Re electromagnetic interference in Eagle:
I suspect the big issue will be neigher component cost nor weight, but rather the size of the engineering task, both for the specification writer and the hardware designers.
An over-complicated spec just makes extra work for everybody. A too-simple spec risks a last-minute re-design. Paraphrasing Einstein, the specification should be as simple as possible, but no simpler.
Alan
Designing for the worst case probably makes sense for Eagle. The number of units being built is tiny so component cost for the electronics is not a big issue. The big issue is weight.
73,
John KD6OZH
----- Original Message ----- From: "Alan Bloom" [email protected] To: [email protected] Cc: [email protected] Sent: Tuesday, July 10, 2007 06:37 UTC Subject: [eagle] Re: CAN-Do-Too! ??????????
Jim wrote:
- EMI Spec: You put a gun to my head, and I'm going to pull out the
MIL-STD, which is probably overkill, but MAYBE NOT. I'd really prefer that one of you guiys current in the INDUSTRIAL world take on this task and come up with something good enough and reasonable.
I specified several power supplies when I worked for HP/Agilent. I found it really hard to come up with system-level noise specifications in instruments for the same reasons as in the Eagle project. The problem is that it's a complicated multi-dimensional problem.
Dimension 1: Emissions (power supply) vs susceptibility (modules) Dimension 2: Radiated vs conducted emissions Dimension 3: Interference path: Radiated:
- Magnetic (inductive) vs electric (capacitive) coupling
Conducted:
- Power supply bus vs signal lines vs ground loops
- Voltage (open circuit) vs current (short circuit) specification
Dimension 4: Emissions level versus frequency (audio to microwave) Dimension 5: Broadband vs narrowband interference (Some circuits are more susceptible to sine-wave interference, some to pulse-type interference, some to broadband noise.)
If you specified every box in the N-dimensional matrix you would have hundreds of specifications.
Add to that the fact that the power supply must be specified before the subassemblies are designed/built/tested and vice versa, and it becomes an almost intractable problem.
The military's solution is to specify all power supplies and subassemblies for the worst-possible situation you can imagine. While that does improve the probability of success, it still is not perfect and is very expensive.
On the projects I worked on, we basically just specified the power supply as good as reasonably practicable using standard shielding and filtering techniques and then did whatever was needed in the various subassemblies to make the instrument meet specifications.
One point to pay attention to is ground loops. Ideally there would be a single-point ground in the spacecraft (probably in the main power supply). The idea is that each assembly's ground return connects only with a separate wire to the single-point ground.
It's not practical to do that 100% (think of the Can-Do power bus for example) but any module that can draw large current transients from the power supply should probably have separate power and ground wires that connect directly to the power supply. Also, any sensitive modules should not have any additional ground returns on the module side of their power supply filters. (That is, the ground return should go through the on-board filter.)
Al N1AL
On Mon, 2007-07-09 at 18:30, Jim Sanford wrote:
All: I'm excited to read all of the emails on this topic. Several thoughts, which I'd really like to discuss in more detail with all of you tomorrow night on TeamSpeak.
- As an experiment, and data collection exercise, I'd like to see
the inductor Juan identified as the source of radiated EMI replaced with a shielded inductor or toroid. As I mentioned to some of you, I saw a similar noise problem (source & victim were reversed) in my day job solved with exactly that one change. I think this would tell us much about how big a problem we have (radiated and conducted), and be a useful piece of data for posterity. As Bdale said almost two years ago, "Our legacy may very well be our documentation."
- Second, I'd like to see if we can, with a few ccomponent changes
on the existing widget board change the switching power supply frequency as Juan suggested. Another useful data point, with minimal effort.
Request for action: Can one of Chuck, Lyle, Stephen, or Bdale take on obtaining a shielded/toroidal inductor of the same value and provide to Juan for testing, and then take on researching a change to the switching PS frequency? IF so, Juan, would you and your team please make the changes and repeat testing?
- Regarding Chuck's comment: If we wind up with Can-Do Too, I'd ask
Juan if his accomplished project oscar team could take on component selection/purchase and construction of a couple of prototypes? This could happen in parallel with John's redesign of the U-band receiver. I know you guys have done a lot, but, YOU"RE GOOD!
