Post by thread:[PIC]: ETS on a PIC ??? Yes

>I've been playing around with equivalent-time sampling on a PIC, >and have put up some info about this on the web. >http://www.oricomtech.com/ets.htm Good stuff Dan! Isn't your first example sort of a non-test, since it violates the Nyquist criterion? In other words, isn't it already known that you can't sample a 120KHz signal at 50KHz rate? I.e., it requires a minimum of 240KHz rate. Or am I missing something? I'm trying to do something similar, but at much lower frequencies, using the A/D at 10KHz using a 4MHz crystal. By the way, that World's Smallest Testbench looks amazing. Is there any possibility of an option to reduce the number of input channels for the logic analyzer to increase the sampling rate? It would be nice to get a 2- or 4-channel analyzer running at 5-10MHz instead of the default. I want one when it comes out :) --Andrew _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

>Sounds about right 20:4 = 5:1 = 50khz:10khz. You would need to select >AD clock = 8 Tosc. >============= Yes sir, works well. Interestingly, I'm curious as to which frequency I would select at 20MHz. The datasheets say that 32Tosc violates the minimum A/D time, but the only option for 20MHz, then, is RC, in which case the Tad is 2-6us (4us typ.)! I'm not sure what's happening here, and I haven't tried it at 20MHz to find out. Any info? >Ha - noooooo way with a PIC76, 2.5 Mhz is tops, although 18Cxxx w/10mhz Exactly what's the limiting factor here? >xtal with PLL enabled might do 5Mhz. Scenix will do 20 Mhz all day long. Hmm, sounds like a project for the WFT (World's Fastest Testbench) :) --Andrew _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Drew wrote: >>Sounds about right 20:4 = 5:1 = 50khz:10khz. You would need to select >>AD clock = 8 Tosc. >>============= > >Yes sir, works well. Interestingly, I'm curious as to which frequency I >would select at 20MHz. The datasheets say that 32Tosc violates the minimum >A/D time, but the only option for 20MHz, then, is RC, in which case the Tad >is 2-6us (4us typ.)! I'm not sure what's happening here, and I haven't >tried it at 20MHz to find out. Any info? > Umm, my datasheets for both 'C76 and 'F876 show 32Tosc is ok at 20 Mhz. 8Tosc does violate min Tad. Forget RC, you can find a way to do it using the clock. BTW, using 'C76, you can get reliably sampling right at a 50 khz [~20 usec] rate with 20 mhz xtal - although there may be a slight LSB resolution loss - at least according to the datasheet. Looking at it on the screen looks fine however. OTOH, you cannot go quite this fast on the 'F876, due to its having a 10-bit A/D. However, you can slow down to 40 Khz [~25 usec] sampling, and all is ok. ============= >>Ha - noooooo way with a PIC76, 2.5 Mhz is tops, although 18Cxxx w/10mhz > >Exactly what's the limiting factor here? > How fast you can execute an instruction. You will notice, with a 20 mhz xtal, sampling at 2.5 mhz allows you "exactly" 2 instructions to do everything --> ie, linear code. =========== >>xtal with PLL enabled might do 5Mhz. Scenix will do 20 Mhz all day long. > >Hmm, sounds like a project for the WFT (World's Fastest Testbench) :) > WFST - World's Fastest Smallest TestBench. Been there, done that, too :). - dan -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Hi, At the risk of being a labeled a heretic... I've recently worked with an ADuC812, which is an 8052 variant by Analog Devices with a 12-bit, 8 channel ADC. It's maximum conversion frequency is spec'ed at 200kHz. It has a DMA mode that can write data to external memory. That's pretty much required at 200k samples per second, since only 5 usec are available to process the data. Like most classic '51s, Tosc is divided by 12, so that doesn't leave much time for processing your data. However, the DMA mode leaves the core free to do its normal thing: e.g. you get 200k SPS at 12 bits written to memory, and get an interrupt when the data transfer is done. Perhaps useful for those wishing to build a faster PC-based oscilloscope? ETS could be done on the external memory by the uC or the host PC. The processor was around $8.30 for 1000 (working from memory), direct from ADI's website. I would think it's a good fit for this type of application. {Original Message removed}

Don Hyde wrote: {Quote hidden}>Snip > >> Good stuff Dan! Isn't your first example sort of a non-test, since it >> violates the Nyquist criterion? In other words, isn't it >> already known that >> you can't sample a 120KHz signal at 50KHz rate? I.e., it >> requires a minimum >> of 240KHz rate. Or am I missing something? >> >Snip >> > >Read the chapter on Nyquist again, and ponder. > >What it says is that signal components between f/2 and f (the sampling rate) >are folded back into the frequency band 0 to f/2, and their frequencies will >be inverted i.e. a signal at (f/2 + a) Hertz will look like a signal at (f/2 >- a) Hertz. It goes on to say that frequencies between f and 1.5f will get >folded so that they look like frequencies between 0 and f/2 Hertz. If you >sample a 120 KHz signal at 50 K samples/second, it will look just like a >signal at 20 KHz. Exactly - the 1st figure Drew was referring to illustrates this in the time domain. Combined with the 2nd figure, the two show that the practical limitations of normal sampling by the PIC A/D can be overcome using the ETS mode. Basically, the normal mode with max 50khz rate will adequately display a waveform of only about 5 khz max, while ETS mode pushes this out to 120 Khz. Who'd a thunk it on a PIC? 120 khz analog signal acquisition. - dan -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

