• Content Count

  • Joined

  • Last visited

About gasstationwithoutpumps

  • Rank
    Frequent Visitor

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. The "headers as comment" button on the oscilloscope no longer seems to add the "#" character to the notes. Also, the metadata for the oscilloscope does not seem to report the amount of averaging done. Incidentally, why is averaging limited to 1k? We should be able to average any number of traces, since the storage for summation does not take up more space as the number of traces being averaged increases. With 14-bit values from the ADC, even 32-bit integer arithmetic should allow 256K (262,144) traces to be averaged, not just 1k.
  2. Using Waveforms 3.9.1 under Mac OS 10.14.2, in DEMO mode with Analog Discovery 2 being emulated, selecting the wrench icon for the compensation menu results in the program hanging and needing to be killed. The same action when an Analog Discovery 2 is attached performs correctly—this seems to be just an error in the emulation in DEMO mode.
  3. Ah, I had forgotten that the compensation has to be re-run whenever any parameters are changed. I wonder, though, how critical that is compared to the errors generated by using the wrong voltage range. At the very least, the user should be warned that measured voltages are out of range (or at the limits of the range) and that the setting for channel 2 should be changed.
  4. If the default setting is just double the amplitude and the same offset, then it should not be called "auto", which implies some sort of intelligent choice. Picking a frequency in the middle of the frequency range and making one measurement to determine the output voltage range would be the sensible way to set the range for "auto".
  5. Shouldn't the auto setting be measuring the DC offset of the channel, rather than guessing?
  6. In Waveforms 3.8.17 (beta), I have encountered a bug in the impedance analyzer when the offset voltage is not zero. I was measuring the impedance of 10µF capacitors with the adapter board at 100Ω, and I got perfectly fine readings when I used amplitude=1V, offset=0V. But when I switched to using amplitude=1V, offset=4V, the impedance was reported as essentially infinite (dropping down to maybe 100MΩ around 100kHz). I got this result with several different capacitors, so it is unlikely that poor connections were responsible. The short and open compensations were redone for the new of
  7. I have used the 10Ω setting and found it useful, albeit a bit risky unless the voltage is reduced to stay within current limits. The resistor too-large/too-small warning could be based on the computed magnitude of impedance, compared to the known magnitude of the reference resistor. If the resistor is not within a factor of 100 of the measured magnitude of impedance, then it is too big or too small. That isn't as good a test as looking at the measured voltages, but should come fairly close.
  8. I do not understand the algorithm being used for reporting "resistor too high!?" in the impedance analyzer. It seems to pop up whenever the phase is close to 0°, which makes no sense to me. The resistor is too large when the magnitude of impedance measured is less than about 0.01 times the reference resistance—the phase isn't really a relevant parameter. A slightly more sophisticated algorithm might look at the voltages of the two measurements and report the resistance being too large or too small when the relevant voltage (across the resistor or across the DUT) gets down to only a few
  9. I requested this feature a year or two ago, and they implemented it! I keep pushing for as much automatic metadata as possible in the files, and I've been quite happy with the responsiveness of the Waveforms developers.
  10. I've needed to use non-linear devices to model loudspeakers (for example, an inductor-like device with impedance (j 𝜔)^𝛼 M (instead of j 𝜔 L). I'm sure that Octave, Matlab, and SciPy can be set up to optimize parameters for complicated functions, but I've found it simpler to use (and teach) gnuplot's model fitting command.
  11. Steps = (stop-start)/step_size samples=1+ (stop-start)/step_size steps = (log10(stop) - log10(start)) * steps_per_decade samples = 1 + (log10(stop) - log10(start)) * steps_per_decade This works exactly as expected, if someone requests an integer number of decades, which is a fairly common use case. The main question is how to handle non-integer results. For example, if someone asks for 100 steps per decade from 20Hz to 400Hz, there are 130.1 steps desired, which rounds to 131 samples. If you start at 20Hz and make the ratio exactly 10^0.01=10.02329, you ge
  12. When sweeping frequency in the network or impedance analyzers, the default number of samples is 100, but this usually results in awkward values, as it corresponds to 99 steps (fencepost error). It would be better to make the default be 101 samples. It might also be nice to have a way of setting the step size, rather than the number of samples. For linear spacing, this is just a value in Hz (for example, 10Hz–100Hz in steps of 1Hz is 91 samples). For logarithmic spacing, specifying a ratio is awkward, but steps/decade is simple. For example, the 1 kHz–1 MHz at 50 steps/decade is 151
  13. Although there are no specs, I have made some measurements of the Impedance Analyzer board, reported at It seemed to me that the Impedance Analyzer was built using 0.2% tolerance resistors (or that they got lucky with 1% tolerance resistors). The 10Ω reference had lower accuracy than the others, possibly because of contact resistance in the relays. If you are careful to pick the right size reference resistor and do open/short compensation, you should be able to get 1% accuracy from the measu
  14. Thanks for making a Mac beta. I've downloaded 3.8.8 and will try it out this week.