Fri 6 Jun 2008
This weeks entry is the INA2143 differential amplifier from Texas Instruments. This is a dual diff-amp and the partitioning is pretty apparent. It is also apparent that most of the device is constructed using a bi-polar process, which matches well with the decreased quiescent current with increased temperature in the datasheet.
del.icio.us |
digg
Wed 4 Jun 2008
While looking through Make Magazine’s blog, I noted a link to the Great Internet Migratory Box of Electronics Junk project and quickly realized that it would be a great opportunity for me to off-load some of my accumulated electronics junk treasures before I move to a different place in State College, PA. The rules are pretty simple: add some components if you keep some components; document your box on your site/blog or flickr; choose an individual who wants a box; send it off after two weeks! I am going with a theme of microcontroller related treasures so I will include the following contents:
I will pick someone from the list of individuals on the Box Request page, so add yourself there if you are interested. Also, keep in mind that this may be a good opportunity to do some spring-cleaning yourself.
See: the microbox!
del.icio.us |
digg
Tue 3 Jun 2008
Time has not been kind to my isolated stimulator design (above) which has accumulated some additional circuitry over the past few years. The original specifications called for a circuit that would receive a program voltage on the non-isolated side and then drive a load with the same voltage off the isolated side using batteries (not pictured) and return the actual drive voltage and current back to the non-isolated side. The requirement for isolation was pretty relaxed and it was assumed that the load’s common mode potential would be on a few volts from the non-isolated ground.
It was then decided that we needed a voltage-programmed, current-output stimulator, so the same design was slightly modified to have a current follower as the output stage. Due to electrochemical properties of the load, we wanted to clamp the potential between the stimulating electrodes (in current-output mode) to some safety window, so one of the extra circuits was added for this function. As all real electrodes, our electrodes could also polarize, so a small (<1uA) over days could cause some problems with poor performing electrodes. This was a problem due to the single-ended nature of the isolation system. To address this issue, an effective high-pass circuit was carefully added to the isolated side with a time constant of minutes to return the output current to zero, which lead to another small PCB added.
I decided that it was finally time to redesign the stimulator since the kludges were beginning to cause the whole device to fail and were a problem to troubleshoot. The points that I wanted to address specifically are:
The first issue is a big one. Since I am expecting a program voltage which can be positive and negative, the operation of a linear optocoupler requires that the driving LED have some fixed intensity for zero program voltage, with higher and lower intensities for higher and lower program voltage respectively (can be opposite). This means that a power failure on the non-isolated side would shut off the LED completely and would cause the perceived program voltage on the isolated side to rail to one of the supplies, thereby providing maximum stimulation. Looking through several other application notes [1] [2] [3], I did not see a clear solution to this issue. However, using a pair of optocouplers in a differential mode should take care of the problem. That is, still require that both driving LEDs are biased to a certain intensity for zero program voltage, however, only use one to transmit the program voltage and use the other to transmit effective ground. The voltages (or currents) generated by the photovoltaic cells on the isolated side can then be subtracted, and with a little tuning, should provide proper isolation during normal operation and the difference should go to zero when the non-isolated side loses power. Using this differential scheme should also provide a means to control for small offset voltages and should remove the need for a built-in high-pass filter.
The question of full optical isolation is also a tricky one. In the current scheme, the isolated side measures current and voltage and feeds them directly into CMOS inputs of voltage followers on the non-isolated side. This would work fine if the isolated and non-isolated sides are at the same potential, however, any difference creates non-linear current paths through the ESD suppression mechanisms that are built into the CMOS op-amps that we use. I am thinking of using a two or three optocouplers to send the voltage and current signals back to the non-isolated side, however, I will have to work out if I can use the ground from the above paragraph somehow to remove any offsets in their transmission and work to limit the drive current to minimize power usage. The Clare LOC110 (the optocoupler we are using) can easily take 25-50mA across the driving LED, which is more than the rest of the isolated circuitry.
Finally, there needs to be an indicator for battery failure on the isolated side. It can be argued that proper tracking of the stimulation voltage and current would be a good indication of battery condition, but this will not work if the program voltage stays at zero (i.e., if we request zero volts, we will get zero volts out even if the batteries are dead). A compounding problem is that the isolated side is driven by a pair of 1.5V batteries (in series) to provide a total of 3V and it should be remembered that the LEDs in the optocouplers require a minimum of about 1V to operate. Hooking them up to the battery power (one for + and one for -) would be quite a wasteful solution as it would draw a substantial amount of current to operate, especially when the batteries are full. My present design idea is to use a pair of optocouplers, each in series with a depletion-mode FET connected to a resistor bridge. This way, when the bridge voltage drops below a certain value, the depletion-mode FETs will conduct and the LED will turn on, warning of battery failure.
I am hoping to get this design finished before August and fully expect another post in a year with new kludges, perhaps bluetooth or something.
del.icio.us |
digg
Sat 31 May 2008
I have been playing around with some wifi networking lately, mostly with the La Fonera, and finally decided to build a directional wifi antenna. Although the cantenna, however, I don’t really like Pringles chips and wanted to make something more interesting. I decided to try and make a simple Yagi antenna with a magnetic dipole as the driving element.
