The truth about the EB-85

[Jack Crossfire]

Well as U probably already know, the EB-85 didn't work like it's supposed to. It appears to interpolate very low quality GPS readings to get 5Hz, much lower quality than what the EM406 generates. Moreover, it doesn't have the capacitor that the EM406 has. Without a backup battery, it has 2 recalibrate every time it's power cycled.

The 5Hz interpolation gives nice smooth plots of large movements, but completely erases the slight movements that the EM406 could detect. News group surfers will defend the EB-85 with their lives, but it simply doesn't do what we bought it for. This is the worst ripoff since Social Security.

[XJet]

One of the problems with the EB85 is that its velocity and heading readings at low speeds is virtually useless.

I attribute this to a lack of precision in its internal processing.

Basically, the heading and velocity are calculated by comparing the difference between where the GPS is *now* and where it was at the last update. At 1Hz, a slow-moving object will move five times as far as at 5Hz -- therefore the magnitude of the vectors involved is five times larger and therefore the calculations are going to be more accurate as the error represents a smaller percentage of the values being used.

In fact, in my experience, the EB85 often doesn't give any valid heading or velocity data until you're travelling at at least 5Kph (3mph).

Chances are that the internal processor just doesn't have sufficient accuracy (not enough bits perhaps) to even register the delta-lat/delta-lon values associated with slow movements.

Despite that however, it does seem to perform satisfactorily at higher speeds when using faster refresh rates.

You get what you pay for folks :-)

[Unterhausen]

my understanding is that the position you get from the EM406 is 1 to 2 seconds old. I think there is an obvious tradeoff between speed and the filtering that they can do on-board the gps unit.

[Jack Crossfire]

The work on EM406 accuracy has already been done and plotted. It doesn't use position difference to determine velocity. It uses an independant doppler shift equation and velocity is slightly more accurate than position differences.

http://www.rcgroups.com/forums/showthread.php?t=739269

It doesn't lag 1 to 2 seconds. The EM406 is slightly ahead of inertial navigation, even with no lowpass filtering.

http://www.rcgroups.com/forums/showthread.php?t=739885

The EB-85 was advertised as doing everything the EM406 did at 5x the rate, but it doesn't. Anyone with the capability of buying an EB-85 and soldering it certainly can achieve the same result with an EM406 and a microprocessor for interpolation, so it's hard to imagine any added value in the EB-85. Yet another case of many newsgroups regarding a topic but virtually no information.

[dmgoedde]

Well as U probably already know, the EB-85 didn't work like it's supposed to. It appears to interpolate very low quality GPS readings to get 5Hz, much lower quality than what the EM406 generates. Moreover, it doesn't have the capacitor that the EM406 has. Without a backup battery, it has 2 recalibrate every time it's power cycled.

The 5Hz interpolation gives nice smooth plots of large movements, but completely erases the slight movements that the EM406 could detect. News group surfers will defend the EB-85 with their lives, but it simply doesn't do what we bought it for. This is the worst ripoff since Social Security.

Won't defend EB-85A with my life - but I'll point out some errors on what you stated...

Photo#1, you wrote "In a manual hover, no velocity was detected at all." Not true, you can't know what what detected, just what was reported. The EB-85A appears to default to a mode of operation that EM406 can be set to, namely reporting zero velocity when it is below some threshold. This is great for use in car navigation systems when sitting at a stoplight - position is locked, velocity is reported as zero, and heading as the last known value. The velocity and heading information by both EM406 and EB-85A are calculated by doppler shift of incoming satellite signals, so I wouldn't expect it to detect 0.5 km/h anyway.

Photo#2 - What are we looking at... units for Y and X axes?

Photo#3 - If I can assume Y axis is km/hr, then why are you making a big deal about RMS noise of about 0.4 km/hr?!? Also, I don't see any real difference in noise from EM-406 versus EB-85A. I see perhaps 3 fliers in EB-85A that amplitude 1 to 1.5 km/h.... what's the big deal? Please send me your data set from plot in photo3, I would like to do a statistical analysis.

Photo#4 - what are the units? I can make assumption, but it is much better if you explicitly state. Looks like Lat and Lon vectors seperated out... in km/h? If so, why are you trying to detect < 1 speed unit? I understand you apparently are using this in helicopter... but I don't think GPS is suitable for your app if you need to know speed down to 0.1 km/hr


Basically, the heading and velocity are calculated by comparing the difference between where the GPS is *now* and where it was at the last update

That is NOT true at all. It is by Doppler shift of incoming satellite signals. that is a standard method used on GPS. Verify for yourself... take the differences in position between time intervals, and calculate the apparent speed, especially for lower speeds. You calc'd value and reported speed won't be exactly the same. Same for heading.


