Saturday, March 20, 2010
Testing the SDR-IP
As there has been earlier posts about the SpectraVue´s approach to listening and measuring the frequencies (see posts on the KingsVillageDX-blog) this is not discussed here rather we try to make a summary of the listening tests by TuA in Lapland, some comparisions by listening to mw-band at Muhos (about 1000 m beverage to the West)on March 5-6th and the analysis of the recorded files by TK afterwards.
General impressions by TuA at ENOX and at Muhos:
TuA tested already IP and compared it to SDR-IQ & 14 during the ENOX DXpedition in Lapland at the end of February. According to these tests no significant differences were found during listening with the IP and SDR-14/IQ but the dynamic range seemed to be better on the IP - there were no problem in clipping with the IP when using amplified antennas (about 1000 m beverages with at least 10-15 dB amplifiers). So when the "clipping lights" were active with the smaller SDRs and the attenuation should be put on (-10 dB ?) the IP had no problem. Tuomo used a dual-core 2.66 Ghz pc, where he had the IP recording at 2 Mhz, one SDR14 doinf the recording + he was listening to the earlier recorded files as well without problems. One strange issue was noted by Tuomo, however - the listening to IP while recording at sample rates of 1.33 Ms/s and 1.143 Ms/s had some problems. According to manufactures it should be ok and the problems should be when using the computer set up there. However, the very same problem occured also at Muhos. Acc. to RF-Space they will check this issue which could be due a “wrong” firmware. Otherwise the tests at Muhos were very simple: we just recorded certain time spans of signals on mw using both the SDR-IP and Perseus receivers at the same time (splitted antennas). The bandwidth was 2 MHz covering the mediumwave bands. Afterwards I listened to a few Top-of-the-hour segments and compared the results.
I tested also IP with my small HP-laptop just to see how it work with minor set-ups. Otherwise a tapletop computer of Pentium4 (2.8 Ghz/1Gb) was used. It took some time to make my small HP-Laptop to begin to work as you have to configure the Ethernet connection. After that it worked but did have some problems as the processor was only a Celeron 1.86 Ghz (1 Gb memory). Also with the P4 2.8 Ghz there was pretty high load when using the IP with 2 Mhz sample rate showing the load around 70-80 %. Usually this big load means problem with USB-connections as it has been noted with the Perseus,but it seems the connection with Ethernet works better. The both computers had XP-operation systems. Afterward listening was done using a tapletop pc dual-dore 3.0 Ghz with 4 Gb memory, which of course has no problems with the load.
The summary of observations at Muhos:
- listening by IP or IQ shows no significant differences when listening with no high RF environment, but using long beverages and/or amplification there are problems with SDR-IQ/14-environment as well as on Perseus (need to use attenuation) while IP seems to work OK without the attenuation
- IP needs a good computer of modern standards, but this is no problem with modern dual core machines (2-3 Ghz speed, 2-4 Gb memory), according to earlier tests by TuA with dualcore 2.6 Ghz pc, a load of about 50 % was noted
- Ethernet connection works flawlessly when configured right and seems to be more reliably than USB
- using modern setups there is no problem using several programs at the same time - however using both the IP and Perseus was not tested now as we did have a bit older pc-setups here
- everything was going OK when recording with the SpectraVue as long as the harddisk was full - just in the middle of morning dxing conditions towards the Western Hemisphere aroud 05.30 utc!
- measuring of exact carrier frequencies was a bit difficult using the P4 2.8 Ghz pc as the 2 M FFTs which is needed for the best resolution did eats a lot of computing power - this can be avoided by listening afterwards by a more powerfull pc - the measurements could be done, but listening at the same time with modest setups is difficult and some stuttering did occur when using the full resolution (at 2 MHz 0.95 Hz)
- trying to listen with full resolution on a HP (Celeron 1.86) was possible, but the load was around 90 % and stuttering did occur
- again it was fun to see all the time the carriers when listening USB/LSB - the fact which has been dissussed earlier compared to Perseus logic not to show the center frequency but sidebands (see earlier posts on blogs)
- it seems the current default file size about 2 Gb means a 4 minute IF-recording (2 Mhz span), which seems to be a bit small although the files can be linked when listening by SpectraVue afterwards – this issue is being developed by the manufacturer though.
