CURRENT ISSUE
Contests
Contact us
|
|
Staff List | About Us | General Info
Reader I/O
Circuit Cellar has always invited readers to send in their comments about the magazine, the Web site, current contests, or whatever it is we're doing at the time. Rather than have the Reader I/O page in the print magazine, we've moved that feature to our Web site in order to expedite the posting process. We hope you'll stop by to drop off your comments, see what other readers are saying, or get any updated Editor's Notes that may pertain to articles in past issues.
CORRECTION
11/19/2011
Circuit Cellar apologizes for the poor visual quality of the JK microsystems advertisement in the December issue (Circuit Cellar 257). Here is the correct ad (click on ad to for high resolution PDF):
READER COMMENT
11/17/2011
A reader had the following comments about Robert Lacoste’s October 2011 article ( "Line-Coding Techniques," Circuit Cellar 255):
I just read your Darker Side column in the October Circuit Cellar.
I make it a point to read your columns for their insight and engineering point of view—thank you for sharing your expertise.
In your October column, you discussed serial channel coding techniques, and one of your examples (in passing) was the Apple II floppy disk.
In particular, you mentioned that the Disk ][ was an example of the use of 8B10B coding. This is actually not the case. One practical reason is that all disk encoding and decoding are done in 6502 software and the 6502 can only conveniently handle 8-bit data.
The Disk ][ actually made use of several different coding schemes. Since they are implemented entirely in software, the particular scheme used is dependent on speed requirements and on software adaptation to the drive electronics.
Address fields are always encoded in RZ code, in which each 8-bit “nibble” on the disk is used to encode 4 data bits, each preceded by a “1.” So the data 0xC would be encoded as 11111010. This scheme (known in the Apple II community as “4 + 4” coding) has the advantage that two successive nibbles can be decoded very quickly into their equivalent 8-bit byte by doing a shift and an AND for on-the-fly address encoding/decoding.
Data fields have been encoded in many ways, but the two “official” schemes employed by Apple software were 5B8B (known in the Apple II community as “5 + 3” because of the way a data byte is broken up for recording), and 6B8B (known as “6 + 2”).
The 5B8B was used in DOS 3.2, and allowed only 13 256-byte sectors per track. Shortly after release of the Disk ][, DOS 3.3 was introduced, which more completely exploited the capability of the drive's bit slicer to encode 6 bits of data in each 8-bit disk nibble (8B6B). This boosted the capacity of a diskette to 16 sectors per track--quite an accomplishment for single-density floppies of the time.
The use of software for encoding and decoding disk data permitted the Apple II to use Group Code Recording, which increased both the data capacity and the data rate of the drive compared to the typical RZ “4 + 4” coding.
This level of versatility and performance was enabled by an extremely clever logic state machine in the
Disk ][ Controller. Steve Wozniak used a shift register and a 256-byte ROM to build a data separator
that was much less expensive (and more capable) than the floppy controller chips of the time. His design
is still one of the finest examples of elegant engineering I have ever encountered. It is fully described
in Chapter 9 of: http://mirrors.apple2.org.za/Apple%20II%20Documentation%20Project/Books/
Jim%20Sather%20-%20Understanding%20the%20Apple%20II.pdf
Michael Mahon
MJMahon@aol.com
http://home.comcast.net/~mjmahon/
7/25/2011
A Canadian reader had the following comments about Robert Lacoste’s June 2011 article (“Stay Cool! A Heat Management Primer,” Circuit Cellar 251):
In the 1980s, I worked for a large building automation company with an international reputation. One project during my time there was to do a cost reduction on a line of Triac and SCR power controllers for voltages 120 to 600 AC and currents ranging from 25 to about 450 amperes, so heat sinks were a major part of the equation. Some of these were multiphase, so 1, 2, or 3 power semiconductors might be used on one or more heatsinks. Fortunately, I did have a little experience with heat flow calculations from previous work.
1) It is worth noting that heatsinks behave differently depending on orientation in free convection. For extruded and (often) stamped fins, vertical is better than horizontal, and parallel to the anticipated airflow is better than perpendicular to it.