ETek EB-85A, when used as I think it is intended (above 0 speed hovering) gives DAMN precise speed and heading....heading down to 0.01 degree resolution. I rely on it for "AttoPilot" to elucidate turn rate. It works great.

I don't see any evidence of mere "interpolation" at 5Hz. It truly is (and must be) 5 Hz data. I say this based on experiments of driving around and having streaming data displayed on my laptop, and how the AttoPilot flies.

[Jack Crossfire]

> ETek EB-85A, when used as I think it is intended (above 0 speed
> hovering) gives DAMN precise speed and heading

What units were DAMN precise speed in?

[dmgoedde]

> ETek EB-85A, when used as I think it is intended (above 0 speed
> hovering) gives DAMN precise speed and heading

What units were DAMN precise speed in?

Fair enough! I got on your case about units.

I don't see what you are complaining about at begining of this post... GPS is an awesome technology but you are trying to split hairs. I have used both EM406 and ETek EB-85A... the ETek is superior for what I need and is no mere 5Hz interpolation.

What I observed is driving on a very straight paved road in Arizona at about 60km/h, heading bounced around less than 1 degree, but I don't have a plot of the raw data, because I didn't capture the stream.

BTW, every instance I know of helicopter control that uses GPS does NOT just use serial NMEA 0183 data stream coming from GPS... that is called a "loosely coupled" approach. A "tightly coupled" approach is used, where raw gps data is fed into the Kalman filter, and the gps latency can be programmed into the model, and normal noisiness in GPS is handled along with noisiness of all the IMU sensors...result is a best estimate of helicoptor state.

[Mark Harris]

Just to throw in my 2c, I've never had any problems with the EB-85A except when using the ones with old firmware. The new firmware ones perform so much better than the EM-406 or EM-411 units. I can see on our OSD when using the EM-406 that the data is clearly over a second behind - usually more - especially with altitude. With the EB-85A it tends to fairly accurate.

I too have done drive tests with it, from 20km/h to 110km/h, taking corners too fast and stopping rapidly/accelerating rapidly (i drive a V6). It was rock solid on everything but rapid deceleration where it would think i was slightly ahead of where i am for a few seconds. This was with full brakes though to the point that ABS was kicking in heavily.

I rather like the EB-85A for the amount of sats it gets a lock on (2-4 more than the EM-411 in Australia) and the speed of its data. If you're finding your speeds are very low try dropping it down to 1 or 2hz.


$PMTK220,1000*1F

[dmgoedde]

Mark - do you have any experience re-flashing the firmware of EB-85A? I have a few units with older firmware. I believe the manual talks about it, but I didn't see a connector scheme to interface EB-85A to a PC.

Dean

[Daniel Wee]

E-Tek says to send the units back to them I think. The good news though, is that the new firmware is holding up quite well. I've tried to confuse it and it seems to be okay, unlike the previous firmware.

I have a scrolling compass and, John, I don't see the lag you are talking about. As far as I can tell from my videos, the GPS is tracking my changes in direction close to real time.

Daniel

[docphi]

Mark - do you have any experience re-flashing the firmware of EB-85A? I have a few units with older firmware. I believe the manual talks about it, but I didn't see a connector scheme to interface EB-85A to a PC.

Dean


I just spoke with E-tek and they said send it back to them for re-flashing.

[dmgoedde]

I just spoke with E-tek and they said send it back to them for re-flashing.

Thanks! I'll be contacting them (I have 6 of the "old" firmware EB-85A units)

[Mark Harris]

Dean,

I've been trying for awhile now to get E-Tek to send me the stuff i need to re-flash, my contact there is trying to get me hooked up with the stuff to do it.

Failing that, Intelligent Flight will be happy to take in all your old EB-85A's and send them to the factory en-masse for a reflash which will take postage fees and any fees etek decide to charge right down ;)

[mlbco]

Jack,

I'm guessing your struggling with some way to get a stable position hover from your heli and GPS feedback alone doesn't have enough damping (or phase lead). A Kalman filter is the best solution but can be very tricky to design and implement. Another possibility is to consider a complimentary filter that combines 2 different sensor inputs over different portions of the frequency spectrum. Your GPS velocity is only good for motions below 1HZ frequency and you need some other info to get velocity estimates above 1HZ.

For a rigid rotor system, the short-term tilt angle of the fuselage is directly related to the vehicle's linear acceleration when hovering. If you take your pitch attitude estimate, integrate it, and then pass this signal through a washout filter with a pole at around 1HZ, you should end up with a short term velocity estimate. This can then be added to your GPS velocity which should be filtered with a low pass at around 1HZ. I've attached a diagram below showing the filters assuming your only sensors are a gyro and GPS. You may need to experiment with the tuning (pole location) and perhaps higher order filters, but this should work. I've used it successfully on a few hovering UAVs.

