OpenServo V4 (was OpenServo V3.x hardware update)

Discussions relating to development and use of the OpenServo hardware.

Moderators: jharvey, Secondary Admin, Admins

Post Reply
robotjay
co-admin
Posts: 225
Joined: Tue Aug 01, 2006 3:58 am
Location: Nebraska, USA
Contact:

Post by robotjay » Fri Apr 23, 2010 5:50 am

From a manufacturing standpoint, it's simply cheaper to place one large part instead of 8 or 9 smaller ones. You should give it a try. Let me know if you have trouble getting samples of the parts.

-Jay
"Nothing is fool-proof; For we fools are ingenious and will find a way."

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Fri Apr 23, 2010 11:05 pm

Don't I recall a document link perhaps posted by Ginge, that noted some short falls with about all off the shelf bridge chips. I seem to recall a slight delay in the h-bridge at the right time, would allowed for significant reductions in electrical noise. I also seem to recall it talked about techniques for reducing power draw as well. Most (perhaps all) h-bridge chips don't allow for these slight delays.

With the individual chips like we have done so far, it allows for these delays, and takes about the same board space.

kbb
Posts: 180
Joined: Fri Jun 01, 2007 1:47 am

Post by kbb » Sat Apr 24, 2010 1:38 pm

robotjay wrote:From a manufacturing standpoint, it's simply cheaper to place one large part instead of 8 or 9 smaller ones. You should give it a try. Let me know if you have trouble getting samples of the parts.
I have looked at data sheets on a lot of motor controller ICs now and so far I have only been able to find the following two devices that would be suitable for OpenServo (physical size, supply range, max cont. amps, rationale, and so on). One is the one Jay has already mentioned.

http://www.freescale.com/files/analog/d ... C33887.pdf
http://www.freescale.com/files/analog/d ... C33926.pdf

Pros
  • May lower the component count (see below).
    May save PCB space.
    Inbuilt “protectionâ€￾.
    For the devices shown there is a current sensing feedback, so we could drop the current sensing components from the OS design, reducing component count.
Estimated component difference between current OSv4 and use of one of the above:
  • Drop: 2 of SO(P)8, 2 of MSOP-8, 4 of SM0402, 1 of SM0603, 1 of RPACK-4, 1 of SOT23-3
    Add: 1 of PQFN-32/36, 1 of SM0806 (or larger), 5/6 of SM0402
    Might need to add back 1 of RPACK-4
Note that the PQFN-32 package of the MC33926 will probably need a footprint of between 9x9 mm to 10x10 mm, however that is without addressing the need to supply some method to dissipate heat. Add on another mm to each dimension for the MC33887.

Cons
  • No pin or pad compatible drop-in replacements.
    Higher RDS(ON).
    For the devices listed “RDS(ON) increases by a further 50% when the supply voltage is below 8Vâ€￾.
    The device itself will run hotter.
    Handling of heat dissipation might be an issue.
    The device will need to be mounted centrally to the OS PCB (rather than being located at the “earsâ€￾, as are the MOSFETs in the current design), which may make it harder to mount heat sinks.
    “Unknownsâ€￾.
jharvey wrote:Don't I recall a document link perhaps posted by Ginge, that noted some short falls with about all off the shelf bridge chips. I seem to recall a slight delay in the h-bridge at the right time, would allowed for significant reductions in electrical noise. I also seem to recall it talked about techniques for reducing power draw as well. Most (perhaps all) h-bridge chips don't allow for these slight delays.

With the individual chips like we have done so far, it allows for these delays, and takes about the same board space.
Are you able to discover where it is?

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Sat Apr 24, 2010 10:49 pm

kbb wrote:Are you able to discover where it is?
There we go, I just found it in this thread.

http://www.openservo.com/Forums/viewtop ... 0&start=15

Down at this part.
ginge wrote:Here is a good paper on basic PWM driving techniques and how we might apply them.
www.drivetechinc.com/articles/SW_BLDCAC5.PDF

robotjay
co-admin
Posts: 225
Joined: Tue Aug 01, 2006 3:58 am
Location: Nebraska, USA
Contact:

Post by robotjay » Sun May 02, 2010 8:21 am

It took me a while (with a little help from my friends) to understand the document Barry linked to.

The impression I come away with is that switching scheme 0 is the least efficient and most noisy of the schemes. It's the scheme that is likely to be implemented in most driver chips, including the MC33926. That being said, the v3 OpenServo is using scheme 0, and we've gotten decent results that way.

It seems like the extra noise and lack of control could be added to the list of cons for using a driver IC. But I suppose it could be justified, if there were enough "pros" to make it worthwhile.


