>>>Wow, you are way out in left field here. Where do you come up with the idea that direct sampling SDRs have poor dynamic range? The Perseus, which is a direct sampling SDR, is in the top 4 receivers of the beloved Sherwood Engineering list for dynamic range. Way above many other conventional analog receivers. It was in the top three until the KX3.<<<
And you don't read.
I will repeat, if you want to sample and very high speeds needed for very high frequencies your have to trade bits in the A->D conversion or limit your frequency span. If you limit the number of bits you limit the dynamic range unless you add RF gain or conversion so you can add agc to the system to avoid overloading the A/D.
It takes 20bits to express any number between and about 1million or 100DB. If you sampling 20 bits at
at least 8x the input frequency that's 160Mbits/S for a 1mhz signal For 100Mhz that is a mere 16Gbits second. That is far faster than most flash A/Ds that are usually under 10bits (o to 1023 or 30db).
Since systems are not quite that fast you either down covert or use fewer bits to to keep the data rate believable.
I do read believe me. You should do some reading before making a fool of yourself here. Obviously, you are a hopeless case, but I will clarify some things in hope that your ignorance does not spread to people who are new to SDRs and how they work.
A sampled signal picks up dynamic range through the process of filtering and decimation. If you have a signal that has been sampled by a 16 bit ADC at 125 MSPS, and you decimate the sampled signal down to 10 kSPS (enough bandwidth for demodulation of most Ham modes) the dynamic range after sampling becomes the 16 bit ADC dynamic range (96 dB theoretical) + the process gain of 38 dB (see equation below) = 134 dB. Of course 16 bit ADCs never reach their 96 dB theoretical dynamic range, the ENOB (effective number of bits) is less. For example the LTC2208 reaches about 75 dB at 125 MSPS. So even in an imperfect world, 75 dB + 38 dB = 113 dB which is still very acceptable.
process gain = 10 * log10( adc sample rate/(2 * final sample rate) )
10 + log10(125e6/(2 * 10e3)) = 38 dB
Where people get confused in between QSD based SDRs and direct sampling SDRs and what determines the dynamic range.
In a QSD based SDR, the RF spectrum is down-converted to the audio frequency range by the QSD which gives you two analog channels, I and Q. The analog I and Q channels are sampled by an audio ADC at typically 48, 96, or 192 kHz (since these are common sound card frequencies). The DSP process typically does not do any decimation of the audio ADC sample rate, so it too runs at 48, 96, or 192 kHz. The dynamic range is determined primarily by the ENOB of the audio ADC used. Typical 16 bit ADCs have an ENOB of only around 12-13 bits, which gives you only 78 dB at the best. The best 24 bit audio ADCs have an ENOB of around 20-21 bits which gives you somewhere around 124 dB which more than adequate dynamic range. There are two problems with QSD based SDRs: The first is the I/Q balance problem which determines image suppression. The second is the widest bandwidth that can be digitized is determined by the sample rate capability of audio ADCs, which is typically 192 kHz. The I/Q signals are generated in the analog domain. It is impossible to have perfect I/Q balance because of component tolerances, etc… so the I/Q inbalance must be corrected after the I and Q channels are digitized by the audio ADC. Unfortunately, the I/Q inbalance is not stable over time, temperature, frequency, impedance, and component aging. You can never have perfect I/Q gain and phase matching between two analog circuits.
In a direct sampling SDR, the RF spectrum is directly sampled at the ADC sample rate which needs to be at least twice the maximum bandwidth that you want to cover. For example, if you sample at 125 MHz, you can theoretically digitize signals up to 62.5 MHz unless used in undersampling mode. Most high speed ADCs have input bandwidths far exceeding their sample rate so they can be used in undersampling applications. For example the 16 bit LTC2208 is responsive up to about 700 MHz. Because of this the ADC is usually preceded by a low pass filter that cuts off at a maximum of ½ the sampling rate. The high speed ADC samples the RF signals directly and feeds them to the DDC, typically implemented in a FPGA. The signal at this point is in the digital realm and is represented as 16 bits. The I and Q channels are generated in the DDC in the digital realm, so the I/Q balance is essentially perfect and not subject to analog circuit component variations. Direct sampling SDRs do not have the I/Q inbalance problems of the QSD that cause poor image suppression. The DDC filters and decimates the ADC’s sample rate down to some sample rate the DSP process can handle. The DSP process can be implemented in the PC or in an embedded DSP processor. Due to the filtering and decimation in the DDC, you pick up dynamic range and the I/Q samples are typically represented as 24 or 32 it samples when they exit the DDC – more than enough bits to handle the dynamic range. Thinking that the number of bits the ADC has is the primary determining factor in the dynamic range is incorrect due to the “processing gain”. Also, because you are not dependent on a sound card ADC like the QSD, the sample rates can be much higher than 192 kHz. The Perseus can process 1600 kHz bandwidth in real time by setting the DDC sample rate to 2 MSPS, and the QS1R can process 2 MHz bandwidth in real time by setting the DDC sample rate to 2.5 MSPS. Both SDRs can record to disk at those rates and be played back later.
