OpenServo.com

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

Andy, I don't think having a board that can support the pinout of the ATtiny45/85 and ATmega is possible. However, I wouldn't want to discourage anyone from trying if they have more insight into this issue.

However, I think the new layout does have some room for N-channel transitors to control the P-FETs. The very small packages you found the other day would likely work well. I'll probably create a version of my new layout to support this to at least experiment controlling all four channels of the H-Bridge rather than two channels in the current design.

-Mike

RifRaf · Joined: 04 Mar 2006 Posts: 23 Location: Australia

hi guys, great to see the work you are doing, the new design looks much cleaner, hopefully will have time to come play again soon but kinda waiting for the design to stabalise a bit as i agree there used to be too many components. hope to get some parts to try make one next week if your new version works ok which i think it will.

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

OK, I completed the OpenServo 2.1 PCB design with the reduced parts count and the IRF7309 MOSFETs.

The PCB layouts can be found here:

http://www.openservo.com/moin.cgi/PrintedCircuit2

The Parts Placement diagrams can be found here:

http://www.openservo.com/moin.cgi/PartsPlacement2

Because we reduced the number of parts nearly in half and spread them out a bit on the PCB it should be much easier to assemble.

I'm going to do more tweaking and double checking of the design, but I'll likely get a new set of PCBs manufactured by the middle of next week. Hopefully I can soon put this hardware stuff behind me and move on to the software.

As always, comments are welcome.

-Mike

ginge · Posted: Fri Mar 31, 2006 4:53 pm Post subject:

Mike, once again, looks great.

Those FET SOIC packages should make this easier to place the components.

Shoud be a dream to solder this, especially now my SMT rework station is repaired.

I will make a start on the Eagle version of 2.1, as the one I have hosted on my site seems to have been downloaded rather a lot.

Barry

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

I updated the OpenServo 2.1 schematic, PCB design and parts placement diagrams with the latest design including the battery voltage measurement.

http://www.openservo.com/moin.cgi/Hardware

I'm getting very close to sending this off to be manufactured. I want to sleep on it a bit and see if there are any other improvements that could be made.

-Mike

andylippitt · Posted: Sun Apr 02, 2006 12:26 am Post subject:

That wide open space next to pins 7 and 8 are just screaming out for optional external crystal and caps pads. Smile

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

I was thinking that where an option dual-N transistor package can go to control the P FETs. I'm sure we'll use it for something.

-Mike

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

Following Andy's suggestion, I went ahead and added pads for an external 20 MHz resonator to the PCB. The resonator will be defined as being optional as it can be omitted and the internal oscillator on the MCU enabled.

I also added a ground plane to top of the PCB, but it will be of marginal use because of how densly packed the components and traces are in most places. However, because it may be of some benefit I'll keep it.

I'll post updated schematics and PCB diagrams on Sunday or Monday.

-Mike

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

On Sunday morning I updated the OpenServo 2.1 documentation at the following URL to reflect the latest changes involving battery voltage sensing and a 20MHz resonator:

http://www.openservo.com/moin.cgi/Hardware

I think this board now has about all that we can put on it. Probably time get the PCB manufactured and order my parts.

-Mike

ginge · Posted: Sun Apr 02, 2006 8:20 pm Post subject:

Hi,

I am glad to see the addition of the resonator. This should increase the accuracy of the position and current samples quite a bit.

A small amount of experimentation should determine how useful this will be though.

Barry

andylippitt · Posted: Mon Apr 03, 2006 4:45 am Post subject:

I know there's not a whole lot that can be done but an appnote linked to in the FETs datasheet (http://www.irf.com/technical-info/appnotes/an-994.pdf) strongly suggests putting as much copper under the FETs as possible for thermal dissapation. While not quite under it, it probably couldn't hurt to extend the blocks on the drains up considerably towards the motor. Perhaps things could be shifted a little directly underneath them to make a touch of room too.

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

I looked into this, but my fear is that the metal case of a vibrating motor could short out the drains if we get too close. Something that could be done is to perhaps use vias to direct the heat to the top side of the board where we can put more copper around the pads to the motor itself. I still consider my next batch of boards to be ""burnt pancakes"" and I'll have to look into noise and thermal issues when I have servos with the new FETs. I can also measure the board against the actual HS-475HB servo to see where shorting problems may occur.

-Mike

andylippitt · Posted: Mon Apr 03, 2006 4:45 pm Post subject:

speaking of vibration, anyone have any thoughts on crystals vs. resonators in this environment? We've got temperature considerations too.

ginge · Posted: Mon Apr 03, 2006 5:01 pm Post subject:

Crystals definately. Ceramic resonators are prone to drift with temperature. They are generally less accurate too.

a quote I found on foxonline sums it up:

mpthompson · Joined: 02 Jan 2006 Posts: 650 Location: San Carlos, CA

Hmmmm. I'm purchasing a handful of the EFJ-C2005E5B ceramic resonators from DigiKey to try out. I wonder how much drift from the 20MHz we are talking about? 1%, 5%, 10% or more? I assume I can measure this with an oscillocsope.

Obviously we want the resonator to be as accurate as possible, but for the servo I have a feeling that the clock could drift as much as 5% or perhaps more without causing a major issue. The servo doesn't interface with any time critical external devices (the I2C clock is driven by the master). The worry would be that PID constants may have to be adjusted if the MCU clock drifts -- but this is a problem as well with the internal clock we are currently using on the MCU. It's probably even more prone to drift than an external ceramic resonator is.

For 62 cents I think the jump in performance from 8 MHz to 20 MHz is worth it even if the clock isn't entirely accurate. It could allow more interesting algorithms to be implemented in software than might otherwise be the case. The AVR is 1 MIP per 1 MHz so we'll have 20 MIPs of processing power available to us -- that's pretty fantastic in such a small device.

-Mike