I am trying to implement equivalent time sampling using an EZCAP dongle configured using gnuradio-companion. Since I am not completely clear about the tasks of the E4000 wrt RTL2832, it could be that I am missing a significant information, so here is my experimental setup aimed at developing a monostatic pulse mode RADAR : * I am using a radiofrequency acoustic delay line to generate 4 echos delayed 1 to 2 microseconds after an incoming excitation pulse is generated by a frequency synthesizer. The excitation pulse is 125 ns long, and repetition rate is 4 microseconds. The carrier frequency at 860 MHz is chopped by a fast duplexer, one side being connected to the frequency synthesizer and the other sizer to the EZCAP dongle * the EZCAP is configured with an LO at 860 MHz, 2 MS/s output, and I plot |I+jQ| after keeping only 1 every 8 sample (2 MS/s=500 ns delay between samples, and 1 every 8 samples means a spacing of 4 us, close to the emitted pulse repetition rate) * equivalent time sampling is obtained by slightly tuning the emitted pule repetition rate off the 4 us: I have checked that I can indeed obtain time stretching by tuning the pulse repetition rate +/- 50 ns away from the 4 us dela, with time stretching factors of up to 10^5 (ie the 4 us pulses are recorded as 400 ms traces (thus including 10^5 points or an equivalent sampling rate of 25 GS/s).
Now of course a stretching factor of 10^5 is overkill and the RF front end definitely does not have such a huge bandwidth, this is just to demonstrate the concept. With a more reasonable stretching factor of 10 (ie sampling every 4.4 us with a 4 us emission pulse repetition rate), I can expect to convert a 2 MS/s sampling rate to an equivalent time sampling of 20 MS/s, which would already improve my monostatic RADAR resolution by a factor 10, and more or less fit my targetted resolution.
So now the questions: * the 125 ns pulse I generate would span 8 MHz. From my reading of the E4k description, this is within the IF bandwidth available. However, although I can observe the emitted pulse (0 dBm), I cannot see the echos (-35 to -40 dBm), whatever LNA gain I use (from 0 to 40 dB). Is the default configuration of the E4k an IF of 8 MHz, or lower. What parameter should I give the rtlsdr gnuradio source block to make sure I have the right bandwidth ? * since I am not clear about task distribution between the E4k and the RTL2832, is there another limitation that will prevent a 8 MHz wide (bandwidth) signal to be recorded through the 2 MS/s I/Q output recorded ?
Thank you, Jean-Michel
The rtl-sdr (and all other software) has only a bandwidth of 3.2MHz maximum available. Only the hardware demodulator inside the realtek chip can process full 8MHz. If you consider the filter setting of the rtl-sdr you effective bandwidth is about 2 MHz. It is hard coded in the tuner code but reasonable for most settings, as it avoids aliasing.
Am 06.07.2012 09:52, schrieb friedtj:
I am trying to implement equivalent time sampling using an EZCAP dongle configured using gnuradio-companion. Since I am not completely clear about the tasks of the E4000 wrt RTL2832, it could be that I am missing a significant information, so here is my experimental setup aimed at developing a monostatic pulse mode RADAR :
- I am using a radiofrequency acoustic delay line to generate 4 echos delayed 1 to 2 microseconds
after an incoming excitation pulse is generated by a frequency synthesizer. The excitation pulse is 125 ns long, and repetition rate is 4 microseconds. The carrier frequency at 860 MHz is chopped by a fast duplexer, one side being connected to the frequency synthesizer and the other sizer to the EZCAP dongle
- the EZCAP is configured with an LO at 860 MHz, 2 MS/s output, and I plot |I+jQ| after keeping
only 1 every 8 sample (2 MS/s=500 ns delay between samples, and 1 every 8 samples means a spacing of 4 us, close to the emitted pulse repetition rate)
- equivalent time sampling is obtained by slightly tuning the emitted pule repetition rate off the
4 us: I have checked that I can indeed obtain time stretching by tuning the pulse repetition rate +/- 50 ns away from the 4 us dela, with time stretching factors of up to 10^5 (ie the 4 us pulses are recorded as 400 ms traces (thus including 10^5 points or an equivalent sampling rate of 25 GS/s).
