[time-nuts] Cutler NAA on 24.0kHz....
csteinmetz at yandex.com
Mon Aug 18 14:34:00 EDT 2014
>Nat Semi App Note 72 page 18, par. 6.4 shows the configuration for
>bandpass active filter. This matches the last LM3900 stage, so you
>would seem to be correct. The shift in filter frequency for 200bps
>is because the higher modulation rate results in a greater frequency
>shift. It's like 50hz instead of the 25hz of the 100bps rate.
>>It's simple, but not obvious. The LM3900 is a Norton amplifier, and
>>while it has differential inputs they are current
>>driven. * * * Both the upper amplifier and the second lower
>>amplifier have 1M feedback resistors, and + inputs fed 10V by 1M
>>bias resistors. That would bias the output at near the supply rail,
>>turning these stages into something like half-wave rectifiers.
>>Since the first lower stage has a 2M bias resistor it idles at
>>about half supply, and behaves as a simple
>>inverter. * * * combining the two outputs produces a negative
>>going full wave rectification of the signal. The fourth LM3900
>>stage looks like an inverting bandpass filter, but I'd have to dig
>>out some reference books to determine its behavior in more detail.
>>As f or the 100-200 switch I'm confused, why would the bandpass
>>frequency be lowered for the higher modulation rate?
The circuit as a whole operates as a frequency doubler using
full-wave rectification and filtering. The rx LO is 100Hz below the
nominal carrier frequency, so in "normal" (non-MSK) mode, the IF
frequency is 100Hz. Referring to the MSK addendum, a received 200
baud MSK signal is 50Hz below nominal, and a 100 baud MSK signal is
25Hz below nominal. With the LO 100 Hz below nominal, this makes the
IF frequency 50Hz when receiving a 200 baud MSK signal, and 75 Hz
when receiving a 100 baud MSK signal. After doubling, these become
100 Hz (200 baud) and 150 Hz (100 baud), so the BPF is switchable
between 100Hz and 150Hz. They used a FET to chop the 150Hz (100
baud) signal with a 50Hz square wave.
I can't say I'm impressed with the design, even for the era. The
whole instrument is built mostly with LM3900s, which makes it
thousands (maybe even millions) of times noisier than it would be if
it had been properly designed with standard op-amps. It may work
more or less, but it's a fugly way to get there. There are other
questionable choices (like the FET chopper, an overall design that
depends on lots of one-shots, etc.). The designers knew about the
LM301 (there is one in the unit), so there was really no excuse for
using LM3900s. Yeah, the 301 was more expensive -- but this was
supposed to be a state-of-the-art measuring device for characterizing
good OCXOs down to PPB or below.
I simulated the MSK board in LTspice. Let me know (OFFLIST ONLY,
please) if you would like the files to play with (662kB ZIP
file). (Note that these won't do you any good if you're not an
LTspice user.) Again, please do not clutter the list with requests
for files -- OFFLIST ONLY, please (check your headers carefully
before you hit "Send").
More information about the time-nuts