Design Question For The LM2937-3.3 Regulator And The micro-SD Card Adapter
idbruce
Posts: 6,197
Hello Everyone
While working on an electrical design, Leon kindly pointed out that I was using an inappropriate output bypass capacitor in combination with the LM2937-3.3. The datasheet specifically states that the minimum capacitance for the output capacitor should be 10 uF, however it also states that this value may be increased without limit. Then in other areas, the datasheet contradicts itself, giving two possible values for the maximum ESR, being 3 Ohms and 5 Ohms, so I have decided to proceed with the lowest of the two or lower. In a round about way, the datasheet kind of suggests the use of a tantalum capacitor for the output capacitor and I have read a post from Dr_Acula about the importance of using tantalums, especially when powering a SD card adapter, but I believe he was using a different regulator. So my question is this, if I am using a LM2937-3.3 regulator, and I intend on adding a micro-SD Card Adapter, what size capacitor should I use?
Here is what I think I know, but I could of course be wrong
[*]Two bi-color LEDs
[*]The Propeller chip
[*]A 24LC256 EEPROM IC
[/LIST]
Any effort that you may put forth into providing me with an appropriate capacitance value will be greatly appreciated and thanks in advance.
Bruce
While working on an electrical design, Leon kindly pointed out that I was using an inappropriate output bypass capacitor in combination with the LM2937-3.3. The datasheet specifically states that the minimum capacitance for the output capacitor should be 10 uF, however it also states that this value may be increased without limit. Then in other areas, the datasheet contradicts itself, giving two possible values for the maximum ESR, being 3 Ohms and 5 Ohms, so I have decided to proceed with the lowest of the two or lower. In a round about way, the datasheet kind of suggests the use of a tantalum capacitor for the output capacitor and I have read a post from Dr_Acula about the importance of using tantalums, especially when powering a SD card adapter, but I believe he was using a different regulator. So my question is this, if I am using a LM2937-3.3 regulator, and I intend on adding a micro-SD Card Adapter, what size capacitor should I use?
Here is what I think I know, but I could of course be wrong
- I will be using a 0.1 uF multi-layered capacitor on the input side of the regulator.
- The board will be powered by a 9V battery.
- The ESR of the output capacitor will be 3 Ohms or lower.
- The output capacitor will be a tantalum capacitor.
- The output capacitor will have a minimum capacitance of 10 uF.
- During normal operation, a maximum of two cogs will be utilized.
- Besides the current consumed by the micro-SD Card Adapter (which is specified at 5ma with a card inserted), the main other current consuming components will be:
[*]Two bi-color LEDs
[*]The Propeller chip
[*]A 24LC256 EEPROM IC
[/LIST]
Any effort that you may put forth into providing me with an appropriate capacitance value will be greatly appreciated and thanks in advance.
Bruce
Comments
It appears that nobody cares to discuss or debate the issue
Here is the capacitor that I decided to try first: http://www.mouser.com/ProductDetail/AVX/TAP106K035SRW/?qs=sGAEpiMZZMtZ1n0r9vR22d%2b8XmbM9QM8YGvHuP81cgA%3d
I just looked up that capacitor and I would think it's OK. 2Ω ESR is less than the required 3Ω, and over the full temperature range.
Don't forget the 0.1uF on the input and I would add 0.1uF on the output too. These should be right at the pins.
The larger 1000uF caps are nice if you are pulse driving motors. But they could be a bit large.
I'm going to suggest a different strategy though:
I understand why you chose the 35V unit because it had the 2Ω requirement.
However, it is in a larger "F" package.
You could use 3 10uF 6.3V 6Ω capacitors.
1 Close to the regulator and the other 2 near the Props VDD pins.
All 3 in parallel will look like 2Ω.
These caps are in the "B" package and are a bit smaller.
Spreading the caps around will still be effective for stability.
BTW, I have been using 5MHz small cylindrical axial leaded crystals by Citizen.
They work nicely and the leads, since they are on one end, are very close to the Prop pins.
You can solder the case to ground nearly eliminate the RF radiation.
Duane J
When I get to a certain point, it is very hard for me to change directions, especially since many of the components are customed designed EAGLE library parts to suit my needs and desires. If you look at the attached image, you can see how the board stands now (As an after thought, I also included the current shematic):
*Please note that the GND polygon fill has been removed for a better understanding.
C5 is an identical capacitor, but it is the input capacitor for the votage regulator.
If you take a peek behind IC2, which is the LM2937-3.3, you can clearly see that I have made room to accomodate the selected tantalum capacitor and it's package size, however it may be a little difficult to squeeze another 0.1 uF in front of it, but I am uncertain without trying to fit the 10 uF first.
I also intend to increase the size of all 3.3V traces to 0.015", but I have not done this yet.
