Shop OBEX P1 Docs P2 Docs Learn Events
FLIP here - Page 3 — Parallax Forums

FLIP here

135

Comments

  • The Prop Mini has analog voltage regulators. I've used them with 12V input and they work, but then 3/4 of the power goes into space heating. Switchers are a lot better for battery and low-power work but the noise can be a real problem. It shouldn't be making harmonics high enough to be noticed by a GPS receiver.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2017-06-01 06:30
    I compared with a couple of other SMPS chips I have. One is a Pololu 5V>12V booster using the SemTech SC4503 chip, the other is one of my own boards that uses a TPS61220 to boost LiPo to 5.5V. Neither affected the GPS lock even at close quarters. There does seem to be something specific going on with the FLIP vs. GPS, but I don't have a sensitive spectrum analyzer or SDR to approach from that angle. The buck circuit layout on the FLIP does look impeccable--The main power loop is short and the traces are as wide as they can be and connect to a large ground plane. The one thing I do wonder about is the 22µF 0805 capacitors, can those be X7R and low enough ESR?

  • jmgjmg Posts: 15,173
    .. The one thing I do wonder about is the 22µF 0805 capacitors, can those be X7R and low enough ESR?
    You could try parallel of something smaller, at both IN and OUT sides ? ( From above Google suggested 20pF region cap is series resonant at 1.5GHz)

  • jac_goudsmitjac_goudsmit Posts: 418
    edited 2017-06-01 14:30
    2017-05-31%2022.46.34%20small.jpg

    I'd like to add that I FLiPped my current project too: I'm working on a Propeller-based S/PDIF receiver/analyzer which will be part of a bigger project (if successful). For more info see here, but that page doesn't have the FLiP in it yet, at the time of this writing.

    I really like:
    • How the module is the same size as a Propeller DIP
    • How the LEDs are bright enough to see but dim enough not to burn your eyes (unlike what most 'Duino clone makers like to put on their boards)
    • How clear the pin markings are
    • The two on-board LEDs for debugging
    • The micro-USB socket (like most people, I have a lot more micro-B cables than mini-B cables)
    • How flexible the power supply options are!

    A few things that I think could be improved in the future:
    • I would have liked the pins on the bottom to be thinner; I'm a little worried that these will wear out my breadboard
    • I would have liked to have pins on top to attach logic analyzer / oscilloscope probes.
    • I would have liked the reset button to be slightly bigger
    • The first time I powered it up (using the USB cable to the computer), it immediately went into overload mode. I had to disconnect the USB cable and disconnect the 3.3V output of the FLiP from the rest of the circuit, then reconnect the USB connector to make the overload go away. Then when I reconnected the rest of the circuitry (which is currently a 74HC04 and a 74HC00 and not much else, definitely nothing that could would cause an overload and there were no short-circuits), it kept working and I couldn't reproduce the problem.

    Thanks!

    ===Jac
    2656 x 1494 - 1M
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2017-06-01 16:45
    @jmg, that was my thought too. I followed your link to
    RF Power Amplifiers for Mobile Communications, and I see the comment (pg.104) about 20pF resonant with its 0.6nH self-inductance at 1.5GHz. But it goes on about other circuit parasitics, so you are left finally with dark arts for the circuit as a whole. My thought was to add anything to break up whatever resonance it is that affects the GPS. I might pile on a 0.1µF X7R, or 100pF or 18pF cog/npo or some combination. Any magic wands?
  • jmgjmg Posts: 15,173
    I would have liked the pins on the bottom to be thinner; I'm a little worried that these will wear out my breadboard
    Where that is a concern, in the past I've used a simple 'sacrificial Dual Wipe DIP' socket : the standard square posts can be forced into that, and then the DIP pins are much smaller (and also more fragile...). You cannot use the socket again for a standard IC, but they are very cheap.
    I would have liked to have pins on top to attach logic analyzer / oscilloscope probes.
    I guess longer tails on the header would allow that ?
    In the meantime, you could solder another header on top, to achieve the same thing ?

  • PublisonPublison Posts: 12,366
    edited 2017-06-02 16:42
  • I like the micro-USB too. A corollary of Murphy's law: If you need a micro-USB cable, you're sure to pick up a mini instead.

    On the FLIP, the µUSB connector is solidly staked to the board at 4 points, not likely to pull off even with heavy use.

