Cheap electric start for inverter generator using a BLDC module revisited

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
 This is more of a follow up, describing where I got up to in my  
exploration of the subject matter. As such, it's turned into another epic  
read. Apologies for that but there was rather a lot of detail to cover.

 Way back in the mists of time (Tue, 17 Jul 2018 18:35:11 +0000 to be  
precise), I created a thread with the subject : "Electric start for  
suitcase inverter generator by using its alternator as a BLDC motor?".

 In fact that post is so old now, all bar my final follow up to the  
original thread is now expired in my news client (Pan). I've saved the  
content of that final follow up for reference (and posterity).

 In essence, I was seeking advice on how best to create an add-on BLDC  
starter motor module to drive the 6 or 7 pole pair outrunner[1] three  
phase circa 300vac[2] alternator (used to generate (after fullwave  
rectification and smoothing) the circa 400vdc which powers the 50Hz  
230vac sinewave inverter module's class D bridged amplifiers) as a dc  
brushless starter motor from a 12 or 24 volt starter battery pack[3].

 Briefly, it had occurred to me that the permanent magnet alternator used  
in my little 99 quid Parkside PGI 1200 B2 inverter genset could quite  
easily double up as a starter motor given a suitable BLDC motor  
controller (a modern electronic form of the classic "Dynastart"(tm) setup  
as used on some small marine engines around half a century back).

 I was just trying to bring this idea up to date, hence my starting the  
original thread to gather opinions and ideas from a group who seemed more  
likely to have had some experience in the art of designing or using BLDC  
controllers to help me in my mission.

 Rather encouragingly, after starting that thread, I came across a US  
patent application for exactly this idea (patent # US7180200B2) with a  
grant date of 2007-02-20 and a priority date of 2002-06-06 which rather  
validates my "Neat Idea" which I'd wanted to test for myself.

 After getting over my surprise at seeing the unpatentable being  
patented, my next thought was, "If these guys (Richard T. Walter, Michael  
K. Forster and Shailesh P. Waikar) had been looking to expand the market  
for their employer's (Black and Decker Inc) rechargeable battery packs  
way back in 2002 and were granted an actual patent way back in 2007, why  
the Hell weren't we at least seeing electric start as a feature (optional  
or standard) of even the most commoditised of inverter gensets by now?

 After all, the concept is such a no brainer enhancement of the utility  
of portable inverter gensets that the likes of Honda should have been  
incorporating this feature a decade ago and had that "patent" rendered  
null and void on the basis of a total lack of novelty (other than for the  
novelty that anyone would think that assembling a bunch of standard  
circuit modules together to sell more battery packs was an actual  
patentable idea).

 Possibly Black and Decker Inc were just hedging their bets and looking  
to get a cross licence agreement with Honda to act as their official  
battery pack supplier should Honda ever decide to offer electric start  
options in their smaller portable inverter genset range. A lot of the  
existing components in the inverter module could pull double duty between  
50/60Hz 230/120 vac inverter mode and 60/30 vdc driven BLDC switching  
mode with the most expensive addition being a 60 or 30 vdc converter  
module powered from a standard 12v SLA or a 3s or 4s LiPo battery pack  
along with an AS5047 magnetic position sensor bolted to the end of the  
genset's crankshaft to drive the PMG as a BLDC starter motor.

 Anyway, harking back to the original thread, I'd hoped to be able to  
achieve this goal without the complication of adding Hall effect sensors  
to the PMG, electing to rely upon sensorless operation of the BLDC  
controller alone. However, this idea was doomed to failure as I  
discovered when I finally got round to buying a cheap Chinese 50v 16A  
sensored/sensorless BLDC controller module with reversing option and  
analogue speed control input to experiment with.

 The 16.8v (4 cell LiPo battery voltage) 20A rated R/C ESC module I had  
been gifted by my SiL to run my initial tests with (I'd made a simple  
servo tester from a 555 timer chip to drive the ESC module) couldn't get  
the generator to spin even with the spark plug removed, I could hear it's  
initialisation beeps and 'growls' as it attempted to drive the PMG but no  
visible motion came of its best efforts.

 I guess the PMG was too much of a mismatch to the small propeller motor  
characteristics the ESC was designed to work with. The cheap Chinese BLDC  
module had better luck driving the PMG but only with the spark plug  
removed. Even using a 48v volt battery pack (51v) borrowed from my  
SmartUPS2000 which produced a nice and steady 480rpm at 400mA load on the  
51v battery failed to make it rotate through the compression stroke with  
just my fingertip 'sealing' the spark plug hole.