By the way, we changed the unit´s small PSU to a bigger one - it seemed the one provided with the unit is not noise-free.
Listening afterwards of recorded spectrum files:
Listening to a IP-file (2 Mhz span) was easy with a Pentium dual-core 3.0GHz/4Gb pc (the load was around 30 %) as was expected.After some comparisions by listening to files at 0200 and 0300 UTC with some common NA and Latin stations some observations were made:
- generally "by ear" there were no differences when listening to stations like CBC 1140, a South American (and Puerto Rico) on 1230 or Boston 1510
- on some channels however, the sound of IP sounded cleaner (like Haiti on 840), although not many comparisions are made so far - this could mean that the dynamic range of the IP seems to be better, overall the DR sounds better on the IP than IQ/14 or Perseus – no measurements, but the IP needed no attenuation
when others did
- Listening with SpectraVue is done as usual either on USB or LSB, but with Perseus it can be done using the AM or SAM when there was no interference near by (for example 1470, but on 1510 it is better to go to LSB)
- on Perseus however you can adjust easily the sidebands of audio spectrum (changing the bandwidth - independently both sides) on the small window as much you want , on SpectraVue it is not that easy as you have the steps of 100 Hz in the band width
- on SpectraVue using USB/LSB you always see the carrier frequency which is not possible on Perseus (measurements has to be done on AM/SAM; the issue widely discussed earlier but not been "corrected" on Perseus yet)
- it seemed the sensitivity of both receivers was rather equal, but this is not the big issue on MW as both are very sensitive receivers
- it sounds that the shape factor of DSP-based bandwith of Perseus is better but that can not easily be confirmed by ear
All in all, both receivers did well on every channel, a conclusion which was already expected beforehand.
Finally I made a file format change by using a nice piece of softaware "sv2perwav" and converted the recorded IP-file (2 MHz span at 02 utc, 4 minutes long file) into a Perseus file. By listening to that the noise blanker, noise reduction and what is most important, the excellent notch of Perseus could be used! There seems to be a bug (?) on this software as it seems to loose the correct time information by not showing the right recorded time on the time bar. As can be seen on the picture(where two processes are going on simultaneously, one demodulating the converted IP-file and another demodulating a Perseus file at the same time on 1580) both receivers shows a bit different frequencies: it seemed that the IP was calibrated showing about 1-2 Hz too much while Perseus showed about 1 Hz too small values depending a little bit where on the band you were. It seemed that IPs “error” was rather same on the different side of the spectrum and Perseus shows a bit different frequencies on the lower and the higher frequencies.
Well, the biggest difference of these receivers is the current price! The IP seems to be more than 2000 euros, so you can get almost 3 Perseus at the prize of it. So from a medium wave dxer´s point of view, the "hobby IP" could have fewer of the options it currently has and just the basics for serious mw-enthusiasts with a bit more developed software with possibly notch and some other usefull features, the Ethernet connection + the ability to record the spectrum (maybe 1.2 MHz is enaugh?) for later analysis. As the manufacturer says the IP is more intended more to business / governmental etc use, not for dxers. RF-Space has informed us that they will be releasing a lower cost option of the SDR-IP with similar high performance later this year using Ethernet, which we welcome of course. They will also look into the current software and develop the demodulators more and some of these improvements have already been done to SDR-Radio software. This program already works with SDR-14/IQ and support the SDR-IP now.
Thank you for RF-Space to let us test the fine IP!
Thanks for the testing team and our host Jari (JPR) and another Jari (JSN) for the food!
PS: Want to test a (big) SDR-IP File?
I will have a file available (1,74 Gb zipped, so be prepared for a long download if you do not have fast speeds) for those who want to get the feeling of listening to a 2 MHz IP-file (center frequency 1290 kHz). By using the latest SpectraVue-software (3.07 at the moment) you can listen to it with a modern pc. It is recorded around 03 utc for 4 minutes at Muhos on March 6th and it contains common Trans Atlantic signals and the Europeans on the MW. You can get the link to the file by sending me a personal message to tk (dot) sdxl (dot) org. The file will be available for about 2 weeks.