μblog: engineering from the trenches

WARNING: This is very dangerous and should not be attempted unless equipped with proper training and equipment.

Some time ago I posted about uncapping integrated circuits and mentioned that newer case materials are more resistant to being etched by sulfuric acid on a hotplate. The reason that I came to that conclusion was that I placed a recent AD711 and LM317 in a Pyrex container with sulfuric acid and heated it without any substantial results. The writing on packages became slightly worn after an hour but that was about it.

I tried to uncap a few chips this week using a dremel motor tool and a round grinding stone but was unable to get close enough to the die without scratching it. I figured that I could grind down the case above the die (while wearing a face mask) leaving only a thin layer of packaging and then use sulfuric acid to get rid of the rest.  To my surprise, the acid worked really well and destroyed the whole package leaving only the die and the metal pins. I used the same acid as before and heated it to 60-80 degrees Celsius (in a fume hood). I think that I used a slightly cooler setting last time, so either the 10-20 extra degrees did the job or removing the surface of the package made it less resistant to the acid.

My next idea is to try to expose the die without destroying the chip. From the experience of using the dremel to expose a few dies, I can tell that if done carefully and under a microscope, it is possible to see the tops of the interconnect wires before you reach the core. That is, either use the dremel very carefully to remove just enough package so the wires only begin to appear, or start with the dremel and then use a fine grit sand paper. Once this is done, the next step would be to carefully time the etch checking periodically with the hope that the package is much thinner over the die and will be dissolved faster. This way, the die will be exposed and the rest of the package will still be in adequate shape, although the pins may have to be resurfaced so they don’t become too brittle.

I am aware that there are commercial places that will take an X-ray image of the chip to get exact depths and then do a controlled etch, and given enough time, I may be able to get access to/build such a system, but the key for me is something that is quick and easy to use. The way things are going, the dremel and sulfuric acid may be a pretty good option.

del.icio.us | digg

I have made several recent posts regarding power measurement, so here is my first average power measurement prototype.  The idea is that I am measuring the current returned from the appliance by measuring the voltage drop across a 0.1Ohm resistor in series with the common mains terminal using an INA117 diff amp. I also measure the voltage delivered by using another INA117 and a resistor bridge to drop the 120Vac down to about 4Vac. I then multiply the two signal using an AD633 four quadrant multiplier and get the instantaneous power delivered. This signal’s offset represents the average power delivered to the device. Since the AD633 also divides the output by 10, I put in an optional gains stage that either leaves it as is for high-power measurements or adds a 20x gain to the signal. Finally, I use a LTC1062 switched capacitor, fifth order, low-pass filter to get rid of the 60Hz signal and return only the DC value, or the value of the power delivered. At this point, a multi-meter can be used to check the power, however, I intend to use an analog voltage display. Right now, I am testing it with RC bridge, however, I plan to put this in an enclosure later (part 2) and provide full schematics. Comments are welcome, as usual.

( ltc1062.pdf ) ( ina117.pdf ) ( ad711.pdf ) ( ad633.pdf )

del.icio.us | digg

AMD and Intel were ahead of their times with their shared-license agreements. These days, no one integrated circuit manufacturer can stand on their own too well, and only multi-partner collaborations can approach the 45nm and 32nm process nodes. I have been putting off looking at some uncapped digital chips because of the die size since we don’t have a low resolution objective that can image most of a 4mm x 4mm chip much less 12mm x 12mm like this. I am trying to avoid using sulfuric acid as much as I can since it takes a long time to neutralize, so the next few IC Fridays will contain larger digital chips. It is admirable that AMD still maintains a decent website for their AM486 line of products.

( ad486dx2.pdf )

del.icio.us | digg

Final exams are over, so I am back to designing things. The first thing that I worked on was this tester for my pre-amplifier set. The pre-amplifiers are differential inputs into INA116 instrumentation amplifiers. I am trying to model very fine electrodes as the source, so I use a high (>100kΩ ) source with a 250μVpp (with a 300mV offset) test signal. What I didn’t take into account was sub-threshold conduction in the LEDs on the board. The minute current added to very high resistance networks created relatively large offsets in the test circuit and made it unusable. Most LED datasheets don’t display the current profiles below 0.1mA and many diode datasheets don’t display them past the μA range. Moral of the story is save yourself a few hours of debugging and remember that diodes conduct even below threshold.

