BLDC controller - Page 102 - RC Groups
Thread Tools
Jan 18, 2007, 03:36 AM
The blade numbers go up to 11
stumax's Avatar
Trevor, do I win the wonderful lounge suite? Do you really need such a large coil for a BEC for a low voltage esc? I wouldn't have thought a model that size would need too much current for the servos? I hit a bit of a stumbling block with my big esc design in that providing 12V for the fet drivers (IRF2110) aint so easy when you have 12S lipo voltage. I've gone for a MAX5033, although I'm not sure if the drivers get the gate drive power from the 12V supply or from the motor supply itself and store it in the caps.

I'd love to hear about soldering the direct fets - I may look at them for my esc, although cost wise they aren't as good as to262 or to220, which have great heatsinking capability. Did you hand solder them? How about using a reflow skillet?

Stu.
Sign up now
to remove ads between posts
Jan 18, 2007, 03:56 AM
Registered User
Trevor_G's Avatar
Quote:
Originally Posted by pentium
But when compiling (AVR-studio) i get an error message saying:
Illegal use of register 'state1' as a label.

Is it right that the source of 3.03 has this bug ? I renamed the label to 'state1_1' and in all the rjmp and calls, and then it compiled with no problems.
I will leave Quax to comment on his board but the error message is easy.

Early versions of the ASM in AVRstudio would allow you to use the same name for a register and a label. Which is how Quax got away with it. However, later versions of the ASM treat this as bad practice and flag it as an error. So your fix is fine and you will need to look elsewhere for your problem.
Jan 18, 2007, 04:31 AM
Registered User
Trevor_G's Avatar
Quote:
Originally Posted by stumax
Do you really need such a large coil for a BEC for a low voltage esc? I wouldn't have thought a model that size would need too much current for the servos?
I would use a different BEC design for a high voltage ESC i.e. one that NEVER lets the battery drop below 10V. With such a ESC a buck regulator would be the best solution and would give a really small inductor. The buck regulator will also give a higher efficiency.

The ESC I have produced should be considered as the low voltage ESC. This requires a flyback convertor employing a coupled inductor (two windings) to which I have added a third winding to power the Mosfet Gate Drivers(MGD's). The core I use is an EP7. I think an EP4 would do the job but I couldn't source one. The flyback convertor is undoubtedly a far more complex design than the buck regulator and should only be used if really neccessary.

In my case the spec. I produced was as follows:
I/P 4.5V->25V
O/P 5V @1A + 12V(+/-2) @7mA

In practise this spec can be met in full for I/P's between 4V and 30 V. Unfortunately the UVLO on the regulator means that an I/P of 4.3V is needed to start it .

Trevor
Jan 18, 2007, 05:43 AM
Registered User
Trevor_G's Avatar

DirectFET Soldering


Before using DirectFET's make sure that you have seen all the documents on the IR site including the reworking video. I particularily like the bit about puting the board in an X-Ray machine to check the joint ( now I know I have one somewhere ).

There are some other points that need to be considered:

1)DirectFETs give direct access to the chip which is glued onto the metal case. I assume they use a silver loaded epoxy which softens at soldering temperatures. While this is
essential to resist thermal stresses during soldering it is also the source of many problems.

2) The metal case is only 0.4mm thick and the device is only 0.7mm high. Apart from the handling problems this means that any other component on the board will be higher in consequence a heat sink across the MOSFETS must not cover any other type of component.

3) The chip is presented as two MOSFET's with common Drain & Gate but 2 Sources. This can be useful during testing.

4) The shape of the metal case means that Drain/Source solder bridges are VERY difficult to avoid .

HAND SOLDERING

Hand soldering is possible but even with two soldering irons very difficult.

Firstly you must mask connections. The only soldermask I can find is the latex type. This is rather thick and I usually thin it with water. The result is messy but it works. I have tried other varnish and conformal coating material but the solder creeps under them.

Next you need to apply solder to the board. As I do not have any silk screen equipment I don't use solder cream. I simply tin the pads and then use solder braid to adjust the level. Too much and the solder will bridge to the case, too little and you don't get a joint (0.1 to 0.2mm is about right).

Wipe a thin layer of flux on the tinned pads and place the DirectFETs in position (they need to be pinned in place with a hold down device).

Then you need to apply a soldering iron to each end of the case until the solder reflows.

