Can We Put Other Instruments on a Chip? Part 2

In part 1, we looked at the function generator (FG) and at the versions available as an IC.

Some of the available devices, such as the Maxim (the sponsor of this site) MAX038, have decent performance. Why is the MAX038 not in a commercial instrument? Perhaps it is because of the THD of the sine-wave, specified at an adjusted 0.75 percent. This is rather high; a typical instrument such as the Tektronix FG502 from the late 1970s is at 0.5 percent from 10Hz to 50kHz. Long ago, working on FG design at Tek, I found that these kind of sine-shapers would produce as low as slightly under 0.1 percent THD and be typically in the 0.2 percent to 0.4 percent range.

Using newer methods based on Barrie Gilbert's multitanh or other translinear sine-converter circuits, this can be reduced to under 0.1 percent and extended to higher frequencies. These newer shapers are eminently integrable and have been in the published literature for decades, yet (as far as I know) have never been used in a commercial FG. The time is ripe for a new generation of FGs, despite the growing emergence of waveform generators (WGs).

Note also that the output amplitude of the MAX038 is 2V, pk-pk. This is a good level for driving an output amplifier, and though the output could be terminated in 50Ω (for impedance-matching to cables), a viable specification for output voltage is more like 10V to 20V, pk-pk. Despite this, the MAX038 has a frequency range that is competitive with other FGs. With DACs for control of amplitude and offset in post-processing, the MAX038 still might have some hope in the FG market. But alas, Maxim discontinued it.

An anticipated MAX2206 would improve the sine converter and possibly add more of an output amplifier which might not be needed with fast, medium-power op-amps such as the TI (NSC) LM6181. Its 100mA output produces ±5V into 50Ω and ±10V into an open output. The MAX2206 would bring the integrated FG into the third generation with a commercially viable FG instrument core.

Will Maxim be first to market with a third-generation FG, or will ADI? The ADI AD9833 12.5MHz (25MHz clock) WG is like the MAX038 in that it shows potential to expand into a full WG. However, at 10 samples per cycle, it is a 1.25MHz WG with <1V, pk-pk output voltage. Maybe Intersil will regain its former status of having been first in first-generation FG ICs. Any of these companies (and others) could do it.

Why do this when the FG is being replaced nowadays by WGs? WGs — digitally generated waveform sources — have some advantages over analog generators, such as accurate waveform values and essentially no drift other than what is in the DAC and amplitude processor following it (which is not negligible).

The main disadvantage over FGs is the resolution-speed tradeoff. The discontinuous nature of WG waveforms requires oversampling to achieve adequate resolution and thus high clock rates. A WG with a resolution of 5 bits (3.12 percent resolution) at 20MHz (which is of comparable waveform accuracy to a good 20MHz FG) requires sinθ/θ sampling compensation (which can precede the DAC and be digital) and then analog filtering. At a 200MHz clock rate, the decade of rolloff of the filter requires 30dB of rejection, a value not hard to achieve. At a higher clock frequency, however, noise abatement becomes a bigger problem. It is more likely that the clock rate will be 100MHz or less and with a filter frequency transition of 0.7 decade (log10 5), the post-DAC filtering begins to be more of a design consideration.

FGs, on the other hand, produce inherently continuous (analog) waveforms and present no such additional complication. The main challenge is in reducing loop delay of the TWG to achieve 20MHz. As the TWG frequency increases and loop delay, td , adds 4·td to the ideal (zero-delay) TWG period, T , then the amplitude also increases at a rate of

This is somewhat like a zero in the dynamic response in that the magnitude increases with frequency (but it is not a zero; it is a nonlinearity) and to maintain some semblance of amplitude accuracy, td  << Tideal  = 1/20MHz = 50ns. Rob Baetke at Tektronix patented a scheme that reduces the peak thresholds of the TWG with frequency, since they are predictable from the above formula. This can allow for more loop delay (or higher frequencies) at the expense of a model-reference adaptive control scheme. With μC-based instruments and DACs to set the (symmetrical) threshold, this is an easy scheme to implement. Consequently, FGs still have life in them and are not ready to be relegated to the electron-tube era.

TPAs present the challenge of wide current and voltage ranges, but SMUs in them need not have ranges beyond what can presently be integrated. This is a topic to consider in a forthcoming article.

Related posts:

1 comment on “Can We Put Other Instruments on a Chip? Part 2

  1. D Feucht
    October 30, 2013

    Ach, the equation giving the rising amplitude of a TWG with frequency should have been

    1 + 4t sub d/T sub ideal

    As the total delay time in a triangle-wave cycle, of 4*t sub d, becomes significant relative to the cycle period without any delay in it, or T sub ideal, the amplitude increases because once a peak is detected (which occurs twice in each cycle) the loop takes some time to shut off the on current source and turn the off source on, to reverse the slope of the waveform. The amount of time taken as overshoot must be “undone” going in the opposite direction until the waveform is back to where the peak should have been. And that is a total of 2*t sub d, where t sub d is the delay time. This happens for both peaks – hence the 4 multiplying t sub d in the equation.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.