Smart Coil 2
Smart Coil 2
Combustion Characteristics
Abstract
The new trends in spark ignited engine technologies (direct injection, lean operation,
cooled EGR, flexible valve-trains, heavy pressure-charging, flexible fuel capability,
CNG…) have increased the diversity of ignition needs for optimum combustion.
Understanding the sensitivities of an engine’s combustion system to ignition
parameters is one key to finding and defining the best compromise of engine
performances and ignition equipment costs. Achieving a definition of this
compromise requires a considerable time against the backdrop of contemporary
technologies, and the need for a truly adjustable ignition tool has been long identified.
Working in the field of ignition since its creation in 1999, APOJEE developed
SmartCoil, an ignition coil emulator capable of generating any desired ignition
characteristics. Extensively used within the automotive and racing fields, this first
product was constantly improved to meet the evolving customer engine
requirements. Ever increasing demands for higher spark energy, longer spark
duration and higher available voltage have recently exceeded the design limits of this
first SmartCoil, requiring the development of a higher performance product.
The capabilities of SmartCoil 2 facilitate rapid ignition research and the definition of
all associated ignition componentry, quickly, safely, economically, and most
importantly in a precise and scientific manner.
Contributors
PhD, Jérémie Darrasse*, R&D team leader, combustion and ignition group, APOJEE.
BEd, Roy Flitton**, product line expert expert, combustion and ignition group,
APOJEE.
Eng, Diego Ismirlian, R&D engineer, combustion and ignition group, APOJEE.
PhD, Paul Tinwell, technical expert, combustion and ignition group, APOJEE.
APOJEE sa.,
29 rue Georges Besse,
ZI du Brezet,
63100 Clermont-Ferrand,
France.
Many equipment sets were sold into quite diverse fields ranging from research labs,
through automotive OEMs and tier-one component suppliers, to motorsport
applications including several of the prestigious F1 teams. Applications have ranged
from extremes such as spark to fuel-spray interaction studies, spark plug
development, catalytic-convertor light-off testing and hunting for that last newton
metre at the crankshaft. In recent years these existing and new customers of
SmartCoil have repeatedly requested an increase of the range of operation of the
spark parameters. Typically needed to investigate the ignition requirements for the
emerging new generation of combustion concepts such as very lean operation,
extreme charge dilution and heavy pressure charging, the original SmartCoil
hardware struggled to meet these requirements. Maximum available voltages up to
and above 50kV, peak discharge currents of up to 2A, single spark energy delivery
exceeding 1J, and often at very high engine speeds proved to be
insurmountable. Many and repeated failures were attributed to extreme EM
interference and internally generated corona which could not be adequately
suppressed. A complete revision of the hardware was considered necessary. The
subject of the present paper is to relate the major aspects of these design changes,
the improved performances yielded and the added functionality achieved. For a full
understanding of the base SmartCoil functionality, readers are directed to the
supporting materials [1,2]
2 SmartCoil Operational Issues and Redesign Objectives
The original SmartCoil hardware philosophy was one of modularity. Creating a spark
characteristic over a huge dynamic range, for example spark durations between 10µs
and 5ms, imposed strong physical compromises on the hardware. This was
originally addressed by the simple expedient of mounting the pulse-transformer onto
a separate ‘high voltage’ card. This facilitated a smaller, more controllable dynamic
sub-range, whilst allowing the card to be swapped quickly and efficiently to access
another sub-range of the overall capability. In practice this amounted to typically two
or three high voltage cards to cover the entire range depending on customer
requirements. A plug-in module to the HV card allowed the polarity of the discharge
to be set relative to the ground of the engine. Extending this philosophy, the control
DSP was mounted on a separate card so its hardware and firmware could be
specifically tailored to that of the high voltage card in use. Another separate card
managed the power supply and high voltage bus. A further separate card controlled
communications with the host PC via RS 422 and USB convertor, together with the
ignition trigger timing from the engine ECU or original ignition system. Offering the
ultimate flexibility this solution enabled the overall range of spark characteristics
detailed in Table 1.