    About the power supply. I sometimes supply USB power from a lab supply, via the USB connector. If you want to do that, you have to remember to short circuit together the two signal lines in the USB cable. That is how the FT231X chip identifies an external power brick and distinguishes it from a computer connection. Without the short circuit, you will get the yellow fault light.



  • I tried adding extra capacitors on the SMPS buck power supply, to see if that would help with the GPS at close quarters. No, that didn't help. The photo shows the extra capacitors in parallel with the stock 22µF 0805. There is lots of space next to the output capacitor, and the stack to the side adds 0.1µF X7R and 18pF COG. The input capacitor is more crowded and has a messy stack of the same values.

    360 x 570 - 97K
  • That was a good test. Interesting it did not fix the problem.

    If you have your soldering iron handy, also try feeding 3.3 directly to the FLiP, bypassing both the external 5-9 input and USB power input, thus disabling the switching regulator completely. AFAIK none of the 40 DIP pins allow you to do this (the pin labeled 3.3 is current limited output not input) so you'd have to tap the PCB at some point, maybe on one of the bypass caps.

    The purpose of the experiment is to solidly confirm that its the switching power supply that's causing GPS interference. I mean, we're all pretty sure. But a test like this would narrow the search.
  • jmgjmg Posts: 15,173
    edited 2017-06-03 03:31
    LeapSecond wrote: »
    That was a good test. Interesting it did not fix the problem.

    If you have your soldering iron handy, also try feeding 3.3 directly to the FLiP, bypassing both the external 5-9 input and USB power input, thus disabling the switching regulator completely. AFAIK none of the 40 DIP pins allow you to do this (the pin labeled 3.3 is current limited output not input) so you'd have to tap the PCB at some point, maybe on one of the bypass caps.

    The purpose of the experiment is to solidly confirm that its the switching power supply that's causing GPS interference. I mean, we're all pretty sure. But a test like this would narrow the search.

    Good idea, it is not easy to disable the AOZ1281 as it has a 87% upper limit on duty cycle.
    Unclear if it likes Vo > Vin ? - Not clear if short Vin-Vout and power from 3v3 stops it switching ? Removal or cuts may be needed ?

    For low cost linear regulators the ancient LM1117 is ok, good thermal package in SOT223 but high IQ and poor VDO

    I found a 'better looking LM1117' in the STm LDL1117, DK : $0.10230/2,500, but strangely they have no data and only dead links at ST's website ?!?

    Ah, octopart finds farnell have stock, and data, so it does seem real device.

    Relative to 'old '1117' the LDL1117 has better PSRR, lower IQ, lower VDO, and tolerable CAP caveats, tho average noise specs.
    If you want very low noise, a RF purposed regulator is needed.

    Seems the LDL1117 could replace the AOZ1281 in most use cases, with a BOM cost saving ? (just runs warmer, only those wanting 9 Vin and > 1A @ 3v3 need a switching regulator)


  • WhitWhit Posts: 4,191
    edited 2017-06-03 13:59
    MikeDYur wrote: »
    Whit wrote: »

    Hi Whit, I was just wondering, I don't think it's a good idea to have the module sitting in anti-static foam while powering up. It is conductive, isn't it?

    Hey - late to the party - it was the foam the FLiP came on in the video test. It is now safely on a breadboard! Wasn't thinking... Luckily the FLip is tough!


  • WhitWhit Posts: 4,191
    These three examples by JonnyMac are really fun to look at and study...

    They are written with the board setting as "other" - but work perfectly on the FLiP. They are on the Demo BlocklyProp Site.
    First one is pretty much where I started and the second and third example get much more interesting!

    http://demo.blockly.parallax.com/blockly/projects.jsp#931
    http://demo.blockly.parallax.com/blockly/projects.jsp#930
    http://demo.blockly.parallax.com/blockly/projects.jsp#945
  • Thanks for the links Whit. I see you and many others have a lot of material there on the Blockly Beta site!

    Let's say I just got a FLIP and I want to run that code. I'm kind of a spin-head and have never run a Blockly program before. All I want to do is a quick look using the leds built onto the FLIP. How do I get started, the nutshell version?
  • I just got mine. Thanks, Parallax! I haven't done much with it yet but I did get the two onboard LEDs flashing. Very cool!
  • WhitWhit Posts: 4,191
    edited 2017-06-05 21:40
    @Tracy Allen

    learn.parallax.com has you covered!

    BP.JPG

    See - http://learn.parallax.com/flip and http://learn.parallax.com/tutorials/language/blocklyprop - START with GETTING STARTED WITH BLOCKLYPROP.

