Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
On 3/23/06, Lee A Crocker [EMAIL PROTECTED] wrote: and Sami Aintila telle me I will get 3,579,545 Hz even if I have ACPI 2.0 running. That's not exactly what I said. I said that performance counter frequency does not NECESSARILY relate to ACPI version. So you can get the ~3.5 MHz reading even if your system is ACPI 2.0. But in most cases you probably should get the CPU clock frequency instead. The explanation that followed was probably confusing. I just described a CPU instruction that provides sub-nanosecond precision regardless of what you get with the performance counter. (And regardless of ACPI version). But the most important thing was to confirm what Bob already said in his first post: This whole high-resolution clock issue was basically bogus. We certainly don't need sub-ns precision for CW timing. The 3.5 MHz frequency should be adequate. So the FM isn't all that clear or is it? Regarding the FM (as Bob and you so passionately call it), I don't find some things you mentioned in my version (1.5.12) of the SDR-1000 Operating Manual. Maybe you have an older version? 73, Sami OH2BFO
[Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
I went to look at the website where perf resides, and this is the authors blurb regarding that applet: perf (download source) (download executable) This is a simple Win32 utility I use to figure out whether a machine has support for the high-resolution performance counters necessary for microsecond-level timers in user land. It just calls the QueryPerformanceFrequency() Win32 API and throws up a dialog with yay/nay and some statistics So I think when perf reports High resolution counters supported that is the main bit of information. As to the manual I am reading the .pdf file I downloaded from the flex website. I have no idea how old or up to date this is. In fact it never occurred to me this might be an obsolete manual. I know this was discussed some months ago on the reflector, before I made my SDR purchase, but since the reflector's server (aka mailman) is not readily searchable (although mailman can be made to have a completely searchable automatically updating archeive since all the reflectors over on contesting.com are set up like this), that information is as good as lost, and I haven't found the page in the manual where it discusses K1EL keyers that use special PTT techniques over serial ports. No I do not have the archeive residing on a PC based email client. So any suggestions I just use the search function in my email client is not a useful suggestion. I use a web based email client and read the reflector from the flex server. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Lee: This is a really good point. One where we can agree completely. We are so long between major releases, but constantly doing interim releases, that we probably should update the documentation as we go along with the code as it is developed. The only reason 1.6.0 is not out today is we have spent two weeks derailed on other things and jhave not been able to finish the documentation. The official download document from Flex is ALWAYS for the last official release and thusyour problem. As such, I added a documentation directory to the svn stuff. Again, this will not be its final resting place but I feel that if Flex is going to tout the latest and greatest, the documentation should try to keep up. Now that open office and word can both write the document, I see no reason not to use doc format. I would make a further suggestion to Flex. That we break the document into sections. It is a trivial matter to print from page X to Y using CutePDF. This will involve work and I do not wish to delay 1.6.0's release with this so I would expect the documentation to be upgraded to this slowly while the monolithic version comes out. This would require changing the pagination to be section based such a I-1, I-2 and then II-1,2, etc. But this would mean you could change an entire section of the manual and it would not impact any of the other sections. As you can see, this implies effort that we cannot allow to stop release of 1.6.0. Your points are very well taken. Bob Lee A Crocker wrote: I went to look at the website where perf resides, and this is the authors blurb regarding that applet: perf (download source) (download executable) This is a simple Win32 utility I use to figure out whether a machine has support for the high-resolution performance counters necessary for microsecond-level timers in user land. It just calls the QueryPerformanceFrequency() Win32 API and throws up a dialog with yay/nay and some statistics So I think when perf reports High resolution counters supported that is the main bit of information. As to the manual I am reading the .pdf file I downloaded from the flex website. I have no idea how old or up to date this is. In fact it never occurred to me this might be an