Hewlett Packard 8640B Signal Generator

 I decided to add a few words to explain a little about the HP8640B that I'm overhauling.

The chassis carries two pieces of test equipment, a signal generator and a frequency counter combined together to give you an accurate readout of output frequency which is very useful and far superior to similar equipments where a separate counter or complicated reference procedures are needed. The frequency counter can also be used independently. The basic 8640B covers the range 500KHz to 512MHz (plus a little extra for the benefit of overlapping ranges) but can be extended (in the factory) to additionally cover the range 512 to 1024MHz. Something which I found at first slightly puzzling, is that the range switch has this extra setting, and although this range can be selected, and the tuning knob shows the higher frequencies on the display, the RF output stays the same as the previous setting of 256-512MHz. I suppose one could use the 2nd harmonic which is present at a lower level.

Many signal generators use loads of mixers, but not this.. which instead relies on digital circuitry to progressively divide the oscillator frequency to match each range. Because these digital outputs are square waves a lot of filtering is necessary to reduce harmonics so, coupled to the range switch are matching filters which are designed to reduce the harmonics and to end up with a respectable sine wave for each range. Clearly this method of operation must employ an oscillator to match the highest range, so HP elected to use a cavity-tuned oscillator covering 256 to 512MHz. The cavity is tuned by a mechanical arrangement covering something like eight turns of the tuning knob and each range has a decent overlap so in fact the cavity oscillator tunes something like 230MHz to 550MHz.

The cavity oscillator output passes to a digital divider chain consisting of a set of nine divide by two stages in series providing x2, x4, x8, x16, x32, x64, x128, x256, x512. The lowest range uses all nine stages and covers about 449KHz to 1074KHz which represents 230/550MHz divided by 512. The next higher range has divide by 256 giving 898KHz to 2.148MHz and so on. The width of each range does result in a major problem in that the filter for a given range needs to perform two contradictory actions. The filter needs to pass say 1MHz but attenuate the second harmonic of 2MHz, but that same filter, when the oscillator is at the high end of the band, needs to pass 2MHz with next to no attenuation. That, alas is impossible unless the filter shape changes with input frequency, which I imagine is theoretically possible, but mighty difficult to implement, so HP just use two filters. One filter known as the Low Filter operates in the lower half of the selected range and attenuates the second harmonic and higher and the High Filter does the same, but starting midway through the range. Coupled to the oscillator tuning control is a potentiometer which signals to the filter assembly that relays should select either the Low or High filter. Every range has a low and high filter but lower ranges additionally add another (the next higher range filter) to help attenuate higher harmonics. This feature is necessary because, as the number of dividers increase the pulse train will have squarer edges and less symmetry. For example the tops of the pulses may be wider or narrower than bottoms resulting in a distorted sinewave (stronger even harmonics) plus the squarer shape of the pulse train will result in more and stronger odd harmonics, so to improve the lower frequency performance more than one filter is employed in series.

What about stability? Ordinarily a design would be based on an oscillator having relatively low frequency because its easier to keep stable but the HP VHF oscillator is inherently very stable at the expense of a rather large mechanical layout. The 8640B does however incorporate a clever technique whereby the master oscillator can be locked to a 5MHz crystal or, if desired, an external reference.

The 8640B has a level control system which normally places at the input to its attenuator an RF voltage which results in precisely 10dBm of RF output with the output attenuator set to +10dBm. This is achieved by an AGC system whose input comes from an RF detector and whose output drives an amplifier. HP call this amplifier a modulator and it takes the form of a hybrid held within a case that looks like a large gold TO3 can. Following the modulator is a second identical-looking hybrid carrying a 16dB RF amplifier. Within this second hybrid is the RF detector whose output is flat from 450KHz to 550MHz. The detector provides a negative voltage whose magnitude exactly relates to the RF output of the hybrid which drives the output attenuator. If the RF output rises for any reason the detector drives it back down via the modulator to reduce its output. The AGC system is straightforward in principle, but has a number of extra features. Firstly a boosted RF output is possible, taking it to almost +20dBm by modifying the loop parameters when the output attenuator is switched to +20dBm. Secondly, one can adjust the output to correspond with meter readings, and thirdly AM or FM can be selected which require changes to the AGC system and to the circuitry which handles the output meter.

 Next, the output attenuator. My example is calibrated from +20dBm to -130dBm in 10dB steps with a centre vernier potentiometer which is incorporated within the AGC loop and allows the user to set the desired output level within the attenuation range. To aid vernier adjustment the better one of two meter scales is automatically selected, and via some clever switching one of a pair of lamps is illuminated to advise the user which meter scale is in use.

 I'll add more once I've checked out and understood other features...

 Return to the HP8640B overhaul