( ina116.pdf ) ( dan235e.pdf ) ( 1n4004.pdf )

del.icio.us | digg

BrandonU from uC Hobby has come up with a clever write-up on how to read a flash chip that is already in a circuit. The idea comes from the realization that Smartmedia cards are effectively flash chips and have compatible pinout with some industry standard NAND flash chips. The device must use 8-bit blocks and share an 8-bit wide I/O for both the address and data. This is by no means an end-all solution to reading flash chips, but it may come in handy. The one part of the write up that may be problematic is inadvertently powering the board by providing power to the flash chip, but that can be looked at on a case-by-case basis.

[Via Hacked Gadgets]

del.icio.us | digg

Since I am taking a CMOS linear integrated circuit design course, I thought it would be reasonable to try to find a few designs of practical operational amplifiers that have been commercially used. Everyone knows the LM741 BJT op amp, but I found it to be difficult to find some complete CMOS designs, even for obsolete parts. I emailed many linear device vendors (TI, Analog Devices, National Semiconductor, On Semiconductor), introduced myself as a student, and then inquired if they had some complete CMOS design that were available to the public. Unfortunately, most of the support people either responded that all of their designs were proprietary information and was not for public viewing, or they sent me a 741 BJT schematic and then said all of their CMOS designs were private. Undeterred from my mission to study complete CMOS op-amps, I went to Google Patent Search. And found designs and explanations straight from the various manufacturers. Since patent information is publicly available, why not provide this information from the start?

( cmos-op-amp-ti.pdf )

( cmos-op-amp-national.pdf )

( cmos-op-amp-mot.pdf )

( cmos-op-amp-analog.pdf )

( cmos-op-amp-ami.pdf )

del.icio.us | digg

This is one of the popular instrumentation amplifiers from Analog Devices. The current datasheet is in revision D, and this chip is in a CERDIP case made in the 1980’s so there may be some discrepancies in performance. Unlike some other analog chips we have looked at, this one contains three sets of initials, if the one above can be considered as one.

( amp01.pdf )

del.icio.us | digg

The swell people over at Analog Devices have created an application note that overviews making a power meter that converts average power into a frequency output that then drives the mechanical counter. The idea is that high-side current is measured resistively and then compared with the applied voltage to the load. The two signals are multiplied giving an instantaneous power. Providing that both the applied current and voltage are sinusoids and vary only in amplitude and phase, the DC component of the instantaneous power signal will be an indication of power delivered to the load and will depend on the power factor (the phase between voltage and current). I am thinking of simplifying the design and using a low-side measurement, a pair of differential amplifiers and a four quadrant multiplier to produce the instantaneous power. Depending on the components available, I can try to use a two-pole Sallen-Key low-pass filter to get the DC out with two decades of ripple suppression.

To clarify, I became interested in building a power meter when I found out that the Wii uses 17W of power in operation and 10W in standby (when Connect24 is enabled). I am wondering what other electronics around the house have such a conservative standby mode.

( an679.pdf )

del.icio.us | digg

While thinking about measuring power delivered to devices, I came across this helpful diagram. Note that this catered to power distribution in the United States. The source is HyperPhysics at Georgia State University.

del.icio.us | digg

I have recently become interested in measuring supplied power to devices from the mains. The pin-out for a U.S. wall outlet included a ground pin, a common pin and a “hot” which outputs the 120V waveform at 60Hz. The common pin is used as current return path and the ground is used as a reference or for protection. I am thinking about using a resistive measurement so it can either be a low-side (in line with the common lead) or high-side (in line with hot lead). There are obvious trade-offs in each, the low-side is easier but introduces extra resistance in the common path while the high-side measurement needs higher-voltage sensing circuits or well-matched resistive networks. The ultimate goal is to measure power delivered to various household appliances so I don’t think putting a 0.1Ω resistor in the current return path will matter all that much. If something blows up, you will be the first to know.

( an39.pdf ) ( low-side.pdf ) ( an746.pdf )

del.icio.us | digg