obsolete manual. I know this was discussed some months ago on the reflector, before I made my SDR purchase, but since the reflector's server (aka mailman) is not readily searchable (although mailman can be made to have a completely searchable automatically updating archeive since all the reflectors over on contesting.com are set up like this), that information is as good as lost, and I haven't found the page in the manual where it discusses K1EL keyers that use special PTT techniques over serial ports. No I do not have the archeive residing on a PC based email client. So any suggestions I just use the search function in my email client is not a useful suggestion. I use a web based email client and read the reflector from the flex server. -- Robert W. McGwier, Ph.D. Center for Communications Research 805 Bunn Drive Princeton, NJ 08540 (609)-924-4600
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Lee, I have put a word in with our reflector host to inquire about the previous email suggestions about making it searchable. I'll post here when I find something out. It is likely that we will be adding our archive to the mail-archive.com website. Eric Wachsmann FlexRadio Systems -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] radio.biz] On Behalf Of Lee A Crocker Sent: Thursday, March 23, 2006 8:30 AM To: FlexRadio@flex-radio.biz Subject: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce --snip-- I know this was discussed some months ago on the reflector, before I made my SDR purchase, but since the reflector's server (aka mailman) is not readily searchable (although mailman can be made to have a completely searchable automatically updating archeive since all the reflectors over on contesting.com are set up like this), that information is as good as lost, and I haven't found the page in the manual where it discusses K1EL keyers that use special PTT techniques over serial ports. No I do not have the archeive residing on a PC based email client. So any suggestions I just use the search function in my email client is not a useful suggestion. I use a web based email client and read the reflector from the flex server.
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
What processor are you using? I have a Athlon 2X (dual core) 4400+ (2.2 GHz) on an Asus A8V Deluxe (VIA chipset) and I get the exact same values you report. If you have a different processor, then there may be a bug in the program. -Tim --- Tim Ellison ( [EMAIL PROTECTED] ) Integrated Technical Services ( http://www.itsco.com ) -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of root [knesbitt] Sent: Wednesday, March 22, 2006 12:47 AM To: FlexRadio@flex-radio.biz Subject: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce Curious if people have run the Performance Counter Query Tool http://lightconsulting.com/~thalakan/perf.exe http://lightconsulting.com/%7Ethalakan/perf.exe to determine whether their motherboard and chipset supports the high resolution clock. I'm particularly interested in the results obtained from users of AMD Athlon64 motherboards using the newer nVidia nForce4 or nForce 430 chipset. My new board seems to be indicating a clock frequency of only 3579545 hz in spite of nVidia's claim to adhere to the ACPI 2.0 standard. My understanding, if I read the SDR-1000 manual correctly, is you must see a clock frequency approaching the clock rate of the CPU, in other words in the Ghz range to be compliant with ACPI 2.0, and thus capable of utilizing high-resolution timing in the PowerSDR cw setup (a must). Are all new boards not created equal? Appreciate hearing which boards do report the higher clock speeds. Kirb, VE6IV -- ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://mail.flex-radio.biz/pipermail/flexradio_flex-radio.biz/ FlexRadio Homepage: http://www.flex-radio.com
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Let me say that I am now of the opinion that all of this high resolution clock business was something of a diversion from the real problems. Anyway, on my machine ACPI 2.0 enabled, I read 3,200,000,000 as the clock frequency and 0.000313 uSec as a tick, etc. using perf.exe What I think really matters is what blocks your keyer? I believe we have removed many of the impediments to good performance in the last six weeks. On my ASUS motherboard, I had to TELL IT to use ACPI 2.0. I did this in the bios setup. The high resolution clock in the setup panel for the keyer means you can safely use 1 msec timer. Otherwise it chooses 10msec. The impact of 10msec is that you really are limited to 20wpm or less with any hope of getting reasonable performance. It is time to revisit the question: are there folks getting reasonable CW performance using the high res timer in the code even though perf.exe reports 3.5 MHz? Bob Tim Ellison wrote: What processor are you using? I have a Athlon 2X (dual core) 4400+ (2.2 GHz) on an Asus A8V Deluxe (VIA chipset) and I get the exact same values you report. If you have a different processor, then there may be a bug in the program. -Tim --- Tim Ellison ( [EMAIL PROTECTED] ) Integrated Technical Services ( http://www.itsco.com ) snip --- -rt the higher clock speeds. Kirb, VE6IV -- AMSAT VP Engineering. Member: ARRL, AMSAT-DL, TAPR, Packrats, NJQRP/AMQRP, QRP ARCI, QCWA, FRC. ARRL SDR Wrk Grp Chairman Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
