2. Is this as it should be? Similarly, I would have expected passband_direction=1 as I have no LO's on the high side and the spectrum is not inverted. However, Linrad sends passband_direction=-1. Is this correct? Linrad FFT Versions --- I have been using FFT Version 0 for the first forward, first backward, and second forward FFTs. This produces floating-point TIMF2 data on the multicast network. To save on memory usage in MAP65 it may be desirable to use 16-bit data for TIMF2. I have effected this by setting first forward FFT to version 5, first backward FFT to version 1, and second forward FFT to version 2. Are these good choices? Is there any downside to their use that I might not have thought about? Also, with these settings I notice that the signal in the high resolution graph (red dB lines) is higher than it was with FFT versions 0. Should I adjust some other parameter to bring this level down be several dB? MAP65 with Windows, Linux, FreeBSD, Winrad, SDRxxx ... ? Like WSJT, MAP65 runs on both Windows and Linux. (It should also run on FreeBSD and Macintosh, but I can't test those here.) In my station I will probably run Linrad under Linux and MAP65 on Windows, on a second computer. In response to questions I have been getting: MAP65 could run with Winrad or with various SDR's if their software is modified to provide the necessary multicasting capability. In itself, this would not be too difficult. However, one of the most important MAP65 features is its automatic Rx-polarization-matching capability. That requires a receiver with full xpol processing, which (as far as I know) is not now provided by Winrad, SDR-1000, SDR-14, or SDR-IQ. -- 73, Joe, K1JT # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: MAP65 and Linrad Network
Hello Joe, I have now established the necessary network connection between Linrad and MAP65. So far, I have found nothing to prevent a happy marriage! MAP65 receives a full 90 kHz of xpol signal from Linrad, finds all the JT65 signals in that bandwidth, matches the linear polarization angle of each one, decodes the messages, and provides the operator with a band map showing callsigns, operating frequencies, polarization angles, and message contents over the past 15 minutes or so. The program provides a full-width waterfall display and another one zoomed in on the frequency of the station currently being worked. Excellent!! Questions about Linrad's TIMF2 Multicast Data - Header information accompanying Linrad's multicast TIMF2 data includes two single-byte parameters, userx_no and passband_direction, about which I have questions. I understand that userx_no should correspond to the number of RF channels, with negative sign indicating floating format. I would have expected the value -4 for my system using xpol antennas and the WSE converters (I and Q for each of two polarizations), but in fact I see -2. Is this as it should be? Yes. You have two RF channels only. You might have used the hardware described here: http://www.sm5bsz.com/pcdsp/hware.htm This uses two audio channels to receive two RF channels but the output from timf2 would have the same format as you have with the WSE units. The difference is that the sampling speed of timf2 is the same as the audio sampling speed with WSE units, but half the audio sampling speed for the other solution. There is no need for MAP65 to know whether the 96kHz timf2 data originates in a hardware using two audio channels with a sampling rate of 192 kHz or four channels sampling at 96 kHz. Similarly, I would have expected passband_direction=1 as I have no LO's on the high side and the spectrum is not inverted. However, Linrad sends passband_direction=-1. Is this correct? You have the RX10700 with LO=13.175, 13.200, 13.225 or 13.250 MHz which is above 10.7. All the other LOs are below so -1 is correct. Linrad FFT Versions --- I have been using FFT Version 0 for the first forward, first backward, and second forward FFTs. This produces floating-point TIMF2 data on the multicast network. On a modern computer you can use version 5 for the first forward fft. It uses the SIMD instructions (single instruction multiple data, now called sse I think) and computes the floating point fft quite a bit faster. To save on memory usage in MAP65 it may be desirable to use 16-bit data for TIMF2. I have effected this by setting first forward FFT to version 5, The first fft is always floating point. It must be because the dynamic range of 16 bits is by far not adequate for the unprocessed A/D input. Version 5 is always a good choice on a computer that will allow it. Pentium II and older