Steve Morris

[Jack Crossfire]

She uses a 14 state Kalman filter for velocity. The velocity reading is well documented and lags significantly. Feeding the integral of the perceived tilt back into velocity feedback hasn't been tried yet. Yet another feedback path on the list.

[Jack Crossfire]

Looks like predicting acceleration from tilt was the silver bullet. Used a neural network to translate tilt to acceleration and position hold just started working. It's not as warm and fuzzy as having direct navigational readings but it works.

[Jack Crossfire]

Sounds like the pros are using WAAS enabled GPS, probably in the 10Hz range.

http://www.canalgeomatics.com/press_release_details.php?news_release_id=31

http://www.roboticstrends.com/displayarticle899.html?POSTNUKESID=b23b1bd8683f72b e22a2ddae9f057eac

[Jack Crossfire]

The GPS unit that Rotomotion first achieved auto hover with was the Ublox TIM-LP. That had 4Hz updates standard velocity deviation below 1cm/sec. Within 2 months of buying it, he had position hold in high wind. Obviously hovering with 1Hz GPS or a 5Hz GPS with >2m/sec velocity error is going 2 B a hard life. The Ublox online store has their current 4Hz unit for $100 but the hunt is on for a retail Novatel 10Hz unit.

[Mark Harris]

Let us know if you find anything :)

[zik]

How about a Chinese uBlox LEA-4S clone for US$20 each. (Ok, that's in 1000+ quantities. But it's cheap!)

http://www.globalsources.com/gsol/I/GPS-engine/p/sm/1003277443.htm

[Jack Crossfire]

According to navtechgps.com, the mighty 20Hz Novatel OEMV-1 is $995. The OEMV-1 can be a base station for DGPS.
https://www.navtechgps.com/Shop/equipment/oem_boards_kits_and_fw_upgrades/oem_boards_by_manufacturer/novatel/oemv-1_card__outputs_rt20__dgps__sbas__20hz_oemv-1-l1.asp

Then of course, there's the 5Hz Novatel Superstar II. That's $115 at
http://www.navtechgps.com/Shop/equipment/oem_boards_kits_and_fw_upgrades/oem_boards_by_manufacturer/novatel/superstar_ii_receiver__5v_ssii-5-std.asp

The Superstar II generates raw carrier phase measurements and is the cheapest thing advertizing carrier phase measurement, but methods to convert that to DGPS input are not published.

It looks like the UBlox 4Hz module is the safest option for after-tax spenders. It's the only thing proven to actually work, in a world of 10,000 web pages per single piece of good information. $100 + $20 antenna + $20 shipping = $140

But the 1 extra Hz & carrier phase tracking of the Superstar II looks intriguing. The Superstar II needs an RF shielding enclosure.

[Jack Crossfire]

Looks like the $115 Superstar II only does 1Hz. The 5Hz model is
$315 and model number SSII-3-5HZ.

https://www.navtechgps.com/Shop/equipment/oem_boards_kits_and_fw_upgrades/fw_upgrades_by_manufacturer/novatel/superstar_ii__3_3v__5hz_pvt_@_9600bps_ssii-3-5hz.asp

[Jack Crossfire]

One more point for the uBlox: adaptiveflight.com uses a uBlox 4Hz unit

http://www.adaptiveflight.com/products.html

[mhacker]

Hi,
I'm new here, but really interested on this thread!

So..? What it's the veredict? :confused:

I'm considering to buy my first OEM GPS for rocket telemetry. The EB-85A deserve to spend USD 30,- more than the ET-406A/411?

Marcelo

[Jack Crossfire]

Without DGPS, the EM406, EB85, and uBlox all had horrible altitude readings. The uBlox in particular lagged altitude by several seconds. That should end any question for rockets.

[mhacker]

Hi Jack,

Thank's for the advice! Did you have any alternative OEM GPS to suggest instead them?

The alttitude issue it's not mandatory because we also had an barometric altimeter -flight computer-, and the main needs of the GPS it's for real-time telemetry to help locate the rocket on it's descend.

We pointed to those GPS because a lot of information and projects on the Internet.

If you have any experience with some of them, we will appreciate your comments.

TIA

Marcelo

[ohararp]

I have been using the EB-85's in my gps loggers for several months in a prototype format and have ordered a number of these units for others. Pricing is $55 and are available here (http://www.ohararp.com/products). We use them mainly for walking and driving navigation and have found them to be very reliable.

