GPS Issues - Do I need a GPS Ground Plane & EMI Ring? - Discussions

Replies

- Jeff Wilson > Gisela & Joe Noci May 4, 2015 at 2:50pm
  
  Good day to you Nampilot,
  
  I just happened to come across your post while doing some research, and being the author of this article, I thought I might give some background here.
  
  In open disclosure, I haven't researched the issue at hand in this post, so I can't say whether any of the design guidelines presented in my article would have hindered or helped. That will take more time than I have available to discern at present. But, let me state that I think our design objectives are different here. You appear to be speaking from an antenna design perspective (and perhaps a passive antenna plus GNSS module design perspective?), while I was primarily concerned with addressing GNSS chipset design into larger multi-processor and multi-frequency system solutions. My limited objective (as given to me by Electronic Design) was to demystify some of the design goals and solutions for the average digital designer who is faced with a GNSS chipset design, not to turn them into GNSS antenna designers (a specialty, which I'm sure you'll agree, cannot be achieved through a single magazine article).
  
  Unfortunately, magazine articles offer limited space in which to reach your objective, so I couldn't get very specific in any one area or else my article would have run over (and would have been edited down by a copy editor with less of a command of the subject). I also had to resort to generalities, that of course would not hold true for all customers, but were generally true for the vast majority of customers attempting a chipset design. Also, as printed, the article was not verbatim what I wrote (there were some edits that occurred after my submission that I could not control), but in general, the concepts were mostly preserved, so I couldn't reasonably object given the medium.
  
  Let me give some context here. I work for a semiconductor manufacture that supplies GNSS chipsets (multi-constellation GPS receivers if you prefer) into automotive, consumer, and industrial accounts. We don't make GNSS modules, only chipset solutions, so I am often faced with customers who are trying to "drop in replace" their current module solution with a dedicated chipset design. This is further complicated by the fact that our solution is a Telematics system on chip (SoC) solution that includes a number of interfaces (e.g. CAN, SPI, I2C, SD-Card, USB, and audio codec interfaces) not typically found on your single frequency GNSS solutiions. We're also the only manufacturer that allows custom software integrations on our solution. All of these high-speed interfaces add many more noise sources than is typically found in the average GNSS module, so some accommodation is needed to keep the edge-noise of these interfaces and subsystems from raising the local noise floor, or worse, producing a jamming signal in either the RF or IF signal paths. To give you an idea of the type of product this solution is used in, take a look at the eTrex 10, 20, and 30s. That solution uses our GNSS processor as the GNSS receiver and the main micro in the system. It does everything, including the map reading and map drawing! This is a highly complex mixed signal solution.
  
  Given that context, I'd like to address your comments (many of which I agree with, but given the limited space of the article in question, I had to make do with a generality that would make life easier for the average designer):
  
  Thank your for your descriptions of the patch antenna. That in itself was outside of the scope of my article (beyond the quick generality about linear versus patch antenna benefits), and it was well presented. Can I suggest you write an article of that for RF Design, GNSS Today, or GPS World? They're always looking for quality articles and I think that one would be a benefit to the readers...please consider it.
  
  Now on to the individual points (I have put the original poster's comment in italics):
  
  “Passive antenna” designs are more complex and can be susceptible to noise coupling into the antenna ground plane if not correctly isolated from other noise-producing components on the PCB.
  
  This is untrue - The antenna is simply that - the actual antenna ( the patch element, the ceramic substrate, the ground plane) is the same for a passive or active antenna. To make the antenna an active one, an amplifier is added - that is all. The antenna is no more or less susceptible to noise pickup - in fact , the active antenna is more susceptible, since the amplifier will now actively amplify any noise within its passband.
  
  For designing an antenna module by itself, I agree with you. But when you have a designer who is now needing to design the antenna, the RF path that is inside the GNSS module (i.e. the LNA, SAW Filters, impedance matching components), in addition to the digital sub-systems of his own design onto a single PCB, the issue becomes more complex. Take as an example where a designer has flooded a digital ground and power plane underneath a passive patch antenna. In this example, digital noise will capacitively couple into the patch antenna's ground plane and (potentially) be amplified by the first LNA (in effect raising the noise floor at the LNA). Since the SNR is fixed by the first LNA in the chain, this would have a disastrous effect on the c/n0 of the system. LNA gain cannot make up for a noisy ground plane before the first LNA stage. This is what I meant by the design for passive antennas being more complex than say an active antenna design were additional LNA gain "may" be able to compensate for a noisy RF path further down the chain.
  
