I see. It is not so much for when the prop is driving the pin as when something external, perhaps with a high output Impedance, is driving the pin.
Yes that’s it. The 64.9K I/O pulldown will form a resistor-divider whenever there is some sort of series resistance between P26 or P27 and the user external circuit. Hence that I/O pulldown may impact voltage logic levels undesirably when those Prop pins are used as inputs (and to some extent when an output driven high).
The value was chosen to be high enough that it shouldn’t really impact most digital uses (on/off type) inputs/outputs, but care should be taken if those two pins are used for analog sampling (such as RCTIME stuff), or high speed signals.
The other scenario I mentioned is when those pins are just unconnected (floating), which would be the case when the Propeller is defaulted to input. If those pins had too high value pull resistor, then you could trigger the pin fairly easily by touch. That sort of thing wouldn't usually be noticed for unused pins, except the FLiP has LEDs on those two pins
So the resistor wanted to be large enough so that it’s possible affect on voltage levels (user external circuits) is minimised, whilst being low enough so as not be become a high-impedance input by itself- susceptible to “picking up” noise voltage!
The LEDs are great for instant gratification, especially for a product like this that is directed toward education.
Here is one thing I've noticed. I went to add a GPS, but it wouldn't lock. So I've started to simplify. If I power the GPS from a separate 5V supply and leave the FLIP unpowered, the GPS locks quickly. But soon after I plug the USB cable into the FLIP, the GPS loses lock. The photo shows the setup on a proto board. I disconnected all the power and signal connections between the FLIP and the GPS. Still loses lock. I probed the RF field around the FLIP with a DF loop. In the photo the DF probe is directly above the 2.2µH inductor that is located on the opposite side of the PCB and not surprisingly that is the strongest field point. The 'scope shot shows the signal there, a pulse repeating at intervals of about 0.72µS and with ringing up to about 80MHz. The signal drops off strongly with distance so I have no measure of what it might be near the GPS. I'll try other arrangements to resolve this. Those white plugboards are notoriously bad for RF work, also, switcher power supplies can be trouble around sensitive RF receivers. Has anyone else tried this?
.... I disconnected all the power and signal connections between the FLIP and the GPS. Still loses lock..
You could try using FLiP like a wand, ie only USB powered and no connection at all to breadboard, then wave it about to see what distance separation in needed for GPS to operate.
Check! I've had exactly this problem with the Waveshare GPS we sell. Switched it out for another GPS of the same make, and had the same problem. Didn't take time to really try to figure out the cause of the problem. Initially it received signal and I connected up an XBee and put the GPS unit outside while I received the coordinates on my desk. The GPS was mounted on a breadboard with the FLiP Module. It worked flawlessly for quite some time until I changed something. Not sure what I changed - maybe the power supply or the actual breadboard - and it never worked again.
... The GPS was mounted on a breadboard with the FLiP Module. It worked flawlessly for quite some time until I changed something. Not sure what I changed - maybe the power supply or the actual breadboard - and it never worked again.
GPS is already somewhat marginal indoors, so adding more sources of RF energy close by, can only make things worse.
The energy in the GPS band will be very low, but it does not need to be much to block the GPS lock.
The SMPS on FLiP will have a broad spectrum....
jmg, tried the experiment as you describe. GPS is fine at 6" distance in air, intermittent at 5", stops at 4". The GPS is still on the white breadboard, but the FLIP is simply plugged into USB and held like a wand.
Then I put the FLIP in a plastic bag wrapped with a sheet of aluminum foil. I had a wire wrapped on the FLIP ground pin and connected to the aluminum foil with conductive-adhesive tape. Now the FLIP can be right next to the GPS and it still locks fine. I also tried it without the ground connection but that didn't work so well. This is a kluged setup, but I would conclude that the interference is electromagetic, and that it can be alleviated with a faraday shield.
So the GPS is tracking, and then you watch it fail to track based on the FLIP moving close? Does the received signal strength degrade based on the proximity of the FLIP? Or is this failure to acquire? I wonder if u-blox has any sort of software spectrum analyzer capability built-in. Some chips include this because it saves the Apps engineers a lot of time.
