Wide range LF to HF and low VHF sweep generator.
Uses a linear sweep with independantly set  DC baseline and peak controls
Using ECL logic and ECL vco chips and   analog computer concepts.

by Ralph Klimek  1998

The motivation for wanting a sweep generator in the first place come from my occassional need to align IF amplifier strips and the occassional wide band filter. I wanted something "simple", a very linear sweep, sweep end points that could be precisely defined. A side feature was to be a sawtooth generator with linear sweep whose DC base line and peak amplitude could be set independantly, unlike simple schemes that define the sweep in terms of offset and gain.

I wanted to satisfy the following design criteria.

The final outcome satisfied some of these criteria!
The design cannot be reproduced because it is now impossible to obtain the MC1658 vco. In any case, modern design would use a combination of PLLs and direct synthesis and microprocessor control.  Its allmost impossible to source ECL logic in this day. This should be a cause for great sadness.

The motivation for this particular implementation arouse after acquiring some discarded PCBs from  old fashioned washing machine sized disk pack drives. The analog boards also contained a large amount of Motorola MECL MC10xxx logic.  There where some MC1658 wide range VCO chips as well.  These remarkable devices can be made to oscillate from VLF to VHF with one control voltage and requiring you to change only one timing capacitor to give over two octaves of vco control. The chip is based on an ECL RC multivibrator and as such has a large noise bandwidth, much larger than an LC and varactor design. The classic L C and varactor design has typically only half an octave span.  The large noise bandwidth was not a concern for me  as I would be alligning only LC bandpass filters.  

ECL is a much maligned logic family. It does not deserve the bad press. Motorola ECL has been around since the early sixties. Sadly they only sold their output to "black" millitary buyers and by the time Motorola has licensed the technology to NEC and others, Texas Instruments had the whole world hooked on their totally inferior 74xxx TTL logic family.  clever marketing arms had it put out that ECL was "too hard" to design with. It is a great pity that early eighties minicomputers were made with 74Sxxx seris logic instead of MECL. Had they used MECL they would have had 10 times the mippage at near zero additional design cost. Seymour Cray had seen the light, the  CRAY1  was made from MECL gates  and the major governments of the world beat a path to his door.  

ECL is EASY TO DESIGN WITH. No ifs and no buts. Simple fan in and fan out rules, every gate output is a line driver, every node is a transmission line (that must be terminated at both ends). A good ECL gate can toggle at least 100Mhz (mid sixties technology here, the best mid sixties commercial logic RTL could barely toggle at one megahertz).  The last generation  MECL111 could easily do 400Mhz in a 16 pin DIP (with a little bit more care to be fair)  The only valid objection to using ECL is that it is very power hungry, that much is true.

I would use MECL to first generate my signals and MECL logic to move it about and switch it. It takes a true tried and tested electronic engineer to get 74Sxxx logic to do anything usefull at 50Mhz but a mere technician/tinkerer like myself can make MECL sing and dance at 100Mhz.
In this design I reused the many MECL gates that I salvaged.

The heart of the design is the linear sweep generator. The ramp is generated by the tried and true method of charging an integrator from a virtual constant current source driven by a voltage reference chip. At the end of the charging cycle the integrator capacitor is discharged by a mosfet.
The integrator reset signal is generated by a comparator that measures when the ramp has reached  the desired level.  This reference level can be switched to directly drive the vco level permitting the span of the sweep to be directly "dialled up" on a an external frequency meter.  A calibrated DC offet signal is added to the ramp before  the reset comparator. This offset voltage may also be switched to the vco so the base line or "zero" frequency of the sweep may also be directly "dialled up" from the ten turn pots on the front panel. There is no interaction in these settings! Unlike the usual mutually interacting adjustments of gain and offset method.  Some designers may have picked out the fact that by doing the sweep endpoints this way I lose a measure of control over the sawtooth period.  As this sweep generator is meant to be used with a compainion oscilloscope, this is not objectionable because the sweep generator also provides a nice horizontal sync pulse, so the actual sawtooth period is not that important.  The comparator, however, provides only a comparison level, not a pulse, and the integrator must be reset with a pulse.  A simple monstable provides a good narrow reset pulse.  For initial debugging and testing dip switches select the reset pulse polarity or a 555 free run pulse .  

I have chosen 5 differant and slightly overlapping sweep ranges. These are created by three MC1658 vco chips, the lower frequency ranges being selected by adding timing capacitance with relays.  The highest ranges are very sensitve to stray capacitance, so relays could not be used, instead a dedicated MC1658 was used . The outputs are selected and routed by ECL logic gates.
A reference counter generates 10M, 1M and 100Khz calibration pips which are mixed with the generator output with an MC10107 XOR gate, the output DC smoothed and displayed on the secondary oscilloscope channel to provide a graduated "pip" display.  Unselected MC1658s are handled by driving an internal voltage reference decoupling line to ECL Vee, this is not the intended use of this pin, the chip has no deliberately designated "chip select" signal. I disable unselected VCOs so that they dont free run and generate spurious signals, a trace of which may reach and contaminate  the output.

Frequency range is determined by a simple rotary switch that feeds a priority encoder to generate a  binary control word. The various logic control signals are derived from this control word and 74151 muxes to create the abitary unary logic control signals from this control word. This arrangement permitted me to corect errors in logic design by simple re wiring rather than by a complete redesign. (also known a microprogramming in more elite circles)

Second thoughts.
The ramp generator could have also been implemented using a variable precision clamp and followed by a  variable gain element. The ramp period could then be independantly set. it would have been difficult to provide a set level for the gain element to permit setting the sweep endpoint with an external counter. Doing so would have required four-quadrant multipliers which are a pain to design , calibrate and use.  The MC1658 VCO is really an RC multivibrator with a controlled current source feeding the internal timeconstant. As a result, this oscillator has a rather high noise bandwidth, it would be much lower if real LC resonators were used in a standard varactor controlled VCO. As it is, this sweep is satisfactory for tuning wide band filters, but useless for narrow band applications like setting up crystal filters. The ramp generator has proved to be very successfull.

individual vco sweep oscillators, signal selection, vco selection and marker pip generation

control logic, rotary switch drives 74148 priority encoder, binary word decoded by 74151 muxes

The ramp generator

misc notes to myself


constructed on wire wrap cards recovered from a Burroughs B340 seris disk pack drive controller