  There’s one thing to note when deciding between chip and patch antennas, though. Patch antennas will provide the best signal performance for their size, as they receive signals on all sides of the patch. Linear GPS antennas (chip or dipole) will generally only receive signals along one of their axes. This results in linear antenna designs being at least half as sensitive (i.e., around –3 dB) compared to patch antennas, and most will probably be around 25% as sensitive as a patch (or about –6 dB).
  
  Again, not true. The radiation pattern for the patch and the chip antenna can be designed to be very similar. The patch does not 'receive on all sides' and the chip not - they 'receive' according to the designed radiation pattern, and a chip antenna with a pattern in only one specific direction would not work for GPS and would never sell...And this is NOT the reason for the -3dB of the chip antenna directivity - the reason is that the Patch is designed for circular polarisation, while it is very difficult to do so for the chip antenna. So the GPS signal suffers the 3dB loss because of cross polarisation. What is important with ALL antenna, regardless of type, is to understand what sort of radiation pattern you need, and what the antenna can provide. For GPS we would like a pattern that is a half sphere upwards, with maybe a -3 to -6dB loss on the horizon ( to eliminate pickup of man-made RF noise from the horizon). Most GPS patch antenna deliver this, if used with a suitable ground plane...
  
  All GPS receiver manufacturers ( of the chip level device) , such as the various Ublox devices mentioned - Neo, Max, etc, have taken care of ALL emc issues related to signal integrity and noise effects to the device itself. External devices, the LNA, etc, are the responsibility of the board level integrator.
  
  Ah generalities and editing. No real disagreement here; of course the polarization is the reason for the signal loss, but also the ends of the linear's antenna pattern present nulls that patches do not suffer. The main point here being that "all things being equal" with a sufficiently large ground-plane, a patch will out-perform a linear antenna. Your version is more correct, but the performance generalization still stands. Also note that the uBlox is not dealing with high-speed mixed signals...more on that later.
  
  It’s common in VHF and UHF RF shielding to connect all points of the shield can to the PCB’s ground plane. This can be a mistake at GPS frequencies, since the open-air wavelengths of a GPS signal is so much shorter than UHF. Depending on the size of the shield can, if there is current flow across the can, the shield can will be able to resonate near GPS frequencies resulting in interference or de-tuning of the GPS RF.
  
  Absolutely untrue. The first sentence is correct, but the concept of a shield can is used on every GPS Chip and many other RF devices - Video TX, RCS TX, etc. The can must connect to the ground plane correctly, ie, the spacing between connections to ground, if not solid, must be a very small portion of the wavelength we are trying to shield against. These small GPS module's cans are soldered every 3 or 4 mm, so the frequencies that may enter' these gaps are in the region of 20GHz.....And if the can could ever resonate ( not possible - it is connected to ground..) the resulting current flow in the can material would be on the outside, and penetrate very little into the material ( skin effect) and would NEVER enter the interior of the can. Imagine a copper sphere, hollow inside. Make a small hole, and stick some coax cable into this hole. Solder the screen of the coax to the sphere. Connect the other coax end to your transmitter and transmit. Now try to measure the rf exiting the sphere...If we could measure any useful signal out of the sphere we could place antenna inside metal fuselages...
  
  I think you may be thinking of antenna shield cans over LNA components and not dynamic current scenarios. In this article, I am talking exclusively about the shield can resonating under the current flow from other digital devices in the system. This is a phenomenon that may be more prevalent in system on chip solutions, but I can tell you it is very real (having measured the resonance with spectrum analyzers in the lab). I had one customer who lost a third of his initial production lot to shield can resonance (the reason it wasn't 100% was due to the variability of how the shield can was placed when soldered down and how complete the soldering was...more complete soldering only fixed about 50% of these failed cases). The issue here is you have several ground planes in use, a GNSS antenna ground-plane, a GNSS RF ground plane, and maybe multiple digital ground planes. Shorting them all together with a shield can allows for noise and current to flow across the planes and the can (which is what this customer did). The easiest fix was to isolate the shield can through an inductor to keep stray currents away from the can. This fixed his issue and he was able to move forward with production.
  