 So, with Lasse Langwadt Christensen's advice about using three HALL  
elements rather than sensorless to overcome the issues of low speed and  
uneven mechanical loading from the engine, ringing in my ears, I  
concluded that only a sensored control that was rigidly locked to within  
a fraction of a degree of the shaft's rotation would be able to cope with  
the job being asked of the BLDC module.

module had been a free gift) on the cheap Chinese BLDC module was  
mitigated by the fact that it did at least possess the necessary Hall  
sensor inputs to allow me to go further with my experiments which left me  
pondering on the viability of adding the necessary Hall sensors to the  

 After searching the internet for more ideas on this, I came across the  
rather wonderful AS5047 magnetic position sensor which, with a suitable  
diametrically magnetised disk magnet on the end of a motor shaft, could  
recreate (after calibration) the effect of having three perfectly aligned  
Hall effect sensors configurable for motors with from one to seven pole  
pairs (as well as absolute position data) accurate to a fraction of a  

 The claim made by AMS, the manufacturer, of "Superior accuracy and  
stability over the traditional three Hall sensor setup" seems to be well  
justified as its use in a powered skateboard adaptation to a VESC module  
demonstrated (just google AS5047 + VESC to track down the relevant youtube  
video) so this would seem to be the way to go forward on this project,  
especially if I can persuade AMS to supply me with a sample of the  
AS5047D-TS_EK_AB, Adapter Eval Kit (with chip and a 6mm dia magnet) for  
free. :-)

 I've just checked AMS's website and it looks like only the chip itself,  
priced at $6.50 (damn Opera's VPN - I forgot to turn it off, I'm sure  

option. The AS5047D-TS_EK_AB, Adapter Eval Kit (with chip and a 6mm dia  
magnet) only seems to be available at a single unit price of $15.75 - ISTR  

 Still, considering the undoubted performance improvement it will bring  
to my project, a modest doubling up of the BoM costs for a relatively  
trivial and the most doable of hardware modification possible, doesn't  
seem all that high a price to pay this close to the end of the day.

 The fact that, without the compression loading (plug removed), I was  
able to get it cranking at a steady 480rpm using a 51v battery pack (just  
about spot on with my initial back EMF to speed ratio estimates) was a  
rather encouraging result. The only fly in the ointment so to speak,  
being that, as per Lasse Langwadt Christensen's suggestion, the project  
is proving to be a lot harder to implement than my initial optimism had  

 The irregular torque loading is proving to be far too demanding for a  
wishy washy sensorless control algorithm that's trying to guess when to  
commutate to the next step from zero crossing detections alone that  
require a minimum (and a not too wildly varying) speed for any success.

 I'm going to try and solve the problem of mixing a 50v rated BLDC with a  
400v PMG by using a couple of small 12vdc coiled 240vac 10A rated DPCO  
contact set with a 25ms release time spec relays to isolate those two  
component parts in a hopefully timely fashion. I'm guesstimating I'll  
have about 50 to 60ms grace before the voltage rises to a dangerous (to  
the BLDC module) level once the engine starts up.

 I also reckon I can speed up the relay release time by some 5 to 10ms by  
using an RC network with a high voltage switching transistor to give an  
operate impulse that will reduce the hold in current and allow the use of  
a high value snubber resistor to speed up the inductive discharge for a  
snappier decay of magnetic pull on the relays' armatures. There'll still  
be good old fashioned mechanical inertia to contend with but at least it  
won't be aggravated by any more electrical inertia than I can minimise  
with my drive circuit.

 Furthermore, I can also switch a low value resistor across the output of  
the 50v dc-dc converter I plan on using in the finalised design to load  
down the PMG voltage being rectified by the body diodes in the 3ph Half  
bridge transistors in my 50 volt BLDC module. Initial tests using a 48v  
SLA battery pack won't need this refinement since the battery will be  
able to handle any such short lived charging events.

 All of this extra complication with isolator relays could be neatly  
avoided as per Lasse's suggestion of using a high voltage 3ph half bridge  
driver module with a simple 800v 8A blocking diode between the driver  
module and the dc-dc converter's output but, unfortunately, such a  
solution doesn't come cheap. The 50v BLDC module and a pair of isolating  
relays is a pragmatic choice aimed at keeping the BoM costs to a minimum  
for what I hope will become a a project that will appeal to the more  
adventurous DIY enthusiast.