Finally you need to test. This is usually followed by desoldering and repeating the above.





Stats:
30mins/MOSFET


50% will die during soldering a further 20% will die as soon as the board is put into service


The solution ( a great many DirectFETs died before I got here):



HOT AIR SOLDERING IRON + Board manufactured with Solder Mask & VIAs

The pads are tinned, fluxed and the DirectFETs positioned as above.

Hot air soldering iron, set at 280C, held above DirectFET until reflow occured.

Test as above.

Stats only one board done this way so far) <1 min/MOSFET only one MOSFET developed a short and had to be reseated. No fatalities . Oh! and I used lead free solder on this board.


TESTING

Visual inspection and continuity checking to the gate & source pin are out! The regime I use is as follows:

1) Check for shorts between Gate, Drain and Source

2) Use a capacitance meter to check between Gate and Source track an IRF6635 will give a reading around 5600pF. A value of 2800pF means one of the sources is not connected. Less than that and you have both sources and/or the gate unconnected.

3) As a final check when the board is completed check the ON resistance of each MOSFET.


A lot of DirectFETs died to discover the above and I have made no attempt to list all the things that failed.

When they are working there is nothing to touch them, certainly not a TO220 device, otherwise I just would not bother. The future is another matter

Trevor
Jan 18, 2007, 06:04 AM
Registered User
pentium's Avatar
Quote:
Originally Posted by Trevor_G
I will leave Quax to comment on his board but the error message is easy.

Early versions of the ASM in AVRstudio would allow you to use the same name for a register and a label. Which is how Quax got away with it. However, later versions of the ASM treat this as bad practice and flag it as an error. So your fix is fine and you will need to look elsewhere for your problem.
Thanks for info.

I think i can handle the rest from here. I am afterall the only one who knows how the controller is soldered.
Jan 18, 2007, 06:25 AM
Registered User
quax's Avatar
Quote:
Originally Posted by pentium
Thanks for info.

I think i can handle the rest from here.
....
I think you should give the newer versions a try, they handle the startup better
http://home.versanet.de/~b-konze/blm...oftware4xx.htm

Ciao
Quax
Jan 18, 2007, 03:26 PM
Registered User
pentium's Avatar
Tried the 4.04 now. Still problems. Like drive with the handbrake drawn.

Looked in the parts list again and saw something written below in small letters. 1% resistors. Ahaaah.
Jan 18, 2007, 03:49 PM
The blade numbers go up to 11
stumax's Avatar
Trevor, thanks for the info! I must admit I'm a little scared to try prototyping with them now! I may design a power stage pcb and get it assembled for me - I have much better things to do than frig around with shorted tracks etc. I wonder how difficult they are in a proper production environment? You would think IR would have their act together well enough not to design a chip that's impossible to use in manufacturing efficiently. Comparing them to a to220, I can get to220 fets with much lower on resistance (I use an IRF2907 currently), and their giant tab is very good for keeping them cool. I remember I did the exercise working out board space vs resistance vs component cost and the to220 won all counts, however they are a pain in production. For a 3kW esc, I can afford the extra weight and bulk of a to220, and the low production numbers (<1K per year) probably tilt it in their favour as well - I don't particularly want to spend a long time just working out how to manufacture a pcb. Having said that, I would like to give them a try some time to see if the benefits are real.

Stu.
Jan 19, 2007, 06:02 AM
Registered User
Trevor_G's Avatar
Stu, I suspect that in a fully equiped production environment the problems go away. But I would not be surprised if even there they cause the odd headache. The real problem is that in a prototype environment they are a complete pain.

On TO220 I suspect our views differ because I am happy with 30V devices while you clearly are looking to run at higher voltages and as you say weight is less of an issue.

As I mentioned earlier their are some devices from Infineon BSC016N03LS and OnSemi NTMFS4833N that are worth a look. These are once again 30V devices. They are in SuperSOT8 & SO-8FL packages which appear to be identical. They look as though they would present no prototype assembly problems. At the moment I have not been able to find a source for either.

Trevor
Jan 20, 2007, 12:03 AM
Registered User

Zero cross sensing while PWM


The PWM control is based on sampling of BEMF and Zero Cross control.
My friend told me the way of the zero cross over sensing while PWM as attached.