Is it possible to load assistance data in these modules? If so, it's something to look into regardless of the FLIP. e.g. see AssistNow in the following:
It might be interesting to change the frequency that the FLIP is using and see if this helps. Maybe some harmonics are falling in band and hurting GPS. Not sure if this module has a SAW either.
..
It might be interesting to change the frequency that the FLIP is using and see if this helps. Maybe some harmonics are falling in band and hurting GPS.
FLiP also includes a SMPS chip, and the SiTime 5.00MHz oscillator, neither are easy to change, and SMPS spectrum will vary with load.
Ken, your milage may vary . It may be orientation dependent. The fields are polarized. It is also possible that the wires out from the FLIP will act like transmitting antennas, so filtering beads or feedthroughs may be called for on wires that go from the FLIP to the GPS at whatever distance, and positioning of those wires may matter. Could explain why it worked for a while and then didn't when something was changed. Don't know. Have to try to build it with the GPS and FLIP actually connected, maybe tomorrow.
@KeithE, when I had it connected to the FLIP, it was returning RMC and GGA sentences, but it didn't lock and showed only empty signal strength and acquisition fields, at least that is what I recall. I haven't been very systematic about it. For the quick checks I was looking only at the lock LED on the GPS board and didn't have it streaming to a computer. The "assistNow" is also available on the MAX-7 u-blox, but I'd skimmed over that feature not understanding it. I took a peek at a u-blox document for assistNow, interesting, could be useful down the line once the basic EMI issue is cleared up.
Can you plug-in an external GPS antenna with LNA? Then you can easily move that around and see if it makes a difference.
I've never used these modules, but assistNow is worth a look if you want to be using GPS indoors. It could be a pain though, because this assistance days quickly expires. (Guessing here, but it's typically days instead of the normal ephemeris data which lasts for hours.)
This module has a build-in ceramic antenna, no option for external/LNA. It may come down to moving the whole module and communicating with it at a distance as Ken suggested. I edited my post after I read the u-blox document about their assistNow options. I hadn't paid any attention to that until you mentioned it. It looks to be useful when there is either network connectivity or a relatively short term deployment. The basic EMI issue does have to be dealt with first.
First, I took a quick look at the WaveShare GPS photo. There are a couple components by the RF jack already that would help protect the GPS antenna input from picking up noise (namely a 10ohm resistor and a ceramic that I'd suspect is an inductor connecting RF to VCC).
That being the case, it would seem places to check are at VIN/GND and IO connections.
The IO cable(s) is the one with switching signals, so can you replace your simple jumper lead with a screened cable? Hook the screen to GND at only one end- probably the FLiP side.
If that doesn't solve it, you could also try:
1. Add a 4.7uF or so cap close to the gps module vin/gnd pins
2. Add a series resistor close to the gps module at each IO. (tx/rx I guess?). Try a resistor value 100ohm to 1k sort of range. Instead of the resistor, you could also use an inductor (20nH to 50nH) or a bead (500ohm to 1k), if you have any of those instead.
... I went to add a GPS, but it wouldn't lock. So I've started to simplify.
Hi Tracy,
In your picture you have those two cables with croc clips passing over the GPS module. The croc clips are also resting above and next to the antenna.
Do you get an improvement if you route those cables away from the GPS module, whilst also making sure there's no large conductors (any sort of metal objects) close to the antenna?
I'm thinking that in the pictured setup, if you "wave" a switching node (DCDC inductor) near those 2 long croc-clipped-cables, then they are quite likely to pickup noise and couple it to the GPS supply rail. In this case, maybe a low value series resistance or bead at the GPS vin, plus a cap, might be sufficient to clean things up. That GPS module can't need much current? 50mA for the antenna (if it's active?) and 30mA for the module perhaps?
Just sharing thoughts here... I'd like to get some time later in the week to do some testing too. I like the look of your hoop-probe (look's like copper pipe! nice!). Do you get stronger signal if you rotate that 90degrees to the inductor, rather than holding it flat on the top of the pcb?
