B
bitrex
Guest
On 7/16/2020 5:13 AM, Michael Kellett wrote:
I see on their site the marketing blurb says \"Unlike conventional
Inertial Measurement Units (IMUs), additional external reference
signal(s) are not required to compensate for drift error.\"
\"Once initialized, ROMOS will experience a maximum of 0.5mm static
variance offset from true position data over its operational lifetime.\"
I\'m not entirely sure what the phrase \"static variance offset\" means in
an IMU-context but I\'m guessing it may not mean the same as \"lifetime
absolute position error.\"
I think it\'s good to be skeptical but there\'s the possibility that the
\"scam\" such as it is, is the usual marketing trick, you say the words in
the top paragraph about no required external references one place, and
the 0.5mm figure the other, and assumption is made that what they\'re
saying is that with no external references at all you can have
performance 15 orders of magnitude better than other state of the art.
But note it says the additional reference signals are not _required_,
which isn\'t the same as saying it could never use them.
they\'re just not telling the whole truth and perhaps letting people
derive assumptions about what they\'re saying that they never quite said.
Like with a lot of semiconductor products you have to dive 15 pages into
the datasheet to figure out the whole truth.
On 15/07/2020 20:12, bitrex wrote:
On 7/15/2020 12:15 PM, Joe Gwinn wrote:
On Wed, 15 Jul 2020 09:18:11 +0100, Michael Kellett <mk@mkesc.co.uk
wrote:
Found this on \"All about Ciruits\" -
https://www.allaboutcircuits.com/news/micron-digital-claims-to-have-eliminated-drifting-in-imus/?utm_source=All+About+Circuits+Members&utm_campaign=a3e890e124-EMAIL_CAMPAIGN_2020_07_09_11_04_COPY_02&utm_medium=email&utm_term=0_2565529c4b-a3e890e124-280503221
It takes you here:
http://www.romos.io/index.asp?FPFHFGFHIRJEIJILIG
They claim an inertial measuring breakthrough:
\"Once initialized, ROMOS will experience a maximum of 0.5mm static
variance offset from true position data over its operational lifetime.\"
With a dose of snake oil.
\"Using higher dimensional computations with back-propagation, Drift is
also eliminated from positional data.\"
This sounds like the standard 6-state or 9-state Kalman Filter. They
do work in big vector spaces.
External references are also provided to this Kalman Filter.
The information to cancel drift is not in the IMU data, so software
can do nothing to cancel drift from IMU data alone.
There is an absurd video too.
It\'s way to good (by many orders of magnitude) to be true - but what\'s
the point ?
How do they make money, are they hoping to trap just one lunatic
venture
capitalist ?
I would think that a direct test would end the game, so I don\'t see
how even a lunatic investor could be fooled for long.
Micron Dynamics claims that the technology is patented, so I sent an
email asking for patent numbers.
Joe Gwinn
I don\'t see what\'s intrinsically bullshit or about claims of an
inertial measurement breakthrough vs. claims of real woo like cold
fusion or machines that run on their own power.
This doesn\'t seem quite in that same vein, and I doubt there\'s anyone
here with the qualifications to make credible commentary as to what\'s
actually possible or isn\'t with whatever machine-learning technique
they say they\'re applying. If they do have patents that\'s some amount
of credibility, I don\'t know whether I\'d trust my own evaluation of
the claims any more than the patent office, they\'re not all just
rubber-stampers who let any old thing fly.
Either they\'ll deliver, or they won\'t, most things in life tend to be
one thing or the other. But it seems like more of a gamble than an
outright \"scam\" that breaks the laws of physics without further
information. it\'s the kind of gamble VC people do day in and day out
and win or lose on, depending.
If they can\'t or won\'t provide references to the patents they claim to
have that would surely make me more skeptical
There are (at least) a couple of reasons to be sceptical:
First the term \"static variance offset\" is not standard IMU talk - if
you Google it you\'ll find one specific hit - under ROMOS !
Secondly, lets take that at face value - they seem to be saying that the
position error over the device lifetime will be 0.5mm.
From school physics you know that s = (a.t^2)/2 (s is distance a is
acceleration.
Lets assume a life of 5 years = 1.576E8 seconds
re-arranging we get a = 2s/(t^2) = 1e-3/(2.48E16) = 4e-20 m/s/s
This level of acceleration bias is not just unlikely but impossible.
Current good parts offer figures like 3ug = 3E-5 m/s/s bias instability.
I see on their site the marketing blurb says \"Unlike conventional
Inertial Measurement Units (IMUs), additional external reference
signal(s) are not required to compensate for drift error.\"
\"Once initialized, ROMOS will experience a maximum of 0.5mm static
variance offset from true position data over its operational lifetime.\"
I\'m not entirely sure what the phrase \"static variance offset\" means in
an IMU-context but I\'m guessing it may not mean the same as \"lifetime
absolute position error.\"
I think it\'s good to be skeptical but there\'s the possibility that the
\"scam\" such as it is, is the usual marketing trick, you say the words in
the top paragraph about no required external references one place, and
the 0.5mm figure the other, and assumption is made that what they\'re
saying is that with no external references at all you can have
performance 15 orders of magnitude better than other state of the art.
But note it says the additional reference signals are not _required_,
which isn\'t the same as saying it could never use them.
Sure, I think the possibility is there they\'re not lying, exactly, butIf Micron Dynamics really had a 15 order of magnitude breakthrough they
would be marketing in a slightly different way.
MK
they\'re just not telling the whole truth and perhaps letting people
derive assumptions about what they\'re saying that they never quite said.
Like with a lot of semiconductor products you have to dive 15 pages into
the datasheet to figure out the whole truth.