If you think that process gain is just some mathematical trick, anyone with a direct sampling SDR can easily see the real results of process gain in the spectrum window by noting the spectrum noise floor. As you decrease the sampling rate, the noise floor also drops by a predictable amount. Each halving of the sample rate results in a 3 dB decrease in noise floor (or a 3 dB increase in sensitivity). This is a direct result of process gain.
The LTC2208 is only 16bits so there must be an agc system to allow a over 100db range as the 16bits is 0 to 65535 or 48db between detection threshold and saturation of the A/D. Oh, that 100DB is the SFDR not the DR which is lower.
This is where you have erred. You have forgotten about the dynamic range gained as a result of filtering and decimation, or process gain as I discussed above. None of the currently available direct sampling SDRs, such as the Perseus, QS1R, Winradio Excalibur, or RF Space SDRs, use any kind of AGC before the ADC. The dynamic range due to decimation of the high speed ADC’s sample rate is adequate without AGC in front of the ADC. Even most QSD based SDRs, like the Flex Radio SDRs and the SoftRocks, do not have analog AGC. The use of a 24 bit audio ADC with about 20 bits of ENOB gives adequate dynamic range.
The LTC2208 ADC's SFDR is around 75-77 dB. But you still must take into account the process gain, which in the example above is 38 dB, so the SFDR is still over 113 dB on the output of the DDC. You have to look at the ADC and the DDC as a system when talking about direct sampling SDRs.
Direct to 63.5mhz and undersampling to 500mhz. What does SSB voice at 432 sound like when you undersample? How do you handle images above 62.5mhz, higher?
In the case of my QS1R, there is an option to bypass the low pass filter in front of the ADC which I have done. When I want to use my QS1R below 62 MHz, I insert an external 55 MHz LPF. When I want to use my QS1R above 62 MHz, I remove the external LPF and insert an external band pass filter that is appropriate for the frequency range I am interested in. Above 62 MHz, the QS1R is used in the undersampling mode. The external BPFs suppress the images of frequencies outside of the band of interest. At 432 MHz, SSB sounds exactly the same as SSB at 30 MHz. The software takes care of the spectral inversion and the frequency readout correction, so that when I want to listen to USB on 432 MHz, the display reads 440 MHz and the mode selected is actually USB.
Now you confirm something I'd figured you didn't understand. Undersampling is downconversion.
How is that done.. well the sample and hold is run at 130mhz so that amounts to the conversion
osc and it will wrap half that or 62.5 mhz down in the direct sampling frequency. And what happens at exactly 62.5mhz or 192.5mhz?
You are the one who does not understand here. If I want to listen to 192.5 MHz, I insert a BPF centered on 192.5 MHz with its lower cutoff frequency somewhere above 62.5 MHz. So signals below 62.5 MHz do not reach the ADC. The same at any other frequency range I am interested above 62.5 MHz and below the max of 500 MHz. So I can listen to any frequency between 10 kHz and 500 MHz in bands of 62.5 MHz, by inserting the correct LPF or BPF filters in front of the ADC, in this case the QS1R RF input connector.
Gene
P.S. I have no gripe with Rob Sherwood. I believe his numbers are accurate. I have only stated clearly and multiple times that the determining decision on what radio to buy should not be determined solely by numbers on the Sherwood list (or can't you read?). To refresh your memory, you are the one who jumped in and took issue with the statements I made about the I/Q balance issue with the KX3.
