Forgot your password?
No account? Register one!
March 12, 2008
Product PRT-12070 |
about 2 months ago
Ditto to what Doug says. The Radio Shack boards 276-150 and 276-168 put 2 power rows down the center where they should be to minimize power and ground inductance (wire length), and to optimize and simplify bypassing. They have provisions for single row headers on all sides, or dual rows if you cut traces between rows. They are cheaper and usually available locally.
The larger RS board has 3 rows of ICs. Having it wider, not longer, fits better in most rectangular enclosures.
I think Radio shack got it right and you guys missed the mark so I’ll continue buying their boards. This is also why I usually go straight to solder breadboard and rarely use solderless breadboads. If I need to plug in components, even R’s and C’s, I use nice machine-contact IC sockets or socket pins. Then when I’m done, I have something solid and permanent, not just to take apart. My projects are at www.djerickson.com
News - Pit of Despair |
about 2 years ago
Thanks for the article, Nate. I agree with you and Kevin. I too have searched for and tried to define the holy grail of a universal hardware interconnect. SPI and I2C are appealing for this since so many chips use them. But when I built real systems with I2C and SPI I learned a few things, mostly the hard way. These interconnects are intended for chips on the same board and with a very good ground interconnect. When run through wires, they radiate and pick up EMI and noise by causing high speed return currents to flow in the grounds. When static (ESD) strikes, you can get errors, also induced across the grounds. And since SPI and I2C have no error detection or correction, you are in trouble when an error occurs. Controlling high current loads (without isolation) can also cause EMI and bit errors.
Another downside of I2C: the rise-time and thus bus speed is a function of the capacitive loading which is a function of the accumulated wire and PCB trace lengths and the number of loads. Many larger I2C systems need to be tuned to get the pull-up resistors right.
For system-level interconnect in large systems, CAN is often used. CAN is differential so it generates far less and is far more immune to EMI. It is impedance controlled so it can use long wires. It has built in error detection and correction. It has built in addressing and slave response prioritization but without the time uncertainty of multiple-retries. 1 megabit per second is typical. I think there are standard CAN connectors, not sure though. The downside of CAN is that peripheral chips don’t talk CAN. Each peripheral pretty much requires a microprocessor and CAN driver, adding a few $$ and some development time to each node.
Product KIT-09612 |
about 2 years ago
Crosstalk can be from a number of sources, but I suspect yours is from grounding. Whenever you use multiple amplifiers from a single power supply, each amplifier’s large ground return current causes a small audio voltage to appear on every other amplifier’s input ground. This appears as input signal and gets multiplied by the amplifier gain (26dB) and so, crosstalk. I have spice simulated both the cause of and a solution to this problem. Typically if you use multiple amps operating from a single supply, the crosstalk is about -40 to -50dB, which you can hear in the quiet rooms.
The brute force solution is to use differential inputs or transformers on each amplifier input. The solution I used is to isolate the input signal ground of each amplifier (input connector, input network, feedback return) from the output grounds (speaker and power supply returns) and connect them only at a single point, back at the audio sources. This only works if the sources (preamps) are fairly close to the amps.
My DIY house-wide audio system uses two LM3886 amplifiers for each room and uses this technique with good success. To see details, go to http://www.djerickson.com/stm32/preamp1.html and search for “ground trick”.
I don’t yet know how to do this for the STA540. It uses a clever internal configuration that may make this technique hard to implement. However the SparkFun board does have a separate input op-amp that may be re-configurable as a differential amplifier.
As far as the noise, the STA540 has about -50db of power supply rejection at 300Hz and higher frequencies so you may hear power supply noise on the output. Try one or two large (10,000uF 25V or 35V) caps across the power supply to see if that helps. Watch the polarity!
No public wish lists :(