- EMI Spec: You put a gun to my head, and I'm going to pull out the
MIL-STD, which is probably overkill, but MAYBE NOT. I'd really prefer that one of you guiys current in the INDUSTRIAL world take on this task and come up with something good enough and reasonable. In general, I'd like to eliminate as much mass and extra "touch" labor in the assembly process (multiple shields) as possible. Bob Davis: Please weigh in here, if the possible milled modules simplify this issue, please enlighten us.
- Chuck: I appreciate your comments on lessons learned. You
obviously have some things in mind, please scribble them down, get them to me, and I'll get placed on the EaglePedia lessons learned page.
- CAN-Do! team: Several folks have opined to me that it appears
that the widget design is final, complete, and we'd better live with it. I'd intended to discuss some of the above with you when Bdale got back. I am EXTREMELY PLEASED AND GRATEFUL to see you stepping up to the plate and acknowledging the need and willingness to do something based on what Juan has learned. Your intellectual honesty, integrity, and willingness to take on more for the team are recognized and appreciated.
- Requirements: Bob McGwier is correct, we really did start with a
very top-level requirements document. It is not perfect (Bob has hated it from the word go), but can be found on EaglePedia under Functional Requirements. It is also in need of updating, after the October BoD decision. On my list to do. Like you, I have finite energy and time, but it is on my list. I think John did an EXCELLENT job of documenting the UHF Receiver requirements based on what he knew. The need for an EMI spec was not obvious, but is now, thanks to Juan's testing efforts and exceptional documentation. Lou: Functional requirements for power supplies? Bob Davis: Functional requirements for structure and thermal performance? etc. etc.....we have much to do, but I think worthwhile effort. By the way, I'm reading (in a few spare minutes here and there) an EXCELLENT book on requirements management. When I finish, look for a review of it on my project management page. I will also be providing suggestions on writing "good" requirements, based on that book and my recent experiences in the day job.
- Somebody brought up the issue of CAN bus vs. discrete wiring
harness. In my mind, the discrete harness is a non-starter. WAY too much room for error and difficulty in testing. Yeah, I know the AO-40 harness was perfect the first time, but the guys who made that happen have made it clear to me that they don't ever want to do it again -- and I agree. So, I think we should remain committed to the CAN bus for many reasons. In my view, the recent "issues" with the CAN-Do! widget are just issues to deal with, not show stoppers. The CAN-Do! team is willing to deal with the issues, and so should the rest of us. Case closed in my mind. As always, open to well-justified better idea, but I think the writing remains on the wall.
I'm sure there's more action following this recent discussion, but specifics currently elude me. Feel free to send me suggestions, update the agenda for tomorrow night's TeamSpeak, and please be there.
Thank you all. Very 73, Jim [email protected]
Robert McGwier wrote:
Bill Ress wrote:
All - -
To add support to Juan's conviction that we need to start
developing
"top down" specifications versus the bottom up activities we've
been
involved with, I would add that in order to develop a "realistic" EMI spec for the satellites power distribution system, we really need
to
know what those circuits will do.
With that in mind, I feel we need to breadboard the key circuits associated with that system and get hard data versus shooting from the hip with assumptions. At some point this power system is needed anyway, so why not focus design attention on that "top level" system now
and
get that issue settled - or at least better understood?
On the issue of housing panel area, and the possible consideration
of
a Mark 2 version, I think if I remember correctly, there were
comments
on wondering why a DB-9 (or even a physically smaller connector
series)
couldn't be used or are all the 15 pins needed?
Regards...Bill - N6GHz
I also support the breadboarding of primary circuits. And we did develop our specifications for the latest incarnation of Eagle top down. We started with what services we wanted to deliver and moved down from there.
When I rejoined the project, and started the software defined transponder movement, and long before I was leadership, the Can-Do
was
"in the can". I had almost no input to it. That said, I really do support the goal of the CanDo. Anyone who has heard of the horror stories of Marie Marr and the wiring harness or seen Lou's
spreadsheet
for interconnections for AO-40 knows that these few words do not do
it
justice. We have found a gremlin. That is normal in any project of this complexity and is not easily remedied by specifications in a project like this where the work and the tools necessary to make the relevant measurements are spread over the globe.
I look forward to our doing better as a team.
Bob
Via the Eagle mailing list courtesy of AMSAT-NA [email protected] http://amsat.org/mailman/listinfo/eagle
Via the Eagle mailing list courtesy of AMSAT-NA [email protected] http://amsat.org/mailman/listinfo/eagle