Table of contents
2 Relevant technical specifications
3 Comparison of the recorded speed graphs
3.1 comparison of the curves in general
3.2 comparison of an interval workout
3.3 comparison of a steady workout
4 Comparison of the measured workout distances
5.1 Recommendations (Speed and Distance)
I experienced that the specific ‘implementation’ of the accelerometer
technology by POLAR influences a lot the specific speed and distance
measurements of the S625X. In order to
show this I compared the performance of the S625X and the Philips Nike MP3RUN on
the data collection and data management.
Both use the accelerometer technology of DYNASTREAM (http://www.dynastream.com/home/).
This report summarizes my findings.
The S625X is the most elaborate running and cycling personal trainer
available in the POLAR product gamma.
Speed and distance features and real live behavior of the S625X and the
accompanying software PPP SW is extensively discussed in http://users.telenet.be/wy/sport/S625X/S625X%20Accurate%20Speed-Distance%20Measuremnt.htm
.
The Philips Nike MP3run is a high-end MP3player-FMradio with a rudimentary
speed and distance measurement facility.
There is also no possibility to program a training (no timers, no
limits, no intervals are available). The
MP3run only measures and displays the stopwatch, instantaneous pace and
distance. There is a vocal pace and
distance feedback. The speed data can be uploaded to a PC (encrypted). In the
very elementary accompanying software (Digital Media Manager, DMM) only the distance,
pace and speed of the uploaded workouts can be shown as function of time. In
fact the major functionality of that software is the management of MP3 and WMA audio
files towards the MP3run.
Both speed measurement systems can not be compared by taking all the
features the S625X and the PPP SW has.
Except from recording the speed, the MP3run simply has no support for
any kind of workout programming and report. The MP3run even does not have heartrate
recording facilities.
I believe the target customer of the MP3run is not the professional or
semi-professional runner. So II feel the speed and distance measurement does
not have to be not as good as the S625X manages this. An example of this is the
fact that the stopping process of the StopWatch of the MP3run takes 2 sec
before the StopWatch and the speed recordings really stops.
Unfortunately the results were really the opposite. Extensive running
tests with the POLAR S1 foot pod on the left foot and the Philips Nike pod on
the right foot show that the performance of the recording process of the Philips
Nike pod was 2 to 5 times better than the POLAR S1 pod. And this for all the
comparable tests I did. In the sequel you will find a summary.
The following table sums up some technical differences having an impact
on the performance of the speed and distance measurement and recording process:
|
kind of technical difference |
MP3RUN |
S625X |
consequence for MP3RUN |
|
Weight (on same balance) |
28 gr empty |
54 gr empty |
1. more light, 2. more easy to attach and detach, 3. less robust? |
|
accuracy of the CF |
up to 1% |
up to 0,1% |
fine calibration not possible |
|
claimed distance accuracy |
95% uncalibrated, 99,6% calibrated |
97% uncalibrated, 99% calibrated |
what's in a claim? |
|
speed crack synchronization delay
on the display |
2 to max 3 sec |
9 to 11 sec |
1. synchronization always within 4 footsteps, 2. very low distance leaks at speed cracks |
|
sampling period |
each second |
each 5, 15 or 60 sec |
1. much smoother graphs, 2. better observation of detail and trend, 3. very low speed crack latency in the data file (2
to 3 sec instead of minimal 10 to 15 sec for S625X) |
|
signal transmission |
Bluetooth with FHSS |
analog, not coded |
1. no speed spikes, no sleeping sensor, 2. robust transmission in any environment, 3. signal separation to other users, 4. transmission distance 5m (compare with 2,9m for
S1), 5. digital transmission hence less power consuming for
the pod (50h instead of the 20h revised POLAR spec) |
The cells with green background are to my opinion the ones with best
performance of that specification of the data acquisition. Especially the
reduction in the data file of the 15 sec speed crack latency towards a maximum of
3 sec is promising.
To validate the influence of the intrinsic better performance, several
tests are done by wearing both equipments simultaneously. The MEM SET of the
S625X was set on 5 sec. Also both systems were calibrated as good as possible
and with the same calibration distance (I used a 16 km steady run to avoid any
kind of systematic calibration error).