From a construction standpoint, the Yagi antenna is made by spacing conducting rods along the directionality axis with a driving element near one end. It is assumed that the incoming radiation is a TEM plane wave, so the direction of the electric field component should be parallel to the conducting rod orientation. The magnetic component is then perpendicular to the rods and to the directionality of the antenna. The rods spacing is then configured so that the coupled EM field generates a magnetic field component (and a curling electric field component) along the directionality axis of the antenna which has constructive interference at the driving loop. Proper spacing then determines the antenna’s gain and directionality in the band of interest (2.4-2.5GHz).
As a first step, I decided to reproduce the Yagi design made available by Andrew Hakman who reproduced the dimensions of a commercial antenna. This first implementation will test the basic operation and is still missing fine tuning and optimization. I am pretty happy with the initial results which demonstrate a 10dBi gain, which is pretty nice given that it took roughly half an hour to assemble. I will use a more precise construction technique (EGX-300 to mill the main beam) and will work out the optimal metal rod length to magnetic dipole ratio. The main idea is that the loop length needs to support one of the resonant transmission modes for the given frequency while the rods should be as long as possible to increase gain, but shorter than the length of the loop. If anyone wants more info on Yagi theory of operation, please post a comment and I will try to write up a post about it.
To construct this, I used a 0.5×0.5 inch piece of wood for the main beam, and 0.125 inch zinc rods for the conductors. I cut the rods to match the lengths in the above design and sanded the ends to remove any pointy spots. I measured out the positions for the rods on a piece of tape and used a small drill press to make the holes. I then gently tapped the rods into place and removed the tape. I cut the loop out of a sheet of bronze, mainly because that is what I had around. It is better to use a strip (versus a round wire) here to make the loop more sensitive to magnetic field components along the directionality axis. Finally, I decided to minimize transmission losses and mounted a USB 802.11g adapter directly onto the loop. I hot-glued everything into place and went to a large set of windows to test out the contraption.
To benchmark the devices performance, I compared signal strengths to the internal wifi adapter on my Lenovo T60. The signal strengths for the same APs were comparable between the internal adapter and the intact USB adapter so any improvement that I saw here was likely due to the Yagi. Although it was sometimes challenging to find the right direction to point the Yagi, I noted a substantial increase in signal power when I switched Netstumbler between the internal and external wifi adapters. Over all, I consider this to be a success since I got better performance from the USB adapter by investing a few dollars and a hour of my time. The next version will be forthcoming in the next weeks and will hopefully display even better performance.
del.icio.us |
digg
Fri 30 May 2008
Today’s IC Friday features the DDC114 from TI. This is a quad ADC, which makes sense given the partitioning of the device. I think a set of sample-and-hold circuits can also be identified.
del.icio.us |
digg
Thu 29 May 2008
When looking at the application notes section of Fujitsu’s site, I came across their FRAM memory guide book. I was surprised as I did not know what FRAM really was and so I flipped throug. Basically, a film deposition process was developed, which is compatible with standard CMOS processing, that introduces films who can maintain their polarization after the applied electric field is removed. We are all familiar with ferromagnetic devices, these are the pieces of metal that can be magnetized when placed in a constant magnetic field. Thanks to some nice electromagnetic research, we can do something similar with thin films and thereby create ferroelectric capacitors that are capable of retaining data without applied power while being as fast as SDRAM. It is clear that half of the Fujitsu guide is a sales pitch for their ICs, however, the other good is a fairly good introduction to the FRAM technologies. The basic technology is discussed along with some typical ferroelectric substrates. A reference list is also attached.
del.icio.us |
digg
Tue 27 May 2008
Paul So, one of my colleagues from George Mason University, has taken some time off from teaching Physics and started an art gallery that focuses on providing practical training to up-and-coming artists. What is interesting is that this gallery aims to improve both the artist’s technique as well as promote a successful career by giving an overview of the business and economics side of things. As the Hamiltonian Artists gallery is a 503(c) non-profit organization, it made sense to create some sort of donation system to supplement other means of income and make the gallery more sustainable.
Since the donation system is a pretty small project, June and I volunteered our time to write the code and integrate it into the existing website. After reviewing my options, I was pleasantly surprised by how easy Google makes it to integrate their checkout system. The first feature that makes Google Checkout attractive to non-profits is that all processing fees are waived “through at least the end of 2008“. I have a feeling that Google added this to protect themselves down the road, however, they will probably keep waiving the transaction fees. The next attractive feature is the availability of example code in various programming languages. I ended up using PHP, however, most other server-side languages are also supported. Finally, the Google Checkout Sandbox makes it really easy to check that your integration system works as expected before any money is transferred. Over all, I was very pleased with the ease of integration and the support provided by Google.
Coding of the system, which includes some hierarchy, has gone pretty smoothly and has taken about 15-20 hours total. The whole system will be reviewed by the intended management users tomorrow and then the last tweaks and polish will be applied. Hopefully everything will be up-and running in a week or two so that the donations can start to come in.