  Specifically to shield can resonance you say:
  
  And if the can could ever resonate ( not possible - it is connected to ground..)
  
  No grounds are perfect (even 30-mil vias impart a 10-ohm impedance mis-match in the signal path) and even if the grounds were perfect, this ignores the self-resonant frequencies of the components and structures. In other words, ground is not absolute. If it helps to understand the shield can resonance problem, maybe you can consider it a grounded loop antenna? It's the AC current flow through the loop that sets up the resonance.
  
  The shielding can is solidly connected to the chipset PCB ground plane, and this ground plane is brought out to many 'pads', on the side of the chipset PCB ( just check any Ublox GPS chipset datasheet for example..)- these pads are then soldered the main PCB ground plane, which forms part of the Patch antenna ground plane in an integrated GPS. If this grounding is maintained, then all is well. And a connection to any part of the ground palne is as good a ground as you will get.
  
  And this approach will get you into trouble in a mixed signal system. The uBlox chipset isolates their digital sub-system inside their module and inside their device package (in effect creating the ground isolations I discuss in the article). System-on-Chip solutions with high speed interfaces (SD-Card, external memory buses, et cetera) do not have that luxury since the return current for all those high-speed edges MUST return to the device package. uBlox can survive with their methodogy because their interfaces are limited in number and are either low rate (RS-232), or are double-ended driven (e.g. USB where the current flowing out is mirrored by the current coming in).
  
  And then the discussion got lost with fast edges, and clocks, etc - nothing to do with fitting a GPS chip-set to a PCB and antenna...
  
  I disagree, it is a crucial point when designing mixed signal, high-speed systems. All those switching signals have to dump into ground somewhere and taken in aggregate, they can severely effect the noise floor of the system. Remember, GNSS signals are already 20dB or more below the open air noise floor; it doesn't take much ground-plane noise from the digital side of the system to raise the noise floor or create harmonics at GNSS frequencies. I once had a customer's asynchronous 4-Mhz clock, via a 392nd harmonic, create 35dB c/n0 jammers in the RF. It doesn't take much. I could tell you more over a beer sometime...
  
  Adding an LNA to increase gain does not increase directivity, and can worsen signal to noise ratio if not done well (shielded, etc).
  
  It also does nothing if it's not helping overcome coax loss or high RF front end noise floors. The signal's SNR is fixed at the first LNA and assuming that LNA has sufficient gain to overcome the losses down-stream, adding additional LNAs downstream will not get you more signal, and in fact may only be adding noise figure losses to the system.
  
  In closing, please do consider writing an article for one of the above mentioned magazines. If you have an interest, I would love to see a comparison of linear-chip, folded-F chips, and patch antennas. One comparison I've been meaning to do is to compare a limited ground plane chip (say 25x25mm) with a patch of limited ground plane (say a 12x12x4mm patch with a 20x20mm ground plane) in terms of directional response, de-tuning effects (think ground planes on wrist watches in close proximity to the human body) and absolute gain. All the antenna manufactures love to give cell-phone sized ground planes (80x60mm) and most of my customers want ground planes 1/4 that size.
  
  Best Regards,
  
  Jeff Wilson
- Paul Meier > Gisela & Joe Noci March 17, 2015 at 9:12am
  
  Nampilot, thanks a lot for sharing your seemingly infinite wisdom...
  
  Always a joy to read from you.
  - Gisela & Joe Noci > Paul Meier March 17, 2015 at 9:32am
    
    AkCopter,
    
    If I implied that a smaller ground plane is better matching, I apologize -
    
    The size of the ground plane is driven by Patch antenna theory, and as I indicated, it should be 1 wavelength long on a side, or around a 1 wavelength diameter disc if round. This is close to optimum for back lobe suppression, and with the correctly designed associated patch radiating element, optimum for directivity.
    
    The impedance matching of such a patch element on such a large plane can easily be achieved.
    
