Discussion in 'Uniden' started by SixPackABS, Sep 26, 2017.
I see what you did there
I guess you need to read up on this board more frequently. I don't remember what it was called but the horn design lends itself to an analog detector, where some other thing would have pointed to a true DSP design.
And again, please provide proof where it's a true analog detector. Just because the Max has DSP written on it's label doesn't make it a true DSP detector either.
Why is scanning in a linear fashion a deciding factor in a digital design, and what do you exactly mean by this?
The Max is very fast due to the FPGA allowing for parallelism. Using a DSP you're still governed by clock rates and sequential tasks. You can try to speed things up by taking advantage of DMA, but you can only do so much. Escort realized this which is why they proved their concept using a DSP but the final design utilizes a FPGA. Probably a lot easier to work with the DSP for R&D purposes.
I think it makes perfect sense that taking in less data with a digital detector design that uses a digital signal processor (DSP) rather than FPGA provides significant benefits.
To me, if the detector uses an ADC and FFT algorithm while completing main signal processing functions in the digital domain then you have a "digital detector". Subsequently it makes the most sense to use a DSP or FPGA for such tasks.
Correct me if I'm wrong here.. the designers of the R series have said they utilize a FFT algorithm. We know that if they're using a FFT then an ADC must be used as well. If we know the R series has a DSP which is a purpose built processor for handling signals in the digital realm then there is a good chance the R series is "digital".
We don't know the sample rate of the ADC, but we can assume the R series will not be as fast as the Max series because it cannot take advantage of parallelism.
To me there's at least reason to believe at this point that it is a digital detector. If you have doubts, then I would say you have to take an agnostic stance on the issue because we haven't seen anything that would surely lead us to believe otherwise. Saying it's not digital without looking at what's going on underneath the horn, which would only be conjecture based on hardware to be fair, is a bit rushed IMO.
Obviously it's not really a spectrum analyzer, but some derivation of what is captured in this link: http://rfmw.em.keysight.com/wireles...s/concepts/content/concepts_types_spec_an.htm
Sent from my Pixel using Tapatalk
My apologies for not replying sooner. I only saw this post two days ago and didn't want to give it a rushed reply. I enjoy talking and hypothesizing about how the designers (both hw and fw) think about and attack these issues.
Not in the same order you mentioned these in but here goes -
Yes, the Uniden designers mentioned using the FFT. Now, I am only mentioning what I'm about to say as an example of one possible use of an FFT outside of having a digital detector. It's possible that an analog detector could just be using an FFT once a signal's been detected to simply get a frequency to display for the user. Nine_C1 commented about this some time ago and I believe in talking with the Escort engineers, he realized - or was told - that the detector didn't need to run an FFT for this purpose. Those engineers explained that they knew the frequency of the incoming signal because the fw has to set (and step) the frequency of the Local Oscillators (LO's) to "tune in" the different frequency ranges captured when scanning. So my mention of using an FFT like this may not be used in real life but it could be.
Both analog and digital detectors would require an A/D converter. If you look back at the M3 patent that Nine_C1 posted a time or two in the past, you'll notice that after the RF section, the signal is finally fed to an A/D. The reason for this in both analog and digital detectors is that microprocessors/DSPs/microcontrollers cannot work directly with analog signals. Sampling the analog signal turns it into a series of digital values. Mathematically, the series of digital values the detector obtains is called a time series.
Now, to get to your question about my statement that "sampling in a linear fashion implies an analog architecture".
First, I will admit that my choice of words there wasn't great.
But what I was trying to say is this - running an FFT takes time. On a DSP, you're only talking a few milliseconds but the amount of time the FFT takes is proportional to the bandwidth of interest. For us, Ka band is 2.6 GHz wide. Some older detectors would sample 10 Mhz at a time so there would be 260 of these samples to perform to scan the Ka band once. If the detector was scanning 100 MHz at a time, there are only 26 of these samples to perform to scan the Ka band once. The issue here is that you need to run the FFT on each sample. So scanning a small section (10 MHz of the band) means you've got to run the FFT 260 times to scan Ka once. If you're sampling at 100 MHz, you've got to run the FFT 26 times to scan the Ka band once.