At 07:27 PM 2/16/01 -0800, you wrote: {Quote hidden}>Don Hyde wrote: >>Snip >> >>> Good stuff Dan! Isn't your first example sort of a non-test, since it >>> violates the Nyquist criterion? In other words, isn't it >>> already known that >>> you can't sample a 120KHz signal at 50KHz rate? I.e., it >>> requires a minimum >>> of 240KHz rate. Or am I missing something? >>> >>Snip >>> >> >>Read the chapter on Nyquist again, and ponder. >> >>What it says is that signal components between f/2 and f (the sampling rate) >>are folded back into the frequency band 0 to f/2, and their frequencies will >>be inverted i.e. a signal at (f/2 + a) Hertz will look like a signal at (f/2 >>- a) Hertz. It goes on to say that frequencies between f and 1.5f will get >>folded so that they look like frequencies between 0 and f/2 Hertz. If you >>sample a 120 KHz signal at 50 K samples/second, it will look just like a >>signal at 20 KHz. > > >Exactly - the 1st figure Drew was referring to illustrates this >in the time domain. Combined with the 2nd figure, the two show >that the practical limitations of normal sampling by the PIC A/D >can be overcome using the ETS mode. Basically, the normal mode >with max 50khz rate will adequately display a waveform of only >about 5 khz max, while ETS mode pushes this out to 120 Khz. > >Who'd a thunk it on a PIC? 120 khz analog signal acquisition. > >- dan > Hi, The signal must necessarily be periodic for this to work.... Lets not forget that restriction! Cheers, Don -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Hi Don, For certain methods of accomplishing reconstruction, you are right, but I don't think that is true in general. You can take any signal bandlimited to delta_f, sample it at 2*delta_f, and reconstruct it from the samples. It doesn't matter what the center frequency is, or whether it is periodic, as long as it is bandlimited. So, for example, you could have a signal which filled the entire band from 0.9 to 1.1 MHz, sample it at only 0.4 Megasamples per sec, and reconstruct it from that, as long as you had the additional piece of info that its center freq was 1 MHz. Sean At 07:41 PM 2/16/01 -0800, you wrote: {Quote hidden}>Hi, > >The signal must necessarily be periodic for this to work.... >Lets not forget that restriction! > >Cheers, >Don > >-- >http://www.piclist.com hint: PICList Posts must start with ONE topic: >[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Bear in mind that the series resistance of the sampling switch (6k at 5V) and the value of the hold capacitor (51.2pf for a '76) will band limit your signal to under 500khz. However, with an external fast sample and hold, one could digitize video at a 50khz rate (3 samples per video line). By sampling progressively later in the line, one could acquire a 640x480 60Hz video frame in 3.8 seconds. (A fully analog version of this technique was used back in the 70's to convert full speed video to the 15Hz line/7.5S frame of slow scan television used in the amateur radio bands (TV in a 3khz BW)). "Sean H. Breheny" wrote: {Quote hidden}> > Hi Don, > > For certain methods of accomplishing reconstruction, you are right, but I > don't think that is true in general. You can take any signal bandlimited to > delta_f, sample it at 2*delta_f, and reconstruct it from the samples. It > doesn't matter what the center frequency is, or whether it is periodic, as > long as it is bandlimited. So, for example, you could have a signal which > filled the entire band from 0.9 to 1.1 MHz, sample it at only 0.4 > Megasamples per sec, and reconstruct it from that, as long as you had the > additional piece of info that its center freq was 1 MHz. > > Sean > > At 07:41 PM 2/16/01 -0800, you wrote: > >Hi, > > > >The signal must necessarily be periodic for this to work.... > >Lets not forget that restriction! > > > >Cheers, > >Don > > > >-- > >http://www.piclist.com hint: PICList Posts must start with ONE topic: > >[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads > > -- > http://www.piclist.com hint: The list server can filter out subtopics > (like ads or off topics) for you. See http://www.piclist.com/#topics -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Hi Robert, Yes, you are right, I should have added the proviso that your sampling system needs to have a high enough bandwidth (and your sampling aperture needs to be very small and low jitter). I'm not talking about the progressive sampling method, though. I'm talking about just taking samples normally. As another poster pointed out, aliasing really only occurs when two "copies" of the signal overlap (in the frequency domain). This will only happen if you exceed the bandwidth allowed by the Nyquist Sampling Theorem, NOT just because the center frequency is very high. Sure, the output samples will be consistent with a much lower center frequency, but who cares if the modulation riding on