All that being said, I am personally leaning back toward individual chips, because I like the control that Barry's new codebase is going to give us. I'm not sure if driver ICs have been discussed at length anywhere else in these forums, but I appreciate taking the time to talk about it, even if we decide not to use them.

-Jay
"Nothing is fool-proof; For we fools are ingenious and will find a way."

nighteagle
Posts: 6
Joined: Sun Aug 29, 2010 10:28 pm

Post by nighteagle » Thu Sep 02, 2010 7:01 am

Hi,

what is with these hardware idea?
Sounds very good.
http://www.01mech.com/supermodified

mfg

nighteagle

kbb
Posts: 180
Joined: Fri Jun 01, 2007 1:47 am

Post by kbb » Thu Sep 02, 2010 8:37 am

nighteagle wrote:what is with these hardware idea?
Sounds very good.
http://www.01mech.com/supermodified
That’s amazing! It looks very professional. Although I have put some time and money into V4, it does kind of raise some doubts about the V4 route in my mind. Anyway, I ought to be at work, so I will have a good look at it tonight.

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Thu Sep 02, 2010 9:40 am

Looks expensive to me. Also to bad the encoder didn't use the centering hole like we did with OE. That chip needs to be very accurately aligned to work correctly. Things like case warping due to structural stressing will cause that encoder to product some funky results. An interesting product none the less.

kbb
Posts: 180
Joined: Fri Jun 01, 2007 1:47 am

Post by kbb » Thu Sep 02, 2010 8:33 pm

jharvey wrote:Looks expensive to me.
You’re right about the cost; it does seem high. In part that is probably due to the exchange rate making it more expensive to produce.
jharvey wrote:Also to bad the encoder didn't use the centering hole like we did with OE. That chip needs to be very accurately aligned to work correctly. Things like case warping due to structural stressing will cause that encoder to product some funky results.
You’re right about the magnetic encoder and the general way that the encoder appears to be installed may be problematical: I have a feeling the device itself is supposed to hold the magnet+shaft in place. I will speculate that with a bit of patience one could possibly glue the pot body to the encoder board and secure the device into the servo with hot glue (but if maintenance was required it might then be a pain). However, their modular approach means that they should easily be able to rectify that.

As far as I can see, there does not appear to be any support for a pot connection, surprising as that should be easy to support. But it could probably be “patchedâ€￾ in easily too, although it might possibly require the encoder board to be removed.
jharvey wrote:An interesting product none the less.
Yes it is - professionally presented too. I cannot see how anyone would actually go about buyng one - there does not seem to be any indication on how or where to order.

I would have hoped that the stacked board approach would give a tad more than 5A for the motor. I have been thinking that OpenServo V4 could do with a two layer stack (although I believe someone already said they didn’t like the idea of stacking); currently I have had to drop a few niceties to get everything to fit (no blind or buried vias you see).
Last edited by kbb on Sat Sep 11, 2010 12:53 pm, edited 1 time in total.

ginge
Site Admin
Posts: 1031
Joined: Sat Jan 14, 2006 2:34 pm
Location: Manchester, UK
Contact:

Post by ginge » Sat Sep 11, 2010 11:54 am

I don't see any problem with stacking as long as we get good mechanical interconnects. Any doubts as to the boards coming apart can be resolved with glue.

That _is_ an interesting product, but I don't really who it caters to.

ginge
Site Admin
Posts: 1031
Joined: Sat Jan 14, 2006 2:34 pm
Location: Manchester, UK
Contact:

Post by ginge » Mon Sep 20, 2010 11:41 pm

I'm gonna stir this thread back up a little.

I want to push forward for a v4 hardware so I can finalise the codebase.

Can we recap on what we are thinking for fets?
Are we fixed on the xmega still? (I note availability is still shocking generally)
Are we happy with the pcbs drafted up?

Baz

kbb
Posts: 180
Joined: Fri Jun 01, 2007 1:47 am

Post by kbb » Tue Sep 21, 2010 7:22 am

ginge wrote:I want to push forward for a v4 hardware so I can finalise the codebase.
ginge wrote:Can we recap on what we are thinking for fets?
I am thinking: ROHM SP8M4FU6TB, but there are other choices.
ginge wrote:Are we fixed on the xmega still? (I note availability is still shocking generally)
Well, one option is to go back to the ATmega328, add a 12-bit ADC chip and bit bang the interface to it. Of course that probably drops some of the other things. But it would probably make the PCB layout easier, as the ATXMEGA MH chip takes up a lot of space in comparison.
ginge wrote:Are we happy with the pcbs drafted up?
Um, not really.

For a sensible fit inside a standard servo, the width of the current PCB layout needs to be reduced by 0.8mm and some package bodies need to be moved in from the sides a bit. The existing design is already at the limits of compactness and it is not proving easy to adjust the existing design in KiCad - which means I will probably have to redo it the next time I have a large amount of time for it.