I believe most detectors will scan the bands a few times a second. This would give a reasonable response time. So this means we've got to scan the 3 bands once every (say) quarter of a second.
Now back to bandwidth. There is something called the Nyquist limit. This limit states that you must sample the signal of interest at at least twice the frequency you're scanning. So if you're scanning 10 MHz at a time, you'd need to sample this at 20 MHz to properly capture the data/information present in that range. If we scanned in 100 MHz chunks, we'd need to sample at 200 MHz to capture the data/information present in that range.
Another issue that comes up is the amount of data generated when we're sampling at the different rates. Remember the analog signal is being turned into digital values for us to feed to the FFT. Something that affects the execution time of the FFT algorithm is the "number of points" the FFT is using. Very, very loosely, this translates to the frequency resolution you want the FFT to give you.
So the frequency resolution we desire comes into play here. Since our detectors often display down to 1 MHz (3 digits to the right of the decimal point), a 10 MHz region is satisfied with 10 points. 100 MHz means we want 100 points. The 100 point FFT takes longer to run than the 10 point FFT.
So a number of things conspire against us regardless of whether we want to take a lot of small samples/chunks (say 10 MHz) versus taking a smaller number of larger chunks (100 MHz). When we scan smaller chunks (say 10 MHz), we have to run the FFT algorithm 260 times to scan all of Ka just once. If we scan at 100 MHz, we need a really fast A/D converter and a lot of memory to store the data the A/D converter is producing but then we only need to run the FFT 26 times and not 260 times for each scan of Ka.
The choice of FPGA or DSP begins to matter now. The FPGA is doing things much faster because it's not executing fw. The DSP is executing fw but must do so slower than an FPGA can run.
But back to my suspicion/statement that the Uniden is probably analog. To begin this part of the discussion, I will go back to Escorts choice of the FPGA. In real life, the A/D converters present in the FPGA can sample much faster (usually) than when the A/D converter is a separate chip. So if we want to scan 100 MHz at a time, we need an A/D converter that can sample at twice that - 200 MHz (this is the Nyquist limit mentioned above). And sampling that much is producing a lot of data ( 200 million digital values per second) which is probably why the Max series had a DDR ram chip for the FPGA to use. But most importantly, the benefit you receive by sampling at such a high rate is that you run your FFT far fewer times per second. And if the FFT on a general purpose DSP chip takes even a millisecond, having to run it only 26 times a scan is very beneficial.
Looking at it the other way, say we only sample at 10 or 20 MHz. Our A/D converter can be slower and we don't need as much ram to store the samples the A/D is producing but the downside is that we need to run the FFT much more often. And even if the FFT only takes a millisecond, having to run it 260 (10 MHz sample rate) or 130 (20 MHz sample rate) times per second begins to eat away at the detectors response time.
Something I've mentioned before is that I can only speculate about how the detectors work. I know a lot about how voice band radios work and my employer markets both analog (older) and digital (newer) radio architectures. The hardware for receiving RF signals, whether they're voice or radar, is remarkably agnostic to the information being transmitted/received. But again, I am speculating about radar detectors. There may indeed be some difference that I don't know of and that difference may matter.
So again, I think a digital architecture would be easier to implement with a higher sampling rate. In Escort's case, they're grabbing a big chunk (maybe as large as 100 MHz) of the Ka band with each sample. This is why when they finally did talk about segmentation, they said the frequencies were still scanned, but some frequencies were just not alerted to. The reason for this is if you're going to scan 100 MHz, you're going to get all the information within that 100 MHz. You can't "not scan" a section of that region. You have to scan it all. You can then decide not to alert to part of it.
Lastly, to be perfectly clear - when I'm talking about scanning 10 or 100 MHz, realize that we're talking about the Ka region between 33.4 to 36 GHz. If we're scanning a 100 MHz at a time, the first scan is getting us the information from 33.4 to 33.5 GHz. The next scan is getting us the information from 33.5 to 33.6 GHz and so on.
I hope this is somewhat clear. As always, feel free to ask questions or make comments.