2) The best location to mount one semiconductor is not always exactly in the middle of the heatsink—on vertical orientations, something below the center often gives better thermal performance.
3) If you are dealing with transistors and seem to need a bigger heatsink, sometimes the best solution is to use two (or more) parallel transistors which are suitably forced to share the current—this divides that part of the total thermal resistance by the number of devices. It can be cheaper than a bigger heatsink if that might force you to use a bigger box.
4) We did make some of our own aluminum extrusions—I thought at one point that we might save some money by not spray painting the exposed raw surfaces black after considering that the radiant component for a convection design might not be too significant. In at least one of the examples under consideration, painting the surface improved performance by something in the order of 5%: much more than I would have expected. Over extended time, black will also look better, since the exposed aluminum surface cannot oxidize—an important visual consideration for a commercial product.
5) At the time, desktop computers were quite new (a couple of IBM PC’s cost us about $5,000 each—10 Meg Hard Drive, 1 Meg RAM, 8 MHz 8088, 13” green monochrome monitor !!!!), and effective thermal computation programs were essentially not available. I did two of “our” heatsink designs by pouring over the designs of other manufacturers for hours, comparing weight and surface area against thermal performance. I was advised that, if I could get closer than 10% between my calculated thermal resistance value and a measured one, I would be doing well—I turned out to be very lucky!
6) Some of our designs were fan cooled. For economy reasons, we had to use the same heatsinks for both cooled and non-cooled designs. Shorter fins and rougher (or spikier) surfaces work better for fan cooling, but we had to live with the compromise, and adjust the length of the extrusion (and sometimes the enclosure, since the heatsink formed a part of the case itself) to meet the maximum current requirements set by the marketing department. (Note: in these designs, the semiconductors were mounted on the inside of the extrusion, so there were no live parts accessible from the outside.)
7) When you are dealing with 100s of amperes, the semiconductor to heatsink interface resistance becomes extremely important—the aluminum extrusions, despite being fairly flat, were "polished" on a circular milling machine to maximize flatness of the mounting interface for higher currents (this is also a standard technique for “hockey puck” SCRs)—mounting torque and maintaining it with compression washers are methods also required to account for thermal expansion and contraction for the latter devices.
8) One option to not using electrical insulating washers, which as you have noted increases the thermal resistance, is to let the heatsink be electrically live. Of course, this only works when there is a suitable way of keeping people from touching the live metal, but this can be acceptable in some designs.
9) Or, some power semiconductors are available in electrically isolated packages, however, a close look at comparative specs almost always shows they are at some thermal disadvantage as a result of this internal isolation.
10) While many power semiconductors use 150 C as the design limit, some can be 175 C or 200 C. This can make a big difference to your design—one should always check the datasheet and know the device.
Keep up the good work.
CORRECTION
4/18/2011
Please note the following correction to page 47 in Robert Lacoste's April 2011 article ("Noise Figures 101," Circuit Cellar 249):
The noise factor F is always GREATER than 1 and not lower (as F = SNRin/SNRout, and SNRin>SNRout). Therefore, its log, the noise figure, is as expected greater than ZERO.
CORRECTION
12/06/2010
Please note a minor typo in Figure 1 on page 67 in Circuit Cellar 245 (December 2010). At the bottom left, “Waterdog timer” should be “Watchdog timer.
SCHEMATIC CORRECTION
9/29/2010
Updated schematic for figure 2, page 47, issue 244.
Unfortunately an error was made when we redrew Brian Millier's Sound Generator circuit in the November 2010 issue. As drawn, bridge rectifier D3 is shown with the outputs reversed. This image shows the corrected power supply section of the circuit which appeared as figure 2 on page 47 of issue 244.
CORRECTION
9/29/2010
”Please note the following update to Figure 7 in Monte Dalrymple’s article, “Calculator Brain Transplant: Update a Vintage Design with FPGA Technology” (Circuit Cellar 243, September 2010). The PWR_UP signal that comes from resistor-capacitor (R24-C6) should be deleted, as it was replaced by the output of the MIC811 chip.
