Analog Shift Register design

[gstopp at fibermux.com]

     Hi DIYer's,
     
     Well another protoboard mass of wires has grown on my bench in the 
     lab, this time it's a four-stage analog shift register. Really it's 
     just four sample-and-holds in a row that are clocked in a 
     bucket-brigade fashion, with a variable shift rate and an external 
     shift input option.
     
     (This is another circuit inspired by taking pieces of other circuits, 
     some from Electronotes, some from databooks, some from my brain, and 
     piecing them together in different ways. My main goal is a reduced 
     parts count for easier building. This time list member Steve Varner 
     asked about some chips, one of them being the LF398 Sample and Hold, 
     and I told him that I had bad luck with these in the past. Well, 
     knowing how I've screwed up in the past, I tried it again in a new 
     circuit, and it works great.)
     
     The circuit consists of four LF398 Sample and Hold chips, arranged in 
     a row output-to-input. The sample triggers are provided by a TTL shift 
     register, clocked by a fixed 2 Khz oscillator. A 555 timer is used to 
     provide the sample command, and it has a switch that determines 
     whether it oscillates by itself at an adjustable rate, or conditions 
     an external trigger waveform.
     
     The fixed 2 Khz oscillator is built around 1/2 of an LM358 low power 
     op-amp, powered between +5 and ground. The traditional linear databook 
     single op-amp "square wave generator" configuration is used, with 
     component values chosen for 2 Khz. The output of this oscillator 
     clocks the shift register.
     
     The shift register used is a 74HCT164 8-stage serial-in parallel-out 
     shift register. Both input pins are tied together and the enable is 
     tied high. A set/reset flip-flop, made from a 74HCT00 quad NAND gate, 
     has its Q output connected to the shift register serial input. The 
     first parallel output of the shift register (pin 3) is tied back 
     around to the "reset" input of the flip-flop, thru an inverter (one of 
     the remaining NAND gates). When the flip-flop is set, a "1" is 
     presented to the shift register serial input, which gets clocked on 
     down the eight parallel outputs. The first output it hits resets the 
     flip-flop, so only a single "1" is sent down the line. Parallel 
     outputs 2, 4, 6, and 8 go to the sample control inputs on sample and 
     hold units 4, 3, 2, and 1 respectively. In this way the output of any 
     of the sample and hold units is passed to the one after it before it 
     gets a new sample command itself. This is how the bucket brigade is 
     implemented.
     
     A 555 timer is set up as an astable multivibrator, and is capacitively 
     coupled into the "set" input of the flip-flop. The oscillation rate of 
     this circuit is variable between 0.1 Hz and 140 Hz. Therefore with 
     every cycle of the multivibrator, a "1" is passed down the shift 
     register and the sample and holds update. The other half of the LM358 
     is used as an input conditioning comparator, with a switch on the 555 
     that allows the internal rate circuitry to be bypassed and external 
     triggers to fire off the shift register.
     
     The LF398 is an 8-pin DIP package with an analog input, an analog 
     output, a sample control input, and a holding capacitor pin. The chip 
     is powered off of +/- 15 volts, and thus the analog input range can be 
     within the full 30-volt power supply range. Since the shift register 
     is being clocked at 2 Khz, the sample pulses are 500 microseconds 
     wide, and a 0.01uF holding cap was found to be a good trade-off 
     between charge time and droop.
     
     A possible drawback of the design is the need to have a +5 supply for 
     the TTL as well as +/- 12 or 15 volts. I have power supplies out the 
     wazoo so it's not a problem for me, but some builders may not be so 
     lucky. +/- 5 volt operation would probably work as well.
     
     If you want a copy of the schematic, email with your fax #.
     
     - Gene
     gstopp at fibermux.com