For a CHE... actually I should say for someone who's not an EE, you've really caught on quickly
Someone posted in another thread that Uniden calls the detector a "hybrid". I don't know what that means but I asked in that thread. I'll have to find that post again to see if my question was answered.
One last note - I haven't been receiving emails when people flag me with the @ sign. If you do post questions to this, maybe please PM me so I know you've done so? I've checked my settings but I've got to look around to see if there's a forum help section. I have no idea why the emails stopped. Others have mentioned that they still get emails in response to the @ sign so I have no idea what's going on.
Really informative Dan and I appreciate the time you took to share all of that. I'm sure it helped everyone learn more about how RDs work and it certainly is helping me gain a better understanding of things. Thanks for the kind sentiment as well, I just try to read what's out there and research into various topics further.
I guess where my confusion stems from is Escort's initial design for their DSP platform is based on 10MHz steps, not something much wider like 100MHz. They sample each 10MHz step 8 times and utilize a A/D converter clocked at 100MHz. This definitely fulfills the Nyquist sampling theorem while providing dB benefits from their averaging technique. That said using a digital signal processor rather than a FPGA doesn't allow you to go much beyond the scope of a design like this if you need to average each step multiple times while remaining cost effective.
Because of what I just mentioned this is possibly why Escort went with a FPGA design - to squeeze more out of the concept without being limited by hardware and costs constraints.
Their processing technique seems a bit rudimentary though, and apparently has issues with detected signals drifting across frequency bins between samples which if I understand things properly washes away most of the dB gains.
Anyway, so to me it seems perfectly reasonable to assume since Uniden is using a digital signal processor rather than a FPGA they have a lot to gain from segmentation. They likely aren't stepping through the spectrum in very wide steps due to sampling frequency constraints, but as we see with Escort's patent that's not pertinent to the detector being "digital". The width of the individual steps is governed by the processing techniques and what's necessary to get the job done.
To me still, a detector is what we'll call digital if it's converting the analog signal data to digital data fairly early on, completing a FFT algorithm on the signal data, and then using signal processing techniques on the data which is digital and in the frequency domain. Because this is done you can also simplify the RF hardware and remove the need for some components that aid in signal conditioning/processing. If you look at the RF board of the R series it seems much more condensed than the LRD/DFR series'. A whole 1/4 of the RF board on the R series is related to the op-amp and whatever else is going on there; the board just looks a lot different than your regular analog detector.
Sent from my Pixel using Tapatalk
@FWGuy I have started to see the term "hybrid" show up in discussions of older sattcomm (think T-1 24 channel) equipment. In order to extend the life of analog equipment from previous generations birds they front end with digital data banks collapsing 24 channels into one and processing all 24 analog channels into 1digital channel. Many are calling this scenario a hybrid configuration.
I suspect the CM folks (some) are calling a analog to digital conversion a hybrid.
Oh yeah, I forgot to comment on that. The hybrid detector has been listed on Attowave's website long before the R series was made aware to anyone.
I actually think it's the technology they use for the LRD/DFR and Radenso XP units. Check this out:
Definitely not the DSP platform that is exclusively designed for Uniden in windshield mounts and NetRadar in remote form. If I understand things from what little had been said by BRD he essentially started this whole thing. He approached Uniden to get back into the game and have a plan to rise to the top, and because of this "partnership" he must have secured rights to future exclusive technology that would be developed. Basically it's my understanding that BRD saw the opportunity to cause a shift in the industry and then partnered with willing parties to make it happen. I'm not quoting him, just putting the pieces together.
Sent from my Pixel using Tapatalk
I was going to make these same points. And while we're at it, for all we know, the word "hybrid" on the Attowave English language page may just be a bad translation into English (anyone read Korean?). Not to mention they never explain what they mean by it. I think it's a leap to necessarily assume it's referring to a mixture of digital and analog given it's been presented that way at least since the LRD950 and Radenso XP came out (and possibly before that). And nobody ever tried to claim either of those detectors are digital (or at least partially so).
I think Brainstorm has a point - the term may be a mistranslated phrase from the Korean language. I just went to the link cihkal provided and Attowave's definition of a hybrid detector is very generic. Their website says (and this is their capitalization and punctuation) -
"Hybrid radar detector is the most advanced Radar and Laser detector in the world. With Ultra high Sensitivity and performance Antenna. Hybrid Radar Detector offers you perfect protection from All-Band Radar and Laser systems."