SmartCoil SmartCoil 2
Table 1: comparison of the major performance capabilities of original SmartCoil and SmartCoil 2.
An interesting and often exploited corollary of such an approach was the capability of
the equipment to be built to service any number of required cylinders. Other than the
requirement for all the cards to be fitted into a single rack-mounted unit, theoretically
any number of cylinders up to and including 16 could be accommodated. In practice,
the mono-cylinder set was of most interest to customers, although several four
cylinder and six cylinder sets were built.
Requests from existing customers for higher yet performances started within the F1
industry. Driven by the search for an ignition solution which ensured the maximum
possible exploitation of the new direct injection, pressure-charged, ‘low engine-speed’
combustion systems, yet which still remained within the framework of the revised
rules, the primary concern was available voltage. Requests of “above 50kV” were
common to ensure reliable ignition with large spark gaps and small ignition advance
angles. Such a figure was beyond the original design capability of SmartCoil.
Similarly, those equipment sets at customers investigating combustion systems
employing extreme lean-burn or charge dilution or downsizing frequently met with the
SmartCoil’s ceiling of capability. Requests of “up to 70kV” were made in these cases
to allow spark ignition studies on equal terms with other ignition concepts such as
corona discharge…
Additional concern was raised over the very complex mechanical arrangement of the
boards which precluded good shielding practices, and which made assembly of the
equipment difficult, particularly in multi-cylinder format. Finally, the potential for
corrosion of internal components and connectors as a result of corona or ozone
generated when operating at the higher limits of available voltage.
Addressing these issues by reworking the existing layout, adding internal shielding
and upgrading connectors proved fruitless. A decision was made to completely
redesign the equipment. Taking the opportunity to upgrade several aspects of the
functionality at the same time, objectives were set as:
The primary concern was the elimination of all inter-card connectors and the
hardening of the respective EMI shielding on critical communications buses. By
definition this required a single PCB, which in turn dictated a very significant upward
step of function and flexibility; one card now needed to embody not only all the
functions served by several cards of the previous generation, but that single card now
needed to be as adaptable as possible to avoid the need for future redesigns as
function evolved. Despite these apparent difficulties, correct function was assured by
adopting two 150MHz processors with dedicated areas of responsibility and bespoke
firmware. All aspects of the hardware were checked with respect to EMI using the
existing mathematical models of the general SmartCoil regulation principle, extended
to incorporate the expected EMI under all operating conditions. Rigorous application
of this simple technique ensured all PCB track routings were optimised and a full
knowledge of the equipment was gained before any prototypes were built.
Figure 1 shows the optimised layout of equipment; Figure 2 the general control
principle. A single PCB mounts all of the control and communications hardware. A
completely separate high voltage board enclosure with RFI gaskets was specified to
ensure the best possible confinement of EMI. All connectors and cables are to RF
specification for maximum resistance to EMI irrespective of the source. An external,
adjustable 1kV power supply was specified for the power bus to ensure ease of
thermal regulation and the facility for rapid changes to SmartCoil 2 output. This
decision ensured a standard off-the-shelf, ‘rugged’ applications power supply could
be used for high reliability with minimum impact on overall equipment development
time. The geometry of the external power supply dictated the case size of the
SmartCoil. Simple thermal modelling specified the cooling fan and duct sizes to
ensure adequate dissipation of the MOSFET switching losses, which become very
significant at the desired switching frequency of up to 5MHz.
The 48V power supply services the control PCB and external communications which
were specified as Ethernet or RS422 are preferred by the end-user. Chosen
because EMI-hardened hardware was readily available off-the-shelf for these
protocols, the appropriate hardware was ‘piggy-back’ mounted to the main control
PCB. The communications cable from SmartCoil to host PC was specified as at least
‘Cat 6’ for good EMI immunity. As default, a standard laptop Ethernet card
(100Mbps) was specified as necessary to communicate with the SmartCoil 2.