Now of course a stretching factor of 10^5 is overkill and the RF front end definitely does not have such a huge bandwidth, this is just to demonstrate the concept. With a more reasonable stretching factor of 10 (ie sampling every 4.4 us with a 4 us emission pulse repetition rate), I can expect to convert a 2 MS/s sampling rate to an equivalent time sampling of 20 MS/s, which would already improve my monostatic RADAR resolution by a factor 10, and more or less fit my targetted resolution.
So now the questions:
- the 125 ns pulse I generate would span 8 MHz. From my reading of the E4k description, this is within the
IF bandwidth available. However, although I can observe the emitted pulse (0 dBm), I cannot see the echos (-35 to -40 dBm), whatever LNA gain I use (from 0 to 40 dB). Is the default configuration of the E4k an IF of 8 MHz, or lower. What parameter should I give the rtlsdr gnuradio source block to make sure I have the right bandwidth ?
- since I am not clear about task distribution between the E4k and the RTL2832, is there another limitation that
will prevent a 8 MHz wide (bandwidth) signal to be recorded through the 2 MS/s I/Q output recorded ?
The e4k does the filtering but has an analog output. The RTL2832 is in effect an USB ADC.
The e4k does the filtering but has an analog output. The RTL2832 is in effect an USB ADC.
apologies if this is obvious to everyone and has already been posted (I have not checked the whole history of this ml), but indeed by connecting a couple of wires to the I or Q differential inputs of the RTL2832 you do get a fine 2 MS/s oscilloscope. This unfortunate AGC means that DC signals cannot be monitored (even after soldering the wires after the common mode cutoff capacitors), but neverless this does provide a 40 fold sampling rate improvement with respect to the use of sound cards as A/D converters (http://sequanux.org/jmfriedt/199999720input_middle.png and http://sequanux.org/jmfriedt/199999720input_top.png for the I and Q measurements when an input signal of 199.999720 kHz is generated by a synthesizer in order to synchronize with the sampled data, ie the sampling clock is off by 1.4 ppm, not too bad !).
JM
Can I say I am impressed with what you are doing friedtj. One thing I have wanted to do with these new Dongles with the e4000 tuners is bypass the RTL2832 chip and piggyback a better ADC on the e4000 tuners output. Since there is so much activity in the "Group" of squeezing the maximum performance from the Dongle Tuners, it may be time to add better hardware. Just to test to see if this is a good route to take, it should be possible to connect a decent PC audio card to the output of the e4000 tuner to see if it suffers any degradation, the AGC debacle can also be checked with this method. If no problems arise then addition of a dual input, 14 bit, 100 Mega-sample, ADC would be the next step to take full advantage of the bandwidth and dynamic range of the tuner that the RTL2832 chip is obviously and purposefully limiting; forget the RTL2832 chip. Will someone publish the e4000 tuner's datasheet and stop all the IP BS.
-----Original Message----- From: osmocom-sdr-bounces@lists.osmocom.org [mailto:osmocom-sdr-bounces@lists.osmocom.org] On Behalf Of friedtj Sent: Friday, July 06, 2012 2:53 AM To: osmocom-sdr@lists.osmocom.org Subject: stroboscopic (aka equivalent time sampling) using EZCAP DVB dongle ?