From the link cihkal provided, click on "product info" on the left side of the page and then you can click "Hybrid Detector". Like I said, I preserved their capitalization and punctuation.
I've mentioned before that I think some companies play fast and loose with terminology. I guess foreign companies do that too with the added interference caused by the language translation.
Me love you long time
I will stick to the terminology in the patent as well as I can. I probably shouldn't have introduced numbers (10 MHz vs 100 MHz sampling rates) but I was trying to convey the constraints imposed by sampling smaller chunks of the spectrum and the benefits provided by sampling larger chunks. If you could sample at 100 MHz, you'd only have to run the FFT 26 times to process the 2600 MHz of Ka bandwidth. If you sample at 10 MHz, you have to run the FFT 260 times to process the whole Ka bandwidth. The FFT takes time to run so we care about how fast it runs and how many times we need to run it to scan through the Ka band once.
As I said, I will stick to the issues/items mentioned in their patent so as not to introduce confusion. Again, I am not an expert in digital signal processing but I've done a fair amount of DSP programming in my life. In my current job, we have a PhD in Mathematics who's responsible for a number of things. One of these things is helping people like me when the math I need to know gets to be a little deep.
If Escort says "10 MHz" sampling in their digital patent, this is probably close to what's being done in real life. But I believe they've also said that their digital platform samples much faster than the analog platforms. I took this to mean that they were sampling bigger chunks of bandwidth in their digital unit and if their digital unit is sampling 10 MHz at a time, it may be their older analog units aren't sampling that much of the Ka spectrum at a time.
But, when they say the digital architecture is sampling faster than the older analog architecture, it could refer to what you pointed out to me via their patent - they could be sampling the same amount of the Ka band (10 MHz) but doing that much faster so they have 8 samples to average together without much of a time penalty.
In an ideal world, noise is random and for white noise, the more samples you can add together, the more the result approaches 0. In other words, random noise cancels out over time. This is the effect Escort is trying to achieve by sampling multiple times and adding the spectrums together. If the noise can cancel out, the signal (if present) will stand out all the more.
You make an interesting point though. It could be in real life, the signal didn't always show up in the same bin. This effect is called "smearing" and it would definitely reduce any benefit they hoped to attain from averaging down the noise. I don't think they'd ever acknowledge why the Max didn't live up to it's billing even if they knew but once they decided to rely on averaging a number of samples together, smearing became a possibility and it would be very hard to eliminate/reduce.
One cause of smearing is that we're really running something called a DFT (Discrete Fourier Transform) and not a real FFT because a real FFT requires the signal we're processing to extend infinitely in time - both into the past and into the future. We don't have this in real life, so the DFT is an approximation to the FFT that we use in real life. I don't want to say more about this because I don't know the complete details and we're getting to the limits of my "off the top of my head" knowledge.
Anyway, let's forget the averaging part of this whole discussion for the time being. Suffice it to say that Escort was trying to reduce the background noise by averaging a bunch of samples.
Now, back to the digital versus analog discussion - I agree that Uniden could very well be using a digital architecture as we're defining it - running the digitized data through an FFT and analyzing the data in the frequency domain - and that they could be doing so using a smaller sample size. I was suspecting that Escort and their digital architecture was sampling at a much higher rate because the benefits to be realized (running the FFT fewer times) are apparent.
But since Escort's patent says 10 MHz, either the analog architectures are sampling much smaller sizes or Escort, with their FPGA approach, can run the FFT as often as is needed even by sampling only 10 MHz at a time. Again, the benefit of the FPGA is that is runs *much faster* than a Digital Signal Processor chip. That doesn't mean the DSP chip doesn't run fast enough.