    1238 x 618 - 108K
    BP.JPG 107.7K
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2017-06-06 06:15
    Thanks Whit. A lot there!

    Here is my spin-head's entry point. I don't need to be convinced that the Prop is better than sliced wonder bread!
    Click on link to JonnyMac's program, single cog blinker
    Click on link for "View project code". Oh, gosh, have to log in, or create account. Turns out I already did that back last October.
    Okay, logged in, try "View project code" again. There it is! Look at available commands on left margin. Hmmm. Want to just run this one.
    Click on BlockyProp BETA image at the top left corner of the screen, doing that leads to the home screen.
    At the bottom of that screen find link to BlockyProp-client, click on it, download the version for my Mac OS.
    Do Mac stuff to unpack and install the client, goes smoothly, now have applications/BlocklyPropClient.
    Double click on that, click the "Connect" button. It tells me the socket is configured and the server started.
    Navigate back to the browser with the program code in view.
    Have FLIP plugged in to USB. See the USB port now listed in the drop-down menu at the upper right on the screen.
    Press the down-arrow icon above the code window to load the program to the FLIP ram (or eeprom).
    Viola! The program downloads and runs. The LED blinks! A terminal window opens and prints "BlocklyProp" once a second.

    Click on the code button at the upper right, see the underlying C code.
    On the drop-down menu at the far right, choose Community Projects.
    Open and load more of Whit's programs for the FLIP (or other), Like Blink Both LEDs. Okay! Very smooth.
    It looks like it is not necessary to do anything more with the BlocklyPropClient, as it seems to load automatically.

  • LeapSecond, Today I did try to feed 3.3V from a current limited lab supply directly into the FLIP. As you know, the 3.3V pin on the module is a protected output only, and external 3.3V there won't power the FLIP. So I soldered a wire directly to the internal 3.3V rail at the top of C102. Result: FLIP still messes with the GPS. I ran a program with RCslow, no change. It turned out that the switcher was still running with lots of high frequency bursts, even with 3.3V connected to its output. Could see them on the 'scope at the pin1 SMPS output, or with the RF sniffer loop. Not too surprisingly, the AOZ1281DI data sheet doesn't say anything about that condition, power fed into its output. The fact that the interference continues even with the Prop itself running RCslow does continue to implicate the switcher.

    My preference is usually a quiet linear supply at relatively low current or power levels, but I think a goal with the FLIP design was to have the full 1.5 amps available at 3.3V, down from from a 5V to 9V supply.


  • Got a FLiP, too - really starting to love the idea of these little compact modules! Anyway, had to take a quick look at what was being received on GPS L1 freq. Attached are two screenshots of CubicSDR. The one with the brighter blue waterfall is with the FLiP unplugged - just the background noise. The darker waterfall is with it plugged in - about 4 inches away from the GPS antenna. There is a pretty energetic spike at 1575.0. I had to turn down the receiver gains significantly before I took these two screenshots - with it turned all the way up, the level of that spike was about +9dB - too close to the receiver's +10dB max input.
    Setup is an Airspy mini powering the VPN 513 GPS Smart Antenna (Parallax #28510).

    I've tried a few other boards (Prop BOE, Prop Mini, Quickstart w/HIB) and none of them emitted anything noticeable like this.

    Cheers,
    Jesse
    2560 x 1366 - 2M
    2560 x 1366 - 1M
  • Jessie,
    That sounds suspiciously like a harmonic from a TV horizontal oscillator which would be 15.75 KHz.
    Jim
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2017-06-10 17:20
    Jesse,
    I'm not used to looking at this kind of diagram, but it seems pretty straightforward. When I look at the plot of amplitude vs frequency, I see the spike at 1575MHz in the one with the FLIP. Also there are what I surmise to be the peaks (much smaller by comparison) at around 1578MHz that show up prominently on the picture without the FLIP. That Airspy looks to be a useful tool! Can you see peaks at submultiples of 1575 using a wideband antenna?

    One app note on the Alpha & Omega site, PIC-005, deals with parasitic loops/oscillations. The AOS1281 buck chip operates at nominally 1.5MHz fixed frequency, but the parasitic they find comes out at ~90MHz. They add an RC snubber in parallel with the Schottky diode to suppress it. The problem could be hard to resolve on the FLIP, if that is what is going on. Easier to move the GPS further away on a cable per Ken's suggestion.