do not have the SIMD instructions so they have to use another version and it differs a little which one will run fastest depending on architecture. On old machines it might be desireable to actually try all versions and pick the one that runs fastest since old machines may be CPU limited. first backward FFT to version 1, and second forward FFT to version 2. Are these good choices? Yes. Is there any downside to their use that I might not have thought about? Also, with these settings I notice that the signal in the high resolution graph (red dB lines) is higher than it was with FFT versions 0. Should I adjust some other parameter to bring this level down be several dB? When the floating point data from fft1 is truncated to 16 bits there will be quantization noise. You should place the system noise floor at least 20 dB above this extra source of noise to make the loss of NF smaller than 0.043 dB. Press A on the keyboard to see the amplitude margins. In case you place the noise floor too high you might find that there is saturation somewhere. The shift parameters for the FFTs will allow you to set the noise floor at the correct level. http://www.sm5bsz.com/linuxdsp/install/dlevel.htm There are many possible sources of rounding errors (quantization noise) and Linrad does not set levels automatically since the criteria for what will be optimum are non-trivial. I hope the above link will be helpful. Under normal circumstances you will not need the maximum dynamic range in the second FFT. Only if you have a carefully calibrated system and want to use the smart blanker saturation on noise pulses will be a problem. Narrowband signals are automatically attenuated to fit within the dynamic range of 16 bits. In case you want a very large size for the second fft and a small size for the first fft, the narrowband signals have to be attenuated to a pretty low level. A perfect sinewave may put nearly all its energy in a single FFT bin.With a
[linrad] Re: Using the Linrad Network
Leif and all, It turns out that my network speed problem had a simple cause. The computer running Linrad is connected to the LAN through a rather old 10 Mbit/s hub/repeater. Therefore it maxes out around 1.2 MByte/s, too slow for 4 channels at 96 kHz in any of Linrad's network modes except mode 4. Looks like I need a new hub, or at least a simple crossover cable between the two computers. I will report further, fairly soon I hope, on my effort to connect Linrad to MAP65. -- Joe, K1JT # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: Using the Linrad Network
Hello Joe, Today I made my first tests of Linrad's network broadcasting capability. I installed Linrad 02.34 and checked it out. All seemed to be well. I selected N on the main menu to initialize networking parameters, and then selected format 4 (16-bit raw data) for output. After toggling network write ON, I found that the program continued to run correctly in Weak CW mode. Ok. Unfortunately, I did not have good luck with any of the other data formats. In particular, formats 7, 8, and 9 caused the program to produce broken sound output -- that is, frequent gaps in the Rx background noise when an active frequency is selected with the mouse. Hitting T while receiving, so as to display timing information, shows an apparent A/D sampling rates as follows: Format A/D --- 4 96 kHz 7~30 8~73 9~49 Obviously, something is seriously wrong! Yes. While writing this mail on my single Pentium 2.66 MHz machine, I have one issue of Linrad running two RF channels at 96 kHz from the WSE units sending fft1 data to the network. I also have two more issues of Linrad receiving the fft1 transforms. All three issues are sending data to the loudspeakers simultaneously. I notice no problems of any kind. This is under Debian with kernel 2.6.18-4-686-bigmem and oss-linux_v4.0-1002_i386.deb When I start the system monitor I get a short break just when the program opens. All three Linrad issues have the second FFT enabled and the total CPU load according to the system monitor is about 40% with occasional peaks up to 70%. When I start a fourth issue of Linrad that reads 154 kHz from an SDR-14 and also sends data (16 bit raw) to the network, things still work fine with a total CPU load of 60% using 1.2 gigabytes of RAM. In this situation, the USB driver for SDR-14 crashes in case I try click on any other window because there is not an appropriate error handeling when data is lost from the USB device. This does not affect the other three issues of Linrad. It seems to me that the problems you see could be caused by excessive RAM allocation that