Going back to something Escort said - because they were sampling "large chunks" - and 10 MHz may be a large chunk compared to what the older analog detectors sampled - they said they couldn't "not sample" a segmented frequency. This is why my initial reaction to "Uniden is providing segmentation" was to say "Uniden is running an analog architecture". Escort still does the conversion on that frequency but they just don't alert to it. This makes sense because of the way sampling works. If the frequency you want to ignore falls in the middle of the 10 MHz, you'd have to break one 10 MHz sample up into 2 or more samples and this would increase the time it takes to process that 10 MHz chunk. So they do the whole 10 MHz sample and then, via the DSP or FPGA or whatever, choose not to alert us to some value that happens to fall inside that sample.
Again, to be clear about something - Ka band extends from 33.4 GHz to 36 GHz. When we talk about sampling the Ka band in 10 MHz chunks, we're saying that the first sample is getting us the frequency range from 33.40 to 33.41 GHz. The second 10 MHz chunk is getting us the data from 33.41 GHz to 33.42 GHz and so on.
So there are about 260 of these chunks (or "channels" as the patent calls them). Add in 25 more 10 MHz chunks for K band and another 5 chunks for X band and we're at 290 chunks. And add it a little extra for oversampling/overscanning the bands and we're at the 300 channels the patent talks about.
So back to my original post that started this whole thing off - that Uniden offering segmentation meant that it had to be an analog architecture. I think I must now take that back because I didn't remember the details of the Escort patent. I though they were sampling much larger chunks than 10 MHz.
But, as Escort says they do, the Uniden could also be sampling the whole spectrum and just not alerting to portions of it. This is the approach Escort came to for the Tesla noise issue they saw in the MaxCi series. I'm sure they could go back and easily apply segmentation to the older digital models like the Max or Max360 if they wanted to.
Turning attention to the pics you posted, yes, it's obvious the R series antenna is a bit more complicated than the older LRD antenna. I suspect the older antenna has some additional circuitry underneath (or somewhere). Remember that when we looked at the part numbers for the chips on the R series antenna, we found a pair of frequency synthesizers/PLL's and the op amp (part labelled C082). I think the older antenna needs those frequency synthesizers because those are the LO's (local oscillators) we talk about. Granted, Uniden (or Attowave) seems to have found a novel new architecture that's giving them some amazing sensitivity.
Why Uniden is benefiting from it, I don't know. I don't know if this new antenna is something Attowave happened to develop and their intent was to offer it or make it available to the first customer with deep enough pockets to pay for it or whether, as you say, BRD came into the picture and had Uniden ask Attowave to come up with something new.
I'd love to know the background story but that's probably not going to happen. BRD (Tom) may have been instrumental in getting Uniden and Attowave together but he may have stayed on the organizational/management side of things. Nothing wrong with that at all but when I say I'd like to know more, it's because I've always been interested in the technical side of things. I'd be much more interested in talking to the Attowave hw and fw engineers in hopes of understanding why and how they made their breakthrough. Side note to Tom - Please don't feel bad. I'm sure talking to you would be interesting too but I'd rather understand the technical details first
Brainstorm made an interesting comment just today I believe - that we couldn't trust some of the info we've been given on other sites. I certainly agree with that but if you consider what happened, you'd be leery of sharing technical info also. By "consider what happened", I'm referring to when I joined ERF. I was interested in the technical details more than most forum members. Larry doesn't know whether I'm a programmer for Valentine or Uniden or whoever. So naturally, I'm not going to get my questions answered. Same if I email BRD or Uniden - wouldn't matter that I just care for personal interest and that I wouldn't breathe a word of it off the site, no one if going to tell me anything about how their detectors work.
And in a way, that's the way it should be. That's some hard won knowledge there.
Ok, this is long enough for now. I do have some more to say about the sampling thing but I'm going to leave that for tomorrow or the next day.
Great stuff Dan. You might find this interesting, the patent for the M3 makes no mention of an analog to digital converter, and it's very detailed. It does mention a digital to analog converter in a couple areas.
What I find really interesting is they have what I guess is a FM demodulation design that takes in 10MHz sections of the spectrum. Im trying to learn how it works, but here's what I'm guessing:
*The various bands are mixed to IF frequencies that are much closer together.
*Because the detector is sweeping for frequencies and mixing them to a range which is close together it would appear like a modulated signal. Say if you mixed everything to the frequencies which make up K band. You could have what appears to be a FM source of that specific frequency range.
