Hi fz2101,

I don't think it's a matter of theory versus practice, I've been working with a full analog correlation receiver. As a simple example for UWB-IR receivers is one of the proposed topologies due to it's simplicity. If you think about the definition of signal space you will see you don't need one branch for "template", just a branch for each space base(of course to make the decision you will need several thresholds of comparation for a QAM for example). I think the decision of making a matched filter(correlation receiver) must be made after a system analysis(i.e. after the correlation receiver you will need a slow ADC it's just a DC level out of each branch, in a heterodyne receiver you will need a ADC at least twice as fast as your highest frequency) to determine whether or not your power budget for the receiver can accomodate a fast ADC or how many power will go to the comparators. Another question is that for example in OFDM systems the number of branches will be so big(one branch for carrier) that it will be unfeasible,then DSP is I think compulsory.
Hope it helps,

Hi frank,
Well I think that many papers published under the term "correlation receiver" are made by people of signal processing(I go to a phd thesis defense one or two months ago and they assume they receiver is an antenna, a LNA and directly an ADC-I made a simple calculation and I estimate something like 80 Gigasamples/second, no comments...-) that are more interested in algorithms for synchronization,demodulation,channel estimation and so on, I know some of this guys at my university and when you ask them "how much power will consume this receiver?" they didn't know what you are talking about. This method will be useful for something like a home-theater were you are connected to electrical network, but for a wireless portable device your battery will die very fast.

About your comments about the integrator or the LPF I think they are correct(I didn't make the system election, I worked in UWB as my master thesis that was part of a Phd thesis of the department).

About the use of gaussian pulses(or derivatives) I think that the question refers to the transmitter,you must use a good filter to do the pulse shaping(I didn't see a generator that directly generates a pulse FCC,Europe or Japan mask compliant), in UWB I think that instead of talking of a power amplifier we must start talking about a "power attenuator".

For the possibility of using a LPF before the mixer(thus reducing the BW,then reducing the ADC BW or the integrator BW) I think it's possible but must be validated by a system simulation seeing how BER degrades. As in IR-UWB there's the possibility of trade-off data rate using processing gain in the form of pulse repetition(sent each bit in more than one pulse) probably it will not be very killer(but you will loose data rate), but I've never done myself the system level simulation filtering the received pulse before the mixer so I cannot tell you for sure the effect.
I'm sorry that I cannot enter into details about the system architecture with I worked(and from time to time I still work) due to the fact that there is a NDA in the middle. I will suggest you to look a book(maybe you know it):
"UWB theory and Applications",Ian Oppermann,Matti Hämäläinen and Jari Iinatti,John wiley & Sons,Ltd. Probably if you've been working a little bit with UWB it's not necessary, but includes a raw(not FCC compliant) transceiver for IR-UWB made with correlation receiver.

Also my research group has published a couple of papers on UWB:
"A Low-Power Template Generator for Coherent Impulse-Radio Ultra Wide-Band Receivers"
"Low Noise Amplifiers for Low-Power Impulse-Radio Ultra Wide-Band Receivers"
Both of them by Enrique Barajas Ojeda and available on the proceedings of "The 2006 IEEE International
Conference on Ultra-Wideband".
Three more articles are accepted for publishing but they are not available this day.
Hope it helps,

I think my confusion comes from the word "receiver" (of the phrase "correlation receiver").

Coming from an analog background, when I hear the word "receiver," I immediately try to fit it under either
heterodyne or homodyne (mixing, downconversion, filtering, all that good business).  However, a correlator is really a detector/estimator,
and if I want to fit it under my meaning of the word "receiver", then it belongs only in the DSP portion of a MODERN DAY receiver, i.e., after
the ADC, in the digital domain.  

For example, B. Razavi's "RF Microelectronics" book discusses correlation receiver" in Chapter 3: "Modulation and Detection",
while heterodyne/homodyne are discussed much later in Chapter 5: "Transceiver architectures".  

I guess because people originally thought that UWB (impulse-radio type) was going to be carrier-less, a.k.a., baseband transmission (
but at very high speed), so they used the word "correlation receiver,"  it kind of conveys the idea that it is very simple (no front-end).  

Thank you all for a great discussion!

frank

Hi everyone,

I came across this very old but interesting discussion about receivers and I have some doubts too. Like fz2101, I have a solid analog background and a very poor DSP background and also for me a receiver is the traditional structure antenna, LNA, mixer, LPF and ADC: beyond the ADC it's still quite a mistery to me.

I'm trying to read Digital Communication by Lee-Messerschmitt, but I have difficulties finding some practical example in real life: the architectures seem to me quite "high level" and I can't see the practical implementation of these concepts. For example, I still don't know in what application a matched filter is used or if it is implemented in the analog or in the digital domain: from your discussion it seems that, for most applications, it is implemented in the digital domain (in fact, in analog design I never saw an analog matched filter during my work in mobile communications).

I'd like to ask you if somebody can explain (to an analog designer) the basic operations/techniques (even conceptually) that are implemented, for example, in a 2G or GSM receiver, both in analog and digital domain: I haven't yet found the answer in classical digital communications book. Fundamental doubts are: there is somehow a matched filter for 2G? How it is implemented and what's its purpose in 2G? I suppose it is not to maximize the SNR of a single pulse... Moreover, we also have a Viterbi decoder? Why it is necessary?

All these doubts and concepts are still quite vague to me right now (and also I have trouble to formulate precise questions about them), but I hope somebody can give me a global vision.

Thank you in advance in any case!