  • jmgjmg Posts: 15,173
    edited 2017-06-10 20:29
    avsa242 wrote: »
    There is a pretty energetic spike at 1575.0. I had to turn down the receiver gains significantly before I took these two screenshots - with it turned all the way up, the level of that spike was about +9dB - too close to the receiver's +10dB max input.
    Wow, that's hit the jackpot...
    What range of signals can this setup expect to see ?

    I wonder which exact part in Flip generates that ?
    The most energetic is the SMPS part, so it seems a good suspect, but as mentioned above getting that to actually stop switching is not easy.
    It has no 100% mode. Change of the STKY could also be worth trying.

    If you heat/cool/load-up the SMPS, does that peak move at all ? I see the next most energetic peak is ~ 1.5MHz away, is that also FLiP sourced ?

  • RS_Jim wrote: »
    Jessie,
    That sounds suspiciously like a harmonic from a TV horizontal oscillator which would be 15.75 KHz.
    Jim

    Jim,

    I suppose it could be, although this signal only appeared when I powered up the FLiP (only code was to alternately blink the P26 and P27 LEDs; no video generation).

    Jesse,
    I'm not used to looking at this kind of diagram, but it seems pretty straightforward. When I look at the plot of amplitude vs frequency, I see the spike at 1575MHz in the one with the FLIP. Also there are what I surmise to be the peaks (much smaller by comparison) at around 1578MHz that show up prominently on the picture without the FLIP. That Airspy looks to be a useful tool! Can you see peaks at submultiples of 1575 using a wideband antenna?

    One app note on the Alpha & Omega site, PIC-005, deals with parasitic loops/oscillations. The AOS1281 buck chip operates at nominally 1.5MHz fixed frequency, but the parasitic they find comes out at ~90MHz. They add an RC snubber in parallel with the Schottky diode to suppress it. The problem could be hard to resolve on the FLIP, if that is what is going on. Easier to move the GPS further away on a cable per Ken's suggestion.

    Tracy,
    I've found it very useful and a quite addictive hobby (started out with one of the rtl-sdr.com units, which would work fine for stuff like this, but much cheaper at $20-25USD). The peaks at 1578 are actually always there, regardless of frequency tuned, so I'm thinking they must be some artifact of the receiver itself. They disappear when the gain is turned up more. I will check below 1575 later today when I get back home. If memory serves, I tried the receiver with a 30-1300MHz antenna + LNA turned up all the way in another room and couldn't see anything from the FLiP but will verify.
    jmg wrote: »
    avsa242 wrote: »
    There is a pretty energetic spike at 1575.0. I had to turn down the receiver gains significantly before I took these two screenshots - with it turned all the way up, the level of that spike was about +9dB - too close to the receiver's +10dB max input.
    Wow, that's hit the jackpot...
    What range of signals can this setup expect to see ?

    I wonder which exact part in Flip generates that ?
    The most energetic is the SMPS part, so it seems a good suspect, but as mentioned above getting that to actually stop switching is not easy.
    It has no 100% mode. Change of the STKY could also be worth trying.

    If you heat/cool/load-up the SMPS, does that peak move at all ? I see the next most energetic peak is ~ 1.5MHz away, is that also FLiP sourced ?

    jmg,

    The Airspy, as well as the less expensive rtl-sdr.com units (which both use the same tuner chip) have a tunable range of 24-1766MHz, or above/below with the appropriate external down/up converter. I will recheck for this other peak and try to find a good heating and cooling source later today and get back to the group.

    Cheers,
    Jesse

  • avsa242avsa242 Posts: 452
    edited 2017-06-11 22:30
    @Tracy Allen,

    With all three gain settings cranked, external LNA on, through wideband (25-1300MHz) antenna...
    (frequencies on the left are the center or "tuned-to" frequency - any peaks reported at their received frequency)

    In another room (approx 15ft away, through one open doorway)
    1575.42: -85dB Noise floor, peak from FLiP, but only about 2dB above noise floor
    787.71: -84dB Noise floor - no peaks
    525.14: -79dB Noise floor - no peaks
    393.855: -80dB Noise floor - no peaks
    315.084: -66dB Noise floor - no peaks
    262.57: -60dB Noise floor - no peaks

    Moved to doorway (approx 5ft away in plain view)
    1575.42: -85dB Noise floor, peak from FLiP: -38dB
    787.71: -79dB Noise floor - no peaks
    525.14: -79dB Noise floor - no peaks
    393.855: -80dB Noise floor - no peaks
    315.084: -66dB Noise floor - no peaks
    262.57: -60dB Noise floor - no peaks