*The FM demodulator takes the incoming IF data and I guess somehow converts this to a voltage signal which can be used with the micro processor. Much like a 0-10v potentiometer or 4-20mA analog source. Since the RD knows what it's sweeping for it can correlate that with the incoming data to determine accurate signal information.
I am completely guessing on what I said above, but it appears they don't use a ADC for the M3 design. That would have been listed on one of the drawings or at least mentioned since just about every related detail is captured. Take a look at the drawing from the patent.
So maybe when Escort says it's taking in large chunks of the spectrum they really mean that in a way to describe taking in 10MHz 8 times which follows their patent and would be a lot more data than these older analog (FM demod) designs can do.
Sent from my Pixel using Tapatalk
That's a lot of reading...
@FWGuy and @cihkal , excellent discussion of the methodology and implementation of the architectures of the Uniden (of which we're not 100% sure?? and the Max and Max 360 series of digital detectors. Dan, we do know from ER themselves that they are taking large gulps (their term, not mine) or chunks as you stated as samples. When we asked about segmentation, we were told that it could not be done as the detector was already "segmenting " the bands by just not alerting to them as they were scanned. So, what you say in bold above is consistent with what ER has to told us the past 5 years with the Max series. However, my question remains then, did they truly segment the Max 360 Ci front horn (M7) or just not alert to those frequencies that were causing trouble with the Tesla (@jdong brought this to our attention early on).
@PointerCone - you get the stuffed Teddy Bear prize (or the stuffed Chinese detector) because you're asking a very good question
For the sake of discussion, let's say the frequency we want to segment (not alert to) is in the middle of a sample Let's also stick with the 10 MHz sample size. And for clarity's sake, remember than a sample grabs the information within some number of MHz out of the whole 2.6 GHz of Ka band that we care about. Ka band is 33.4 GHz up to 36 GHz. So our first 10 MHz sample is capturing the information from 33.40 GHz to 33.41 GHz. The second sample would be grabbing the information from 33.41 to 33.42 GHz and so on. 10 MHz is .01 Ghz so hopefully you see and agree with my math. When I say we're "sampling" the 10 MHz chunk, realize that we're going to "sample" enough chunks that we get all the information in the Ka band from 33.40 GHz up to 36.00 GHz. The term "scanning" is almost synonymous with "sampling". We "scan" the whole band by "sampling" small pieces and collecting/accumulating the results.
Before I go further, I'll say this - I'm not entirely sure what you mean by the sentence I've bolded in the copy of your post immediately above. I'll come back to that later.
The issue is this - once we start to sample, it's easier to let the sampling operation complete. The reason is that this "sampling operation/procedure" is a non-interruptable operation. This means that once it starts, there's no easy way to interrupt (stop) it and then continue from the point we left off at. But the sampling operation is fast. So by letting it complete, we get all the information present within that 10 MHz piece of Ka band. If the frequency (or frequencies) we want to ignore fall within that region, we just don't alert to them even though we have the information/data necessary to do so because we let the sample operation finish.
I'm taking a guess here as I've read your (and others) comments in past posts. I believe things are being confused because you and others have an intuitive (and correct) understanding that sampling/scanning takes time and if you can eliminate some of the sampling/scanning, you save time. This of course is true.
What's also true is this -
Escort's digital approach increased the amount (frequency region) within each sample. Now, the new sample size for the digital unit is (say) 10 MHz whereas before it may have been smaller. So if the frequency or frequency range we want to segment out (not alert to), falls inside this 10 MHz region, we have two options. We can 1) break our 10 MHz sample into two or more samples. The first part of the frequency region would be the start of our 10 MHz region up to the lower limit of what we want to ignore and the second sample would begin at the end of the range we want to ignore and continue up to what would have been the end of our 10 MHz chunk. The second option is what Escort chose to do - sample the whole 10 MHz range and just not alert to it if the user wants to ignore some of those frequencies.
For the most part, sampling a chunk of the Ka region takes roughly the same amount of time as sampling a smaller chunk of the region. So by deciding to break that one 10 MHz region into 2 separate samplings, you've been struck with a time penalty. Sampling twice takes longer than sampling once. By sampling twice, the detector doesn't get the information the user want segmented out. By sampling once, the RD gets the information and the processor must ignore it. This "ignoring" step is far, far faster than the collection (sampling) step.