Figure 1. SmartCoil 2 hardware layout
Following the chief requirements for improved performance, and realised both
through significant modelling effort and extensive experimental work, the maximum
available voltage, dV/dt, and maximum available peak current were all achieved in-
line with Table 1. Verification of these data was undertaken using three techniques
to load the output of the SmartCoil:
2 An adjustable “open” spark gap (Figure 3). This load solution can be equally
configured with non-resistive or suppressor cabling and offers a more realistic
discharge loading to SmartCoil. The required voltage for gap breakdown can be
raised by increasing the spark gap size, and there is naturally a true transition
through all phases of a spark discharge. Such a solution generates the maximum
possible RFI which is excellent for understanding the behaviour of SmartCoil under
such conditions, but there remain shortcomings compared to a real application. The
most obvious being there is little capability to affect adjustment of the glow-phase
voltage. Perhaps less obvious is the unrealistic overall capacitance of the load.
3 A final test bench configuration was to mount a spark plug into a suitable
pressure chamber and pressurise the firing end to pressures typical of the ignition
point in a real engine. The interior of such a chamber is shown in Figure 4, which
may be equipped with video capture equipment and simulated airflow to produce a
reasonable representation of the real conditions within an engine, albeit at room
temperature. The chamber is typically pressurised with dry air or N 2. For this
sequence of tests air (without flow) was used, together with a single spark plug type
of 1.0mm spark gap. Such an arrangement gives an excellent simulation of real
engine conditions, without the need for an engine test facility. An engine test
chamber is indeed available at APOJEE and much experimentation was undertaken
during and subsequent to the period of SmartCoil development. The data collected
are considered outside of the scope of the present paper and will be the subject of a
future publication, with the sole exception of flame ion-signal monitoring discussed in
a later section.
Figure 4: a spark plug mounted into a pressure chamber.
Facilities are available for video capture and imposing an
airflow over the spark plug electrodes.
a maximum of +55kV, available 30µs after the trigger (i.e. dV/dt = 1.8kV/µs)
Figure 6a: 10-spark average resultant output Figure 6b: 10-spark average resultant output
voltage at 500V supply bus: +39kV. voltage at 700V supply bus: +49kV.
500V and 700V respectively. This result showed the sensitivity of the supply bus
voltage, and confirmed the capability of the equipment to reach its target of 50kV as
detailed in Table 1. As a point of note reflecting the accuracy of the SmartCoil 2
measures, Figures 7a and 7b display the voltage and current waveforms of the last
spark in the sequence of 10 and the captured data respectively. As for the initial
version of SmartCoil these data are available for user export at any time, in the form
of an Excel worksheet.
Figure 7a: Voltage (top) and current Figure 7b: Statistical data of the ten
waveforms displayed in the SmartCoil user discharges in sequence. 700V supply bus.
interface for the last spark in the sequence of The majority of the data fields are blank in
ten discharges. 700V supply bus. this example as SmartCoil 2 did not detect
the presence of a breakdown.
• a maximum of +55kV,
The experimental variable for each test sequence was the requested rise-time (TUs
cons) with values of 10µs, 40µs and 60µs, representing dV/dt slew rates of 5.5kV/µs,
1.4kV/µs and 0.9kV/µs. The supply bus was set to 700V for the duration of the tests
and suppressor type ignition cables were employed in all configurations.
At the shortest rise time value shown in Figures 9a (load box), 9b (open spark gap)
and 9c (spark plug at 20bar) it was not possible to achieve the desired 55kV despite
the 700V bus supply. This represents the ultimate limit of the pulse transformer
imposed by its characteristic inductance of 2.2mH. If such a fast rise time were
needed in combination with available voltage higher than the achieved figure of
+34.5kV, a lower inductance transformer would be needed, and could be built to
order. As illustration by Figure 10, a faster pulse transformer with inductance of 1mH
results in a significantly quicker rise time, in this case achieving 50kV in less than 4µs
with a slew rate of 26kV/µs
For the three loading conditions of Figure 10, a voltage slew of 2.2kV/µs was
achieved with the load box; 2.8kV/µs on the open spark gap and 1.8kV/µs on the
spark plug. All of which are significantly below the target figure of 5.5kV/µs,
confirming the need for a faster pulse transformer at this particular test condition.
