Re: Prob. a very basic Q re: capacitors

"Joerg" <notthisjoergsch@xxxxxxxxxxxxxxxxxxxxx> wrote in message
news:u7Ljk.9173$L_.2723@xxxxxxxxxxxxxxxxxxxxxxx
Else it's mostly lack of knowledge. For example it seems that there are
hardly any engineers who understand POR/BOR circuitry anymore, let alone
guys that can design them. It took even seasoned semi mfgs years to figure
that out. Some haven't figured it out yet ...

Atmel's early AVR microcontrollers didn't have BOR included and it was touted
as a significant new feature when it was introduced... despite the fact that
many people felt it should have been included all along, as some of the parts
were a bit flaky unless you built your own out of discretes.

Which, ahem, is one of the reasons I rarely use them. Cost is a bit highish.
I mean, if you can do something with three 5c logic chips and 15c worth of
discrete parts then that's what I am going to do. I had a discussion with a
semi mfg about that and despite the fact that this was a virtual meeting I
could sense quite a bit of shock when I told them what circuits cost when
they are done with discretes.

OK, here's a real-life example of something that it looks like I'll be doing
in the near future -- let me know whether you'd go the discrete route or take
the more "integrated" approach: Hard-switching a 1MHz-6GHz signal between a
couple of inputs quickly (low kHz rate) produces sidebands in the output at
the switching rate, and the hard-switching (<100ns!) causes those sidebands to
be noticeably large. To mitigate this problem, the idea is to stick some
attenuators in-line with the inputs and, when switching from, say, input 1 to
2, first ramp up a variable attenuator to, say, 60dB on input 1 (with input 2
also severely attenuated at this point), hard-switch the output from input 1
to 2, and then ramp down the variable attenuator on input 2; the ramping takes
place over, say, 10-100us. (The hard-switching is still required due to
isolation requirements -- the attenuators alone probably aren't sufficient.)

This is really the age-old "key click" problem Morse operators have and the
"shaped carrier" solution they came up with to prevent it. As with that
approach, while a linear ramp on the attenuator is probably OK, having a
little more control over the exact shape is good too (something like part of a
sinc function comes to mind...). Since the bandwidth is so wide here, the
easy (but not exactly cheap) approach is to go buy some ultra-wideband
attenuator from the likes of Hittite and drive them with a DAC controlled by a
microcontroller. This makes it easy to play around with the attenuation
shape, time, etc., which is handy. By the time you're done, you're probably
looking at upwards of $20 in parts in low quantities.

With discrete components... presumably you first build your own attenuators
using JFETs in a pi- or tee-configuration, although that's perhaps a
non-starter bsed on having to pass 6GHz. Then you build the soft-switching
waveform generation using a few op-amps or something, and finally add a bit of
control logic to make it all happen. Cost is probably <$10? I doubt the
difference would make up for the engineering time, but of course education is
always worth something as well.

Thanks,
---Joel


.

Relevant Pages

  • Re: Prob. a very basic Q re: capacitors
    ... It took even seasoned semi mfgs years to figure that out. ... if you can do something with three 5c logic chips and 15c worth of discrete parts then that's what I am going to do. ... in the near future -- let me know whether you'd go the discrete route or take the more "integrated" approach: Hard-switching a 1MHz-6GHz signal between a couple of inputs quickly produces sidebands in the output at the switching rate, and the hard-switching causes those sidebands to be noticeably large. ... To mitigate this problem, the idea is to stick some attenuators in-line with the inputs and, when switching from, say, input 1 to ...
    (sci.electronics.design)
  • Re: Prob. a very basic Q re: capacitors
    ... guys that can design them. ... It took even seasoned semi mfgs years to figure ... discrete parts then that's what I am going to do. ... attenuators in-line with the inputs and, when switching from, say, input 1 to ...
    (sci.electronics.design)