I am trying to implement equivalent time sampling using an EZCAP dongle configured using gnuradio-companion. Since I am not completely clear about the tasks of the E4000 wrt RTL2832, it could be that I am missing a significant information, so here is my experimental setup aimed at developing a monostatic pulse mode RADAR : * I am using a radiofrequency acoustic delay line to generate 4 echos delayed 1 to 2 microseconds after an incoming excitation pulse is generated by a frequency synthesizer. The excitation pulse is 125 ns long, and repetition rate is 4 microseconds. The carrier frequency at 860 MHz is chopped by a fast duplexer, one side being connected to the frequency synthesizer and the other sizer to the EZCAP dongle * the EZCAP is configured with an LO at 860 MHz, 2 MS/s output, and I plot |I+jQ| after keeping only 1 every 8 sample (2 MS/s=500 ns delay between samples, and 1 every 8 samples means a spacing of 4 us, close to the emitted pulse repetition rate) * equivalent time sampling is obtained by slightly tuning the emitted pule repetition rate off the 4 us: I have checked that I can indeed obtain time stretching by tuning the pulse repetition rate +/- 50 ns away from the 4 us dela, with time stretching factors of up to +10^5 (ie the 4 us pulses are recorded as 400 ms traces (thus including 10^5 points or an equivalent sampling rate of 25 GS/s).
Now of course a stretching factor of 10^5 is overkill and the RF front end definitely does not have such a huge bandwidth, this is just to demonstrate the concept. With a more reasonable stretching factor of 10 (ie sampling every 4.4 us with a 4 us emission pulse repetition rate), I can expect to convert a 2 MS/s sampling rate to an equivalent time sampling of 20 MS/s, which would already improve my monostatic RADAR resolution by a factor 10, and more or less fit my targetted resolution.
So now the questions: * the 125 ns pulse I generate would span 8 MHz. From my reading of the E4k description, this is within the IF bandwidth available. However, although I can observe the emitted pulse (0 dBm), I cannot see the echos (-35 to -40 dBm), whatever LNA gain I use (from 0 to 40 dB). Is the default configuration of the E4k an IF of 8 MHz, or lower. What parameter should I give the rtlsdr gnuradio source block to make sure I have the right bandwidth ? * since I am not clear about task distribution between the E4k and the RTL2832, is there another limitation that will prevent a 8 MHz wide (bandwidth) signal to be recorded through the 2 MS/s I/Q output recorded ?
Thank you, Jean-Michel
-- JM Friedt, FEMTO-ST Time & Frequency/SENSeOR, 32 av. observatoire, 25044 Besancon, France
If no problems arise then addition of a dual input, 14 bit, 100 Mega-sample, ADC would be the next step to take full advantage of the bandwidth and dynamic range of the tuner that the RTL2832 chip is obviously and purposefully limiting;
1) AFAIK, The E4000 has a limited output bandwidth, in the range of 8 MHz max usable IIRC. It's designed for DVB/TV applications and they don't need more so the IF filters available don't go above that.
2) You still need to pipe all those samples to the PC somehow, this will raises the price of all of this way above the original 20 USD and make it a much bulkier setup.
Cheers,
Sylvain
If all you want to do is watch and listen to HAM and Police Trunk Radio, the $20 Dongle is adequate, but if you want to hack the E4000 Tuner, a better ADC is needed.
1) Yes, if all you need is to be able to detect a signal on an analog stream its Nyquist Frequency will do, (2 times the frequency being measured) a 20 megasample ADC would be adequate for the E4000 tuner's I/Q output, but there is more information in the Stream than just the presence of Radio Frequency Energy. If all you need to do is detect the presence of energy or listen to it (Audio Modulation) a 2 times Sample frequency is adequate, but if you need to decode octal phase modulated sideband signals, you need to go to at least 10 times Analog to Digital. This is the case with many signals on Satellite and Cable Streams and it is the case with many phase modulated signals over the air. Instead of just the minimum Conversion being done, you need 10 times the Conversion to reconstruct the modulation, and this may even miss subtle information. It is information we are seeking, information as intelligence modulating Electromagnetic Energy. Cable Modems, for example, Digitize the entire spectrum from 100,000 KHz to 800 MHz. This is because the TV Signals are encrypted with Octal Phase Modulation. You have to be able to digitize with wideband (multi-Hundred Megahertz Converters) to be able to Analyze, Decode, or even just to detect these modulation schemes. So even though the E4000 Tuner has only 8 MHz Bandwidth. You need at least 10 times that bandwidth to accurately reconstruct Phase Modulated Signals. In the vernacular of ADCs, anti-aliasing filters are required to remove unwanted high frequency content. Thus content is discarded, and noise is introduced. To adequately sample a broadband signal without discarding anything, radical oversampling is required. Oversampling allows Signal Processing to take place like Noise Reduction, with is very important for DXing. ADCs are cheaper than ever, driven by Scientific Instrumentation, Satellite, and Cable Industry needs.