would cause active data areas of Linrad to be swapped on the hard disk. It could also be a problem due to interrupt sharing. I noticed a bug. In case I press T when Linrad runs from fft1 input, the A/D rate becomes 192 kHz and an error message is displayed. This does not affect the processing however. I did notice that when using Format 9, although the audio on the master computer has frequent interruptions, watzo running on a second computer displays waterfall data that looks more or less correct. I tried to use one instance of xlinrad to broadcast format 4 data, and a second instance of xlinrad on the same computer receiving the raw 16-bit data. This failed with error #1272 (Write to network socket returned an error) on the master and error #1274 (Network read failed) on the slave. Something is wrong. Presumably the reason is the same as for the other errors you observe. Can you help me to understand what may be wrong in my setup? The computer on which Linrad is running is reasonably fast (2.7 GHz P4, 0.5 MB cache, 1.25 GB memory), and is around 20% busy when running xlinrad. Check for memory usage first. 73 Leif # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Re: Using the Linrad Network
Hi Leif, Thanks for the reply. Something is wrong. Presumably the reason is the same as for the other errors you observe. ... Check for memory usage first. No swapping is taking place, so I don't think it's a memory issue. I will study the problem more and report back if I learn anything useful. -- Joe # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Using the Linrad Network
He Leif and all, Today I made my first tests of Linrad's network broadcasting capability. I installed Linrad 02.34 and checked it out. All seemed to be well. I selected N on the main menu to initialize networking parameters, and then selected format 4 (16-bit raw data) for output. After toggling network write ON, I found that the program continued to run correctly in Weak CW mode. Unfortunately, I did not have good luck with any of the other data formats. In particular, formats 7, 8, and 9 caused the program to produce broken sound output -- that is, frequent gaps in the Rx background noise when an active frequency is selected with the mouse. Hitting T while receiving, so as to display timing information, shows an apparent A/D sampling rates as follows: Format A/D --- 4 96 kHz 7~30 8~73 9~49 Obviously, something is seriously wrong! I did notice that when using Format 9, although the audio on the master computer has frequent interruptions, watzo running on a second computer displays waterfall data that looks more or less correct. I tried to use one instance of xlinrad to broadcast format 4 data, and a second instance of xlinrad on the same computer receiving the raw 16-bit data. This failed with error #1272 (Write to network socket returned an error) on the master and error #1274 (Network read failed) on the slave. Can you help me to understand what may be wrong in my setup? The computer on which Linrad is running is reasonably fast (2.7 GHz P4, 0.5 MB cache, 1.25 GB memory), and is around 20% busy when running xlinrad. -- 73, Joe, K1JT # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Network and bits
Hi all, The network is now working fine although several things are still missing. I have decided to abandon the idea of using 1024 byte fixed block size because the network in itself is too much CPU load to make it attractive on old computers at high data rates. As it turns out the size of the packet does not matter at all for the CPU load and it does not matter when two computers are tied directly to each other. When the packets have to pass a switch it is different however. At high rates the network does not work through the switch if the packet size is a little above 1500 bytes, but it works well when the size is a little below. Besides allowing MAP65 and other external softwares the Linrad network is intended to encourage experimentation. I have decided to not freeze the protocol at an early stage. With routines that copy data from a buffer of an arbitrary size (it has to be a multiple of 12 however) the size can be adjusted in case there will be future needs. The current payload is 1404 bytes plus a 24 byte header. One of the obvious experiments is to process the data with different number of bits simultaneously. The result is presented here: http://www.sm5bsz.com/linuxdsp/run/network.htm There is also some more info about the network on this page. 73 Leif / SM5BSZ # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] AW: [linrad] Network speed problems.
