Leif, Chavdar, Grant, What a great information guys. These are one of those conversations that I will definitely print out to my SDR project map and read it every since and then :) It is now obvious that the QSD responses to a certain outside of band frequencies and should have better filters in front that for example the SDR for the masses articles succest to use. This all began from when I was testing my SDR transceiver at the club station on 20 meters. The club station has an 20 meter 3-element yagi 42 meters ABS, so the signals are quite huge compared to my GP antenna at home. According to my simple tests my SDR has quite good dynamic range, especially when looking at the on the band signals, but last time at the club I noticed this spectrum background pumping phonenom and couldn't find the source of the signal that was causing it. Obviously it was an outside of band signal... Then I found some talk about 1/2 of the LO signal response and realized that there are really strong signals at 14.200MHz-14.300MHz/2 = 7.1MHz-7.15MHz .... BC stations, and at this location, especially strong russian stations. My BPF filters are quite wide (direcly from the SDR for the masses design) and I'm using the lowpass filters of the PA on RX too, but didn't think about/or knew about the LO/2 response at the time I was designing the QSD and QSE. Now, since the good good convertion - shedding a light to this issue, I decided that I'm not trying to achieve general coverage on my RX anymore and will definitely use narrower filters on RX. Any advices guys? Thank you really really much! 73 de Janne Pulkkinen, OH1GTF
--- On Tue, 8/12/08, Leif Asbrink <[EMAIL PROTECTED]> wrote: From: Leif Asbrink <[EMAIL PROTECTED]> Subject: [soft_radio] Re: QSD response to the half LO frequency. To: [email protected] Date: Tuesday, August 12, 2008, 7:06 PM Hello Chavdar, > By the way I am using "single tone" method to obtain IP2 figure. I > feed the DC RX input with a strong pure sine wave signal (X-tal > generator & 7th order LP filter to reduce the 2nd harmonic) and > measure the 2nd audio harmonic amplitude with PC spectral analyzer > carefully testing whether the square law for the amplitude is valid. > The input frequency of the test signal is 5 – 20 KHz above or bellow > LO frequency so that the second audio harmonic spur must be in the > baseband. I do not know where exactly the 2nd order distortion takes > place since in the chain there are mixer, audio pre-amplifier and > ADC. This method measures IP2 of the total system and does not > depends from input filters. OK. Have a look at fig. 6 here: file:///home/ bsz/sm5bsz/ linuxdsp/ rxhfa/imd2_ hfa.htm Two generators at 7.080 and 7.085 present 0 dBm to the input of the test object. The sum frequency 14.165 is 6 dB above the each of the second harmonics at 14.160 and 14.170. (A third generator at 14.164 is used as an amplitude reference) Surely it is enough to use a single tone for IM2 measurements on "normal receivers" but in case you use a Tayloe mixer it could be quite different. You may find that two signals that have a frequency separation that is much larger than one over the time constant of the Tayloe mixer give a significantly worse result just as you may find that the Tayloe mixer is excellent for close range IM3 but much less so for wide range IM3. > One of the articles (in Bulgarian language) at my home page is > named "Measuring the parameters of SDR. Single tone method." At > least the figures and photos can be understood. > I can not say at the moment how single tone method compares to the 2- > tone method since I do not have access to high quality signal > generators. Also different origin of distortions might be assumed > and comparison should be taken with care. To compare correctly the > IP2 figures, the following level correction formula must be applied: > IP2 (2-tone) = IP2 (1-tone) – 6 dB > So the before mentioned IP2 figures from my measurements should be > reduced with 6 dB in order to compare them with standard 2-tone test. > The measured IP2(corrected) of my DC RX is between 49-59 dBm. MDS is > around -126 dBm @ 500Hz @14MHz. The mixer I am using is a double > balanced Tayloe QSD mixer with s&h capacitors in the switch output. > 74HC4052 is used and it works fine up to 14 MHz. The audio pre- > amplifier is TS462 (4nV/Hz) and Audigy SE 24bit sound card. I can say > that for a high grade RX this figure is not sufficient , specially > for AM detection problem. Figures above 70 dBm are desirable. > > I tried to use single tone method to measure IP3 but the results were > unreliable. The cubic low requirement was not fulfilled. > Can you give some IP2 figures for your mixer ? The difficult problem is not linearity, but spectral purity of the LO. Therefore my design consists of cascaded converters with low noise crystal oscillators. For this reason the I/Q mixer operates at a fixed frequency so it is easy to make sure that it will never experience any signals anywhere as low as half the frequency. Therefore I can not give any IM2 numbers. The inband second order intermodulation is a little better than -100 dB for a signal that is 0.5 dB below A/D saturation. http://www.sm5bsz. com/linuxdsp/ rxiq/rx2500lynx. htm 73 Leif / SM5BSZ