CORRECTION
9/7/2010
A reader posed a question regarding the schematic that appeared in Figure 3, page 27, of David Ludington’s Issue 242 article, “Precision Temperature Control Circuitry.” Please note that there should be a black dot at the juncture of U2A-pin1, R8, R1, and U3-pin4 showing that all of these components are connected. It was included in the author’s original submission, but was missing in the finished article. Circuit Cellar apologizes for the inconvenience.
NOTE
7/23/2010
Circuit Cellar author Dev Gualtieri informed us that a reader noted an improvement to his X10 controller circuit ["Build an X10 Controller (Part1),” Circuit Cellar 240, July 2010]. Dev Gualtieri wrote:
"Although my X10 controller circuit (Circuit Cellar, Issue 240, July, 2010, pp. 28ff.) functions as published, Curt Terwilliger has noted one improvement. A 1 k-ohm resistor should be inserted at the base of Q3. This prevents excessive currents at the Q4 collector and the Q3 base. The circuit functions as published, since the duty-cycle on these transistors, which are active only when an X10 command is issued, is low. The additional resistor is better engineering practice."
The Circuit Cellar editorial staff thanks both Dev and Curt. This close interaction between a reader, an author, and editors sets a great example for everyone in the embedded design community. It's great to see everyone reading so attentively and working so hard, all in the name better engineering practices!
CORRECTION
6/2/2010
A typo appeared in the preview of Stefan Siegel’s Issue 239 article. The preview reads: “system an for.” That should be “system for an.” Thus, the correct preview is:
The first part of this series introduced a home automation system for an energy-efficient house. This article covers the embedded hardware and software for the nodes. You can use similar parts and design techniques to automate the building of your choice.
The typo was not the author’s mistake. We apologize to Stefan and to our readers.
CORRECTION
1/28/2010
Regarding page 42 in Evgeni Stavinov’s Circuit Cellar 234 article titled “A Practical Parallel CRC Generation Method,” please note that:
P(x) + Q(x) = x^3 + x^2 + x
Should be:
P(x) + Q(x) = x^3 + x
CORRECTION
11/23/2009
In the preview on page 14 of Dave Tweed’s Issue 233 article “iMCU W7100,” there is an extra word (“the”) in the last line. The word should not be there. It was inadvertently added by the editor, not the author. I apologize for the mistake. – CJ Abate
CORRECTION
10/21/2009
On page 9 in Circuit Cellar Issue 232, the URL for ZMDI should be www.zmdi.com. We inadvertently ran an incorrect URL.
UPDATE
9/30/2009
Regarding Issue 231 page 24, the author
Arnold Stadlin has updated Figure 2 in his Issue 231 article, “Frequency Sensing Made Simple.” The correction for the Reset Switch section is:
Updated files are also available in his code folder on Circuit Cellar FTP site.
CORRECTION
May 26, 2009
In Tom Cantrell’s article “Easy (E)mbed” (Circuit Cellar Issue 227, June 2009), the NXP part number should be “LPC236x” (not LPC263x) . The editorial department apologizes for the confusion. (Click here for an updated version of the article that reflects the proper part numbers.)
NOTE
March 6, 2009
Circuit Cellar thanks Smart Robots Inc. for the SR4 robot image that appears on the cover of Issue 224 (March 2009).
CORRECTIONS
February 2, 2009
The following changes/comments apply to Robert Lacoste’s article, “Microstrip Techniques” (Circuit Cellar Issue 223, February 2009).
A reader pointed out that the following sentence may be misunderstood: “the power transfer between a source and receiver is maximized when impedances are matched.”
Robert Lacoste replied:
“If you have a 50-ohm source, then the power delivered to the load will be maximized with a 50-ohm load as explained in the article. However, if you have a 50-ohm load and are allowed to choose any source impedance, then the power delivered to the load will be maximized with a 0-ohm source impedance, simply because the voltage on the load terminals will be higher. Unfortunately, in the RF world, if you match a 1-ohm source with a 50-ohm load, you may have other significant issues as 92% of the energy will be reflected back to the source, giving nasty mixing effects in the source if it is an active component.”