Ok - maybe if someone has the same issue with the Waveshare then they can try the external antenna. It's definitely a good idea to deal with the root cause though.
5 MHz is a not a great choice for a signal to be running around a GPS receiver, so if there's anyone that can sniff for that and its harmonics, or change it would be good. You don't want anything near GPS L1 @ 1575.42 MHz.
jmg, tried the experiment as you describe. GPS is fine at 6" distance in air, intermittent at 5", stops at 4". The GPS is still on the white breadboard, but the FLIP is simply plugged into USB and held like a wand.
Then I put the FLIP in a plastic bag wrapped with a sheet of aluminum foil. I had a wire wrapped on the FLIP ground pin and connected to the aluminum foil with conductive-adhesive tape. Now the FLIP can be right next to the GPS and it still locks fine. I also tried it without the ground connection but that didn't work so well. This is a kluged setup, but I would conclude that the interference is electromagetic, and that it can be alleviated with a faraday shield.
What happens if you mount the Flip on a breadboard with the faraday shield between it and the breadboard? (grounded of course) Is that enough or does it need more?
Jim
I did another quick setup at my desk this morning. No measurement RF tools - just physical separation and the Waveshare GPS immediately locks and provides data (inside my office, where it certainly wouldn't work before especially when near the FLiP). You can all take this for what it appears to be. As Tracy reported "your mileage may vary" by moving things around. But it works here!
For what it is worth, here is another data point.
I have a uBlox GPS Module which has always worked just fine indoors.
I have a FliP running all eight cogs active. The GPS loses lock (the blue LED stays on) if I get the FliP within about 4-6 inches.
No electrical connection between the two.
For what it is worth, here is another data point.
I have a uBlox GPS Module which has always worked just fine indoors.
I have a FliP running all eight cogs active. The GPS loses lock (the blue LED stays on) if I get the FliP within about 4-6 inches.
No electrical connection between the two.
Thanks for testing that Tom. I'll have my FliP tomorrow to test with two GPS modules.
Ken, that's reassuring. Is it a stock cable? There's a lesson there.
I realized I didn't have my 'scope set at 50Ω input impedance in a test I did yesterday. Attached is another shot. The 80MHz ringing is gone which had been due to the 1M termination at the 'scope. The power supply is unloaded except for about 28mA drawn by the FLIP itself running one Spin cog. There is the main 0.72µs period with a sharp rise time and a ringdown at about 8 MHz. The feature highlighted and expanded in red came up intermittently, more often than not, a ringing with a period of about 4ns.
I like the look of your hoop-probe (look's like copper pipe! nice!). Do you get stronger signal if you rotate that 90degrees to the inductor, rather than holding it flat on the top of the pcb?
Google "shielded loop probe". Also known as DF, for direction finder. It is made out of a length of solid-shield 50 ohm coax. Yes to the question. I can find a deep null if the loop is held perpendicular to the pcb.
If it helps the discussion, I've had a Prop Mini in continuous use within inches of a WaveShare GPS for a month without trouble.
A few days ago as I brought the FLiP nearby to measure its 40 MHz, GPS reception went from erratic to zero. My guess was that the RFI of the FLiP is off-the-charts compared to the Mini.
I avoided the problem by moving the FLiP a foot away and surrounding it in aluminum foil.
If it helps the discussion, I've had a Prop Mini in continuous use within inches of a WaveShare GPS for a month without trouble.
A few days ago as I brought the FLiP nearby to measure its 40 MHz, GPS reception went from erratic to zero. My guess was that the RFI of the FLiP is off-the-charts compared to the Mini.
I avoided the problem by moving the FLiP a foot away and surrounding it in aluminum foil.
If it helps the discussion, I've had a Prop Mini in continuous use within inches of a WaveShare GPS for a month without trouble.
A few days ago as I brought the FLiP nearby to measure its 40 MHz, GPS reception went from erratic to zero. My guess was that the RFI of the FLiP is off-the-charts compared to the Mini.
I avoided the problem by moving the FLiP a foot away and surrounding it in aluminum foil.
Sounds like it is less the Prop, and more the Switching regulator then.