The middle rise-time condition of 40µs (1.4kV/µs) gave a much more satisfactory
result (Figure 12). With the same three load conditions, the requested peak voltage
was achieved on the load box with a corresponding slew of 2.0kV/µs. On the open
spark gap the gap was a little small, breaking down at +46kV, but offered a slew of
3.3kV/µs. Similarly, on the spark plug again pressurised to 20bar, only +34kV was
possible before breakdown yielding a slew rate of 2.3kV/µs.
Figure 9a (left), 9b (centre) and 9c: Voltage waveforms and peak values reported from the
oscilloscope at the 55kV in 10µs (5.5kV/µs) rise-time condition for the load-box, open spark gap
and the spark plug in a pressure chamber at 20bar. All tests at supply bus of 700V.
60
50
1,0mH Pulse transformer
40
30
30kV
Voltage (kV)
20
10
0
-5 0 5 10 15 20
-10
-20
1,16µs
-30
Time (us)
Figure 10: Voltage waveform and slew rate available with a 1mH pulse transformer with a supply
bus of 700V.
The data of the slowest rise time condition, 60µs to 55kV (0.9kV/µs), are presented
in Figure 12. Here, the same load conditions resulted in a somewhat bizarre voltage
characteristic as the SmartCoil2 attempted to control the voltage rise over an
unrealistically long period. Most apparent on the load box (Figure 12a), the
consequence of the long requested rise time and control difficulty was only 39kV
being achieved, realising an overall slew rate of 0.6kV/µs. The open spark gap case
fared much better (+39kV before the initial breakdown, and 0.95kV/µs slew), and
likewise the pressurised spark plug with a very clean characteristic (+38kV and
0.85kV/µs).
It should be apparent these data indicate the target rise time was missed by some
margin at each load condition, with the exception of a few cases. These data are
presented in the first instance to illustrate the often overlooked influence of the
ignition system capacitance. Each load condition presents a quite different level of
capacitance. The load box will present the lowest characteristic capacitance,
whereas the spark plug in the pressure chamber will likely be the highest.
Additionally, any variations in the routing or length or type of the ignition cable will all
change the system capacitance case to case. Even the use of the high voltage
probe will significantly add to the system capacitance and effect the voltage rise time.
Care must be taken to ensure a minimum of changes between tests to avoid
problems, although in a real test environment this rarely is of concern with the
exceptions of cable position or spark plug capacitance.
Figure 11a (left), 11b (centre) and 11c: Voltage waveforms and peak values reported from the
oscilloscope at the 55kV in 40µs (1.4kV/µs) rise-time condition for the load-box, open spark
gap and the spark plug in a pressure chamber at 20bar. All tests at supply bus of 700V.
Figure 12a (left), 12b (centre) and 12c: Voltage waveforms and peak values reported from the
oscilloscope at the 55kV in 60µs (0.9kV/µs) rise-time condition for the load-box, open spark
gap and the spark plug in a pressure chamber at 20bar. All tests at supply bus of 700V.
Figure 14: SmartCoil 2 user interface displaying a typical ‘linear decrease’ current profile with a
peak control value of 220mA. The resultant discharge (on a load box) follows the desired
triangular current profile with a peak at 233mA.
Figure 16: SmartCoil 2 user interface displaying a quasi-linear current profile with a peak
control value of 1000mA, and a spark duration of 1000µs. The resultant single discharge (on a
spark plug pressurised to 20bar of air) follows the desired current profile with a peak at 970mA.
Note the very low glow-phase voltage (604V) in this case and the remarkably low resultant
delivered energy (173mJ).
Table 2: An example worksheet, or MAP, used in this instance for multiple discharges. The
characteristics of each spark in a train of eight is defined, together with the requested spacing between
each spark. Up to 20 individual sparks may be defined in this way and the entire train will be delivered at
a single SmartCoil 2 input trigger.