The sample schematic is:
https://www.rcgroups.com/forums/show...postcount=1239

Takao
Last edited by Takao Shimizu; Jan 28, 2007 at 06:55 AM.
Jan 22, 2007, 08:50 AM
Registered User
Trevor_G's Avatar
Quote:
Originally Posted by Takao Shimizu
The PWM control is based on sampling of BEMF and Zero Cross control.
Takao,

The more usual way of doing this is to switch on the high side and hold the low side on. That way when the PWM signal is OFF the BEMF passes through O volts. Which is much easier to handle than your system which has BEMF passing through Vcc.

I am not clear how your system operates when the lowside switch is off as the LM393 needs its inputs to be <Vcc-1.5V. In your circuit you will not meet this condition.

Trevor
Jan 22, 2007, 06:59 PM
Registered User
Trevor,

>The more usual way of doing this is to switch on the
>high side and hold the low side on. That way when the
>PWM signal is OFF the BEMF passes through O volts.
>Which is much easier to handle than your system which
>has BEMF passing through Vcc.

I should try it. Thanks.

>LM393 needs its inputs to be <Vcc-1.5V. In your circuit you will not meet this condition.

You are right. LM393 uses PNP inputs configulation.
I use a comparator in MPU with divide resisters as the schmatic shows.

Takao
Last edited by Takao Shimizu; Jan 22, 2007 at 07:10 PM.
Jan 23, 2007, 02:48 AM
Registered User
Trevor_G's Avatar
Quote:
Originally Posted by Takao Shimizu
I use a comparator in MPU with divide resisters as the schmatic shows.
Takao,
Are you sure you this is the correct schematic? The one shown clearly shows an LM393.

I assume that in this design the motor supply voltage is the same as the MPU Vcc. Otherwise you would need a resistive divider for the +ve i/p to the comparator.

Trevor
Jan 23, 2007, 03:57 AM
Registered User
quax's Avatar
Quote:
Originally Posted by Trevor_G
Takao,

The more usual way of doing this is to switch on the high side and hold the low side on. That way when the PWM signal is OFF the BEMF passes through O volts. Which is much easier to handle than your system which has BEMF passing through Vcc.
....
Normally the high side FETs have a higher rds-on and it's better to switch the low side for better efficiency. There is no problem at all, to let the power of the high site on, I did it, beginning with my first version. I can scan the comparator (intern or extern doesn't matter) all the time, even if the low site FET is switched off. Only a little disturbance of the signal during on/off switching happens. With a simple double scan, this problem can be solved.

Ciao
quax
Jan 23, 2007, 06:37 AM
Registered User
Trevor_G's Avatar
Quote:
Originally Posted by quax
Normally the high side FETs have a higher rds-on and it's better to switch the low side for better efficiency.
With a p-chan/n-chan design one side may switch faster than the other giving a small efficiency gain. This is not a direct effect of the value of RDSon but the lower gate charge associated with high values of RDSon. Better gate drivers would probably achieve the same results on both sides. With n-chan/n-chan designs the FETs are usually the same type so there is no difference in switching speeds.

My point was that using high side switching measurements are made relative to the Ovolt rail, rather than the + volt rail. This reduces noise problems and simplifies comparator circuitry.

To take some of the problems:
1) Most of the comparators and op-amps available use pnp i/p stages and as a result i/p can go below OV but will cease to operate well before they reach the +volt rail.
2)If the +volt rail is greater than the Vcc of the MPU then some sort of level conversion is required.
3) The +volt line has a lot of noise on it which needs to be dealt with.

All these problems can be overcome but designing a BLMC is difficult enough why add to the complexity, and component count, when there is no advantage ?

Trevor


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
For Sale Quadframes 2 Axis BLDC Gimbal V2 w/RCTimer Alexmos Controller ramdmc Aircraft - Electric - Multirotor (FS/W) 1 Dec 09, 2016 11:57 PM
Help! Motor control for BLDC - Gimbal the_g Multirotor Drone Talk 1 Sep 09, 2016 09:36 PM
Help! DIY Sensorless BLDC Controller gdelf DIY Electronics 0 May 16, 2016 12:04 PM
Help! Control BLDC gimbal motor with arduino an L6234 slu123 Power Systems 0 Mar 06, 2016 09:43 AM
For Sale Quadframe v2 BLDC gimbal w/ Centerpiece Gimbal controller Mactadpole Aircraft - Electric - Multirotor (FS/W) 0 Jan 31, 2016 04:28 PM