Teaser Speed Controller

[gwhy!1]

Thanks nieles, At a 20A battery limit on the bike it was fine, I have changed the gate resistors values and the bootstrap cap I have also lowered the pwm frequency a tad this have improved things a little , I really dont think its phase amps killing the fets at the moment the 4 fets that I have blown are all low side fets and all have failed in the manner.

i was only using 4 220uF caps at the time, and no ceramic/film caps so that could have very well been the problem! the rest was stock, as JH schematics
in the next iteration i will also be adding 10k at the gate-source, and a 15v zener diode for extra protection

Snap this is what I am planning to do. When the controller was on my bike it had the full compliment of caps ( as JH schematics ) across the power lines so yes this may well be reason I had no issue with the 20A .

 

[gwhy!1]

JH original circuit diagram:
View attachment Controller board.pdf

View attachment Power board.pdf

 

[Lebowski]

Looking at the datasheet of the MC33033 it's a controller which is very sensitive to hall positions and such.

I'm wondering whether the halls are correctly placed. Maybe something is wrong there but the motor is still
able to start. Power at low rpm, the fets won't survive then but maybe above 4krpm the motor inductance
limits the currents and helps to safe your fets ?



In the plot the three square waves are the hall signals, the 3 sinewaves the motor back-emf voltages.

The 33033 works by using the hall sensors to divide one 0-360 degrees period into 6 slots. For your motor to
run good and to be able to accept power in every slot the highest backemf (sinewave) must be (via the fets)
connected to battery + and the lowest backemf to battery -. The one in the middle is left unconnected (both fets off).
The 33033 has no clue about backemf and depends on the circuitbuilder to connect everything in the correct order
(this is where the 'try all hall combinations' comes from).

Looking at the first slot in this picture (0-20 on x-axis, here 0-127 means 0-360 degrees), it's easy to identify
the motor terminal with the lowest backemf but the one with the highest, this changes right in the middle of
the slot. Choosing either one will make the motor run, but very crappy (and noisely) and wil blow fets (it did ).
Moving the hall sensors 30 degrees would solve all issues.

So, question is, how good is your hall sensor placement ?

 

[bigmoose]

Looking at the timing of the FET driver chip and the way you level shifted the inputs, IMHO your top FETs are still on when the lower FETs turn on. I would recommend delaying the lower FET turn on, but you don't have this capability in software. You are going to have to work an analog fix to delay the turn on of the lower FETs, but have them shut off as quickly as they are now.

If you have a dual channel scope, just look at both gate signals simultaneously. Or a high frequency current probe on the lower FETs source and you should see the shoot thru signature.

 

[Lebowski]

Looking at the timing of the FET driver chip and the way you level shifted the inputs, IMHO your top FETs are still on when the lower FETs turn on. I would recommend delaying the lower FET turn on, but you don't have this capability in software. You are going to have to work an analog fix to delay the turn on of the lower FETs, but have them shut off as quickly as they are now.

If you have a dual channel scope, just look at both gate signals simultaneously. Or a high frequency current probe on the lower FETs source and you should see the shoot thru signature.

would this effect get worse at high motor rpm ?

 

[gwhy!1]

Looking at the timing of the FET driver chip and the way you level shifted the inputs, IMHO your top FETs are still on when the lower FETs turn on. I would recommend delaying the lower FET turn on, but you don't have this capability in software. You are going to have to work an analog fix to delay the turn on of the lower FETs, but have them shut off as quickly as they are now.

If you have a dual channel scope, just look at both gate signals simultaneously. Or a high frequency current probe on the lower FETs source and you should see the shoot thru signature.

I checked the signals ( hi and low drive ) initially after I built the circuit and there did not appear to be a overlap throughout the output range and the signals had very crisp on/off transitions, but I will re-test again and pay a little more attention to what is going on.

Looking at the datasheet of the MC33033 it's a controller which is very sensitive to hall positions and such.

I'm wondering whether the halls are correctly placed. Maybe something is wrong there but the motor is still
able to start. Power at low rpm, the fets won't survive then but maybe above 4krpm the motor inductance
limits the currents and helps to safe your fets ?


In the plot the three square waves are the hall signals, the 3 sinewaves the motor back-emf voltages.

The 33033 works by using the hall sensors to divide one 0-360 degrees period into 6 slots. For your motor to
run good and to be able to accept power in every slot the highest backemf (sinewave) must be (via the fets)
connected to battery + and the lowest backemf to battery -. The one in the middle is left unconnected (both fets off).
The 33033 has no clue about backemf and depends on the circuitbuilder to connect everything in the correct order
(this is where the 'try all hall combinations' comes from).

Looking at the first slot in this picture (0-20 on x-axis, here 0-127 means 0-360 degrees), it's easy to identify
the motor terminal with the lowest backemf but the one with the highest, this changes right in the middle of
the slot. Choosing either one will make the motor run, but very crappy (and noisely) and wil blow fets (it did ).
Moving the hall sensors 30 degrees would solve all issues.

So, question is, how good is your hall sensor placement ?