    Same spot, FLiP held in hand about another 2ft closer:
    1575.42: -85dB Noise floor, peak from FLiP: -33dB
    787.71: -79dB Noise floor - no peaks
    525.14: -79dB Noise floor, peak from FLiP: -24dB at 525.0MHz
    393.855: -80dB Noise floor - no peaks
    315.084: -66dB Noise floor - no peaks
    262.57: -60dB Noise floor - no peaks

    About 1 inch away:
    1575.42: Skipped - too close for comfort (EDIT: This line had an inadvertent copy/paste from the section above)
    787.71: -79dB Noise floor - no peaks
    525.14: -79dB Noise floor, peak from FLiP: -10dB at 525.0MHz
    393.855: -80dB Noise floor - peak from FLiP: -40dB at 395.0MHz (?)
    315.084: -66dB Noise floor - peak from FLiP: -10dB at 315.0MHz (?)
    262.57: -60dB Noise floor - no peaks


    Also, just a caveat: I wouldn't take absolute numbers too seriously - neither the software nor the receiver are "instrumentation-grade", or at least the software hasn't been calibrated against a trusted signal generator, or anything. I think the relation of the peaks to the NF is more important.

    @jmg,

    There is a slight effect from temperature:

    For this I zoomed into the peak as far as the software would allow. To read the approximate center of the freq, I placed the demodulator (green highlighted reticle) as close to the visible center as I could see.

    The center freq of the stabilized peak is more or less 1574.997822 (screenshot "Stable.png")
    Blowing on the PS (don't laugh!) raised the freq to about 1574.998132 (Screenshot "Cooled.png")
    Holding a cigarette lighter about 1/2inch above it lowered the freq to about 1574.997143 (Screenshot "Heated.png")
    The return to the stable freq was pretty swift. Zoomed in this far, for some reason the software slows down the waterfall in spite of whatever speed it is set to. I'd say about 10 lines per sec? So return to stable freq in certainly less than half a second I would say.
    EDIT: Don't know if this is sufficient, but I loaded the prop with something relatively taxing, I think - Linus' "Turbulence" demo. I'm not sure if loading the prop with something like this is putting any noticeable load on the PS. The peak was stable.

    Cheers,
    Jesse
    2560 x 1366 - 1M
    2560 x 1366 - 1M
    2560 x 1366 - 1M
  • Jesse - have you tried disabling the Propeller's crystal oscillator to see if it makes a difference? Maybe someone else already did that in this thread?
  • @KeithE,

    Ok, initially it looked like RCFAST shifted the peak down, but after trying RCSLOW and the crystal+pll again to confirm, it didn't change from there, so it might've just been colder on initial startup (A/C'd room), then once it warmed up it stabilized to a slightly lower freq. Good suggestion - hadn't thought of that.
  • 525.0 MHz is almost exactly one third the GPS L1 frequency of 1575.42 MHz. This is surely the source of the GPS interference.
  • Then where is the fundamental? 1/3 of 525 is 175 MHz, then 58.333, 19.444, 6.4814 ,...

    I tried it with RCSLOW too, but now recalled that FLIP uses the ST8918 5MHz oscillator, not a crystal, and it is running alll the time. It seems highly unlikely that it could be the source of the parasitic. It claims less than 2.5ps RMS period jitter. But that possibility can't be eliminated by putting the Prop into RCSLOW.
  • jmgjmg Posts: 15,173
    Then where is the fundamental? 1/3 of 525 is 175 MHz, then 58.333, 19.444, 6.4814 ,...

    I tried it with RCSLOW too, but now recalled that FLIP uses the ST8918 5MHz oscillator, not a crystal, and it is running alll the time. It seems highly unlikely that it could be the source of the parasitic. It claims less than 2.5ps RMS period jitter. But that possibility can't be eliminated by putting the Prop into RCSLOW.
    - hmm, actually, I had overlooked the ST8918...
    Those Synth oscillators do work by starting from some (much) higher oscillator, and dividing. (eg Si5351 is 600~900MHz)
    If 525 MHz is the lowest spectrum, maybe it starts from about there ?

    How well is it decoupled ? Probably not the easiest to remove, but maybe the Vcc can be cut ?
    Maybe it needs a small Cap, nominally series resonant at 525MHz, right at the package ?
  • FWIW, 262 was just a relatively arbitrary stopping point. I can delve down to about 1MHz if anyone thinks it's worth while (though not tonight :) )
Sign In or Register to comment.