January 12, 2009
The following changes/comments apply to Robert Lacoste’s article, “PID Control Without Math” (Circuit Cellar Issue 221, December 221).
1. At the top of page 60, the article reads: “At equilibrium, the heat loss to the ambient air is exactly equal to the 3°C error multiply by Kd.” This is a typo. It should read “Kp,” not “Kd.”
2. A reader wrote: “The article uses Error = actual - target equation, but this seems to be opposite to the simulator code downloaded. Am I right that what appears in the article is just a typo?”
Robert Lacoste replied: “In fact, the sign of the command is just a convention. For example, if the controlled system is a motor-driven car, and if the sensor is a position sensor, then you could either need to send a positive or negative command to the motor for a positive error, depending on the wiring. Effectively, with the example the command (heating power) must be positive if the target is above the current temperature, so Error = target - actual with Kp > 0. But you could also have the opposite ...”
July 11, 2008
A mistake was found in Figure 1 in Robert Papp’s article, “Sound Effects Processing” (page 71, Circuit Cellar 216, July 2008). The corrected section of the schematic is now available at:
ftp://ftp.circuitcellar.com/pub/Circuit_Cellar/2008/216/Papp214Fig1b_corrected.pdf
April 22 , 2008
Regarding R7/R8 in Robert Lacoste’s April 2008 column, “Low-Power Techniques” (Circuit Cellar 213), note that R7 impact is probably very small as polarized only when the button is pushed. But R8 impact is not minor at all: 2 V/10 kohm = 200 microAmp. So, R8 must be far higher, 1 Mohm or so.
April 1 , 2008
In Tom Cantrell’s April column “More Than A Core” (Issue 213, p. 86), it was stated that the STM32 Primer uses an ARM7 MCU to provide a USB debug interface, but the part used is actually an 8-bit ST72651 MCU.
January 30 , 2008
Please note the following correction to Aubrey Kagan’s article, “Resilience in Embedded Designs (Part 3),” which ran in Issue 208 (November 2007). In Figure 2 on page 59, the decision box in the background routing the test is:
If cOldCheck0=cCheck0
The marked branch should be "Y" and not "N."
November 21 , 2007
In Figure 6 in Robert Lacoste’s issue 209 (December) article, an arrow is missing from "Output frequency" to the "Divide by N" block.
August 16 , 2007
In Robert Lacoste’s "Let's play with EMI" column published in issue 205 (August), he stated on page 67 that an FFT spectrum analyzer based on a 8-bit digitizer could provide a dynamic range of 48 dB. This is true if the signal is a pure sine, as when the signal amplitude will be lower than one LSB, the output of the ADC will be zero. However, as a reader pointed out, this is very pessimistic in real life. If you analyze a very small amplitude signal summed with some white noise, then the FFT calculation can bring signals out of the noise even if their amplitude is far below one LSB. This is due to the fact that the noise is "shared" between all FFT frequency bins, so each time you double the FFT size, the noise floor is lowered by 3 dB. And luckily noise is eveywhere.
January 3, 2007
During the final proofing stage of the editorial process, the editorial staff inadvertently inserted three typos into Tom Cantrell’s January 2007 article, “Hot Chips 18” (Circuit Cellar 198). Please note the following corrections:
Page 78, column 2, paragraph 3: “10 Kbps” should be “10 KBps”
Page 78, column 3, paragraph 1: “12.5 Mbps versus 10 Kbps” should be “12.5 MBps versus 10 KBps”
Editor’s note: According to IBM, the RAMAC transfer rate was “8.8 Kbytes/s” (http://www.research.ibm.com/journal/rd/255/ibmrd2505ZC.pdf).
Page 83, Table 2: "Xilinx Vertex-4" should be "Xilinx Virtex-4.
Our apologies go out to our readers and Tom Cantrell.
|
