Version 01.01, Nov. 06, 2005 - Additional probe notes.
This page discusses "lessons learned" trying to make low-cost, easy to install
A perfect sensor not required to get usable data provided
mitigate the risks.
The perfect magnetic loop is a toroidal core
but the secondary winding would have to be hand
threaded through the core as well as the primary, MG1 or MG2 power lead.
This is a lot of labor and in the case of the MG1 cables, not easy without
disconnecting and reconnecting a lot of cables and tubing.
Fortunately, a perfect core is not a hard requirement since many
commercial current probes use a 'clamp on' sensor, a sectioned core.
Furthermore, the area of the magnetic material is not critical since we
are after voltage levels, not pulling a significant power load.
So the following experimental cores have been attempted:
- Coat-hanger wire loop - was the first test core. However, the stiffness
of the "C" shaped wire made it difficult to install in the limited access space of the
existing MG1 and MG2 cables.
- Pipe hanger loop - was the second test core.
It worked better but was still difficult to install, especially for
Problems included the threads tended to engaged the wire bobbin causing the
wires to stretch and eventually break.
Also, the mounting "U" provided a path for adjacent current to induce noise.
- Hinged, snaping key ring - works but the large diameter makes it suseptible
to stray magnetic fields. Still, it did generate usable signals and if a smaller,
hinged key-ring could be found, would be a good alternative.
- Regular key ring - can not be opened large enough to go around the MG1 or MG2 power
It might be possible to cut one, flatten the ends, Drimel tool shape a latch, drill
a hing hole and use a small nail to make a small hinged keyring.
But this is a lot of work.
- Chain shackle - was at first looking very good. Unfortunatley, the test units
were made with stainless steel which led to no signal at all!
It must be a ferous material with softer steels doing better.
- U-bolt with shackle - not test, yet.
The goal is to generate audio line voltage, ~.775 VAC into a 600 ohm load
to provide full range, 16-bit audio samples.
Using a coat-hanger wire loop, we measured voltage levels of ~5-7 mVAC for
This projected the following turn ratios for the secondary coils:
- 100 turns - MG1 secondary
- 55 turns - MG2 secondary based upon the ratio of MG1, 18 kW, to MG2, 33 kW.
Currently single pairs of Cat-5 cable are bringing the secondary signal into
the cabin and the audio jack.
The software suggests we are seeing ~6-7 bits of noise, 128 units, out of the 16-bit sample
size, 65,536 units, or two and a half orders of magnitude.
We would like to reduce this to 2-3 bits.
First we will investigate quesent noise, the noise when all systems are powered on but the
engine, MG1 and MG2 are off.
The laptop normally runs off of an inverter and we will disconnect it to see if it
is inducing noise.
Other noise risks:
The DC impediance measures 4 and 2 ohms through the coils.
This suggests a very low value, 5-50 ohms, would more closely match the secondary impediance.
Alternatively, a 600 ohm value would more closely match the standard audio input impediance.
- Cat-5 cable pairs - may be picking up noise from other sources in the engine
compartment or cabin.
One mitigation would be to use two-pairs in parallel to bring the two audio signals
into the cabin.
The different twists will hopefully cancel out any stray magnetic or electric fields.
- Low-pass filter - failed in the first experiment using an RC network.
In fact, the noise appeared to double.
- Termination resistor - adding a termination resistor at the audio input jack may
attenuate the non-signal noise. However, the value is not known.
Audio Signal Sampling
We are using the open source, "Audacity", audio recording software.
Once saved, the data is exported as a ".WAV" file and a program passes
through the data to analyze and extract the signals.
But there is still a lot of work that remains.
As we improve the data quality, we will be able to reduce the sample interval,
currently 22 kHz, and look at "live" data reduction.
Also, we will add USB CAM image capture to look for correlations between
traffic conditions and mileage.
Adding a magnetic compass combined with wheel turns will provide a dead-reconing
navigation capability leading to route-optimized speed management.
However, we have found some interesting, software detectable patterns:
- The MG1 starts to spin up and that induces a substantial amount of
noise on the traction battery bus.
This 'common mode noise' is evident in the signal on the traction motor
- As the vehicle begins to accellerate, we see a double-pulse start for form.
- Finally, as the ICE is running and MG1 noise tails off, we see the clear
double-pulses from the control unit.
MG2 Transition Traction to Regeneration
In this sample, we see a transition from traction to regeneration.
The sailent characteristics are:
- A double-pulse identifies the low-power traction mode.
- A sine-wave with high noise just before the peak, is the signature of
MG2 Traction Waveform
- The low-level traction signals are double-humped
- The high-power traction signals are nearly triangular from the
Pulse Width Modulation
MG2 Regenerative Waveform
- Regenerative waveforms have a sinusoidial shape with distinct impulse
noise on the rising slope to the local peak.