2) Remember back a in the 1990s when USB first came to be. It was expensive and a luxury on high end computers. Now we have things like the Tuner Dongle with USB and virtually every computer with USB as a Standard; now there is USB 3; even faster. Apple came up with FireWire to solve the problem that USB had. FireWire is capable of real-time wideband serial communication. Better yet, even though it is still expensive, ThunderBolt will become the new "USB." At Multi-Gigabit Streaming rates (800 Mega Bytes per second). You will see in the near future, ADCs with USB 3 and ThunberBold built-in. There is just too much money to make. Current High Definition Cameras use SDI to stream their signals. There are already available Video equipment servicing the High Definition Television Industry with both SDI and ThunberBolt Side-by-Side. You may pooh-pooh this as too expensive, but so was USB when it first rolled onto the scene.
-----Original Message----- From: Sylvain Munaut [mailto:246tnt@gmail.com] Sent: Saturday, July 07, 2012 3:34 AM To: Jay Salsburg Cc: osmocom-sdr@lists.osmocom.org Subject: Re: stroboscopic (aka equivalent time sampling) using EZCAP DVB dongle ?
If no problems arise then addition of a dual input, 14 bit, 100 Mega-sample, ADC would be the next step to take full advantage of the bandwidth and dynamic range of the tuner that the RTL2832 chip is obviously and purposefully limiting;
1) AFAIK, The E4000 has a limited output bandwidth, in the range of 8 MHz max usable IIRC. It's designed for DVB/TV applications and they don't need more so the IF filters available don't go above that.
2) You still need to pipe all those samples to the PC somehow, this will raises the price of all of this way above the original 20 USD and make it a much bulkier setup.
Cheers,
Sylvain
- Yes, if all you need is to be able to detect a signal on an analog stream its Nyquist Frequency will do, (2 times the frequency being measured) a 20 megasample ADC would be adequate for the E4000 tuner's I/Q output.
This is IQ signal you don't need twice the bandwidth, you only need once. So a 8 Msps IQ would be adequate to represent a 8 MHz bandwidth.
[snip ....] You need at least 10 times that bandwidth to accurately reconstruct Phase Modulated Signals.
No you don't. You want a concrete example: Most GSM phone are demodulating a phase modulated signal by sampling at exactly one sample per symbol (so they're even sampling _lower_ than the actual signal bandwidth)
And although oversampling is useful in certain case to analyze signal, it's essentially useless if the signal has _already_ been filtered like it is in the E4000 case. If the IQ output has only 8 MHz of useful content at the output, you can oversample all you want, it won't bring back what the E4000 discarded ... Those antialiasing filter you mention, they're in the E4000 !
2) I didn't say bigger / better SDR would be useless ... I'm just saying if you're spending hundreds / thousands on better ADC and better PC interface you might as well ditch the E4000 and use a better RF front end as well. The spectral output of the E4000 is far from being "clean", there is distortions / LO leakage / random spurs / 28.8M clock leakage / ... etc ... Oh and Elonics declared bankrupcy so the E4000 supply is fading ...
Cheers,
Sylvain
The E4000 Tuner offers a stepping stone into the world of SDRs. It does the heavy lifting, significantly reducing the size, cost, and complexity of a SDR. It does not do any programmable signal processing, which is where SDRs are strongest. Quantization has been left out of the discussion because the RTL2832 is so bad at it (to conserve size and cost). The RTL2832 hamstrings the usefulness of the E4000 significantly narrowing the application of DSP. If you want high fidelity reception, you must apply a high fidelity ADC. The Tuner Dongle offers a low cost opportunity for an enterprising hacker to jump ahead and produce a product worth pursuing. As is it is very limited.