Since you are much better at this then me, as it stands now, does it look like a pretty sound design (with the exception of the missing 10 uF capacitor and the narrow 0.010" 3.3V trace widths)? I am hoping that everything will be okay when I add the micro-SD Card Adapter.
Anyhow Duane, thanks for your input, although perhaps a little late on convincing me to change directions
Bruce
There are quite a few PDFs that list a number of facts and print idealized schematics AND then contradict themselves in text or just omit details that one wants to know.
In this case, you may just be expecting too much and just need to put what you prefer to build under testing to demonstrate it meets your expectations of quality, stability, and design.
In my own situation, such testing is last thing I can afford to do. I don't own the equipment or want to acquire such expenses. So I just consider myself a hobbyist.
A careful reading of all the entries about by-pass capacitors in "The Art of Electronics" by Horowitz and Hill will generally tout the benefits of tantalum capacitors in bypass situation, but then one arrives at page 457 and discover the authors actually warn that ceramic and tantalum bypass capacitors may in some cases begin to oscillate.
So what should one do
a. Take the risk?
b. Replace the tantalum?
c. Repalce all the ceramics?
d. Wait for someone else to write a compelling solution that may have no basis in reality?
The main point is the regulator you refer to is a good regulator with a substantial proven track record AND requires substantially more capacitance on the output side than the LM7803.3
What do I do? I usually use a 20uf electrolytic. But I got hooked into thinking that I best avoid tantalum by Horowitz and Hill -- maybe I just should have decided to avoid ceramic. After all they have piezoelectric problems and can have microphonic problems in some situations.
Capacitors are the bad boys of electronics - endless variety, not known for longevity, and so on.
There are no perfect solutions. As a general contractor and construction engineer, I have had to build houses that I desire to be perfect in snow storms, torrential downpours, and termite swarms. Idealization can just lead to frustration, endless frustration.
I really don't like Tantalums. Yes, it's a gut feeling. They have nice specs. But they have warts too.
I've done some long term leakage current testing on them. Basically running at 1.5 * the working voltage rating. I periodically measure the leakage current by measuring the time to discharge to 0.3679 * applied voltage or 1 time constant.
Tantalums have lower leakage currents but the leakage changes over time in unpredictable ways both increasing and decreasing.
Electrolytics, while having higher leakage current, behave quite predictably. Always increasing with time but nothing surprising. I can trust them.
Tantalums are by nature solid lumps so if a defect occurs it can't repair itself except by fusing a small portion of the lump. Kind of violent but I guess it works OK in low impedance power rails.
Electrolytics are by nature liquid so if a defect occurs they can repair themselves.
Tantalums should never be used in high impedance timing circuits where the fusing thing can't happen.
Ceramics can be trusted, but they are big and ugly.
Ya, tantalums are kind of the bad boys of the capacitor world. Sometimes you have to use them but you never get that warm feeling. Their compact and have low impedance, but can you really trust them????
Duane J
Please don't feel you have to argue your own personal solution as nearly everything I build is just cheapo electrolytics and ceramics.
At one time, I actually thought that Thailand was getting rich by exporting tantalum, but after visiting the country twice - I now see they export more rice than any other country in the world and a wide variety of product
The MOST interesting thing about new LDOs is not which capacitor they require, but that the LDO in combination with polarity reversal protection, and rugged protection for spikes to about 60V plus or minus makes them really useful in automotive environments.
With the GND traces gone you can tidy up some of the signal traces.
Duane J
In the post above, there is a link to a post in another thread, and within that post, there is a ZIP file which contains an array of PDFs with images. The file "PropProgTH_Msk.pdf" contains the photomask for the bottom layer of the board and I believe it should show the problem as resolved. If it doesn't, please let me know.
Bruce
Ya, I understand, I remember you saying this is a single sided board.
So the fill is on the top side only.
Yes, plated through holes can be done with single sided designs.
However my board guy really complains about this. Although he can do it.
The problem is without a pad on the second side the plating is not as reliable.
The way they do it is to drill the holes and chemically deposit a bit of tin through the holes.
Put resist on all areas where you don't want copper on a board with very thin copper on it. Maybe 1/10th mil or less.
Then plate copper and tin on top resulting in thick copper conductors and holes.
Remove the resist and etch the thin copper away quickly. The tin is actually the resist for this.
( I might have missed something but this is generally how they do it.)
Is this what you are doing?
Are you using that really low cost phenolic board?
Duane J
After a lot of research, I came to the realization a couple weeks ago, that plated through holes were going to be out of my reach for a while, at least until my patent sells. So this board was designed, with the sole intention of being a single copper bottom layer, without any plated through holes. The traces shown in the image above are actually the bottom layer of copper, so for this project, I will be attempting to use single-sided, 1/32" thick, 1 oz., positive acting pre-sensitized PCB material from Philmore Datak.
Bruce