So back to the sentence from your post that I put in bold - I suspect Escort is trying to tell you that it's better for them to sample/scan the whole (or portion) of the band and then just ignore any frequencies the user wants filtered out. The other way of doing it - not scanning the frequencies the user doesn't want alerts for will actually take more time than grabbing all the information and then not alerting to some of it.
So on the 360Ci, I believe they're scanning/sampling the whole band and the processor is just not alerting to the frequencies the user wants eliminated. This is actually the faster approach even though you're (the detector is) scanning a larger frequency range than in the old analog way.
In the past with analog detectors, not scanning saved time. It turns out that the new approach - scanning big regions and not alerting to them is faster than scanning a number of small regions so that you don't get the frequencies the user wants to ignore.
I hope this helps. Always feel free to ask questions. Remember that I don't get emails anymore when someone uses the @ sign. I've checked my settings, etc. and haven't found a reason for not being emailed.
@cihkal - I will reply to your post in the next day or so. I have to get off for the evening in a little while.
While they don't specifically mention an A/D converter, I believe it or something along the same functional lines - you mentioned a pot or 4-20 ma source - has to be there. Remember there are several meanings to the term "DSP". One of those is "the area of mathematics whereby analysis of a continuous (analog) signal that has been sampled is made".
You're correct that the 3 different police radar band frequencies are converted to lower and much closer together frequencies. Before going to the processor, they pass through the 10 MHz FM demodulator. So what started out as X band or Ka band ends up being a 10 MHz signal just before it gets to the processor. Notice in the block diagram you posted that there's a "Detection MicroProcessor" block prior to the "Host MicroProcessor" block. And also notice that before the 10 MHz FM Demodulator, there's a 10 MHz BPF (Band Pass Filter) block. No importance of the band pass filter except to filter out the noise introduced by all the frequency conversions needed to get from 10-36 GHz down to 10 MHz.
As for them mentioning an A/D converter, I don't know. It is a very common part since a lot of our (human beings) environment is analog - voice, video, music, etc. and a lot of over the air transmission protocols are digital. Signals at the source need to be changed to digital and then once we've received them, they need to be changed back to analog for our ears (D/A).
On a different note, something I've mentioned a time or two was that to me, I was less concerned about whether the detector is analog or digital, but I was far more interested in the performance they've managed to achieve. To me, the performance is all the more impressive if the detector turns out to be analog. Digital done right should give better performance over an analog unit but for another version/spin of an analog detector to add 3 or 4 dB over and above what other companies have been doing the last 15-20 years is REALLY impressive to me.
As I mentioned, I'd love to know more about how the R series came to be. Right now, it looks like it'll go down in RD industry history as being a jump in performance over the competition - as well it should.
Kudo's to BRD for whatever role he played. He should be very proud as the father of a bouncing baby radar detecting king!
It's VERY early for me, since I've just started running the R1, but this is BY FAR the only series able to match and beat the M3 consistently on 33.8, and 34.7. Next closest...LRD950(on 34.7) and XP and 360 on 33.8.
This line so far has shown incredible performance. More testing and real world to come, but man it's awesome.
It's a breath of fresh air, especially for those in my terrain. Finally something to run that can cover my ass without a doubt.
Sent from my iPhone using Tapatalk
I've always followed your posts because you're someone who has 33.8 locally. I believe some of the guys out west do also but it's convenient when you talk about it since I'm familiar with some of the roads up your way. I grew up on Long Island and spent a fair amount of time is NNJ.
Thanks for commenting
Don't follow just mine! Lol.
Thanks Dan appreciate it. Love it here. Lots of radar. That's for sure
Where in NNJ?
@NPark - I grew up in Astoria Queens and Syosset N.Y (out on Long Island). In Astoria, I was only a mile or so from the old Shea Stadium (now called Citi).
I have cousins in Phillipsburg and Freehold.
Haven't been up in more than a year so a trip is definitely due.
Separate names with a comma.