Figure 18: The spark waveforms from the SmartCoil 2 user interface (left) and an oscilloscope
(green trace is current), resulting from the request parameters of Table 2 and a single input trigger.
A spark plug pressurised to 10bar air was used as the load.
5 On-board Protection
The hardware of the original SmartCoil equipment did not support the facility for on-
board monitoring of its function. Other than the obvious control of spark formation
and discharge, and the spark function diagnostics such as flagging a partial spark or
a complete misfire, there was no internal function ‘watchdog’. Later software
included a basic temperature monitor for the MOSFET switches, with a progressive
shutdown in the case of overheating. This proved to be perfectly adequate for the
realitively low power capability of the original SmartCoil. With the new SmartCoil 2
handling significantly higher peak currents, higher voltages and with a higher internal
regulation frequency, the consideration of on-board protection became quite
important.
In-field testing of this solution demonstrated good capability, but occasional failures
attributed to thermal problems were still experienced. Work to understand the
phenomenon highlighted spurious behaviour and repeated failure of either the
MOSFETs or the pulse transformer, often when requesting a single spark by manual
trigger! All examinations of the damaged components confirmed the issues were
thermal, yet increasing the thermal inertia of the cooling system had little effect.
Math models of all related components were examined more closely and
substantially increased in complexity to reflect a more realistic solution. Issues with
unregulated switching causing a catastrophic, runaway thermal event within a few
hundred microseconds of the start of a spark quickly became apparent. Modifying
the control circuitry to supress these resonant problems gave unacceptable
performance degradation, so an alternative solution of avoiding operation at any of
the troublesome conditions was pursued. Unfortunately this task was somewhat
complicated by the need to consider the instantaneous thermal response of the
components, together with their inductive and capacitive interactions, as well as the
power switching requirements of the requested spark parameters. To this end, a
series of control functions and look-up tables was devised to evaluate the safety of a
parameter set before the SmartCoil 2 is permitted to generate a spark at the
requested parameters. Two models are run in the processing background
continuously: one based on the proximity to saturating the pulse transformer, the
other on the instantaneous thermal considerations of the MOSFETs and the pulse
transformer. If the requested spark parameter set is unsafe by either analysis, the
request is rejected and an appropriate user message is displayed. Importantly, even
if a parameter set passes this appraisal process, the global thermal monitoring by
thermistors described above is still active.
Figure 19: A visualisation of the on-board SATURATION security monitoring calculations of SmartCoil 2.
Data are plotted at the fixed conditions of a linear decrease current, Vglow of 1000V and spark duration
control (left) and energy control (right). Away from this condition the green, safe-operation area changes
substantially. SmartCoil 2 will not allow operation within the black area of the curve.
phase voltage, for a linear decrease current characteristic and when operating in
spark duration control mode (left graph) and energy control (right). Clearly the extent
of the safe green operating area substantially changes from one condition to the
other despite their ostensible similarity. Other conditions further alter the safe
operating regions and illustrate the necessity for continuous on-board monitoring.
7000rpm
7000rpm
9000rpm
11000rpm 15000rpm
13000rpm
14000rpm
Duration control 15000rpm
Energy control
6 Communications
The communications protocol of the original equipment set used RS422. Quoted as
being suitable for use in balanced digital multipoint systems, networks employing the
standard can be used effectively over long distances and in electrically noisy
environments and has found much favour in industrial environments [3]. Data
transmission speeds of up to 35Mbit/s over distances of 10m are not uncommon and
proved to be completely suitable for the original SmartCoil. To negate the need for a
suitable communication port on the control PC, a RS422 to USB convertor was
specified. Although this was susceptible to physical damage when transporting
equipment, and occasionally in use, the data throughput was more than sufficient to
ensure the PC screen refresh rate of only 5Hz was accommodated, typically
transmitting 1Mbit data packets at each update. Occasional problems were found
with noise ingress at the convertor, especially once the higher available voltages
were being pursued before the decision to redesign the equipment.