 I'm sure that if the inverter genset manufacturers (including those who  
make commodity versions) see the light, they could easily integrate the  
required high voltage gubbins into the existing inverter modules for very  
little cost, leaving only the fitting of a AS5047 magnetic position  
sensor onto the end of the generator's crankshaft as the single most  
costly BoM item (with the 50 to 60 volt DC-DC converter integrated into  
the main inverter module being the second most costly item).

 If the manufacturers modify the inverter modules, they can save the cost  
of the sensor module by offerng recoil start only economy versions to  
retain the cheaper end of the market whilst offering their customers the  
option to purchase an electric start upgrade kit[4] at a nice profit.  

 If they, the manufacturers, play their cards right, they might also be  
able to offer their customers a LiPo or LiIon battery pack upgrade to the  
time-you-need-it' SLA battery. :-)


[1] I have yet to determine the actual pole pair count. All I know at  
this stage is that a pole pair count of 6 or 7 is fairly typical of the  
class of alternator used by portable inverter petroleum (gasoline)  
powered generators. It's a fairly trivial (though a bit of a fiddly)  
exercise in "Pole Counting" which I haven't yet gotten round to doing.

[2] The actual rms voltage required depends on the waveshape which might  
be more trapezoidal than sine in character (trapezoidal would be more  
suited to producing a DC voltage after full wave rectification with  
minimised ripple content). The important point being, regardless of wave  
shape, that the controller/inverter module lands up with a circa 400vdc  
rail to power its 50Hz 230vac inverter stage.

[3] Even 24 volts may be insufficient. Testing with a 51 volt battery  
pack provided a speed of 480rpm with the spark plug removed using a cheap  
(12 quid delivered from China) 50v 16A rated sensorless/sensored BLDC  
module designed to drive electric scooter motors and the like, rated up  
to 360W.  However, just blocking the spark plug hole with a fingertip was  
enough to stall the controller when relying on its sensorless control  

 I'm pretty confident that adding an AS5047 magnetic position sensor with  
suitable magnet attached to the end of the crankshaft of the engine/
alternator assembly to synthesise the required Hall sensor commutation  
signals (more accurately and reliably than the typical Hall sensor setup  
on sensored BLDC motors) will allow that cheap BLDC controller to pull  
the engine through the initial compression stroke (even if perhaps rather  
arthritically if it happens to be stopped in the wrong position) allowing  
it to reliably accelerate to a respectable cranking speed in time to  
crank through subsequent compressions with reasonable alacrity.

 The AS5047 magnetic position sensor modification is the most practical  
way to upgrade the alternator to an accurately commutated hall effect  
sensored BLDC motor. I just need access to the pull starter end of the  
crankshaft rather than have to pull the flywheel/outrunner off the  
crankshaft in order to wedge three hall effect sensors into the stator  
assembly and just hope for the best with fingers crossed.

[4] The newer models of electric start gensets will, of necessity,  
include the mechanical modifications required to readily accommodate the  
sensor board and magnet. The cheaper pull start only variants will simply  
be lacking the sensor module, requiring that the customers who later opt  
for the upgrade merely have to bolt the module onto the end of the  
generator housing and plug a multi-core fly-lead into the waiting  
connector on the inverter module (and perhaps install a starter switch/
remote control sensor assembly to complete the upgrade).

 The manufacturer gets to both eat and have their cake - the potentially  
gone to waste unused components in the inverter module would be a very  
small cost investment against the very strong possibility of selling a  
highly profitable electric starter upgrade kit to their customers.

Johnny B Good

Re: Cheap electric start for inverter generator using a BLDC module revisited

Quoted text here. Click to load it

use thyristors to short-circuit the generator if the voltage gets too
high. generators arw inherently current limited, so nothing will explode
(but you might stall the motor if you leave it shorted for too long)

Quoted text here. Click to load it

they'd put sensors on the end of the stator where they can see
the rotor's field

Quoted text here. Click to load it

they might put two sets of pads and install different parts on the
cheap model.

  When I tried casting out nines I made a hash of it.

Re: Cheap electric start for inverter generator using a BLDC module revisited
On Wed, 31 Oct 2018 06:45:07 +0000, Jasen Betts wrote:

Quoted text here. Click to load it

 I was considering that sort of approach using a BFO "zenner" as a  
backstop protection against 'failure to release in time/ever events' with  
the isolating relays. The BFO zenner being a hefty power transistor  
driven by a regular zenner with a high wattage resistor as a collector  
load - I'd used that circuit over three decades ago to replace the 12v  
power zenner on a heatsink used by the later Triumph Bonneville  
motorcycles, to good effect.