First some graphs, generated in the DMM and PPP SW, are compared. This
is done by taking screen shots and rescaling these graphs in Paint Shop Pro to
be able to overlay both curves correctly. This was the only way to compare the instantaneous
recorded speed and distance values since DMM does not have facilities to export
data (the workout data of DMM is also stored in binary files). Besides the
scaling there was no other manipulation done on the graph data.
3.1 comparison of the curves in general
Figure 1 visualisation of a
short steady run.
One observes the MP3run graph is smoother (less
spiky, less nervous). This results in a better trend observation of the sampled
speed values.
3.2 comparison of an interval workout
Figure 2 visualisation of a short interval workout.
One observes the MP3run graph tracks more
correctly the speed cracks. The result is a graph with a better resolution. During
the test there were also 2 signal errors (a zero speed) of the S1 (one at the
location the S625X label is put on the graph); I don’t know the reason for
that.
3.3 comparison of a steady workout
Figure 3 visualisation of a steady workout.
Again, the recorded speed signal of the MP3run is
less nervous, so the observation of the trend of the speed measurement is easier.
At the location where the MP3run label is put on the graph, one observes
a fast decay and rise (an inverse speed spike) of the body speed. At that point
I ran around a sharp 90degrees corner. In the remainder part of the workout I
went another 3 times around a sharp corner. This shows the high
resolution of the MP3run speed data.
Clearly the acquisition
process of the MP3run is very much designed towards the reception of the
tangential (forward) body speed, without making the curve as a whole nervous.
The result is a smoother graph with a higher resolution.
The following table
shows the recorded workout distance for both running computers. For the S625X I
took the displayed distance at the event the StopWatch is paused and stopped,
and the LAPdistance at that stoptime. In order to avoid the LAPdistance bug, no
intermediate laps were taken. For the MP3run I took the displayed distance when
the StopWatch is stopped. The results shown are typical (multiple tests were
done to confirm this, except for the last two data rows for which only 2
workouts were done (*)).
|
speed profile |
S625X |
MP3run |
|||||||
|
displayed result |
hrm LAPdistance |
displayed result |
performance with
respect to S625X |
||||||
|
absolute |
accuracy |
absolute |
accuracy |
absolute |
accuracy |
||||
|
reference distance |
1,075 km |
||||||||
|
interval |
30s walking, 30s stand still |
1,17 km |
108,8% |
0,769 km |
71,5% |
1,015 km |
94,4% |
5 times better |
|
|
30s running, 30s walking |
1,01 km |
94,0% |
1,009 km |
93,9% |
1,043 km |
97,0% |
2 times better |
||
|
reference distance |
8,775 km |
||||||||
|
steady |
running |
8,73 km |
99,5% |
8,736 km |
99,6% |
8,763 km |
99,9% |
3 times better |
|
|
walking |
8,39 km |
95,6% |
8,350 km |
95,2% |
8,965 km |
102,2% |
2 times better (*) |
||
|
reference distance |
3,142 km |
||||||||
|
steady |
running on different surface |
3,19 km |
101,5% |
3,165 km |
100,7% |
3,135 km |
99,8% |
3 times better (*) |
|
In the first two data
rows the influence of speed cracks is verified (In the first data row one also
notices the Display bug of the S625X). With this the influence of speed cracks
is measured.
In the data row 3, 4
and 5 steady speed profiles are compared. In order to fade out the effect of
the single LAPdistance bug, the workout distance is always larger than 3km. For
these speed profiles the same CF is used (the one obtained by running at about
12km/h on the asphalt track of data row 3.). This makes it possible to measure the
influence of a different stride and a different running surface.
The columns in light
blue and green background show the obtained accuracy with respect to the
correct distance. The last column calculates from these colored columns the
fraction that the MP3run performs better than the S625X (e.g.
abs[(1-0,715)/(1-0,944)]=5 ).
From this one observes
that, depending on the kind of speed profile, the MP3run is 2 to 5 times better
than the S625X. As already stated, the conclusions on this matter could be wrong
for the last two data rows (*).