For SmartCoil 2, a much higher screen refresh rate was desired to ensure the display
within the user interface reflected conditions at the spark plug as accurately as
possible. Together with the higher resolution data captured by the SmartCoil 2, a
higher transmission speed was desired. As Ethernet communications had emerged
as the favoured protocol for industrial databus applications, EMI hardened hardware
was readily available, there was immediate compatibility with both desktop PC and
laptops without additional equipment, and up to 1Gbit/s was very practical, this
technology was adopted as standard for SmartCoil 2.
Using a standard 100M Ethernet card in the control PC, a screen refresh rate of 22Hz
is now comfortably realised with this increased communications bandwidth. As an
interesting aside, and noting that the Ethernet system used is ordinarily intended for
use for point-to-point communications, the SmartCoil 2 could be directly connected to
the user’s network for remote control of the equipment from any site. Remote
diagnostics at APOJEE is also possible by this route too, if permitted.
As an associated point of note, despite the significantly higher screen refresh rate
achieved, not every spark issued by the equipment to the engine will be captured and
displayed. Only really apparent at high engine speeds, this is considered of relatively
low importance; the characteristics and all measures of every spark issued are
captured and stored in the SmartCoil 2 data buffer.
Any ignition cable type may be adapted for use. If a suppressor type is chosen, there
will be considerable loss of energy to the cable impedance, especially at high peak
ignition currents. SmartCoil 2 can compensate for this loss to ensure that the energy
arriving at the spark plug is truly represented by the data in the user interface.
To effect a calibration of the chosen ignition cable, SmartCoil 2 has the facility to
account for cable (and spark plug) resistance. Adding a measurement to the ‘Cable
Losses’ tab found in the ‘Option’ menu and shown in Figure 21 accounts for cable
loss accordingly.
Correction in this manner ensures the delivered energy correlates to the requested
energy ± 5%. If the spark plug is a suppressor type, a measurement of the overall
cable plus spark plug resistance will ensure the closest approximation to energy at
the spark gap.
8 Ion-sensing: Combustion Feedback
Figure 22: An overview of the concept of ion-sensing for detecting flame presence. Examples of
intermediate combustion reactions show the liberation of free electrons and positively charged
ions. In the presence of two electrodes with an applied DC bias voltage, charge movement may
be measured as a current flowing in the resultant circuit. Simplistically, the magnitude of the
measured “ion current” is proportional to flame intensity.
Using the spark plug as the ion-sensor electrodes is a well known solution, and
indeed the original SmartCoil included a basic function of ion-sensing to extend it’s
utility. Following specific customer requests this function was extended on the revised
equipment.
Typical bias potentials are in the range of 50V to 400V. With typical automotive
ignition hardware the resultant ion-currents are of the order of some hundreds of µA.
Many practical circuit designs exist for automotive- (and other) application ion-
sensing. Some achieve remarkable performance from very simple concepts, such as
providing the bias voltage by charging a capacitive storage circuit during coil
discharge. This technique was unfortunately not compatible with the SmartCoil 2
hardware architecture and carried with it certain inflexibilities. For example if a higher
bias voltage were needed for very large gap spark plugs or fuels other than gasoline,
and the delivered spark energy were requested to be very small, there was a strong
possibility the required bias potential could not be realised. Indeed the only sensible
way to implement the technique was considered to be the use of a separate,
configurable DC power supply. A ‘floating’ supply was perfectly feasible, but
protecting the supply and the sensitive ion current measurement circuit from the high
transient voltages of up to and above 50kV in the spark discharge presented
challenges. Incorporating suitable spark gaps with well-controlled breakdown
characteristic were not permitted by local regulations and high voltage diodes were
considered insufficiently durable.
The chosen solution was to incorporate a suitable MOSFET and simply disconnect
the bias supply and ion current measurement circuitry during the period of the
discharge. With an ideal transistor, switching would be instantaneous; in reality two
main constrains were encountered: high and unpredictable residual charge on the
ignition line immediately following the end of the spark, and speed of switching.
