OpenServo.com

[ginge]

Hi all,

I am starting this thread with the aim of opening up the Lynxmotion - OpenServo project to the world. I will start the ball rolling by dropping in the first schematic I have.



This schematic was not designed for the Lynx board, but is more than suitable for that purpose. It will need tweaking to the Lynx specs, but it shouldn't take much. I have not labelled many of the components as yet, as I have not calculated all of the parameters.

The design uses a quad discrete N FET arrangement with the high side driver being an LTC1155 ad the low side an IR4427. The NOT gates control the bridge driver switching to make sure there is no shoot-through issues.

Question/comments/improvements welcome.

EDIT: It has just come to my attention that this design is derived from Andras Tantos of Modular Circuits. More specifically this:- http://www.modularcircuits.com/h-bridge.htm
As this design was resurrected from my design archives, I feel it only fair that I attribute the correct sources to bridge portion of the design to the rightful owner. Also note that the uModule is covered under a non-commercial license, and subsequently this design can not be used verbatim. The changes that Cliff has outlined below mitigate us from any potential trouble inferred by the original license.

Barry

[Cliff]

Hi Barry,

Your schematic is a very good start towards a new Lynxmotion design. Now that I have spent some time reviewing the design, I have a few comments and suggestions. I am assuming that, at this time, a number of component values are just place holders and as such, if I comment on circuit values, I'm not criticizing your schematic values. (That will come later. )

a) My first knee-jerk reaction to the LTC1155 high-side driver chip was: that won't work - its too bloody slow. Then I realized that you had moved the Enables to the high-side switches, which makes the speed a non-issue. However that change does make BEMF measurement more complex. The thought in my OSvX2 design was to be able to stop PWM, leaving the Enable on, make the measurement and then restart the PWM - all without messing with the PWM ports. While BEMF is still doable with your change, it would be better if the PWM / BEMF routines could be the same as for OSvX2. By adding the bootstrap circuit (shown in the data sheet), the PWM could be moved back to the high-side and the same PWM / BEMF code could be used. The only down side I can see is that the PWM would be limited to 10KHz. max., but I don't think that will be a problem, because the larger motors used with this design will have larger inductance. Also, I noticed that the ST Microelectronics motor control chip is limited to 10KHz. max. as well.

b) I have an alternate to suggest for the bridge N-FETs, which is: IRLR7807Z. This part has about half the package size and about half the cost. The on resistance is about the same, but the gate charge and diode reverse recovery time are much smaller. The smaller gate charge requirement, will mean a faster turn on time given the same gate drive. The faster diode reverse recovery time (23-35nS) means the body diode can be used and no external bridge diodes will be needed, because the body diode is as fast as many fast rectifiers and faster than most (25-250nS). I have wondered about use of external diodes in bridge circuits, but recently ran across an old National App. Note (1988) that explains:

[ginge]

Hi Cliff,

[ginge]

Cliff,

I have taken most of you design suggestions on board and whipped up a quick schematic.



Notes:
a) I still have not calculated all of the values. Notably the BEMF divider and the LTC1155 current sense resistor. I am not sure of the best way to proceed on the Back EMF potion yet. I really need to read over your docs again to make sure we get the optimal configuration from a varying voltage source.

b) I have switched back the enable and pwm from the previous inverted setup to satisfy the BEMF requirements.

c) I have added the LTC1155 bootstrap for the PWM conversion. As you stated, we are limited to 10khz maximum now.

d) Changes:- Added the bulk capacitor, but unsure as to package to choose for the layout. I have added the missing Lynx requirement of a trim pot. Changed the xtal for an integrated smt type

I think that is all... for now.

I drafted this up fairly quickly, so let me know if I made any obvious errors. If you think we are proceeding in the right direction, I will check the the layout very carefully.

Barry
_________________
http://www.headfuzz.co.uk/
http://www.robotfuzz.co.uk/

[Cliff]

Barry,

The resolution on your second schematic post is too low to be readable when enlarged. Could you bump the resolution up to that of the first schematic?

Looks like progress - I'll get more comments to you, when I can see the details.

Thanks,

Cliff

[ginge]

Cliff,

The images on my server are 2000x1000px. This small image hyperlinks to the larger one on my server. It is identical in size to the first image.

Barry
_________________
http://www.headfuzz.co.uk/
http://www.robotfuzz.co.uk/

[Cliff]

Barry,

[ginge]

Cliff,

Great work. I have incorporated your calculations into the schematic, and I have taken care of some others too. The BEMF dividers could probably do with readjusting, but the components are available in those values.

Here is the 0.4 version


(once again linked to larger image due to phpbb resize restrictions)

EDIT:
noticed I probably created a problem in the back EMF by clamping the output from the Back EMF switch portion so I am not taking those component limits into account. I will check over the datasheet and set appropriate limits.