Thanks Lebowski,
This is a very good observation and Im assuming that the timing is correct with the motor I am testing with (80-100) ( it appears fine no load and runs quiet throughout the range with a 6fet standard controller ), I am using external hall sensors spaced at the 17.14 degree mark but now that Im thinking about it I know thud and others commented that these bigger motors do not run to well with this arrangement ( Maybe this is causing the problems ). I think I will dig out one of my smaller motors to test with and maybe set up the timing with a mechanical 120degree spacing, then this should rule out any possibility that its the motor timing causing the grief.

 

[bigmoose]

would this effect get worse at high motor rpm ?

In my opinion yes, that said without having the hardware to test. As speed increases, commutations/sec increase. If there is a shoot thru each commutation, then that poor FET can only take so much abuse. Then he just can't hold it together anymore and the smoke comes out...

I checked the signals ( hi and low drive ) initially after I built the circuit and there did not appear to be a overlap throughout the output range and the signals had very crisp on/off transitions, but I will re-test again and pay a little more attention to what is going on.

Now I don't want to come off as an elitist, but was the measurement made with a dual channel scope with a bandwidth at or above 200 MHz? You are after 10's of nSec to 50 of nSec overlaps here. I've seen 20MHz scopes "expand" to this range, but the results aren't "real" or "real enough." Also there is a slight delay from the gate condition to the cessation of conduction, hence the reason I suggested a direct current measurement on the source lead. Just some more thoughts, and hope they help.

 

[gwhy!1]

The scope I have is a dual trace realtime digi 60Mhz but to be honest I never have really got on with it, it has all the bells and whistles for the sort of work I used to do many years ago but I also think it is lacking for anything serious with this sort of work. I will see what I can come up with to test and try to get some better measurments.
Thanks BM.

 

[Lebowski]

The scope I have is a dual trace realtime digi 60Mhz but to be honest I never have really got on with it, it has all the bells and whistles for the sort of work I used to do many years ago but I also think it is lacking for anything serious with this sort of work. I will see what I can come up with to test and try to get some better measurments.
Thanks BM.

With a dual trace you can already check the hall setup, especially on an RC motor. Put the motor in a drill press to spin it, then use
your scope to measure a hall signal and the backemf at the same time. Then you should be able to figure out whether the halls
are placed correctly.

I would sya 60 MHz is fine... Moose is right in saying that it won't show fast transitions but I think you can still estimate the delay between two signals, no ?

 

[nieles]

i checked the schematics for the hall board for Phase current sensing, and if you feed it 5v it will work.

it will give you a 0.1-5v signal for 0-max current

the circuit will need an another opamp for the current limit to be adjustable

what was your plan for using the signal? a voltage divider with a 50:1 ratio?

 

[gwhy!1]

i checked the schematics for the hall board for Phase current sensing, and if you feed it 5v it will work.

it will give you a 0.1-5v signal for 0-max current

the circuit will need an another opamp for the current limit to be adjustable

what was your plan for using the signal? a voltage divider with a 50:1 ratio?

Hi nieles,
This would be ideal, yes it would probably have to be a 50:1 ratio , but I dont really know yet what sort of currents we are talking about. Im toying with the idea of mixing the battery and phase current (voltages) then using this resulting voltage as the overall current limit input to the chip, but at the moment I am try to get my head around what effects this will have for the protection of the fets and how the chip will behave.

 

[deVries]

Karma modded his controller using Mineral Oil for cooling & was able to over volt & over amp well beyond the stated specs w/o any failures or apparent visual damage upon inspection. I know this won't solve design flaws, but it might tweak it just enough to save the controller/design if the tolerance to failure is borderline & close enough this might be the trick & treat.
http://www.endless-sphere.com/forums/viewtopic.php?f=6&t=25286&start=45#p467013
[youtube]G_KYIhDHHMg[/youtube]

 

[gwhy!1]

Karma modded his controller using Mineral Oil for cooling & was able to over volt & over amp well beyond the stated specs w/o any failures or apparent visual damage upon inspection. I know this won't solve design flaws, but it might tweak it just enough to save the controller/design if the tolerance to failure is borderline & close enough this might be the trick & treat.
http://www.endless-sphere.com/forums/viewtopic.php?f=6&t=25286&start=45#p467013
[youtube]G_KYIhDHHMg[/youtube]

I must have missed that thread, Looks good. I have just taken order of one of these controllers to add my power boards to ( dont know when I'm going to get round to trying it out tho ) if it works out, it will end up a reliable 10kw controller with 12fets And now it looks like I have a active cooling method ( should I need it ) which may make it a 14kw controller

 

[NeilP]

All you guys in USA...Oz...you are to far away for Gwhy to send them to you for beta testing. I am a lot closer..I'll volunteer on your behalf..just drop a couple to the Serviceair ground crew up at Bristol airport..I'll pick them up from there..I am in there twice a week

 

[gwhy!1]

Ok its been a while but I have finally stacked 2 power boards, I am waiting for a new 12fet controller to turn up as I am going to graft this onto it. I haven't added insulators or the rest of the hardware yet to the stacked boards but you should get the general idea.
View attachment 2

size comparison upto a old broken 12fet controller


The tall caps on the 12fet will have to be either layed down or put onto the rear of the board.