I wonder if the Switching regulator can be improved ?
Google finds this : "20pF in series with an inductance of 0.6nH will be resonant at 1.5GHz"
ie some quite small caps may help.
Addit: also varying the SMPS diode would be easy to try, I see others with lower Cj and lower Trr like SS24FL
At least the SOD-123FL package is popular, DK shows 109 stocked hits.
Anyone with a Flip also have a spec-an or an SDR? (even a cheapo rtl-sdr might be able to give a better idea of how bad the situation is - nice at-a-glance tool for things like this)
Several of you were interested in how the well on-board FLiP oscillator worked over time (drift), or over temperature (tempco). Here's a plot showing the past couple of days. I used Tracy Allen's 40 MHz generator code and measured against some high-end hp gear.
Since the temperature coefficient is negative I inverted the scale so that the two plots match up and you can see how they correlate.
In this plot the oscillator was within 1 or 2 ppm in accuracy (40 Hz out of 40 MHz) and the tempco is well under 1 ppm per C. Nice. I'm impressed with the SiTime oscillator that Parallax chose (SiT8918).
Several of you were interested in how the well on-board FLiP oscillator worked over time (drift), or over temperature (tempco). Here's a plot showing the past couple of days. I used Tracy Allen's 40 MHz generator code and measured against some high-end hp gear.
Since the temperature coefficient is negative I inverted the scale so that the two plots match up and you can see how they correlate.
In this plot the oscillator was within 1 or 2 ppm in accuracy (40 Hz out of 40 MHz) and the tempco is well under 1 ppm per C. Nice. I'm impressed with the SiTime oscillator that Parallax chose (SiT8918).
Thanks for the update!
You have some pretty neat stuff on your website. A man with a passion about time. http://leapsecond.com/
It is made out of a length of solid-shield 50 ohm coax.
Gosh, my eyes may need testing. That looked like a piece of ~3mm copper tubing, such that might be found by a small fridge compressor or ice maker. Even the welding looked authentic to that use. That's why it caught my attention; shame it's a coax wand after-all, but the thought still lifted my day!
Thank you for that photo, and all your other feedbacks.
@LeapSecond,
Interesting result (and web site). I see you came out with a tempco of about 2.5Hz per °C. Nice. It looks like there is a slightly larger response to the dT/dt component, the rapid temperature shifts. For the cleanest signal at 40MHz, I should have used frqb := negx instead of frqb := posx. A small quibble. With posx the pulse train will drop a pulse once every 53687091200 counts (53.687,,. seconds). With negx, it never drops a pulse.
@Ken, You aren't kidding -- that is a bundle!
@Michael, You can find this 1.25mm (0.047") semi-rigid coax from RF suppliers like this one. A refrigeration loop, hmmm, could be.
Comments
Yes that’s it. The 64.9K I/O pulldown will form a resistor-divider whenever there is some sort of series resistance between P26 or P27 and the user external circuit. Hence that I/O pulldown may impact voltage logic levels undesirably when those Prop pins are used as inputs (and to some extent when an output driven high).
The value was chosen to be high enough that it shouldn’t really impact most digital uses (on/off type) inputs/outputs, but care should be taken if those two pins are used for analog sampling (such as RCTIME stuff), or high speed signals.
The other scenario I mentioned is when those pins are just unconnected (floating), which would be the case when the Propeller is defaulted to input. If those pins had too high value pull resistor, then you could trigger the pin fairly easily by touch. That sort of thing wouldn't usually be noticed for unused pins, except the FLiP has LEDs on those two pins
So the resistor wanted to be large enough so that it’s possible affect on voltage levels (user external circuits) is minimised, whilst being low enough so as not be become a high-impedance input by itself- susceptible to “picking up” noise voltage!