[ Picture is of the construction going on in the Hamiltonian Artists building. ]
del.icio.us |
digg
Fri 23 May 2008
This weeks images are of the LMC6442 dual op-amp from National Semiconductor. The devices organization makes it somewhat easy to read with the top and bottom halves dedicated to the two op-amps. Looking at the bottom half, the two inputs are at the bottom center and the op-amp output pin is in the bottom right corner. The circular structures that look like BJTs near the in/out pins are not for driving, but for ESD protection. It looks like the main differential transistor pairs are above the input pins around the center of the chip. The left portion of the device looks to be biasing/current-source circuitry while the right side two-three gain stages. Two metal capacitors can also be seen per op-amp which are likely used for frequency compensation.
I am going to try to focus the next few weeks on more analog ICs as well as an ADC and DAC, these will hopefully be easier to “read” and will give some insights into IC layout.
del.icio.us |
digg
Thu 22 May 2008
One of the circuits that I am working on has an optically-isolated sub-circuit which runs on batteries. This is an analog device where the non-isolated input uses LOC110 optocouplers to send some program voltages to the isolated side which then drives a load accordingly and sends back the applied voltage and current back to the non-isolated side. One of the problems with the initial design is maintaining proper signal gain and offset across the optocoupler while the batteries on the isolated side drain.
The LOC110 is a nice device that has an infra-red LED that is driven and a matched pair of photo-resistive elements so that one can be used to control the LED and thereby set the current through the second device. One solution to the problem above was to use a pair of matched voltage references and track the current through the photo-resistive elements while driven at the reference voltage. This proved to be mostly successful, however, it did entail some research regarding voltage reference design.
The most simple voltage reference design involves only one diode operating in a controlled reverse breakdown mode. The exact reverse breakdown voltage depends on the device design (see this post about breakdown mechanisms), however, the general idea is that the current through the device increases very quickly when a certain reverse voltage bias is achieved. One only has to put a resistor in series with the reverse-biased diode that will ensure that the reverse-breakdown criteria are met given the power supply and a voltage reference is born. The devoted fan of μblog will surely exclaim: But Nick! Didn’t you use a reverse-biased diode as a temperature sensor once? What good is a voltage reference that varies with temperature? I would agree completely and exclaim that datasheets for voltage references display the equivalent circuit model of a diode in reverse breakdown, however, the actual designs are a bit more complicated.
This application note from Analog Devices provides some good insights into designing single-technology voltage references. The basic idea in many cases is to create a differential voltage by building mismatched devices or by introducing resistances, such as R3 above, and then using an amplifier to generate a reference voltage. To maintain good stability over a temperature range, the amplifier can be designed to increase gain in order to compensate for the beta degradation at higher temperatures. It is still possible to do all of this using a bi-polar process by designing biasing current sources that change appropriately with temperature. It is also possible, although more difficult, to introduce FET devices which can have opposite thermal effects as compared to BJTs. In this case, it would be fairly straight forward to increase the gain of the difference amplifier through increased FET trans-conductances with increased temperatures.
In the end, I still have some drift and offset problems that are not associated with temperature, however, the design process has been a good introduction to and has developed an appreciation for voltage reference designs.
del.icio.us |
digg
Tue 20 May 2008
Since Penn State University in in the middle of Pennsylvania and has adequate amounts of farm land, there is also an on-campus creamery which contains various dairy products. I typically visit for the cheese, not the ice cream. During my last visit, I had the novel idea of trying to determine if rodents have a predisposition to cheese, as we have been taught by movies, cartoons and rat-trap vendors.
One of the groups at CNE that I design electronics for is looking at navigation and memory in Long-Evans rats. These rats look like little cows as they have both black (or dark brown) and white splotches covering the whole body. The rats typically run on a maze where they have to make appropriate path selection based on training and one of the rewards that they receive are pieces of a high-sugar kids cereal like Fruit Loops. At first, the rats are hesitant to eat them as they have no idea what the strange loops are and they see that they are different from their typical “Feed 5000″. After being given the fruit loops a few times, the rats become more accustomed to them and visibly prefer them to their normal feed. With this observation in hand, we set out to roughly see if rats ad an a priori preference for cheese.
We bought a fresh package of white cheddar cheese and broke off a few small pieces to offer to several rats that were not participating in any experiments. We only did one trial, however, we fed the cheese to several animals. The results were very similar to the fruit loops where the rats were hesitant at first to this novel object but eventually warmed up to it and also demonstrated visible preference over their normal feed. The important result was that none of the rats habitually went for the cheese after identifying it as though they had some kind of built-in fondness for cheese. Some took a few nibbles and then discarded it until a later time. From a practical standpoint, cheese may be a good reward snack for a rodent provided they have enough exposure to it to appreciate it, however, this poorly-controlled study leads me to believe that cheese would not attract rats like it does in cartoons (unless, of course, the rat had already been habitually eating the dairy product).
( Image is from the Spittoon. )
del.icio.us |
digg
