[quote author="ian"]Our volume is probably not 0.01 of either of you guys, but I have never (yet! and that I know of) had a similar problem in a DP design. I just moved our standard > 1uF to ceramic instead of tantalum where ever possible, hopefully it is less of a concern now :)[/quote] Two things:
1) My client is not really high volume. They do runs of 100, no more. Despite that, they've had several failures. It could be pick-and-place, bad hubs, or just poor parts selection.
2) While researching this, I came across an article rating various capacitor types specifically for USB bypass. Ceramic has the worst performance, and tantalum is most likely to fail. The winner in that article turned out to be aluminum electrolytic, which seem to only have the disadvantage that they're larger. The article talked about differences in performance that you might not expect based on the the specifications. In other words, the filtering ability of the same value capacitance differed according to the dielectric type. What really bums me out is that I cannot find that PDF again, so far. It was open in Adobe Reader when I rebooted my machine, and now it's fallen off the recent documents list. It's a really bad sign when you download so many electronics PDFs that you can't keep track of them all... Anyway, if someone has seen an article like the one I just described, please reply with a link.
What I ended up doing is modeling the entire power circuit in SPICE. That includes the noise filtering caps, the ferrite beads, the Zener, the LDO, and the two regulator input caps. At first, I had about 30 A or 40 A! But when I changed the caps to lower values, I got close to 500 mA maximum.
The last iteration was to replace the 100 uF cap with 1 uF, and replace the 47 uF cap with 0.68 uF. That kept the in-rush current below 500 mA at all times, at least according to the step response simulation in LTspice. I checked the current through both beads and the LDO, and all were were close (makes sense, all the rest of the parts were shunt, but those three made up the "circuit").
The only questionable factor was that the ferrite beads wanted an inductance, but the data sheet for the part I cited above has no inductance specified. I had to plug in a guess of 1 uH. I have not gone back to check, but maybe I should try other values for the inductance to make sure that it does not affect the in-rush current.
Maybe I'll try to upload the LTspice file here - if I can get it transferred from XP to OSX to DP.
As for your time constant equivalent of a 10 uF capacitor, I was thinking along the same lines but decided not to bother with that. Thus, I have not checked your math to see whether it's all on track. It seems more direct to simulate the in-rush current directly and make sure that doesn't exceed 500 mA. But I do appreciate your effort and may come back later to look at it from the rise time point of view.
[quote author="bearmos"]I remember measuring the value of unknown caps with a signal generator, resistor and oscilloscope. Here's something similar that could be implemented in spice:http://http://michaelgellis.tripod.com/68hc11/cap.html - the capacitance is proportional to the output frequency.[/quote] I have a Fluke 87 that includes a capacitance meter. Unfortunately, it does not measure really small or really large capacitances. The minimum is 5000 pF and the maximum is 5 uF. That makes it difficult for me to measure the equivalent capacitance of my circuit, since it is probably > 5 uF. The Fluke 87 does have a frequency counter, so maybe there's something I could do with that.
I'll take a look at the SPICE solution as well - thanks!
Quote
Other than effectively taking a measurement of tau (with a known resistance), I'm not sure how else you'd find capacitance with SPICE - but I don't use it on a daily basis either.
I suppose that if I use the same resistance for both circuits, then I can compare my complex USB power input to a simple 10 uF cap and see how the step response looks. I was worried about finding a precise resistance (and I will still search the USB specifications to see whether there is a power output impedance associated with that 10 uF requirement), but I guess it does really matter so long as the resistance is close to the typical value and the same in both circuits.
I also think that SPICE can tell me the current through the diode, which would be the in-rush current, but I guess this gets back to the precision of the output impedance of the voltage source. Seems like if I model the output impedance of the USB power source with the wrong value, then the in-rush current could be completely different.
Regulators are TPS70145 for 3.3 V and 1.2 V, MAX1797 configured for 5 V.
The whole power circuit from Vusb starts with a 0.1 uF cap to Vgnd, followed by a series Murata BLM18PG600SN1D EMI Ferrite bead (60 Ω impedance, 500 mA rating, and 0.1 Ω DC resistance) on both legs, then a 0.01 uF shunt cap and a 6.2 V Zener to protect against overages. The LDO is an ON Semi MBRS130LT3. The big regulator caps come after that. It's very similar to the Spectrum Digital DSK5509A Schematic with a few additions (and deletions, since I have no alternate power source).
By the way, I found an article which seemed to rate aluminum electrolytics much higher for USB input. Ceramics supposedly performed poorly, and Tantalum might catch on fire. I'd probably have to revise the SMD pads for the aluminum electrolytics although they might fit on the existing 1206 / 3216-18 pads. The MAX1797 evaluation board uses ceramics for bypass, but they're not feeding it from USB power.
[quote author="sqkybeaver"]the inrush current could exceed the usb specs when first plugged in with high capacitance and the voltage drop when pluged in should not fall below the 4.something volts required for usb.[/quote] Is there a way in SPICE to simulate this? Seems like you'd need a switch to simulate plugging in the USB cable and you'd need to know the maximum allowed output impedance of the USB power source. I suppose the latter might be in the USB specifications, but it's not one of the pieces of information that I've memorized (there are a ton of details).
I recall an impulse response tutorial for LTspice which built the step/impulse input signal descriptively. Seems like that could be used to simulate plugging in the USB power, but I don't think it would be quite the same as making a connection.
I have a USB-powered board that needs to feed a couple of regulators (one standard, one boost). Example circuits show either 47 uF or 100 uF on the input of each regulator to help keep the voltage up (combined with 0.1 uF to filter higher frequency noise).
The problem is that the USB specification calls for no more than 10 uF of capacitance on the input of a device between the VUSB and GND pins. The parallel combination of 47 uF and 100 uF would be 147 uF, which is well over that limit.
What I'm wondering is whether the LDO diode ahead of these caps effectively hides them. Doesn't a diode act like a capacitor? ... or is that only under certain conditions? Since series capacitance is equivalent to less than the minimum capacitance, then it seems like the diode could hide the 147 uF capacitance so long as the diode itself is 10 uF or less.
I tried to build the equivalent circuit in LTspice, but I don't see a "show me the equivalent input capacitance of this circuit" feature. I've graphed the step response, and it seems to reach full voltage by 20 us, but I'm not sure how to interpret that. I also simulated a simple 10 uF capacitor for comparison, and the rise time also looked to be about 20 us, but the problem there is that I put in a non-zero output impedance for the voltage source, but it seems fairly obvious that changing the output impedance will change the frequency of the simple RC filter and thus alter the rise time. I guess I could scan through the USB specifications again and see if there is some maximum resistance between the power source and the USB Device that I could plug in for a worst case rise time with a 10 uF capacitor, but then I thought it might be simpler to just ask here.
Is there a simpler way to determine whether my USB Device circuit violates the 10 uF maximum equivalent input capacitance for Vusb? Maybe there's something in LTspice or TINA-TI that I could use.
Note: A dead simple solution would seem to be to drop the bypass capacitors ahead of the regulators to 4.7 uF each, or less, such that their parallel combination is guaranteed to be below 10 uF. But I still want to figure out whether that diode ahead of the caps hides them effectively or not.
Another tidbit, I found an article about USB power that mentioned a few guidelines. They seem to recommend 50 V ratings. Ceramics had the worst performance in terms of slow response while aluminum electrolytics were the best. After reading about the problems with tantalum, I did not want to use them ahead of the voltage regulator, but I'm happy maintaining tantalum after the regulators.
The disappointing catch here is that while you can put 1206 pads on your PCB and easily switch between ceramic and tantalum, it's not quite so easy to swap in an aluminum electrolytic without changing the layout. I'm planning on checking the dimensions to see whether the same pads could accommodate either. I think that's doable, but the aluminum cans will need more clearance around the pads for the square base compared to the 1206 SMD parts.
[quote author="bearmos"]I really hate blaming things on ESD, because I view it as the scapegoat of choice for EE's :) [/quote] I hear ya. Seems like there's a new scapegoat: tantalum electrolytics catching on fire!
Here are a few choice quotes from the article linked above:
Quote
Tantalum capacitors manufactured per military specifications (MIL-PRF-55365) are established reliability components and have less than 0.001% of failures per 1000 hours (the failure rate is less than 10 FIT) for grades D or S, thus positioning these parts among electronic components with the highest reliability characteristics. Still, failures of tantalum capacitors do happen and when it occurs it might have catastrophic consequences for the system.
Quote
A risk of using of an electronic component, and a tantalum capacitor in particular, can be defined as a product of the probability of failure and consequences. In this regard, tantalum capacitors can be considered as low failure rate parts with a high risk of application.
Quote
A specific feature of tantalum capacitors is so-called surge current or turn-on failures when the board is first powered up. The mechanism of surge current failures has not been understood completely yet, and different hypotheses were discussed in relevant literature.
Quote
The existing system of screening and qualification of tantalum capacitors per MIL-PRF-55365 is supposed to assure that the parts would operate reliably at two operating environments: steady-state and surge current conditions. [...] In spite of the obvious importance of SCT, this test is optional per MIL-PRF-55365. [...] Also, there is only scarce experimental data on the effect of surge current events on the reliability of tantalum capacitors.
That's a lot to consider. I hope to finish reading that article soon.
[quote author="bearmos"]As a brief aside, I searched for a while for a suitable TVS to deal with ESD on USB power/data lines - it's tricky because of the minute amount of capacitance you're allowed on the data lines. . .I wound up settling on a ESDA6V1L, from ST Micro. They seem to work well from a cursory look. . .never interfered with data transfer at full speed. Since the product never went into full production, I never got a chance to take it through it's paces, ESD gun, eye diagrams, etc - but from a capacitive standpoint, it meets the required spec's for full speed USB 2.0.[/quote] This is the sort of conversation I hoped to spark. Have you had many problems with ESD? ... or is the problem that you can't really tell what the cause is when such failures occur?
[quote author="sqkybeaver"]do you have a hi res photo of the area?
does the power jack have a ground switch? and is it connected correctly? i have run into this myself where an equivalent part had a different pinout.[/quote] Great questions. Users seem to be pulling the caps off after they burn out, so the photos I have show more amateur soldering work than anything else.
The power jack is not populated, so it shouldn't be causing any problems, but that's a good place to look. The second diode is superfluous when the power jack is not populated, but if the DC power were grounded then the diode should be reverse-biased.
Further research has revealed that the Nichicon F931C335MAA is actually a tantalum electrolytic rated at only 16 V, and with a 4.5 Ω ESR. I'm not absolutely certain that they're all being installed with the correct polarity since the original design specified a non-polarized ceramic. It still seems like 16 V should be enough, but someone pointed out an article about tantalum failure modes that I haven't read yet. Effect of Surge Current Testing on Reliability of Solid Tantalum Capacitors, Alexander Teverovsky, Perot Systems, NASA
This is one of those cases where Eagle makes it easy to think you have polarity specified because the schematic shows the typical asymmetric symbol, but in reality the board layout has no clue for the assembly operator.
A secondary issue is that nobody double-checked the "ceramic >=30V" spec on the bill of materials, and let the 16 V tantalum slip through in production. I recently read an article by a designer who pointed out that one person does the schematic, another does the board layout, and then someone else selects parts. It's kinda challenging to wear all of those hats, especially when nobody else can check for oversights.
Bottom line: I think this has been solved. I've not heard any reports of failures with a 30 V ceramic.
I have a PIC board that I designed to take power from either a 5 V jack or USB. There are ON Semiconductor MBRS130LT3G diodes, rated at 1 A and 30 V, from each power source to a common voltage rail where a 3.3 µF bypass cap is placed. The cap is supposed to be rated at 30 V, too, and is a 1206-sized ceramic. Part number is Nichicon F931C335MAA. The only other part connected to this node is a SHARP PQ1M335M2SPQ (a cheap $0.33 3.3 V regulator with 500 mA capacity!)
For some people, this cap burns out, and it might have something to do with questionable USB hubs. I'm trying to figure out how this is possible. Seems fine with computer USB ports.
What baffles me is that the cap is rated for 30 V, so I don't see how it could burn out from a mere 5 V supply, even with a really cheap USB hub.
One major problem is that I cannot be certain that the specified cap is the one being assembled, because I don't really have control of that. At one point, a sample board ended up with a cap that I did not specify, a Panasonic ECJ-3FF1H105Z. But that one is rated for 50 V.
Note that USB recommends 10 µF or less in total Vusb capacitance, to minimize inrush current. That's why I aimed for 3.3 µF for the main cap between the jack and the regulator, assuming that any stray capacitance wouldn't possibly throw the total over 10 µF. I suppose 1 µF wouldn't be bad.
Does anybody have any idea why this cap might be burning out?
Do any of you have stories of USB Devices (that you designed) burning out any of the first few components that connect to the Vusb input jack pins?
I don't expect you guys to necessarily help me with this, but I thought that discussion of USB powered Devices, interesting parts, and common problems might be of general interest.
[quote author="bearmos"]So the assembly house is specifically asking for a *.CAD file?! odd. . .Mine is fine with DXF or even PDF (I think they program their PnP from the gerbers, since we get "turn-key" from them). They use the drawing mainly for visual reference, I think.[/quote] Yes, I've had two assembly houses request a CAD file. They even know to suggest that I use gencad.ulp in Eagle to produce the file, and one of them send the ULP to me in an email.
What's really strange is that the second shop is the one who had the problem, but only on the second run of the same board! Apparently, the first team looked at the .plc file from Eagle and got the electrolytics oriented correctly. The second team looked at the .CAD file and got them backwards. I found out when I nearly burned my fingers on the caps! I thought I had somehow managed to shock myself with 3.3 V or 5 V, then I realized that it was heat, not electricity, that was causing the discomfort. I took the boards back and made them flip the caps around (on four boards), and I hope they weren't too damaged from heat on the one board that I tested.
Obviously, I now give these guys special instructions to check the .plc file instead of the .CAD file, but I would prefer to fix the problem. Even previewing the .CAD file as I'm seeking in this topic is less than ideal, because it would be oh, so nice if it all just worked.