Addressing the first of these problems required the inclusion of a ‘security time buffer’
initially set in firmware to a pessimistic 100µs following engine tests under many
spark request conditions. This could be reduced for low energy or short duration
sparks, but a user control is not currently implemented to ensure maximum
protection. As the MOSFET must be floating, a Li-ion battery with a finite life is used,
and this life is a function of switching speed. As such, driving the transistor very hard
to effect a very rapid change of state was not preferred. An alternative of modelling
the switch opening behaviour and subtracting the ion current characteristic from the
final measured data was considered. Using a simple first-order low-pass filter
technique with a cut-off frequency of 7kHz, engine testing showed the result to be
very effective. This choice of cut-off of course precluded the use of the technique for
use as a knock indicator, but the solution was considered acceptable as an initial tool
to gain experience of the equipment’s
The technique is illustrated on a single spark example in Figure 23, taken from a
series of experiments on a single-cylinder gasoline engine behaviour in real test
conditions.
The total delay following the end of a short, but not unusual spark of 200µs duration
in this case was found to be of the order of 450µs. During this period no ion current
data are available and the technique is clearly of limited use, especially in the case of
high speed engines. However, as the solution offered clear and stable real-time
results, efforts were made to find an improved, faster switching MOSFET that would
not overly compromise the important battery life.
At the present time, development in this direction has yielded a solution with a greatly
improved bandwidth. Targeting a cut-off frequency of at least 20kHz without
substantial attenuation is the desired direction for this work. Hardware is currently in
prototype form and is soon to be implemented into the SmartCoil 2 architecture.
Figure 23: An example ion current treatment applied to a single spark. Data collected from a
single cylinder gasoline engine operating around stoichiometric mixture, 5000rpm, 35% throttle
opening and 30°BTDC ignition timing. Sensible ion current data are available 450µs after the end
of the spark. 7kHz cut-off, low-pass filter technique.
9 AC Discharges
At least one coil manufacturer follows this argument [8], having recently published
details of a multiple discharge, alternating polarity ignition system. Reporting an
improved ignition robustness to GDI variables, and an extension of the lean limit of
operation, together with an improvement of the misfire rate under conditions of very
high mixture turbulence, it is clear such a solution would offer this bridge.
Unfortunately, few (if any) of the published ignition systems are available to
customers, which has further driven the need for this function in SmartCoil 2,
especially if the usual flexibility of delivery were to be encompassed as part of the
package. Targets were set.
The four-MOSFET drive described earlier, with its capability for instantly driving the
spark with either polarity, makes the task little more than a software exercise for
SmartCoil 2. The first example of the control achieved, in this case requesting a slow
sinusoidal oscillation of the glow-phase current, is illustrated in Figure 24. Details of
the SmartCoil 2 user interface request and the resultant output by oscilloscope with
external probes are shown. No current probe data were available at the oscilloscope,
so the user interface data are included for reference. The load was a pressurised
spark plug at 10bar. Despite requesting a negative breakdown, immediately followed
by a positive glow-phase current, the desired profile is realised with only minor
difficulties.
The requested current waveform, and therefore the frequency and shape of the
discharge is fully user configurable as for other control strategies of SmartCoil 2.
Work to understand the resultant behaviour of an engine under difficult to ignite
conditions when using this ignition solution is on-going and will form the subject of a
subsequent paper.
Figure 25: 5kHz triangular current waveform AC discharge generated on a spark plug pressurised to
10bar. Spark duration is requested and delivered at 5ms, peak discharge current at 500mA. The
total delivered energy is over 1.1J.
References
[5] V Heise, P Farah, H Husted & E Wolf; High Frequency Ignition System for
Gasoline Direct Injection Engines; SAE technical paper 2011-01-1223, 12 April
2011.
[8] W Piock, P Weyand, E Wolf & V Heise; Ignition Systems for Spray-Guided
Stratified Combustion; SAE Technical paper 2010-01-0598, 12 April 2010.