    Reducing the ground plane size will reduce the antenna performance in terms of back lobe suppression, and directivity to some degree, but the patch element can still be properly matched .The antenna in my pictures shows the connection pin - the impedance match is made by moving this pin closer to or further away from the element center - center is 0 ohms, edge is high impedance, and 50 ohms somewhere in between.
    
    So the manufacture may choose to use a smaller than optimum ground plane, for size reasons, and then match the patch element for that size plane. The patch in my pictures was designed for a 30mm plane , from the mnfr specs.
    
    So the bottom line is, the plane should be the size recommended by the patch designer/manufacturer, since the impedance was matched for that set of conditions. The smaller the plane the poorer the back lobes suppression, and potentially the poorer the directivity.
    
    Changing an existing design without recomputing the affected elements merely messes up every design aspect...
    
    My tests show that the plane had to be small for optimum impedance match, as was intended by the manufacturer...
    
    Hope that clears it up.
    
    The Nampilot...
    - Doug Walmsley > Gisela & Joe Noci March 17, 2015 at 10:13am
      
      Thanks for the insightful data and test results. So at this point I see no need to alter my existing boards from CSG Shop.
      
      So at this point I suspect EMI caused by ESCs, Motors, and Flight Controller my affect the signal strength if antenna is located too close to them.
      
      So my questions are
      
      1. What reasonable distance is optimal for mitigating the noise from the electronics? Is there a formula that allows us to determine an optimum displacement for the GPS antenna from the electronics?
      
      2. Are there other options to help reduce EMI?
      
      3. Is there a correlation between the increasing NAVErrors folks have seen lately with respects to the types of GPS units on the market and the Pixhawk/firmware?
      
      After reading you write up, I feel I do not have the right knowledge or equipment to build and test a larger plane for the GPS antenna and have to trust that my manufacture performed all computations and settings needed for the ground plane I have on my 25mm and 40mm antennas.
      
      Lastly, You mentioned that us who fly "..inverted lawnmower things" (funny) should use as large a ground plane as possible due to EMI from ESCs, What size would you recommend assuming size doesn't matter?
      
      Thanks again for shedding light on this topic.
      - Gisela & Joe Noci > Doug Walmsley March 17, 2015 at 1:05pm
        
        Hi Doug,
        
        Firstly, what GPS module do you have? Is it one of these two?
        
        Is the Taoglass patch antenna parts number one of these :
        
        CGGBP-356A 02
        
        CGGBP-184A 02. ]
        
        If so, the the ground plane should be 70mm x 70mm. This is quite clear in the Taoglass antenna specs. It certainly does not look like that in the images... I think your second last sentence must be modified - I think the mnfr did NOT design as per the antenna requirement Specs...
        
        Remember, that does not mean the GPS will not work - it means that the impedance match is wrong, so power is lost, which they make up for with the LNA, but the bad side is the loss of directivity and subsequent pickup of radiated noise from below.
        
        I strongly suspect that radiated noise is the biggest culprit in this situation. The best way to discover this would be to attach a similar antenna with ground plane to a spectrum analyzer and measure the detected emissions with the craft at full power.. not easy without the equipment.
        
        Regarding your questions -
        
        1. No there is no easy way to estimate this without measurement. The radiation decreases with the inverse square of distance ( a simplification, but good enough for this), but you have no idea how high the levels are to start with.
        
        2. Yes - by shielding the noise generators, which is always the best approach anyway. There are some metallic sprays which can be used to create a 'shielded can' from your AP enclosure, etc - not totally effective as the cables still exit unfiltered, so the cables conduct the energy out from the enclosure. Cables can be fitted inside braided shields. ECS should be as close to the motors as possible - the 3 phase wires to the motors are the big culprit - hi current hi frequency square wave PWM signals with harmonics into the many hundreds of MHz. In essence, you want the electronics inside a shielded can, the same as that little can over the GPS chipset. That is very difficult, if not impossible, to achieve, but all the hi current, hi frequency cables should be shielded and grounded at a common point - place a disc of copper foil, maybe 60mm diameter, at the base or some convenient position in the craft, and connect all shield, braids and grounds to it. Braid shield all signal leads, servo leads, power cables. A lot of work, can add weight, but it does work.
        
