Making a long story short, the LM2940 does not like running with a 1uf output cap. It needs at least 22uf. This was a bugger to track down (I thought it was related to the new switcher circuit I had designed, and incorrectly focused my efforts there.)
Yellow trace is input voltage (from the switcher), blue trace is output. The oscillation was making my A/D's go haywire. It seems such a simple thing to miss.
Originally I was P.O.'d because this LDO should be able to handle 0.5v dropout, and the scope clearly shows that I was above that.
My lesson for the week: Read the "f-ing" data sheet!
Based on the capacitor ESR stability graph on page 10 of:
I would use a low ESR 22 uF capacitor in series with a .33 or .47 ohm resistor to make sure. If you can generate a pulse load for your regulator, you can easily see the improvement in stability the series resistor gives you.
Forgive my stupidity, but why should I bother with a low esr cap if I am adding .33ohm series resistance? The difference in ESR between say low ESR tant and say a generic electrolytic will be minimal compared to the .33ohm resistor you mention. For that matter a real cheap cap, say 22uF at 16V is going to have an ESR around around that point anyway. A Nichicon PW series (my cheap client loves these for cost) is 22uF 16V = 0.60 ohms @100k @ 20c
The low ESR capacitor plus resistor guarantees a stable RC zero over the life and operating conditions of the system. Getting the zero from some poorly controlled and unstable resistance from a high ESR capacitor is a ticking time bomb. By the way, "low ESR" in this contest means a small (1/4 to 1/10th, say), compared to the added series resistor, so that the fixed resistor dominates the total resistance, even if the ESR varies quite a bit.
Does the client expect the circuit to work for several years and over a range of ambient temperature beyond normal room temperature? If so, proving that the cheap capacitor is stable enough to expect success may cost more than using more predictable parts.
If this circuit is part of a greeting card or toy with a one month expected life time, then the cheap capacitor is definitely the way to go.
What is important is to understand the problem and the stability and cost of its possible solutions. In this case, the solution not only includes the capacitor and series resistance at the regulator, but the additional parallel bypass capacitance distributed at the loads. The data sheet isn't really helpful about setting up this series parallel combination inside the stability box. But it has been my experience that additional parallel capacitance is tolerated better if you are in the higher resistance half of the graph shown on the data sheet. But a pulsed load test of the alternatives is still a good thing to do. The one that rings the least (in radians of oscillation) is most stable.
On 2/28/2007 12:01 PM, The digits of John Popelish's hands composed the following:
Gotcha. Do you have any data to show trends of ESR change over time and temp. In the case of the regulator in question it can use a fairly wide range so if you hit in in the middle I wonder if it is overkill to assume it could fall outside that range.
Oh I am with you on that but you gotta understand clients.
I do contract engineering and while I can advice I have to accept what my clients want in the end. Most clients are reasonable in there desires although some want me to over analyze things beyond nessasary and some the other way. Usually the other way is because they are bean counters, or non technical or because they can't "see" theory.
In the case of this "cheap" client it has been a real struggle to try to get them to understand why just because I power it up on the bench and it appears to work does not mean the design is ready to ship. I can't tell you how many times they have built a product with parts other than I specified and even if I find it has issues with those parts (that I can prove on the bench) if they don't see it then it isn't real to them.
Reality is these kinds of clients are not uncommon. They are not technical and run by bean counters. I get so sick of "well the capacitance and voltage are the same - who gives a !@#$ about other specs".
It varies a lot from manufacturer to another, and from one product line to another. But you can rely that almost any change from new and room temperature ambient will increase the ESR. With most LDO regulators the problem is that all but the lowest cost capacitors are too low in ESR, so either you pick an unstable cheap one, or a little more costly, more stable one and add a 2 cent resistor to make it work, now and over the long term. If the load includes lots of little bypass capacitors in parallel to the one at the regulator output, having a lot more than the absolute minimum resistance in series with the main filter capacitor may be essential for initial stability. Then you don't have much margin for having the ESR of the capacitor go up as that capacitor ages or gets cold.
I have given up on hitting a good zero in the response just by selecting a capacitor for LDO regulators. Fro the price of a resistor, I sleep better and get less aggravated.