Wednesday 1 March 2017

EGT Instrumentation & Dyno Calibration

EGT Instrumentation

To try and close out the base fuel & spark calibration once & for all, a deadline of mid February was imposed by booking a track day at Mallory Park. That provided some additional incentive to get the bike on a dyno in January to do the base calibration.

First, a few small loose ends on the electronics side needed to be tidied up. One of these was to try and implement EGT measurement to provide a second metric to determine optimum fuelling and make sure exhaust port temperatures were under control.

Initially the plan was to simply read EGT values and display them on a screen live and then keep an eye on them while running the bike on the dyno.

However, there would be several advantages to being able to feed the data into the ECU so that it could be logged with all other engine parameters. Unfortunately, due to the limited I/O capacity of the Microsquirt ECU, the only way of getting four additional data streams into the ECU would be via CAN-Bus. That meant using a microcontroller to read the EGT data from thermocouple amplifiers, arrange the data into a CAN message in a format that the Microsquirt could read and load the message onto the bus. 

To prove the concept, an Arduino Uno and CAN Shield were used to get the data processing & CAN code working together properly. The Megasquirt CAN-Bus protocol was the most difficult part of the task as it is quite different from the standard 11-bit or 29-bit header protocol that I am familiar with.

Once the hardware and code combination was proven, a single PCB was designed and manufactured which used an Arduino Micro as the controller and contained the thermocouple amplifiers, CAN controller, CAN transceiver and other peripherals that were required to make the board work on the bike. The finished board is quite large due to being conservative about component spacing to allow hand soldering of components and only used one side of the board for component placement. For future PCBs, component spacing could be tightened up, a dual layer PCB and a microcontroller with inbuilt CAN controller could be used to significantly reduce the board footprint. 
A protective case for the completed board was 3D printed.

EGT CAN-Bus Module Board
Cased Board

On the instrumentation side, 1.5mm K-Type thermocouples were installed as close to the exhaust ports as possible using compression fittings without interfering with the fitment of the radiator or making exhaust fitment difficult. The small diameter probes would allow shorter temperature stabilisation but at the expense of thermocouple life. This was determined to be a reasonable compromise as the main purpose of the thermocouples was to monitor temperatures during dyno runs and short road tests only. As they would not be used for long term control, durability was not deemed critical.

Thermocouple Tip Position In The Exhaust
Thermocouples & Flanges In Place


Exhaust Header Flange Upgrade

While the exhaust was removed for drilling & welding, the opportunity was grabbed to rectify the much annoying issue of the exhaust header flanges bending while the nuts were being torqued up on the studs.

The problem with the Honda design is that the flanges do not clamp between the nut and the cylinder head but rather the flange is designed to be clear of the cylinder head when fully torqued up so the flange has a tendency to bend around the header collar as the nuts are being torqued up. This may not be an issue if the specified torque was always adhered to but on both sets of headers I have owned (OEM & TSR), the flanges had been bent previously. Bent flanges mean that more preload on the stud is required to achieve the same torque which only amplifies the issue and risks stripping the threads in the cylinder head. The studs also bend as the nut tries to sit flush with the flange which can make removing the flanges difficult.

To try and combat this issue, a new set of flanges were laser cut from 316 stainless steel plate. Compared to the TSR exhaust flanges, the new design uses thicker material (8mm vs. 6mm) and also adds c.2mm of material around the outer profile. These changes have the effect of making the new flanges approx. 3 times more resistant to bending at their weakest point compared to the TSR flange.

Exhaust flange comparison, Redesign (left) vs. TSR (right)


Dyno Fuel Calibration

With the EGT instrumentation in place it was time to get the bike on the dyno and complete steady state fuel calibration.

The dyno was an eddy current braked dyno which allowed any speed and load to be held for a period of time to ensure engine conditions could stabilise at each calibration point. The only downside was that as it was a car dyno designed for much higher power vehicles, it was impossible to hold the engine at the lower speeds and throttle angles. Therefore only the area above 4,000rpm & 10% throttle angle could be successfully mapped. As this represents the area where the majority of riding is carried out then that was not much of a problem.

After the fuel mapping was carried out, an attempt was made to see if there was any additional power to be had in the ignition timing. The bike seemed to be very insensitive to part throttle, steady state ignition timing changes although this can be very difficult to judge on a chassis dyno, especially one with quite high inertia.

A few full throttle pulls from 8,000rpm to 16,000rpm were carried out with varying ignition timing to determine what effect it had on engine power. Baseline timing was the Bluefox ECU timing. It was found that a global timing offset of -2° produced a negligible change to engine power. Both +2° & -4° global ignition timing offset produced measurable power losses across the engine speed range. Given that the Bluefox ignition map is c.1.5° more advanced than the OEM Honda curve in that region, it suggests that Honda did a pretty good job of mapping the engine to MBT timing from the factory.