Here is one thing I've noticed. I went to add a GPS, but it wouldn't lock. So I've started to simplify. If I power the GPS from a separate 5V supply and leave the FLIP unpowered, the GPS locks quickly. But soon after I plug the USB cable into the FLIP, the GPS loses lock. The photo shows the setup on a proto board. I disconnected all the power and signal connections between the FLIP and the GPS. Still loses lock. I probed the RF field around the FLIP with a DF loop. In the photo the DF probe is directly above the 2.2µH inductor that is located on the opposite side of the PCB and not surprisingly that is the strongest field point. The 'scope shot shows the signal there, a pulse repeating at intervals of about 0.72µS and with ringing up to about 80MHz. The signal drops off strongly with distance so I have no measure of what it might be near the GPS. I'll try other arrangements to resolve this. Those white plugboards are notoriously bad for RF work, also, switcher power supplies can be trouble around sensitive RF receivers. Has anyone else tried this?
The GPS is Reyax RYN25AI.
Check! I've had exactly this problem with the Waveshare GPS we sell. Switched it out for another GPS of the same make, and had the same problem. Didn't take time to really try to figure out the cause of the problem. Initially it received signal and I connected up an XBee and put the GPS unit outside while I received the coordinates on my desk. The GPS was mounted on a breadboard with the FLiP Module. It worked flawlessly for quite some time until I changed something. Not sure what I changed - maybe the power supply or the actual breadboard - and it never worked again.
Ken Gracey
Ken and I talk about the problem last week. Just awaiting hardware. (Hate when the holiday throws and extra day into shipping).
The energy in the GPS band will be very low, but it does not need to be much to block the GPS lock.
The SMPS on FLiP will have a broad spectrum....
Then I put the FLIP in a plastic bag wrapped with a sheet of aluminum foil. I had a wire wrapped on the FLIP ground pin and connected to the aluminum foil with conductive-adhesive tape. Now the FLIP can be right next to the GPS and it still locks fine. I also tried it without the ground connection but that didn't work so well. This is a kluged setup, but I would conclude that the interference is electromagetic, and that it can be alleviated with a faraday shield.
Is it possible to load assistance data in these modules? If so, it's something to look into regardless of the FLIP. e.g. see AssistNow in the following:
http://www.waveshare.com/w/upload/5/5e/NEO-7-DataSheet.pdf
It might be interesting to change the frequency that the FLIP is using and see if this helps. Maybe some harmonics are falling in band and hurting GPS. Not sure if this module has a SAW either.
Or some distance separation, right?
Ken Gracey
@KeithE, when I had it connected to the FLIP, it was returning RMC and GGA sentences, but it didn't lock and showed only empty signal strength and acquisition fields, at least that is what I recall. I haven't been very systematic about it. For the quick checks I was looking only at the lock LED on the GPS board and didn't have it streaming to a computer. The "assistNow" is also available on the MAX-7 u-blox, but I'd skimmed over that feature not understanding it. I took a peek at a u-blox document for assistNow, interesting, could be useful down the line once the basic EMI issue is cleared up.
I've never used these modules, but assistNow is worth a look if you want to be using GPS indoors. It could be a pain though, because this assistance days quickly expires. (Guessing here, but it's typically days instead of the normal ephemeris data which lasts for hours.)
First, I took a quick look at the WaveShare GPS photo. There are a couple components by the RF jack already that would help protect the GPS antenna input from picking up noise (namely a 10ohm resistor and a ceramic that I'd suspect is an inductor connecting RF to VCC).
That being the case, it would seem places to check are at VIN/GND and IO connections.
The IO cable(s) is the one with switching signals, so can you replace your simple jumper lead with a screened cable? Hook the screen to GND at only one end- probably the FLiP side.
If that doesn't solve it, you could also try:
1. Add a 4.7uF or so cap close to the gps module vin/gnd pins
2. Add a series resistor close to the gps module at each IO. (tx/rx I guess?). Try a resistor value 100ohm to 1k sort of range. Instead of the resistor, you could also use an inductor (20nH to 50nH) or a bead (500ohm to 1k), if you have any of those instead.
Hi Tracy,
In your picture you have those two cables with croc clips passing over the GPS module. The croc clips are also resting above and next to the antenna.
Do you get an improvement if you route those cables away from the GPS module, whilst also making sure there's no large conductors (any sort of metal objects) close to the antenna?