Barry
_________________
http://www.headfuzz.co.uk/
http://www.robotfuzz.co.uk/

[Cliff]

Barry,

Here are suggested values for the BEMF circuit:


These values will give the same response time as OSvX2.

The bus switches will clamp the output to the ADC - the diode clamps are to protect the switches from positive voltage spikes greater that 7V at the inputs. The switches already have internal negitive spike protection.

Cliff

[Cliff]

Barry,

Here are some suggestions for the gate bootstrap circuits:


Links to parts used:
BAS40W-06
MMDT2222
MMDT3906
MMBZ5248BS

BTW, it looks like the bootstrap capacitor got dropped on your schematic - you might want to look around on the floor, it must be somewhere near by.

Cliff

[Cliff]

Barry,

A few more notes (what do you mean 'too many notes'?).

a) The bulk capacitor needs to be moved to the battery voltage.

b) I don't know the part numbers for the battery and motor connectors used on Lynx v1, but looking at similar connectors of the same size, I find the current rating to be 6A max. - we're going to have to use a larger version or something else. Ideas?

c) Putting battery voltage on the I2C/Serial and PWM input connectors begs for someone to connect their battery there, which when running a large motor will result in a melted connector. If the battery voltage is desired for output on these connectors, then we should limit the current to what the connectors will handle - poly-switch? fuse?

d) Why do you have a separate AVR-ISP connector? I think that it would better to have a connector set-up that is consistent with, if not the v3 connector, at least the OpenServo v2 connector. I'll let you guess what my preference would be.

e) While I understand the desire for a lower noise measurement, connecting the centering pot to the AVCC voltage will mean that a larger inductor will be needed and if the pot becomes intermittent, it will generate noise that will affect all the ADC measurements. Since this is a pretty static input, I think software filtering and limit (delta) checking would be a better way to deal with this input. That is, until a new value is stable for a period of time, the old value is used. I would recommend putting this pot on +5V.

f) I notice that ADC6 is free and that there is no temperature sensor on board. Even though this board will not be stuffed into a servo case, make no mistake, this board is going to get hot. Jim Fry's vision for mounting, to date, is to encase the entire board in heat-shrink and tie-wrap it somewhere . IMHO that we need to come up with a better plan, but until we do, it would be nice to know when to shut down without waiting for plastic to melt. I would recommend adding the temperature sensor - its zero impact to remove it later if cost is an issue.

g) The low side gate drivers have internal pull-downs, but to be on the safe side you might want to add pull-downs on the PWM signals (no internal pull-downs on the high side drivers).

Enough for now,

Cliff

[ginge]

Hi Cliff, you have been busy

Thanks for the scaling on the BEMF divider.
I found the bootstrap caps ( on the floor ) and added them back in

Component suggestions:
MMDT2222
MMDT3906
MMBZ5248BS
These dual transistors/zeners are difficult for me to get hold of without splitting my suppliers. I prefer using the SOT23 as it doesn't add much to the board space.

a) Done. Any thoughts on a package for this cap?

b) I was planning on using a Molex 5.08mm screw terminal.This one handles 10A and would look up to the job, There are quite a few manufacturers to choose from with this current rating and pitch, so I can't see a major problem there.

Molex 0399800302 - Farnell

The Tyco MTA-156 range supports up to 12A contact and are smaller than the above 200 pitch component.

c) Yeah, it was something I wanted to talk to Mike about. The lynx v1 design had a jumper switch bank to control the power input (output?), and I was hoping to be able to drop this entirely. I just needed to confirm that is was not needed. If not needed we can roll the I2C PWM and serial into one connector.
If it is needed I think a fuse would be best.

d) I got the impression it was a requirement of the design. I would love to have a OSvX2 connector on there as well, but I guess we need to see if we can justify it to the powers that be. I think it is a good idea to have a standardised connector on there for flashing to make it easier on the masses.

e) No problem, I understand what you are saying. IIRC the ADC is already filtered in software so this should not be a problem.

f) Good idea. I think just a couple of holes and provision for an off-board sensor would work best here. That way you can take the thermistor off to the motor if you wanted to. The voltages will need scaling, so I will look over the datasheets for a suitable device. I think something like the LM35 might be a good choice. We will need to work out what temperatures we are dealing with to get a good voltage divider value.

[Cliff]

Barry,

Looking pretty good. We are quickly getting to the point where input from Mike and/or Jim will be needed to continue. Mike was going to send me e-mail describing the requirements / desires, but I guess a hungry spam filter ate it. We really need to have a plan for mounting and heat-sinking this board - heat-shrink and tie wraps are clearly not the right answers. I have some ideas on mounting, but until I get a better feel for Mike and Jim's 'product specification', I can't really take those ideas anywhere.

[ginge]

Cliff,

[Cliff]

Barry,