I had hoped to use a smaller version of the TC4428A which would have helped, but they are only available in large quantities. Although that situation has improved since I last looked, the minimum is still high for anyone wanting to build small numbers of boards.

Possibilities if space proves too tight:

1. Dropping the MCP9701A temperature sensor, the XMEGA appears to have an internal temperature sensor which appears to be available. Although the external sensor might be seen as preferable.

2. Do something with the capacitor array (C1, C2, C3 and C4), for example maybe we can do without C2.

3. Board stacking.

RasmusB
Posts: 4
Joined: Sat Dec 15, 2007 12:13 am
Location: Södertälje, Sweden

Post by RasmusB » Mon Feb 07, 2011 8:06 am

kbb wrote:
jharvey wrote:Looks expensive to me.
You’re right about the cost; it does seem high. In part that is probably due to the exchange rate making it more expensive to produce.
jharvey wrote:Also to bad the encoder didn't use the centering hole like we did with OE. That chip needs to be very accurately aligned to work correctly. Things like case warping due to structural stressing will cause that encoder to product some funky results.
You’re right about the magnetic encoder and the general way that the encoder appears to be installed may be problematical: I have a feeling the device itself is supposed to hold the magnet+shaft in place. I will speculate that with a bit of patience one could possibly glue the pot body to the encoder board and secure the device into the servo with hot glue (but if maintenance was required it might then be a pain). However, their modular approach means that they should easily be able to rectify that.

As far as I can see, there does not appear to be any support for a pot connection, surprising as that should be easy to support. But it could probably be “patchedâ€￾ in easily too, although it might possibly require the encoder board to be removed.
jharvey wrote:An interesting product none the less.
Yes it is - professionally presented too. I cannot see how anyone would actually go about buyng one - there does not seem to be any indication on how or where to order.

I would have hoped that the stacked board approach would give a tad more than 5A for the motor. I have been thinking that OpenServo V4 could do with a two layer stack (although I believe someone already said they didn’t like the idea of stacking); currently I have had to drop a few niceties to get everything to fit (no blind or buried vias you see).
Hello!

I'm building a robot as a hobby project, which will use a large number of servos. I ordered a superModified to try out (just send an email to buy one), and I can confirm the issues mentioned previously. The encoder alignment is almost impossible to get right. The price is also too high - I'll be needing 24 servo controllers - but the biggest issue is that the i2c works in multi master mode only. This causes some really big problems for me, since i want to control the servos from a linux-based embedded system. The linux kernel i2c drivers can only work as master OR slave, not both. Switching modes require a kernel recompile...

So my next idea was getting a OpenServoV3 combined with a batch of openEncoders, but the v3 seems to be sold out, and I can't find a way to acquire a openEncoder kit (pcb, components, magnets etc). So right now I'm kind of stuck waiting for the v4 to be available. I'm working as a hardware design engineer in automotive electronics, and would be happy to help out where I can if needed!

In the meantime, can someone point me to a place where I could get a v3 and an openencoder? Assembling them myself is no problem as long as i can get a components kit...

Best regards,
Rasmus
/ Rasmus

jharvey
co-admin
Posts: 362
Joined: Sun Mar 15, 2009 12:06 pm
Location: Maine USA
Contact:

Post by jharvey » Mon Feb 07, 2011 9:29 am

I do need to get around to making OE stuff more available. I should inquire with ginge about doing that some time. I have a pile of SLA shafts, PCB's, and magnets used with the MG995/6. I have sold them in the past such that when I'm sold out I can get more parts. I'll PM you the price for that. I've also assembled a batch of these for a company in Spain. If lead time and $ permit, I might be able to do that again.

I built some OSv3's these aren't the same quality as what was sold at robot fuzz, but there is a potential small batch of these out here. I'll see how many I have laying round. I found some components a bit hard to install, as they are quite small. So I didn't place the components for the BEMF circuit. That has minimal support in software any how, so it was really just extra parts.

One item that has been brought up is integrating OE with the v4 spin. I've been tempted to start a v4 design with KICAD, but time has been an issue, so I haven't been able to poke the right bears such that we start with it.

I'll PM you pricing, if you don't get that shortly, bug me and I'll get to it.

RasmusB
Posts: 4
Joined: Sat Dec 15, 2007 12:13 am
Location: Södertälje, Sweden

Post by RasmusB » Mon Feb 07, 2011 11:52 am

Sounds great! I'll just need a few OSv3:s and openEncoders to get me started, for the main build I can wait for the v4 :)

Also - if there is anything I can do to help with the development of the v4, please let me know.
/ Rasmus

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests