Here's a brief update. After around a week and a half Seeed offered a small discount for the inconvenience. They would have been quicker, but my email spam filter pushed their first two responses to my junk mail and I missed them. Overall, I think they handled the situation professionally, and reasonably promptly for a hobbyist oriented business.
In the mean time, I couldn't wait for new boards, so I went ahead and assembled a single board to assess how big of an issue the missing solder mask would be. My paste application was pretty clean and the board reflowed well, so I assembled four more in a batch. Overall, I had about one-two solder bridges per board on 0.5mm QFN packages, that probably wouldn't have occurred with the solder mask. They are easy enough to fix with some flux and and iron so it isn't a big deal. On the four follow up boards, my paste application was somewhat sloppy, but things turned out well.
That being said, not having the solder mask is a minor inconvenience, but not enough to motivate me to change my processes and switch to a new vendor.
Just a quick glance at the schematic, I would guess around 20-40 uA idle current for everything. The battery monitor voltage divider will draw around 15uA. If the 3v regulator isn't specifically designed to have a low bias current it could be 10-30 uA, if it is it could be 1-5 uA. All the other peripherals likely have good sleep modes down to 1uA or so. With a 200+ mAh battery with included protection, overdischarge won't be much of a problem for you.
The uCurrent's have been out of stock for a long time, it appears Dave may have decided to stop manufacturing them. Unfortunate, as I could use one.
Very cool project. I've been working on something pretty similar for a client. Do you plan on using a LiPo with built in over-discharge protection? We opted out for lower cost and better integration of the cell into packaging. One issue we've had is leakage currents over-discharging the cell. Once the main processor falls below PoR it gets worse, as some of the peripherals get switched back on.
Have you designed in a circuit to cut off the cell output when the battery gets heavily discharged, or are you able to shut off enough peripheral circuits to get the leakage low enough to wait out a low battery condition?
First product lacks the precision, and the I2C interface is better as a current monitor within a product. The uCurrent looks really good.. very close to what I'm looking for. 2Khz is a tad slow for looking at potential 100uS current pulses, but it should work. May not be able to see pulses like that anyway due to power decoupling (haven't ever estimated tRC for the board). It also may not be possible to do much better, considering the high gain necessary, the GBWP would be high.
Here's an idea for a fairly simple breakout board, or also just a product I'm looking for but can't find (maybe you know of one). I've recently found myself in need of a way to characterize current consumption of a battery powered device in various operating modes in order to hit an aggressive battery life budget (6 months+ off a coin cell). The best way I can think of to do this is to insert a current shunt with amplifier to convert current to voltage, so the current can be monitored and correlated with software events using an oscilloscope. I don't particularly want to splurge on a differential scope probe, especially when this kind of amplifier could be made fairly easily.
The only product of the sort I could find were breakouts for ICs with internal magnetic coupling. They're intended more for high current measurement, and are too inaccurate down to the uA range. In a low power or battery application, the isolation would be unnecessary.
So what I'm describing would be a current shunt + instrumentation amplifier + possibly a low noise voltage regulator to power it. Amplifier can be externally powered (or maybe switched to battery powered), and the ground can be tied to the DUT ground. The shunt would be inserted between the battery positive and the device. I would want a usable range of ideally 100nA to 10mA, which is quite a few decades. That could be achieved using a logarithmic output voltage, or having switches/jumpers/etc to change the sensitivity range, or even with different assembly variants to provide a mA and uA version (and could be hooked up in series in practice). To avoid having an expensive (accurate) reference source on board, the user could measure the output voltage vs. input current using multimeters at two different points and extrapolate. The accuracy would be pretty decent if the amplifier has good linearity.
I haven't done much digging on whether it is technically feasible or practical. Things like, can such a high gain be achieved from shunt voltage to output without noise problems. My expertise is a bit removed from this kind of analog design.
Thoughts? Has anyone seen this kind of product or test tool?
I've got all the original Eagle files, and the original gerbers I sent to Seeed. Same design rule file, same footprint. In my gerber viewer the pads look identical in the soldermask layer.
I'm guessing Seeed has switched suppliers, and there has been some miscommunication about soldermask expansion between them. The Seeed design rule file applies a 2mil spacing around the pads in the mask layer. Maybe the manufacturer is not expecting this, and applies their own additional margin for their process.
I'll report back with an update when I get a response from Seeed.
[quote author="Sjaak"]I also noticed the spacing in the copper pour is wider (in the GND-polygon?).[/quote]
Good eye. Looks like I specified a different width for the copper pour between the two, 6mil vs 8mil. Unfortunately, not due to design rules.
For both batches I've ordered from their four layer service the production numbers have been printed in silkscreen. The numbers are quite different between the two, again suggesting they have switched suppliers. The ID code on the poorly toleranced board is quite similar to two layer boards I've received through their service.
I have submitted feedback to Seeed regarding this issue. However, their response time and the resolution are unknowns, and I have a development schedule to meet. So I'm searching for alternatives for this contingency.
Here's a shot of the difference in ID codes, the poorly toleranced board is on the right.
That service does look promising. I was checking them out after finding matseng's post about them in another thread. The stencil service is especially nice, steel for $20.
In working with them did you get any additional information about their process? In particular I'm interested in their options for layer stack up. I have a couple transmission lines that are a sensitive to dielectric and spacing. Seeed specifies that 1-2 and 3-4 layer spacing is 0.2mm.
Their 20 mil feature to edge spacing may be a problem for me, as I've already committed to space constraints with this spec at 10mil. I think I may submit my design with the 10 mil spacing and see what I get...
Has anyone experienced swings in tolerance from the Seeed PCB service? I ordered some four layer boards a couple of months back with great results. My board had several tight pitch QFNs and the boards came back with great soldermask coverage. Given the cost of the service is 1/5th to 1/10th of using an American supplier, I was amazed and excited by the results. I recently ordered another board with similar design features, but the soldermask coverage is going to make assembly by reflow impossible.
Has anyone else had a similar experience? I'm trying to trace this back to an issue with tools on my end, or a change on their end. Here's a photo of the difference, the footprint is from the same Eagle library. The process from 3 months ago is on top, the recent order on the bottom. The first set of boards came shrink wrapped in plastic, and the second set came loose in a green ziploc style bag.
[attachment=0]
Any recommendations for 4-layer services that can beat two hundred USD for 10-20 quantity boards that are ~10cm x 10cm would be welcome. :)
Found this steal when I was in Tokyo last December. I was in need of an LCR meter for a project, and had been researching my options online. I was in Akihabara and saw one that looked mighty familiar. The ODM that makes the handheld LCR meter for IET Labs seems to have sold a lot to one of the best uC stores in Akiba: Akizuki Denshi. IET list price, $330, Akiba street price, 4700 Yen, or about $60. It can do measurements up to 100khz, which is uncommon in these meters under $300. Anyway, here's a link: http://akizukidenshi.com/catalog/g/gM-06264/
Wireshark is a popular open source network sniffing tool. It has a lot of commercial support, and thus has a really rich set of plugins for different protocols. Examples of LA relevant protocols already supported: I2C, CAN, Bluetooth HCI, ZigBee, USB. It supports protocols higher in the stack for BT, ZB, and USB. Would it be possible to leverage their plugin formats for the OLS client? If support for their format could be enabled, the client would inherit a rich set of protocols.
In a recent project I was working on, it would have been invaluable to decode bluetooth packets from an analyzer sniff of a UART bus.
I spent some time working on another part of my current project. Here is an issue I ran into, the solutions I've come up with, and which one I'm likely to choose.
In short, you could call the issue "High side, medium voltage current switching with a low voltage microcontroller." Essentially, I'm building a large LED array, and need to switch 5v at reasonable current (500mA). Because my uC runs at 3.3 volts, it is difficult to shut off the switching device without some form of level shifting.
Here is the row driver circuit: [attachment=0]
The relevant part of the driver is the PNP BJT between Vcc (5v) and the row. R1 is your standard current limiting resistor, and R2 is a pull-up to prevent inrush current at power up (haven't sized it yet). There is a protection diode and simple filter in there as well, that we can ignore for this analysis (I'll discuss them in a later post).
To turn on the driver, you simply need to pull ROW0EN_ low. The problem comes when shutting off the row, because a 3.3v logic signal cannot completely shutoff the transistor. With 3.3v on the enable, Vbe is -1.7 so the device will conduct.
Here are the ideas/options I considered:
Option 1: Use a shift register or decoder as a level shifter.
Devices may be available that can operate at 5v Vcc, which accept sub 3.3v Vih
Cost is low (was considering such a device before realizing this shutoff issue)
Option 2: uC output driver float for shutoff
To shutoff, drive en_ to 3.3v then float the pin. Base and pullup current will charge the base until the device shuts off
It would require 5v tolerant uC pins.
Can't use a shift register or decoder, because pins can't be individually floated.
Option 3: Discrete inverter/level shifter
Use discrete devices to shift from 3.3v to 5v
Requires two extra devices per channel
Bias current may be high in low output state (though only one channel is on at a time)
Performance may be an issue for driving high speed control signals
Option 1 or 2 would be best, as the cost is fixed compared to my baseline design. I found that the 74HCT series supports TTL inputs (2.0v Vih, good noise margin), so using those devices would level shift for free. The uCs I've been looking at appear to have around 40% of their IOs 5v tolerant, so this option is a possibility. I'll be going with option 1 or 2, and deferring the decision until I can do a cost tradeoff and look into the details of the uC I'll use.
How would/have you solved this?
Off topic interesting read, how does an IC implement ESD protection with a 5v tolerant input? Don't they usually use clamping diodes to Vcc and Gnd? Many IC designs depend on bipolar snapback of NFETs caused by impact ionization. Essentially at high Vds, a parasitic BJT (think NPN) can switch on. Current will then flow through the device, providing a clamping effect similar to what a diode would. The voltage where this occurs is process dependent, but independent of Vcc in the way that a diode is. We can have 5v tolerant input and ESD protection! More info (warning, academic): http://http://ekv.epfl.ch/files/content/sites/ekv/files/mos-ak/stuttgart/Vassilev-mos-ak-STR04.pdf
[quote author="bearmos"]I was recommending the COTS supply because I thought you were more interested in having someone else design an inexpensive fully-featured bench supply (rather than do it as a personal project). I was thinking this because this thread is in "Project Development Ideas, and Suggestions", which I interpret as DP project suggestions.[/quote]
Yeah, this is along the lines of what I was thinking. After seeing the ATX breakout, which I think is a brilliant idea, I was thinking how useful it would be to have a spare adjustable power supply. It would be a small step that would add more utility to a hobbyist's tool bench.
It also seems to me that $70-80 may be over a critical threshold stopping makers from buying bench supplies. Though, I'm not really in touch with the market, so I may be completely off here. Thoughts? It could just be that I'm cheap..
I wish I had the bandwidth to take on another project, and this would be a worthy contender for my time. Analog design, power electronics, both are areas that I could gain expertise in.
[quote author="arakis"]As of now there is no plan to build a 3+A power supply by us, we have a ATX BB v2 design for which the PCBs might come soon, but that will be only free PCB, no production is planed. V2 featurs 1.25-10V 0.5A and -1.25--10V 0.5A adjustable PS in addition to the normal ATX rails.. [/quote]
This could be a good option.
[quote author="arakis"] Building a 0-11V 3+A linear power supply powered from 12V would require some bad ass cooling, as at it's maximum you are looking at 36W of dissipation on the regulators....[/quote]
So the initial numbers are a bit unrealistic. De-rate to 2A, and design the PS to switch to the 5V source when the output drops below 4.5V. Now the worst case is around 15W. Even a one amp design would be useful on most benches.
[quote author="bearmos"] As soon as you start trying to match the feature set of something available commercially, the savings starts to disappear very quickly because you don't have the economies of scale.[/quote]
Point taken. Though it would be interesting to see how much of the BOM cost is tied up in the transformer and HV conversion circuitry. This is where an ATX based design can beat the commercial version, given that most makers have a spare power PC supply lying around. Some people will pay extra to DIY, and who wants to send an old PS to the landfill?
As a hobbyist/maker/hacker/etc, it would be useful to have access to a cheap variable constant voltage/current supply. Something similar to the functionality you get with the entry level bench top power supplies. Having something similar to your ATX breakout, that provides said variable voltage/current source would be useful. An ATX supply would have enough power/voltage to build a current/voltage source from 0 to 11ish volts, at 0 to 3+ amps with only a linear regulation scheme. A design of that type would also provide a lower noise source than just an ATX breakout. Put me on the pre-order list for a two-channel version!
I haven't searched the web to see if said project/product exists, but entry level power supplies appear to be around ~$70.
