Get the SoR Robotics Android App on Android Market for FREE. See this forum post for details.
0 Members and 1 Guest are viewing this topic.
I suspect that my filter needs a much highers LC value to get a nice and flat voltage for my electronics.Am i right or am i missing something?
The filter you're showing also doesn't put a coil on the ground wire, so perhaps there can be some leakage there.
Finally, a 7805 is a very old linear regulator. You can find other regulators with much better noise immunity, such as an LM350 or a LF50ABV.
You also want a significantly bigger capacitor in front of the regulator -- say, 100 uF electrolytic.
Another 10 uF after the regulator would also be good.
Since when was Kirchhoff's second law terminated?
Quote from: jwatte on December 17, 2012, 01:06:04 PMFinally, a 7805 is a very old linear regulator. You can find other regulators with much better noise immunity, such as an LM350 or a LF50ABV.LM350 is a very old regulator as well.By which parameter do you get to that conclusion? (Hint:All 3 of them have close to equal "noise immunity", although the LF50 is slightly worse than the other two).
Quote from: jwatte on December 17, 2012, 01:06:04 PMYou also want a significantly bigger capacitor in front of the regulator -- say, 100 uF electrolytic. Not unless there's half a meter or more of wire between battery and regulator (which there shouldn't in the first place.
Quote from: jwatte on December 17, 2012, 01:06:04 PMAnother 10 uF after the regulator would also be good.22µF would be aces.
We changed the capacitors, removed C1 and C2, C3 was chosen to be 47 uF.This way I built a pi-filter. Still i got spikes, from 5 volts down.
The addidtion of a second coil in the ground leads solved this problem.
I do not understand why the coil in the ground lead helps.That is a problem, because we want our students to built a similar filter, therefgore we must understand how it works ourselves :-)
I was thinking of common mode interference, which may end up screwing with the ground of the logic side. This assumes that wire inductance and circuit capacitance allows such problems to propagate -- for example, if you don't use a star connection to the battery. This not being the most likely source of problems, I added "perhaps." Clear enough?
I remembered the 7805 as only having 60 dB rejection. I remembered LM350 having 86 dB, and the LF50 80 dB.The 7805 (just looked it up in the data sheet) is specified with 56 dB minimum rejection and 72 dB typical: http://www.ti.com/lit/ds/symlink/lm340-n.pdfLM350 has 66 minimum and 86 dB typical, with 10 uF output capacitor: http://www.ti.com/lit/ds/symlink/lm150.pdfLF50ABV has 80 dB typical rejection: http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00000546.pdfSo, if you go with "typical," my recommendation still stands, but the difference is not as dramatic as I recalled.
Sometimes, you have no choice in the distance between your logic and your power source.
More is often better!
Also, the traditional advice of doubling up a ceramic with an electrolytic also makes sense, as the ceramic has much lower ESR and better high-frequency performance, and the electrolytic can provide bulk capacitance for lower frequencies (such as seen at the fundamental frequency of DC brushed motors.)
One place you have to be careful with not using too much capacitance is when inrush current matters. This turns out to be the case for motor terminal capacitance. Depending on the ratings of the motor controller, as little as 10 nF may cause a fast motor controller to detect overcurrent and go into shutdown (I had this happen with a Pololu controller for small motors.) Unconditionally putting 100 nF across the motor terminals might cause motor controller problems. My advice would be: Try it, and if it works, then keep it.
the noise from a motor is differential mode
If a motor controller flatlines over a 10nF (or even a 100nF) cap, I have only one thing to say about it: Extremely poor design.
And look at the parameters of the 7805 in a datasheet for the 7805, rather than in one for an LM340!
I really wish you'd stick to not guessing or presenting misconceptions as facts when you post
Yet, the OP reported that adding an additional inductor on the GND line solved the problem. (Exactly why this is is yet TBD -- it's quite possible that other changes, not involving the GND line, would have had the same effect.)
Yet such things are made by popular chip manufacturers and sold on popular robotics web sites.
You can call the 78xx (and LM1117 and friends) pet peeves of mine. There is no reason to use them. For low-current applications, there are more modern devices which have *much* lower drop-out (0.3V instead of 2.0V)
and for high-current applications, the LM350 does a lot more (if you heat sink it) and you should probably go with a switching DC DC converter anyway. My real goal in saying that the 7805 is old is to do my part to try to get it eliminated from all the beginner electronics designs out there. You may or may not agree with this goal.
QuoteI really wish you'd stick to not guessing or presenting misconceptions as facts when you postI have seen you rely on memory and end up being off on this very board. It's a human mistake, and I admit I'm both a human and make mistakes.
But I try to help people who post questions, with the experience I have. For electronics, that's hobbyistlevel. I appreciate whatever statements you contribute about the solutions, as in engineering, there rarely is only one workable solution.
Quote from: jwatte on December 22, 2012, 12:23:15 AMYet such things are made by popular chip manufacturers and sold on popular robotics web sites.This is just revealing a great lack of knowledge (a good reason not to answer questions in that category).People able to read and interpret the datasheet you linked to can see for themselves, that a 10nF cap is unable to kill this chip. I don't know why you insist that this is so, I can only inform you that it ain't.Based on the "typical" numbers, a 10nF cap can result in a (very short) surge of slightly below 1.76A peak
Why do you think they're still produced if LDOs were the universal solution?
And if you wanted to be true to that cause, then get rid of the old LM350 as well, it's a weakling when compared to eg. the Micrel 7.5A LDO (w. a current limit of 9.5A to 15A) regulators I use a lot.
And I'm not talking about a slip of memory, rather about your weird claims - like the 10nF that killed your motor driver.I wonder how you came to that conclusion btw. so feel free to comment!
There's a world of difference between a memory slip and spreading misinformation.