Also, GSM is very narrowband voice. Data is wideband and complex, Data is where the Gold is, mining it takes better tools which the RTL2832 prevents application to the reception.
Typical low price ADC ...The ADC12DS105 is a dual 12-Bit, 105 Msps A/D Converter http://www.ti.com/product/adc12ds105 Price $31 http://www.ti.com/tool/adc12ds105lfeb Design Kit $495, Accessories include a USB Interface
http://www.ti.com/solution/software-defined-radio-sdr-diagram http://www.ti.com/lit/an/slaa407/slaa407.pdf
How to Select the (Best Minimum) Sampling Rate The selection of sampling rate is also important for ADCs. Fast encode clocks offer the advantage of making analog filtering significantly easier. Given a fixed signal bandwidth, a higher encode rate increases the allowable transition band. This allows lower order filters to be used to reduce cost, or, if higher order filters are still used, greater stop band rejection can potentially be realized. Another advantage with increasing clock speeds is processing gain. Processing gain is achieved when the signal of interest is oversampled and then digitally filtered. Processing gain occurs when noise outside the band of interest is digitally removed, which results in improved in-band SNR. Processing gain can be easily determined using Equation 1. This equation determines the increase in SNR over and above the SNR of the converter alone. Processing Gain = 10log10 (Half of the sample rate/ filter BW) Also, when selecting the sampling rate, the user must consider the second and third harmonics (HD2 and HD3). These harmonics must lie outside the band of interest. These can be filtered using the digital filter inside the DDC/DSP. That means, frequency planning also becomes important when designing a SDR system.
-----Original Message----- From: osmocom-sdr-bounces@lists.osmocom.org [mailto:osmocom-sdr-bounces@lists.osmocom.org] On Behalf Of Sylvain Munaut Sent: Sunday, July 08, 2012 10:29 AM Cc: osmocom-sdr@lists.osmocom.org Subject: Re: stroboscopic (aka equivalent time sampling) using EZCAP DVB dongle ?
- Yes, if all you need is to be able to detect a signal on an analog stream its Nyquist Frequency will do, (2 times the frequency being measured) a 20 megasample ADC would be adequate for the E4000 tuner's I/Q output.
This is IQ signal you don't need twice the bandwidth, you only need once. So a 8 Msps IQ would be adequate to represent a 8 MHz bandwidth.
[snip ....] You need at least 10 times that bandwidth to accurately reconstruct Phase Modulated Signals.
No you don't. You want a concrete example: Most GSM phone are demodulating a phase modulated signal by sampling at exactly one sample per symbol (so they're even sampling _lower_ than the actual signal bandwidth)
And although oversampling is useful in certain case to analyze signal, it's essentially useless if the signal has _already_ been filtered like it is in the E4000 case. If the IQ output has only 8 MHz of useful content at the output, you can oversample all you want, it won't bring back what the E4000 discarded ... Those antialiasing filter you mention, they're in the E4000 !
2) I didn't say bigger / better SDR would be useless ... I'm just saying if you're spending hundreds / thousands on better ADC and better PC interface you might as well ditch the E4000 and use a better RF front end as well. The spectral output of the E4000 is far from being "clean", there is distortions / LO leakage / random spurs / 28.8M clock leakage / ... etc ... Oh and Elonics declared bankrupcy so the E4000 supply is fading ...
Cheers,
Sylvain
On Sun, 8 Jul 2012 12:30:05 -0500 "Jay Salsburg" jsalsburg@bellsouth.net wrote:
The E4000 Tuner offers a stepping stone into the world of SDRs. It does the heavy lifting, significantly reducing the size, cost, and complexity of a SDR. It does not do any programmable signal processing, which is where SDRs are strongest. Quantization has been left out of the discussion because the RTL2832 is so bad at it (to conserve size and cost). The RTL2832 hamstrings the usefulness of the E4000 significantly narrowing the application of DSP. If you want high fidelity reception, you must apply a high fidelity ADC.
I do not understand how you think to arrive at this conclusion. The RTL2832/E4000 dongels allow superior quality demodulation of FM signals.
The suppression of neighbouring channels is limited due to aliasing, but it is not bad and interference can be avoided by changing the LO frequency. There is also a mirror image but if you use Linrad you can apply the calibration procedure that eliminates image spurs.
Now that AGC is eliminated these dongles allow a pretty good general purpose SDR at small cost. http://www.sm5bsz.com/linuxdsp/hware/rtlsdr/rtlsdr.htm
The high fidelity reception is in no way limited by the ADC in these dongles as long as the desired signal is not much weaker than surrounding signals. (Much means 50 dB or so.)
Regards
Leif / SM5BSZ
Well, this is interesting. Think out of the box. Ask yourself if a 8 bit NFM is high fidelity. I do high fidelity; 6 channel, 24 bit Wav files, that is high fidelity. When you have high conversion rates and high bit depth signals you get high fidelity. Audio NFM signals are only meant to satisfy walky-talky radio communication (Trunk Radio). The high value targets in Radio technology today is Data Services. Look in the (Data Stream) Mine for the Gold, Data is Gold. Data is wideband, multiple channel, and frequency diverse (not NFM), then there are all those wideband RADAR Signals. With high fidelity conversion you do not need anti-aliasing. Anti-Aliasing is a technique to filter, the very word means to take something away. Data services count on the fact that the Data is difficult to intercept.
The object of any great effort is to rise above all the rest. To get the most out of the Tuner the ADC must be improved. All you guys have so much to offer and you have spent so much of your time and money to get the most out of this $20 Dongle, all because the Manufacturer refuses to open Source the Chip. Jump ahead and make better use of it. Go to the next level and hack the circuit, not just the Operating Code.
The most extensive effort should be in acquiring intelligence from the Either, not listening to high power squawking Police channels. This requires very high granularity, the RTL2832 is low granularity. If you want to drop in on intelligence from over the horizon, very good selectivity is required and the RTL2832 is just not that; highly selective, intentionally by design.
-----Original Message----- From: osmocom-sdr-bounces@lists.osmocom.org [mailto:osmocom-sdr-bounces@lists.osmocom.org] On Behalf Of Leif Asbrink Sent: Sunday, July 08, 2012 2:29 PM To: osmocom-sdr@lists.osmocom.org Subject: Re: stroboscopic (aka equivalent time sampling) using EZCAP DVB dongle ?
On Sun, 8 Jul 2012 12:30:05 -0500 "Jay Salsburg" jsalsburg@bellsouth.net wrote:
The E4000 Tuner offers a stepping stone into the world of SDRs. It does the heavy lifting, significantly reducing the size, cost, and complexity of a SDR. It does not do any programmable signal processing, which is where SDRs are strongest. Quantization has been left out of the discussion because the RTL2832 is so bad at it (to conserve size and cost). The RTL2832 hamstrings the usefulness of the E4000 significantly narrowing the application of DSP. If you want high fidelity reception, you must apply a high fidelity ADC.
I do not understand how you think to arrive at this conclusion. The RTL2832/E4000 dongels allow superior quality demodulation of FM signals.
The suppression of neighbouring channels is limited due to aliasing, but it is not bad and interference can be avoided by changing the LO frequency. There is also a mirror image but if you use Linrad you can apply the calibration procedure that eliminates image spurs.
Now that AGC is eliminated these dongles allow a pretty good general purpose SDR at small cost. http://www.sm5bsz.com/linuxdsp/hware/rtlsdr/rtlsdr.htm
The high fidelity reception is in no way limited by the ADC in these dongles as long as the desired signal is not much weaker than surrounding signals. (Much means 50 dB or so.)