Hello, here is what we found when we had some network performance problems in the office: se a well known search engine and look for the phrase net.core.rmem_default for example described in http://wwwx.cs.unc.edu/~sparkst/howto/network_tuning.php Please do not ask me any special details, I only read it and took it. During some earlier network optimizations we also found that just looking at the raw amount of data beeing transfered is only partly correct. Some network stacks reach the limit first by the number of packets, not the amount of data. So it might be useful to make transport blocks as large as possible. Just like a Layer4, just inverse. 73, Günter, DL4MEA -Ursprüngliche Nachricht- Von: Linrad mailinglist [mailto:[EMAIL PROTECTED] Im Auftrag von Leif Asbrink Gesendet: 07 January 2007 00:28 An: Linrad mailinglist Betreff: [linrad] Network speed problems. Hi all, Having been sucessful with raw data at speeds up to 96000*4*(16+18+24)=22 Mb/s I sort of believed it was possible to run at rates reasonably near 100 Mb/s when only sendind data in one direction since the machine says: Link is up at 100Mbps, full duplex Now, as it turns out I did the above test with only one RF channel so the speed was only 11 Mb/s. What is more of a problem is that the maximum throughput through one socket is actually just a little below 96000*4*16 = 6.1 MB/s. It is perfectly ok to send 5MB/s through four sockets simultaneously but sending fft1 transforms through a single socket is impossible. Presumably this is well known, I an just a newcomer to networking and I could not guess it would behave this way. Is there some settings I should change? One possibillity seems to be to open several sockets on different ports simultaneously to increase the throughput. I can not send even 16 bit noise-blanked time domain data from Linrad to MAP65 over the network on a single socket. 14 bit data will be fine however because the Linrad processing removes all strong signals when the MMX routines are selected. Is there something wrong in what I do? Does Windows behave the same way? To be able to run at 96 khz with fft1 transfer it seems I will have to use at least 6 sockets in parallel with different port numbers. It is interesting to note that the old Pentium MMX at 200 MHz uses very much more CPU (under an old kernel) but that the throughput through one socket does not differ much. It can not send more than a 16 bit channel because there is not enough CPU time available. Trying to find some hints on the Internet I came across this: http://www.ncsa.uiuc.edu/UserInfo/Resources/Hardware/IBMp690/IBM/usr/share/man/info/en_US/a_doc_lib/aixbman/prftungd/2365c93.htm * Do not change the default (and maximum) MTU of 1500 bytes. * Set the application block size in multiples of 4096 bytes. * Keep socket space settings at the default values. * If the workload includes extensive use of services that use UDP, such as NFS or RPC, increase sb_max to allow for the fact that each 1500-byte MTU uses a 4096-byte buffer. It seems to indicate that I should send much larger packets??? (a multiple of 4096 bytes 'header' included) 73 Leif / SM5BSZ # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED] # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
[linrad] Network speed problems.
Hi all, Having been sucessful with raw data at speeds up to 96000*4*(16+18+24)=22 Mb/s I sort of believed it was possible to run at rates reasonably near 100 Mb/s when only sendind data in one direction since the machine says: Link is up at 100Mbps, full duplex Now, as it turns out I did the above test with only one RF channel so the speed was only 11 Mb/s. What is more of a problem is that the maximum throughput through one socket is actually just a little below 96000*4*16 = 6.1 MB/s. It is perfectly ok to send 5MB/s through four sockets simultaneously but sending fft1 transforms through a single socket is impossible. Presumably this is well known, I an just a newcomer to networking and I could not guess it would behave this way. Is there some settings I should change? One possibillity seems to be to open several sockets on different ports simultaneously to increase the throughput. I can not send even 16 bit noise-blanked time domain data from Linrad to MAP65 over the network on a single socket. 14 bit data will be fine however because the Linrad processing removes all strong signals when the MMX routines are selected. Is there something wrong in what I do? Does Windows behave the same way? To be able to run at 96 khz with fft1 transfer it seems I will have to use at least 6 sockets in parallel with different port numbers. It is interesting to note that the old Pentium MMX at 200 MHz uses very much more CPU (under an old kernel) but that the throughput through one socket does not differ much. It can not send more than a 16 bit channel because there is not enough CPU time available. Trying to find some hints on the Internet I came across this: http://www.ncsa.uiuc.edu/UserInfo/Resources/Hardware/IBMp690/IBM/usr/share/man/info/en_US/a_doc_lib/aixbman/prftungd/2365c93.htm * Do not change the default (and maximum) MTU of 1500 bytes. * Set the application block size in multiples of 4096 bytes. * Keep socket space settings at the default values. * If the workload includes extensive use of services that use UDP, such as NFS or RPC, increase sb_max to allow for the fact that each 1500-byte MTU uses a 4096-byte buffer. It seems to indicate that I should send much larger packets??? (a multiple of 4096 bytes 'header' included) 73 Leif / SM5BSZ # This message is sent to you because you are subscribed to the mailing list linrad@antennspecialisten.se. To unsubscribe, E-mail to: [EMAIL PROTECTED] To switch to the DIGEST mode, E-mail to [EMAIL PROTECTED] To switch to the INDEX mode, E-mail to [EMAIL PROTECTED] Send administrative queries to [EMAIL PROTECTED]
