High-G gps tech testing results [cross-post]

[jderimig]

This was also posted in the Vendor Forum but thought I would share it here also. We (RFS-Marsa) have been developing some new tech over the last year or so that we expect to start appearing in products soon. Here is a link to a PR post on the website to describe some great results we are seeing.

High G gps test result

 

[plugger]

Isn't that still within the limits of uBlox 8 or below? I thought below 50km and 500m/s those units would keep lock and I've seen this in application as well.

To my knowledge there's not a high G lockout on those chipsets?

 

[rocket_troy]

I'm sure there's never been a high G lockout on any chipset. It's still a bit mysterious why there's always been a universal 4g limit spec on virtually any cheap off-the-shelf commercial GNSS module. John has mentioned in the past it's a doppler shifting limitation with overhead sats for high vertical accelerations.
The issue I see under my current (albeit naïve) understanding is that you can possibly maintain lock under high vertical acceleration from sats lower to the horizon with good antennas and the correct coupling of such to the receiver; however it's the sats of high elevation that provide your altitude accuracy. So, I think the trick is to utilise the lock from the lower elevated sats to regain a faster lock with the high elevation sats for the coasting phase.
Anyway, we are also in the phase of testing all this stuff out with a package housing multiple GNSS modules including a Novatel 20g specd module coupled to a MTI-7 for INS support. That should provide some answers.

TP

 

[plugger]

The issue I see under my current (albeit naïve) understanding is that you can possibly maintain lock under high vertical acceleration from sats lower to the horizon with good antennas and the correct coupling of such to the receiver; however it's the sats of high elevation that provide your altitude accuracy. So, I think the trick is to utilise the lock from the lower elevated sats to regain a faster lock with the high elevation sats for the coasting phase.

TP

Please correct me if I'm wrong Troy, but assuming you have the VDOP and HDOP values during high vertical acceleration you should be able to confirm if you've had a degradation of accuracy?

 

[rocket_troy]

Please correct me if I'm wrong Troy, but assuming you have the VDOP and HDOP values during high vertical acceleration you should be able to confirm if you've had a degradation of accuracy?

Well, you'd think so, but hey, given I had to look up what those initialisms were, I'm probably not the greatest authority to answer that

TP

 

[plugger]

Just had a look at my most aggressive flight I have with both good Raven and TeleGPS (uBlox 7) data. According to the TeleGPS exported data, I had lock for the first 1 second of boost, I lost lock for ~6 seconds after the first second before I regained lock.



According to the Raven, the time I was above 500m/s was ~6 seconds. For the first couple seconds of the flight I was doing 40Gs Axial Acceleration and only lost lock once I exceeded 500m/s.

 

[jderimig]

The chipset firmware does not have a logical acceleration limit. The limit come into play from the analog and rf side of the entire system.

 

[UhClem]

It's still a bit mysterious why there's always been a universal 4g limit spec on virtually any cheap off-the-shelf commercial GNSS module.

Not mysterious at all. It's the tracking loops.

 

[rocket_troy]

Not mysterious at all. It's the tracking loops.

Thanks David.

TP

 

[plugger]

Hi John, looking at your blog post are you sure you had lock for boost? The straightness of that line in the KML pic looks a little too straight imho.



Here's the KML visual for the flight data I provided above.



Normally a straight line indicates a loss of lock.

 

[jderimig]

The boost was the other side of the plot. Your arrow is pointing to the descent where I truncated the data for the plot.

 

[FredA]

It's my understanding that some GPS modules lose lock under boost due to their on-board crystal changing frequency due to the G-load which can do all manner of things including causing their radio's to drift off of frequency.
Some "cuts" of XTAL's do better than others - as well as some orientations of the XTAL provided some added immunity.
Hard to tell what, if any, impact this has on an individual module unless you really dig into what is inside.

 

[UhClem]

It doesn't matter if the local oscillator changes frequencies so long as it is stable over the (very short) time scale of the measurement. Vibration can cause horrible problems because that results in clock jitter.

Something I ran across long ago: Size Up Acceleration Sensitivity On XOs.

 

[Grog6]

No. The frequency will shift due to doppler effect, and the capture in the receiver is done by a phase locked loop that tracks the frequency shift. Typically in most radios we use that shift for music or the gps signal, here it is a shift proportional to the velocity between the xmitter and rxvr. When it hits a limit, it loses lock.
This is a big problem for sattelites like casinni, when it was booking around jupiter, trying to keep it locked.
They were predicting what the shifts would be when, and read in those numbers to the pll settings, when they got there.
Thiis probably won't help, but wtf:
https://en.wikipedia.org/wiki/Phase-locked_loop

 

[jderimig]

No. The frequency will shift due to doppler effect, and the capture in the receiver is done by a phase locked loop that tracks the frequency shift. Typically in most radios we use that shift for music or the gps signal, here it is a shift proportional to the velocity between the xmitter and rxvr. When it hits a limit, it loses lock.
This is a big problem for sattelites like casinni, when it was booking around jupiter, trying to keep it locked.
They were predicting what the shifts would be when, and read in those numbers to the pll settings, when they got there.
Thiis probably won't help, but wtf:
https://en.wikipedia.org/wiki/Phase-locked_loop

Yes, and the issue with acceleration is the doppler frequency shift rate or how fast is the frequency shifting and can the PLL track that shift rate. At 1g the doppler frequency is changing at around 51.5Hz per second. The 4g spec means the PLL can track a shifting frequency at least to 206Hz per second with a good signal. I am sure that this 4g is conservative and difference chipsets will have different actual upper limits. But there are tradeoffs with extending PLL frequency response.

However tracking a shifting doppler shift will get more difficult with low s/n signals. So in order to achieve the chipset's PLL entitlement good old blocking and tackling RF best practices needs to be done. As the acceleration goes higher the receiver is going to lose the low S/N sats with the high accelerations first. Starting with a robust constellation increases the probably of maintaining lock throughout the boost. There is a reason good GPS antennas cost 3-10X than GPS receivers. Controlled impedance PCB design is still somewhat of an art.