______________
Ryan O'Hara
www.ohararp.com

[kbosak]

Without DGPS, the EM406, EB85, and uBlox all had horrible altitude readings. The uBlox in particular lagged altitude by several seconds. That should end any question for rockets.
There are a few amateur altimeters based on Motorola MPX series sensors, designed just by model rocket amateurs. Google, will find. Try also searching for MPX2100 in this context.

[kbosak]

Hi all,
I have some strange problem with EB-85 from sparkfun. Using their devboard.
Initially using no backup battery.
One day the unit locked on 8 satellites over Poland, indoors, in 1-story house, concrete roof, a few inches from very large window (1.5x2m), far from any interference except wolves (facing a large forest).
Later that day, it locked none in identical situation.
I started believing I have damaged the unit.
2 weeks later, the unit locked on after a few minutes, and status led on sparkfun devboard started working, status led started blinking from time to time! then afte some 20 minutes it lost all 7 (or more) satellites), status lead off, nothing locked despite a few hours of power-on.
Patch antenna always facing up, connector looks fine...

should I solder the connector?
power supply works fine, recent tests showed that adding backup battery is making setup persistent, but it doeasnt locks the sat.
i have reheated solder bubble on patch antenna to make sure it contacts well. no change.
oscilloscope shows clean power supply, tested different, including separate battery.

This is very difficult to diagnose, because even if this is a connector or some soldering problem, I will have to keep it in 'good' state for a few minutes to get any lock.
Help!

P.S. RS232 always working perfect, here you see capture when locked:

$GPGGA,095826.600,5102.9808,N,01723.8771,E,1,3,2.1 6,149.8,M,42.4,M,,*5B
$GPRMC,095826.600,A,5102.9808,N,01723.8771,E,1.12, 167.57,050308,,,A*67
$GPVTG,167.57,T,,M,1.12,N,2.08,K,A*37

$GPGGA,095826.800,5102.9808,N,01723.8771,E,1,3,2.1 7,149.8,M,42.4,M,,*54
$GPRMC,095826.800,A,5102.9808,N,01723.8771,E,1.10, 167.57,050308,,,A*6B
$GPVTG,167.57,T,,M,1.10,N,2.03,K,A*3E

$GPGGA,095827.000,5102.9808,N,01723.8770,E,1,3,2.1 6,149.8,M,42.4,M,,*5D
$GPRMC,095827.000,A,5102.9808,N,01723.8770,E,1.08, 167.57,050308,,,A*6A
$GPVTG,167.57,T,,M,1.08,N,2.00,K,A*34

$GPGGA,095827.200,5102.9807,N,01723.8769,E,1,3,2.1 6,149.8,M,42.4,M,,*58
$GPRMC,095827.200,A,5102.9807,N,01723.8769,E,1.10, 167.57,050308,,,A*66
$GPVTG,167.57,T,,M,1.10,N,2.04,K,A*39

$GPGGA,095827.400,5102.9807,N,01723.8768,E,1,3,2.1 6,149.8,M,42.4,M,,*5F
$GPGSA,A,2,32,11,28,,,,,,,,,,2.38,2.16,1.00*05
$GPGSV,3,1,09,19,75,244,,32,72,172,16,22,59,071,,0 3,52,175,*70
$GPGSV,3,2,09,14,32,124,,18,21,052,,11,19,281,21,0 1,16,150,*7D
$GPGSV,3,3,09,28,10,332,23*48
$GPRMC,095827.400,A,5102.9807,N,01723.8768,E,1.12, 167.57,050308,,,A*63
$GPVTG,167.57,T,,M,1.12,N,2.08,K,A*37

...and here without lock:
$GPGSA,A,1,,,,,,,,,,,,,,,*1E
$GPGSV,1,1,00*79
$GPRMC,111456.065,V,8960.0000,N,00000.0000,E,0.00, 0.00,050308,,,N*7A
$GPVTG,0.00,T,,M,0.00,N,0.00,K,N*32
$GPGGA,111456.265,8960.0000,N,00000.0000,E,0,0,,13 7.0,M,13.0,M,,*44
$GPRMC,111456.265,V,8960.0000,N,00000.0000,E,0.00, 0.00,050308,,,N*78
$GPVTG,0.00,T,,M,0.00,N,0.00,K,N*32

[Jack Crossfire]

That was one which had to be taken outside for the lock & then it would maintain the signal indoors. It couldn't always lock indoors. If new satellites appeared or disappeared between sessions, the EB-85 & uBlox recalibrated from scratch. People win or lose a lot of money by what is said about these sensors on the internet, so obviously we can't give away the whole story.