Regards
Leif / SM5BSZ
Hello Jay,
Well, this is interesting. Think out of the box. Ask yourself if a 8 bit NFM is high fidelity.
You have to specify what you mean by this. The NFM signal carries information by the instantaneous frequeny of the signal. When sampled with 8 bit frequently enough the frequency can be determined with great precision.
I do high fidelity; 6 channel, 24 bit Wav files, that is high fidelity.
This statement is basically false. There is no way that you could have 24 bit accuracy in a .wav file (unless the waveforms are synthetized in the computer from mathematical functions.)
A good 24 bit soundcard like the Delta 44 provides 18 bit data. Bits 19 to 24 contain only noise and can be truncated without any effect on the recorded information. The best modern A/D converters may provide nearly 20 bits, but providing 24 bit is not possible. The noise floor is set by the thermal noise of any resistor at room temperature.
The 8 bit data in RTL2832 is from two separate A/D converters and together they provide 8 bit at 4MHz sampling rate. That is the same as 9 bit at 1MHz or 10 bit at 250 kHz or 11 bit at 62.5 kHz.
I was thinking you were referring to FM broadcast in which case the FM detector would improve S/N by about 20 dB. (3 bit)
When you have high conversion rates and high bit depth signals you get high fidelity. Audio NFM signals are only meant to satisfy walky-talky radio communication (Trunk Radio).
Hmmm, "Audio NFM"? do you mean the audio signal delivered by an FM detector? They would have about 13 bit accuracy when produced from the 8 bit 2*2 MHz data stream from a rtl2832. In real life the S/N would limit performance and it would be identical to what you observe on a better radio unless you have really strong interferring signals.
The high value targets in Radio technology today is Data Services. Look in the (Data Stream) Mine for the Gold, Data is Gold. Data is wideband, multiple channel, and frequency diverse (not NFM), then there are all those wideband RADAR Signals. With high fidelity conversion you do not need anti-aliasing. Anti-Aliasing is a technique to filter, the very word means to take something away. Data services count on the fact that the Data is difficult to intercept.
It does not matter at all whether we do anti-aliasing in analog hardware or whether we sample at a higher rate and do it by a digital filter. The result is identical.
The effective bandwidth is what matters.
The object of any great effort is to rise above all the rest. To get the most out of the Tuner the ADC must be improved. All you guys have so much to offer and you have spent so much of your time and money to get the most out of this $20 Dongle, all because the Manufacturer refuses to open Source the Chip. Jump ahead and make better use of it. Go to the next level and hack the circuit, not just the Operating Code.
Well, I did that and came to the conclusion it would be a good idea until the AGC problem was solved. Now, without the AGC, the RTL2832 chip matches the performance of the E4000 tuner pretty well. Adding a much better A/D converter will not improve performance significantly. We would need a better tuner to motivate a better A/D converter.
The most extensive effort should be in acquiring intelligence from the Either, not listening to high power squawking Police channels. This requires very high granularity, the RTL2832 is low granularity.
Are you sure you know what you are saying here?
The RTL2832/E4000 dongle provides a dynamic range of 80 dB. You might tell us what the received power levels are from the stations you want to hear - and what is the signal level of the stations that might interfere (high power squawking Police channels...)
If you want to drop in on intelligence from over the horizon, very good selectivity is required and the RTL2832 is just not that; highly selective, intentionally by design.
The RTL2832 is just a dual A/D converter. (As operated for SDR) Replacing it with something 2 or 3 times faster would allow us to remove the alias signals that we see in the DVB dongles. It would not improve performance much in other respects however because the minimum noise floor from the E4000 is about 1 bit on the RTL2832 and running the E4000 at higher output levels would make the harmonic distortion unacceptable.
Surely performance of the DVB dongle is limited, but it is not bad. Quite respectable actually. (But you need the latest software - or maybe even a little later than what you can download today. I have not evaluated the latest changes so I do not know.)
Regards
Leif / SM5BSZ