I'm thinking that in the pictured setup, if you "wave" a switching node (DCDC inductor) near those 2 long croc-clipped-cables, then they are quite likely to pickup noise and couple it to the GPS supply rail. In this case, maybe a low value series resistance or bead at the GPS vin, plus a cap, might be sufficient to clean things up. That GPS module can't need much current? 50mA for the antenna (if it's active?) and 30mA for the module perhaps?
Just sharing thoughts here... I'd like to get some time later in the week to do some testing too. I like the look of your hoop-probe (look's like copper pipe! nice!). Do you get stronger signal if you rotate that 90degrees to the inductor, rather than holding it flat on the top of the pcb?
Michael.
5 MHz is a not a great choice for a signal to be running around a GPS receiver, so if there's anyone that can sniff for that and its harmonics, or change it would be good. You don't want anything near GPS L1 @ 1575.42 MHz.
Jim
Ken Gracey
I have a uBlox GPS Module which has always worked just fine indoors.
I have a FliP running all eight cogs active. The GPS loses lock (the blue LED stays on) if I get the FliP within about 4-6 inches.
No electrical connection between the two.
Thanks for testing that Tom. I'll have my FliP tomorrow to test with two GPS modules.
I realized I didn't have my 'scope set at 50Ω input impedance in a test I did yesterday. Attached is another shot. The 80MHz ringing is gone which had been due to the 1M termination at the 'scope. The power supply is unloaded except for about 28mA drawn by the FLIP itself running one Spin cog. There is the main 0.72µs period with a sharp rise time and a ringdown at about 8 MHz. The feature highlighted and expanded in red came up intermittently, more often than not, a ringing with a period of about 4ns.
Google "shielded loop probe". Also known as DF, for direction finder. It is made out of a length of solid-shield 50 ohm coax. Yes to the question. I can find a deep null if the loop is held perpendicular to the pcb.
A few days ago as I brought the FLiP nearby to measure its 40 MHz, GPS reception went from erratic to zero. My guess was that the RFI of the FLiP is off-the-charts compared to the Mini.
I avoided the problem by moving the FLiP a foot away and surrounding it in aluminum foil.
Thanks for the feedback!
Sounds like it is less the Prop, and more the Switching regulator then.
I wonder if the Switching regulator can be improved ?
Google finds this : "20pF in series with an inductance of 0.6nH will be resonant at 1.5GHz"
ie some quite small caps may help.
Addit: also varying the SMPS diode would be easy to try, I see others with lower Cj and lower Trr like SS24FL
At least the SOD-123FL package is popular, DK shows 109 stocked hits.
Since the temperature coefficient is negative I inverted the scale so that the two plots match up and you can see how they correlate.
In this plot the oscillator was within 1 or 2 ppm in accuracy (40 Hz out of 40 MHz) and the tempco is well under 1 ppm per C. Nice. I'm impressed with the SiTime oscillator that Parallax chose (SiT8918).
Thanks for the update!
You have some pretty neat stuff on your website. A man with a passion about time.
http://leapsecond.com/
I'm so glad you asked! Sales opportunity right here!
https://www.parallax.com/product/751-00010
How many packages might you need? I have 32 in stock and can get another container of them if needed
Ken Gracey
Gosh, my eyes may need testing. That looked like a piece of ~3mm copper tubing, such that might be found by a small fridge compressor or ice maker. Even the welding looked authentic to that use. That's why it caught my attention; shame it's a coax wand after-all, but the thought still lifted my day!
Thank you for that photo, and all your other feedbacks.
Interesting result (and web site). I see you came out with a tempco of about 2.5Hz per °C. Nice. It looks like there is a slightly larger response to the dT/dt component, the rapid temperature shifts. For the cleanest signal at 40MHz, I should have used frqb := negx instead of frqb := posx. A small quibble. With posx the pulse train will drop a pulse once every 53687091200 counts (53.687,,. seconds). With negx, it never drops a pulse.
@Ken, You aren't kidding -- that is a bundle!
@Michael, You can find this 1.25mm (0.047") semi-rigid coax from RF suppliers like this one. A refrigeration loop, hmmm, could be.