top of them is the narrowband signal you are looking for? You really don't need to do any special processing (such as progressive sampling), as far as I can see. If you want to reconstruct the signal exactly, you will need to know what the original center frequency was, which you can obtain either by progressive sampling or by just using a frequency counter (like the Bitscope does). Sean At 11:05 PM 2/16/01 -0700, you wrote: {Quote hidden}>Bear in mind that the series resistance of the sampling switch (6k at >5V) and the value of the hold capacitor (51.2pf for a '76) will band >limit your signal to under 500khz. >However, with an external fast sample and hold, one could digitize video >at a 50khz rate (3 samples per video line). By sampling progressively >later in the line, one could acquire a 640x480 60Hz video frame in 3.8 >seconds. (A fully analog version of this technique was used back in the >70's to convert full speed video to the 15Hz line/7.5S frame of slow >scan television used in the amateur radio bands (TV in a 3khz BW)). > >"Sean H. Breheny" wrote: > > > > Hi Don, > > > > For certain methods of accomplishing reconstruction, you are right, but I > > don't think that is true in general. You can take any signal bandlimited to > > delta_f, sample it at 2*delta_f, and reconstruct it from the samples. It > > doesn't matter what the center frequency is, or whether it is periodic, as > > long as it is bandlimited. So, for example, you could have a signal which > > filled the entire band from 0.9 to 1.1 MHz, sample it at only 0.4 > > Megasamples per sec, and reconstruct it from that, as long as you had the > > additional piece of info that its center freq was 1 MHz. > > > > Sean > > > > At 07:41 PM 2/16/01 -0800, you wrote: > > >Hi, > > > > > >The signal must necessarily be periodic for this to work.... > > >Lets not forget that restriction! > > > > > >Cheers, > > >Don > > > > > >-- > > >http://www.piclist.com hint: PICList Posts must start with ONE topic: > > >[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads > > > > -- > > http://www.piclist.com hint: The list server can filter out subtopics > > (like ads or off topics) for you. See http://www.piclist.com/#topics > >-- >http://www.piclist.com hint: The list server can filter out subtopics >(like ads or off topics) for you. See http://www.piclist.com/#topics -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Hi Sean, The method you describe is often used in direct digital detection receivers. The IF (or tuned RF) output is in a known band and an appropriate choice of sample rate achieves a downconversion to 'base band' without additional RF stages. I believe Plessey Semiconductor and/or Thomson CSF had digital TV chip sets and app notes using this method. It is actually quite instructive to take two function generators (one as signal, the other as sampler) and feed the first through an analog switch (4066) which is driven by the 2nd as a gate for a sample and hold, to a scope. As one sweeps the frequency of the input signal past Fs/2 one easily sees the changing wavelength of the signal, and the folding effect (and zero beat as the frequency approaches Fs). Try it. It will make aliasing a very 'intuitive' quantity. [10Khz is a nice sample rate which gives you stable bright traces] Robert "Sean H. Breheny" wrote: {Quote hidden}> > Hi Robert, > > Yes, you are right, I should have added the proviso that your sampling > system needs to have a high enough bandwidth (and your sampling aperture > needs to be very small and low jitter). > > I'm not talking about the progressive sampling method, though. I'm talking > about just taking samples normally. As another poster pointed out, aliasing > really only occurs when two "copies" of the signal overlap (in the > frequency domain). This will only happen if you exceed the bandwidth > allowed by the Nyquist Sampling Theorem, NOT just because the center > frequency is very high. Sure, the output samples will be consistent with a > much lower center frequency, but who cares if the modulation riding on top > of them is the narrowband signal you are looking for? You really don't need > to do any special processing (such as progressive sampling), as far as I > can see. If you want to reconstruct the signal exactly, you will need to > know what the original center frequency was, which you can obtain either by > progressive sampling or by just using a frequency counter (like the > Bitscope does). > > Sean > > At 11:05 PM 2/16/01 -0700, you wrote: > >Bear in mind that the series resistance of the sampling switch (6k at > >5V) and the value of the hold capacitor (51.2pf for a '76) will band > >limit your signal to under 500khz. > >However, with an external fast sample and hold, one could digitize video > >at a 50khz rate (3 samples per video line). By sampling progressively > >later in the line, one could acquire a 640x480 60Hz video frame in 3.8 > >seconds. (A fully analog version of this technique was used back in the > >70's to convert full speed video to the 15Hz line/7.5S frame of slow > >scan television used in the amateur radio bands (TV in a 3khz BW)). > > > >"Sean H. Breheny" wrote: > > > > > > Hi Don, > > > > > > For certain methods of accomplishing reconstruction, you are right, but I > > > don't think that is true in general. You can take any signal bandlimited to > > > delta_f, sample it at 2*delta_f, and reconstruct it from the samples. It > > > doesn't matter what the center frequency is, or whether it is periodic, as > > > long as it is bandlimited. So, for example, you could have a signal which > > > filled the entire band from 0.9 to 1.1 MHz, sample it at only 0.4 > > > Megasamples per sec, and reconstruct it from that, as long as you had the > > > additional piece of info that its center freq was 1 MHz. > > > > > > Sean > > > > > > At 07:41 PM 2/16/01 -0800, you wrote: > > > >Hi, > > > > > > > >The signal must necessarily be periodic for this to work.... > > > >Lets not forget that restriction! > > > > > > > >Cheers, > > > >Don > > > > > > > >-- > > > >http://www.piclist.com hint: PICList Posts must start with ONE topic: > > > >[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads > > > > > > -- > > > http://www.piclist.com hint: The list server can filter out subtopics > > > (like ads or off topics) for you. See http://www.piclist.com/#topics > > > >-- > >http://www.piclist.com hint: The list server can filter out subtopics > >(like ads or off topics) for you. See http://www.piclist.com/#topics > > -- > http://www.piclist.com hint: The list server can filter out subtopics > (like ads or off topics) for you. See http://www.piclist.com/#topics -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Robert Rolf wrote: >Bear in mind that the series resistance of the sampling switch (6k at >5V) and the value of the hold capacitor (51.2pf for a '76) will band >limit your signal to under 500khz. Hi Robert, Ah good, I see someone else is familiar with the PIC A/D internals. Based upon the published info, I had expected to see a fairly good BW in the A/D, something somewhat commensurate with F = 1/(2*pi*7K*51pF) = 440 khz However, I was surprised to measure only about 120 khz BW [external influences being taken into account] for the 'C76. And the 'F876, which has a 120 pF sampling cap, registered only about 60 khz BW. In addition, the A/D gain internally falls off at about -60 dB/decade, which makes a really nice anti-aliasing filter when you are doing ETS at 1 mhz, but I was hoping for a little more BW. Interesting that the measured values are so far off the 1st order approximation. Got any clue as to why? best regards, - dan michaels http://www.oricomtech.com ================== -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

>In addition, the A/D gain internally falls off at about -60 dB/decade, >which makes a really nice anti-aliasing filter when you are doing ETS >at 1 mhz, but I was hoping for a little more BW. Maybe I'm barking up the wrong tree here, but my electret microphone input on a project (single-stage amplification of 50 nominal) into the A/D results in absolutely no amplification at all, really. Not being very familiar with all this audio stuff, does the above statement have any relation to this problem? I have the mic biased to 2.5V, but I can't get any AC swing. What I ended up doing, only haphazardly, was to crank the gain up to something outrageous, in the 1300 range. Oddly, the noise is only about 2 bits on an 8-bit A/D. The overall range still only goes from ~20d to 225d on the A/D result, and only saturates if I speak DIRECTLY into the mic. If I'm about one foot away (actually, several feet is my eventual goal), it requires me to really shout. Any idea why the output needs such high gain? Is this even possible? --Andrew _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Andrew Vasallo wrote: >>In addition, the A/D gain internally falls off at about -60 dB/decade, >>which makes a really nice anti-aliasing filter when you are doing ETS >>at 1 mhz, but I was hoping for a little more BW. > >Maybe I'm barking up the wrong tree here, but my electret microphone input >on a project (single-stage amplification of 50 nominal) into the A/D results >in absolutely no amplification at all, really. Not being very familiar with >all this audio stuff, does the above statement have any relation to this >problem? > Nothing whatsoever. The A/D doesn't fall off until about 120 khz, audio is only 20 khz tops. ========= >I have the mic biased to 2.5V, but I can't get any AC swing. What I ended >up doing, only haphazardly, was to crank the gain up to something >outrageous, in the 1300 range. Oddly, the noise is only about 2 bits on an >8-bit A/D. The overall range still only goes from ~20d to 225d on the A/D >result, and only saturates if I speak DIRECTLY into the mic. If I'm about >one foot away (actually, several feet is my eventual goal), it requires me >to really shout. Any idea why the output needs such high gain? Is this >even possible? > Nothing unusal, you just need a lot of amplification when using an electret mike. The key, I believe, is to DC bias the mike, and use AC-coupling to all successive stages: mike --> amp --> amp --> A/D. Probably best to use at least 2 amplification stages, as a single stage at gain = 1300 is not gonna have much BW. LM324 has GBW = 1 Mhz. If setup with gain = 1300, the BW of the stage is only 1mhz/1300 = 770 hz. This is not enough BW for a typical audio signal. - danM -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.