These results have
nothing to do with the PPP SW bugs, since only the single recorded and
unmodified LAPdistance is taken into consideration. Also the LAPdistance bug was eliminated as
much as possible. So the real origin of this different performance, with use of
the same accelerometer technology of
Dynastream, is most likely a different blueprint of the implementation of the
Dynastream technology.
This clearly has the
following key results for the MP3run:
1. internal latency for speed cracks minimized
from 9-11 sec (S625X) to 2-3 sec
(MP3run), or within four footsteps;
2. recording rate is reduced from 5sec to 1sec.
This has the following
consequences:
Ř
With
regard to long distance running, the performance on the over all distance
accuracy is already good for the
S625X. The accuracy of the MP3run
however is extremely good.
Ř
With regard
to interval sessions or workouts where speed cracks have some relevance the difference
on the accuracy is enormous: 72%-94% (S625X) to 94%-97% (MP3run).
Ř
My
experience with regard to instant speed feedback performance is summarized as
follows:
|
|
TIMEX BODYLINK |
S625X |
MP3run |
|
Steady pace feedback |
Bad (jumping values) |
good |
extremely good |
|
Reaction time towards speed changes (cracks) |
Not explicitly tested |
Bad (to long latency, not responsive enough) |
Good (always within four footsteps) |
It is shown that the today’s
technology is available to obtain an accuracy of 97% (probably even 99% -with
two pods?-) for all types of workout profiles.
The actual design
limitations of the S625X make it not feasible to measure relative correct over
all speed and distances for interval sessions. In http://users.telenet.be/wy/sport/S625X/S625X%20Accurate%20Speed-Distance%20Measuremnt.htm it was already shown that the ‘expected’
accuracy for common (i.e. no stand still) interval sessions is expected to be
94,5% and better.
The MP3run beats the
S625X on the basic running speed and distance recording (factor 2 to 5 times
better) and profiling (less nervous graphs).
POLAR helpdesk
informed me that the S625X is designed for “steady endurance”. Feedback of long distance runners show that
long distance runners are satisfied by the quality of the S625X, especially on
the accuracy and instant feedback for steady runs. But how long the same long distance runner will accept all the
effects of a less performing speed and distance data acquisition and management?
5.1 Recommendations (Speed and Distance)
1. to (potential) customers:
a.
The MP3run
is not designed as a full featured personal trainer (no programming possible,
no heartrate recording), although the speed and distance measurement is
extremely accurate. For walking and recreation running this might be a good choice.
b.
The S625X
is only accurate for steady endurance workouts. Supplementary, as can be found
in http://users.telenet.be/wy/sport/S625X/S625X%20Accurate%20Speed-Distance%20Measuremnt.htm,
there are serious design flaws (LAPdistance recording) that harm the accuracy,
also for steady endurance workouts. If speed and distance accuracy is not
prevalent for you running, this might be a good choice.
2. To Philips Nike:
a.
Provide
also speed feedback on the display (km/h or mi/h). Actually, the pace-only
feedback on the display (min/km or min/mi) is not really adapted to the
potential customer.
b.
Provide
also multi-language audible feedback
c.
Make the
Calibration Factor with one more significant digit.
d.
Make the
DMM database more versatile
3. To POLAR:
Redesign the speed and distance acquisition. The S625X must bridge the
gap with the today’s MP3run and even go beyond its actual performance. It must have promising features for those who
are not only working on steady endurance. This redesign will have benefits that will be
really not only for the interval runner, but also for the long distance runner:
·
Numerically: the accuracy on the overall measured distance
could be improved for steady runs, and
largely improved for workouts containing speed cracks;
·
Graphically: less nervous speed curves (less spiky) will
show a more correct and more easy to observe trend of the body speed, but also
with a better resolution towards the real forward body speed (speed cracks,
corners, …).
·
Instant
feedback and response speed: almost
no speed feedback latencies at speed cracks, and a better fluent instant speed
and pace feedback;
·
Improved
signal transmission: channel separation
to other users, no speed spikes any more, no electromagnetic interference any
more, longer battery usage, …