 The one ohm collector resistor consisted of a yard or two of pvc single  
core (0.5mm I think) wrapped around one of the steel frame tube sections  
to create an effectively heatsinked high power load resistor. The idea  
being that when shunting a full 12A (more than the simple single phase PM  
alternator and rectifier pack could deliver) the transistor would only  
have to handle the 12A with maybe half a volt or so Vce drop (6W) with  
the maximum dissipation being at half that current, around the 36W mark  
in theory.

 In this case, I'd have been looking to use something like a couple of BFO  
6 ohm 100W rated resistors each in series with a 2N3055 or similar  
plastic power transistors in parallel across the BLDC module's power  
rails with a 56v zenner to drive them before realising I could forget the  
zenner altogether since I'd be shutting the 50v dc-dc converter down -  
detecting the rising back emf voltage within a millisecond or so is easy  
enough to initiate the starter abort process when I won't be requiring  
output from the converter anyway.

 Just switching a BFO resistor across the BLDC module's supply rail will  
be all I need at that point to save it from excess voltage should the  
relays get stuck or I've overestimated that 50 to 60ms grace time for the  
transition from motoring back emf to that of an accelerating PMG hitting  
the danger point voltage threshold. It is, after all, just a safety  
backstop measure to ensure that the cheap 50v rated BLDC module remains  
protected from such overvolting events which should normally never arise.

 By my reckoning, I estimate a load resistance value of 10 ohms will be  
enough to gently stall the engine before it can generate 50v of back emf.  
I don't want to be trying to short out the 50 volt charge on a 470?F  
capacitor without some form of current limiting and I also don't want to  
be rapidly stalling the engine when I might just want to briefly limit  
its rpms during transients from delayed contact opening events which may  
prove to be an inherent (if low probability) feature of the system (that  
genset may well have a better 0 to 60 volts acceleration figure than my  
best guess).

Quoted text here. Click to load it

 Yes, that's a common method but it does leave the sensors exposed to the  
distorting effects from the stator phase currents. It's this which  
probably accounts for the vastly improved commutation accuracy provided  
by the AS5047 sensor setup. The better design of HE sensored BLDC motor  
use a completely separate set of magnets, mirroring the ones used in the  
rotor, simply to eliminate the polluting effects of the stator's magnetic  
fields produced by the drive current pulses.

Quoted text here. Click to load it

 Ah, yes; the rather sad exercise in "Cutting one's nose off to spite  
one's face" approach to pissing your customers off in the (sadly for us,  
not so vain) hope that they'll stand still for this trickery to get them  
to 'upgrade the expensive way'; so sad and so frustrating. :-(

 Unfortunately for their more discerning customers (the likes of you and  
me), but fortunately for the beancounter mentality that drives such  
companies' marketing policy, there is a more than ample supply of the  
"one born every minute" 'mass consumer' available in their targeted mass  
market demographic for them to make this policy pay.

 As far as most manufacturers of mass consumer goods are concerned,  
they've pretty well all come to realise that it doesn't pay to offer  
excellence in product design when barely adequate mediocrity will  
suffice. A fact of 'manufacturing life' that seems to have become swiftly  
apparent to Microsoft once they had gained the attention of a wordwide  
mass consumer market demographic with their blinged up "Idiot son of  
win2k" version of windows, aka winXP.

 It's easy enough to moan about modern manufacturing and sales practices  
such as this, since it's everywhere around us. It is, rather sadly,  
simply an annoying and irritating fact of modern life.

 I'll freely admit that the reason I'm "wasting my time" on this project  
is for my own pleasure in achieving something that seems to only exist  
elsewhere as an 11 year old patent application and in the mind of a few  
other enquiring minds but which has never been implemented by the usual  
suspects (Honda et al) and because I believe I may actually succeed where  
others have seemingly failed or else declined to publish their success.

 The "Doing it for the benefit of all mankind" factor is there as a  
secondary consideration in looking to publish the project details for  
consumption by/inspiration for dedicated DIY tinkerers to play around  
with. TBH though, this is more to be able to bask in the warm glow of  
admiration from my peers for publishing the details for all to use and to  
ponder on the absence of genuine innovation in modern day  
manufacturing. :-)

Johnny B Good

Site Timeline