No, they're not DXF. I tried eDrawings, changing the extension from .CAD to .DXF, and there was an error reading the file. Looking inside the file, it is ASCII, and looks a little bit like DXF, but according to an online description of the DXF formatting it is quite different.
So, I'm still looking for an application, Mac OS X or Windows XP, that will open and view a .CAD file created from Eagle's gencad.ulp
I guess I can retroactively define my goal in starting this topic as: to learn how to get the most out of a battery by minimizing the losses due to dissipation. I naively assumed that the hardest part of the challenge was to get buck/boost working at reasonable efficiency when dropping from 9 V to 5 V, and that after that was solved there would be little else. But it makes perfect sense as you've explained it that you still need to customize the solution to the voltages and current levels needed in a particular design.
I had hoped that there would be a generic design, suitable for just about anything (given 9 V input and 5 V output, or maybe 3.3 V output). I suppose I'm not really too disappointed to learn that I'm probably not going to find a generic solution with maximum efficiency.
That said, I still learned a lot from this topic, so thanks for all of the suggestions and pointers.
To narrow the focus of the topic from here on out, I think it's safe to say that I'm mostly interested in powering Dangerous Prototypes circuits from a 9 V battery. I assume that most Dangerous Prototypes products pull less than 500 mA because so many of them are USB powered. I haven't really measured the current draw, but perhaps they're all under 300 mA, although I just don't know. I'm left with the impression that I want a pure buck regulator in this situation; one that has a maximum of maybe 500 mA.
If anyone has more gems to offer, I'm still reading...
[quote author="bearmos"]I would say these regulators are doing a good job, accurate to within +/- 0.5%. Do you really need anything more accurate than this for a power supply rail?[/quote]No, not at all. I was merely quoting these numbers so I could get a handle on what I should consider "good enough" when selecting the resistor divider tolerances. Thanks for pointing out that this is a really good performance.
Quote
If you're trying to use the power supply as a voltage reference and the app calls for high accuracy, a dedicated voltage reference might be a better fit. A quick search for something with better than 0.4% accuracy turned up this, it's accurate to +/- 0.1% Of course better references are available - their price usually being proportional to accuracy :)
Yes, I made the mistake of trying to get by on a budget product by using a regulator instead of a voltage reference. Texas Instruments support were kind enough to answer my questions and point out that their A/D chips which do not incorporate an internal reference will need an external reference of fairly high quality.
So far, in those applications where I need both a regulator and reference, I used the reference to feed the A/D and an op-amp in parallel, then the output of the op-amp was used to drive several other circuits using power from a higher rail. The reason I did this is that references generally provide very little output current, especially when compared to a regulator. Sometimes it doesn't make sense to have both a regulator and reference, especially if the voltages need to match, so the unity-gain op-amp should work great, provided that the op-amp specifications are sufficient.
Thanks for bringing up another angle to consider. I totally see what you mean because I tried running the -10 V supply for a preamp in my car stereo from one of these 9 V batteries, and it was emptied overnight!
My problem is that I started this topic with the cart before the horse. I do not actually know what current or power will be needed. I have never designed a circuit to be battery-powered, only for USB, wall wart, or full-on 120 VAC transformer+rectifier+capacitor+regulator supplies. I was merely thinking of the Flash_Destroyer as a 7-segment LED display, and thought that batteries might be more convenient.
As a user of electronics, though, I've used plenty of devices that run on a 9 V battery, and they all seem to last a long time. Now that you mention it, I suppose the only reason for my impression that these batteries last a long time is because manufacturers only specify them for low-power electronics. These batteries are most convenient, though, because they snap right into a circuit without requiring a holder, and that is an attractive feature to me. I have never seen rechargeable ones, though, so perhaps that's already a sign that I should look elsewhere.
Despite my lack of experience designing for batteries, my instinct was to think that dropping from 9 V to 5 V with a standard regulator would be very inefficient, and the dissipated voltage would basically just run down the battery needlessly. Thus, my very next thought was a switching power supply to bring the 9 V down to 5 V without wasting any power.
So, after reading your comments, I'm thinking that if 300 mAh or less is enough for whatever I end up designing, then the 9 V battery would be a fine choice. I'm assuming that the output would be 5 V or 3.3 V for most things I might build, given the popularity of USB-powered devices. Obviously, though, if I need more than 300 mAh, then I should have a totally different battery pack. Some of the spring-loaded cell carriers have the 9 V snap connector, so that might make for an easy interchange.
Every assembly shop that I've worked with asks for a .CAD file. Eagle can generate these from your project files by running gencad.ulp
I've recently had some problems with electrolytic capacitor polarity not matching between the.plc file and the .CAD file, so I am looking for a free viewer that I could use to double-check things. It would be preferable if this worked on Mac OS X, but I also have XP installed on my Mac.