On 3/22/06, Robert McGwier [EMAIL PROTECTED] wrote: Let me say that I am now of the opinion that all of this high resolution clock business was something of a diversion from the real problems. That was my opinion all along, but I didn't want to touch the subject, since I have never used the keyer myself. The high-resolution performance counter frequency doesn't necessarily relate to the ACPI version. For example, Tim's system probably has the dual core performance fix applied (with /usepmtimer in boot.ini), so he will get the ~3.5 MHz reading even if his system is ACPI 2.0. PMTimer is the ACPI power management timer. Regardless of ACPI version, it always runs at 3,579,545 Hz (which is closely related to the original 4.77 MHz PC clock frequency). If your perf counter freq == 3,200,000,000, then the system is obviously not using ACPI PMTimer. Instead it's reading the CPU's time stamp counter: http://faydoc.tripod.com/cpu/rdtsc.htm You don't even need inline assembler for this operation, because recent Microsoft compilers have it as an intrinsic function __rdtsc(). Getting back to the more important point: It is time to revisit the question: are there folks getting reasonable CW performance using the high res timer in the code even though perf.exe reports 3.5 MHz? Unfortunately I can't test this myself, but 3.5 MHz really should be fast enough for CW. If there still are some problems, you need to look elsewhere for the reason. 73, Sami OH2BFO
[Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
CW performance is the holy grail for me. In general the SDR-1000 has passable CW performance. I bought a IBM G41 thinkpad to use with my SDR. It runs 3.3 ghz and is a P4 (not a M class) processor. It has the precision clock and HT. I checked perf.exe and got the 3.5... mhz response, but it also told me: High resolution performance counters supported which I interpret to mean it is telling me it recognizes I have the high precision clock. In PowerSDR I have Hi Res ticked under the keyer screen, and I use an external idiom press keyer with standard weighting (in other words no changes to play around with the character weighting, first character timing or all that) I can do 60 wpm (the outer limit of my keyer's speed range) with no problem using this set-up with perfectly transmitted code as monitored in another RX. I'm sure I could go a lot faster if I had a keyer that would go faster. The performance killer is the transmission lag, that is the time between the last keypress, and when the radio is fully back in the receiving mode. If you run any speed and work DX this is a problem. Operators develop a back and forth rythm associated with pile-ups it is a tick-tick kind of thing, sort of like professional table tennis. If their radio will do a given turn around from TX to RX they expect your radio to behave in kind. The faster you go the more precise the tick-tick needs to be. The extra lag it takes to go from last keypress to RX is often enough to miss your callsign comming back from the DX station in a pile-up. It's like playing table tennis with a 50lb weight strapped to your wrist. The performance can be improved if you use some sort of MOX control like a foot pedal, but the radio should be smart enough to go from T to R without the operator being the controlling timer. I don't think QSK is the real issue to most CW operators as much as being able to maintain the tick-tick relationship you establish with your contact. QSK rigs as a byproduct of their design do this automatically. So I would look at strategies that minimize the last key press to RX timeframe as much as possible if you want to enhance the CW performacne of the radio. I modified my keyer's sidetone so I could mix it into my headphones and completely disable the sidetone in PowerSDR. This is a perfectly acceptable solution for me. Frankly if it would speed up the T_R of the radion you could do away entirely with any sidetone generation, internal keyer in the software etc, and just let me use an external keyer to do these functions, or at least make this an option to be turned on and off. That being said, I love my SDR-1000 even if CW is somewhat compromised. 73 W9OY __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Tim, I'm running/testing a AMD Athlon64 (3200+). The mobo uses the new nVidia nForce 430 chipset. This board replaced an ASUS A7V (XP3100+) running the VIA chipset similar to yours. Both boards report the 3.58 mhz clock on perf.exe :-( Can you or possibly others tell me how well your system handles cw (transmit) at speeds in excess of 30 wpm? Curious to know if setting the high resolution clock in the set-up panel has any effect one way or the other. Cheers- Kirb, VE6IV --- --- --- --- Tim Ellison wrote- What processor are you using? I have a Athlon 2X (dual core) 4400+ (2.2 GHz) on an Asus A8V Deluxe (VIA chipset) and I get the exact same values you report. If you have a different processor, then there may be a bug in the program. -Tim --- Tim Ellison ( [EMAIL PROTECTED] ) Integrated Technical Services ( http://www.itsco.com ) --
[Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
On Mar 22, 2006, at 11:27 AM, Lee A Crocker wrote: The performance killer is the transmission lag, that is the time between the last keypress, and when the radio is fully back in the receiving mode. If you run any speed and work DX this is a problem. This is a timely post for me, Lee. I experienced this exact situation last night working 20 M cw. I had the turn around times down to a minimum with my SDR but with the delays I felt the timing with the DX station was off. The contact needed a couple of unnecessary repeats to complete it. I have also tried using MOX for cw in these situations by programming one of the buttons on my Shuttle for the MOX function. But there is still too much of a lag switching back to rx. Frankly if it would speed up the T_R of the radion you could do away entirely with any sidetone generation, internal keyer in the software etc, and just let me use an external keyer to do these functions, or at least make this an option to be turned on and off. I second this request. 73 Rick ve3mm
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
You mean like the one that has been in the radio for over a year? Dale Boresz and I both use the SDR-1000 with an external keyer and no sidetone. It works perfectly. RTFM Bob N4HY Richard Stasiak wrote: On Mar 22, 2006, at 11:27 AM, Lee A Crocker wrote: The performance killer is the transmission lag, that is the time between the last keypress, and when the radio is fully back in the receiving mode. If you run any speed and work DX this is a problem. This is a timely post for me, Lee. I experienced this exact situation last night working 20 M cw. I had the turn around times down to a minimum with my SDR but with the delays I felt the timing with the DX station was off. The contact needed a couple of unnecessary repeats to complete it. I have also tried using MOX for cw in these situations by programming one of the buttons on my Shuttle for the MOX function. But there is still too much of a lag switching back to rx. Frankly if it would speed up the T_R of the radion you could do away entirely with any sidetone generation, internal keyer in the software etc, and just let me use an external keyer to do these functions, or at least make this an option to be turned on and off. I second this request. 73 Rick ve3mm -- AMSAT VP Engineering. Member: ARRL, AMSAT-DL, TAPR, Packrats, NJQRP/AMQRP, QRP ARCI, QCWA, FRC. ARRL SDR Wrk Grp Chairman Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Dale Richardson wrote: The delay of the signal through the receiver is noticeable. I put a Yaesu FT-920 next to the SDR on receive and transmitted cw with another transmitter and the SDR is always a few milliseconds behind. The lag is noticeable enough that it is easy to lose the sync of a dx contact. In essence the SDR is still processing the other stations signal after he has stopped transmitting. That's exactly what's happening, and it's a *totally inevitable* consequence of the quality of the filters: the lag is in the length of the buffers. That's how FIR filtering works. The only way to eliminate this lag is to change the hardware/software configuration from being essentially half-duplex to full-duplex. This is not fundamentally a software fix, and no amount of ingenuity will change that. 73 Frank AB2KT
[Flexradio] High-Resolution Clock on AMD64 / nVidia nForce
Curious if people have run the Performance Counter Query Tool http://lightconsulting.com/~thalakan/perf.exe http://lightconsulting.com/%7Ethalakan/perf.exe to determine whether their motherboard and chipset supports the high resolution clock. I'm particularly interested in the results obtained from users of AMD Athlon64 motherboards using the newer nVidia nForce4 or nForce 430 chipset. My new board seems to be indicating a clock frequency of only 3579545 hz in spite of nVidia's claim to adhere to the ACPI 2.0 standard. My understanding, if I read the SDR-1000 manual correctly, is you must see a clock frequency approaching the clock rate of the CPU, in other words in the Ghz range to be compliant with ACPI 2.0, and thus capable of utilizing high-resolution timing in the PowerSDR cw setup (a must). Are all new boards not created equal? Appreciate hearing which boards do report the higher clock speeds. Kirb, VE6IV --