        WRT your last sentence - I would recommend a ground plane ( with appropriate antenna of course) of at least 70mm x 70mm, placed maybe 200 to 300mm above the avionics. This is almost guaranteed to work in even the most stubborn cases. In less stubborn cases, both dimension can be reduced, eg, in a light craft with lower power to the motors in flight. I would not go less than 40mmx40mm antenna designs, maybe 100mm min above the craft. Any smaller and the antenna sees as much noise from below as GPS signal from above.
        
        As an aside, you may be able to extend the ground plane on this module ( assuming it is one of the two in the pictures?)
        
        Scrape the ground plane copper clean around the PCB periphery, and cut strips of double sided PCB , same thickness, and butt them up against the GPS PCB plane. Solder all along the seam, making sure to cover the joint. A thin copper foil over the joint would be great. The strip width should be such that the 'new' dimension ends up 70mm x 70mm. Don't leave big bumps of solder along the joint. Don't short out other signal tracks to the ground plane in this process..
        
        Then raise the PCB to say 150mm above the avionics and see what improved..
        
        Your question 3 I cannot answer - we have our own autopilot and software so it all works differently..However, with respect to the 'types of GPS on the market' - they are all pretty much the same - all have similar sensitivities, all have similar numbers of tracking channels, all use GPS engines from similar manufacturers, so there is not a lot of difference among them. Where the BIG difference lies is in their integration into a complete GPS unit, ie, when mated with an antenna, the 'correct' ground plane, etc.
        
        A simple check is to obtain the antenna mnfr and part number, get the data sheet, and see what the mnfr recommends as a ground plane. If the GPS unit you are looking at complies within say 10% of this size, go for it. Else avoid like the plague...Unless you are flying normal craft, or your lawnmower is not noisy...
        
        I will be back next week to chat some more - on the way to the Namib Desert tomorrow - we will be darting 5 Desert Lions and fitting GPS SAT tracking collars for a conservation project.
        
        Have a look here just for fun:
        
        http://desertlion.info/news.html
        
        Cheers
        
        The Nampilot..
        
        MR60
        
        Hugues > Gisela & Joe Noci March 17, 2015 at 1:48pm
        
        Thx Jor for your excellent, as usual, explanations.
        
        For the sake of simplicity and clarity, can you graphically point out, on the ceramic patch antenna pictures you posted, where is the "RF ground" : is it the top silver layer , is it the bottom pin ?
        
        When people say to solder the RF ground to a ground plane extension, what part of the ceramic patch antenna must then be soldered to the ground plane extension ?
        
        Gisela & Joe Noci > Hugues March 17, 2015 at 8:58pm
        
        Hi Hughes,
        
        The ceramic patch in the photo - the top shows the pin soldererd to the active element. The bottom shows the pin protruding - that is the RF connection to the antenna = goes to the GPS Chipset RF input, either directly, or via the center of your coax, or via the LNA. The underside of thes type patches, where the silvered 'ground' is, does not connect directly to your PCB copper plane - it connects capacitively, and the RF received 'sees' a ground due to the low impedance of that capacitor at those frequencies. So, your ground reference connection is your PCB copper plane, nowhere else.
        
        I do not understand the concept of a 'ground plane extension' - any connection to this ground plane other than to get power to the GPS from your battery source, serves no purpose at all. This ground plane is the GPS chipset ground, and the Patch antenna ground plane and reference only.
        
        Sorry, back on Monday....
        
        The Nampilot...
        
        MR60
        
        Hugues > Gisela & Joe Noci March 19, 2015 at 1:13pm
        
        ok, if I understand correctly what you're saying, if I want to extend my M8N ground plane from the 35x35 size to a 70x70 size (what seems recommended in the specs of the manufacturer of the chipset), I just solder pieces of copper to extend the PCB size to a 70x70 rectangle (soldered on "scratched PCB" borders to reveal the ground copper layer).
    - AKcopter > Gisela & Joe Noci March 17, 2015 at 9:48am
      
      OK so building a custom ground-plane is definitely out... coz' I don't have the equipment or the know-how to pull this off without making things worse...for now it would be best to focus on the INAVerr and get some dev help on the issue of missing GPS samples
- AKcopter > Gisela & Joe Noci March 17, 2015 at 8:49am
  
  Man!!
  
  WHO IS THIS GUY ;) :)

of 15