For anyone who cares about these things, the final figure on that particular dyno was 37 bhp measured at the rear wheel at 14,700rpm. No gains over the standard bike were expected with EFI and given the engine is in an unknown state of repair, it was considered a reasonable result. The bike was never dynoed before the EFI conversion and even if it had been, the comparison wouldn’t have been possible on the same dyno and therefore not comparable.

It became apparent from studying the power & torque curves was that the original 18,000rpm rev limit is totally unnecessary as power drops off quite sharply after the peak. A graph of driving torque in each gear shows that there is no point in going faster than 16,000rpm in any gear as, above 16,000rpm, there will be more torque available in next gear up. As such, following the dyno testing, a soft limit at 16,500rpm and a hard limit at 16,700rpm was imposed with the aim to shift up at 16,000rpm.


Calibrating The Bike On The Dyno

The dyno run completed base fuelling & spark for a given barometric pressure and manifold air temperature. All other starts and runs from here on out will help to apply appropriate air temp & barometric pressure corrections and dial in the transient fuelling corrections.



 
Trackday

As planned, the bike was taken on a trackday at Mallory Park after dyno testing to see how it would manage. Unfortunately the chosen day turned out to be cold, wet & snowy but still proved to be a great test of the bike’s rideability with the EFI system.

It was also hoped to use the track day as an opportunity to get a lot of data logging done to help dial in the transient fuelling but as luck would have it, the brand new lambda sensor installed died within minutes of the bike being started so no fuelling data was recorded throughout the day.

Despite this setback, the bike rode really well on the track and throttle response & power were as good as could have hoped given the lack of any form of transient fuelling corrections.

Trackday

Tuesday 21 February 2017

The Search For Electrical Overhead

Due to the additional electrical load being placed on the bike from the EFI system, the charging system needed some attention. It was suspected from the very beginning of the project that the original reg/rec may not able to cope with the constant drain on the system from a high pressure fuel pump and twin headlights. This became very apparent as the bike was being used more and more recently.
The battery could not be kept charged when either the dipped beams or main beams were switched on. There was a definite drop in battery voltage depending on whether the lights were switched on or not and it was noticed that the reg/rec was getting quite hot after night-time rides so it was likely that the reg/rec was going to fail sooner rather than later.

A new Shindengen FH020AA mosfet reg/rec had already been purchased for an upgrade further down the line but given the circumstances, the upgrade was moved to the top of the priority list. The installation of the new reg/rec was relatively straight-forward. The FH020AA is quite a bit bigger than the OEM reg/rec but still fits under the fairing in the same place. Just about. One additional hole had to be drilled in the mounting plate to accommodate the wider hole spacing of the FH020AA and the standard connector was chopped from the loom and replaced with the dual connectors required for the FH020AA.

FH020AA vs OEM mc22 reg/rec

An immediate improvement in voltage regulation and stability was observed with the FH020AA fitted and the reg/rec merely ran warm to the touch as opposed to hot. Although, since the reg/rec can only improve regulation efficiency and can not force the stator to provide any extra power, the bike was still a little short of electrical headroom. It was not bad enough to risk running the battery flat during a long night-time ride but rather meant that battery voltage with the lights on at idle tended to hover around 12V as opposed to the 14V which was observed at idle with no lights. When a current clamp was placed on the battery earth strap, it became apparent that with lights on, the predominant flow of current was out of the battery. As such, a way of either a) boosting the generated power or b) reducing the electrical load on the bike had to be found. Of the two, option b) was clearly the preferred route. Given the headlights represented the single largest current draw on the electrical system, they were the chosen target for modernisation.

After much deliberation, research and changing of mind back & forth, it was decided to install LED bulbs. The options were either HID or LED but eventually the LED won out due to being actually able to package the ballasts in the front of the bike, being more adjustable for beam pattern and having lower overall current draw although they were almost twice the cost of the HIDs for a quality set from a reputable manufacturer.

Each H4 LED bulb is rated at 20W compared to the 60W/55W rating of the standard halogen units so, in theory, a set of two should have freed up c.6A capacity which would be more than enough to cover the additional load the fuel pump, lambda heater & injectors placed on the system. Before installing the LEDs, a charge current of -7A (draining battery) at idle with dipped beam on and -6A at idle with main beam on was measured. With the LEDs installed, charging current was measured at +3A at idle with both dipped and main beams on, leaving ample headroom on the charging system.

The physical installation of the bulbs was reasonably straight-forward. The heatsinks on the back of the bulb made it all a bit bulkier than the standard setup but the heat sinks are well hidden given they are black in colour. The two small ballasts found a nice place to sit either side of the clock stay bracket where they do not interfere with anything else. Getting the beam pattern right took the most time but the adjustment in the bulb housings worked a treat. There is also the advantage that the LED headlamps are quite a bit brighter than the halogens which makes night-time riding quite a bit more enjoyable.