An Instrument on a Chip? A Look Back

The story of analog integration in test instruments is long, storied, and rife with names renowned in the annals of analog design who innovated upon each others' design to realize the foundation of today's circuits.

So what is the state of integration of the analog circuitry in test instruments? Measurement on a chip started long ago and continues with Intersil, Maxim, and other companies offering integrating ADCs with auto-zeroing capability in 24 or 40 pin ICs.

They have been around for decades as the core of digital panel meters (DPMs) and digital multimeters (DMMs). DMMs as instruments lead the integration effort; cheap ones sell new for as low as $10 US. To achieve the voltage range required of a typical instrument, discrete resistive dividers are used, along with off-chip capacitors, fuses, trim-pots, and switching. Cheap DMMs do not use the ADC in a DIP package but apply board-level chip-on-board integration, as shown below. The circular black epoxy dot encapsulates the DMM IC on-board.

The most non-integrated aspect of these cheap, handheld DMMs is the electromechanical switching. If you have taken apart a handheld DMM, you will have noticed that the big, many-position rotary switch also has many tracks (shown above as seen through the board) and several rotated metal contacts to connect adjacent tracks, sometimes on both sides of the board (or on multiple boards). This switch performs two functions: it is the main element of the human interface (or front-panel ) and it also embodies the encoding logic for the circuit interconnections that configure the instrument settings .

Another instrument to receive some integrative attention is the function generator (FG). Exar and Intersil came out with FG ICs as long ago as the mid-1970s. Exar's XR2206 and Intersil's ICL8038 integrated most of the transistor circuitry required except for the output amplifier. That portion of the circuitry is better implemented with an op-amp or discrete transistors.

These two FG versions have quite different circuit designs, each with its relative merits. An early, semi-custom IC foundry, Interdesign, offered Monochip Application Note APN-8 (by James Knapton) titled “Waveform Generator”. It included, as did the Exar and Intersil FGs, a triangle-wave generator (TWG) — the core of a FG – and sine shaper. The TWGs all produce a square-wave in digital form. None of these three FG TWGs had the same design, though all of them were capable of sloppy 1-MHz waveforms. The sine-wave at the high-frequency end of the instrument is cleaned up by bandwidth-limited circuitry, but the triangle-wave loses its sharp peaks and the square-wave starts looking more like a sine-wave.

Maxim introduced a next-generation FG IC in the 90s as the MAX038. (Apparently, someone at Maxim was more impressed with the 8038 than the 2206.) It was a major improvement and had instrument-quality capability. It had a maximum frequency of 20 MHz and also included a phase detector for phase-locking the TWG to an external frequency source. TWGs are a kind of voltage-controlled oscillator (VCO) — or rather, a timing-current controlled oscillator — and can be swept in frequency, making them easy to accommodate in a phase-locked loop. For reasons inexplicable to me, Maxim dropped this part from its repertoire of available products.

Back in the 70s, two talented engineers at Tektronix — Barrie Gilbert (now a Fellow at Analog Devices (ADI)) and Art Metz — each devised different schemes for converting triangle-waves to low-distortion sine-waves. These schemes are readily integrable. They are not (to my knowledge) being used in commercial FGs though they are superb schemes.

For a while, ADI had an FG-like device in their portfolio, the AD639 trig function generator. It was designed by Gilbert after he moved there. The device has since been dropped from their portfolio — also inexplicable.

The 8038 used diode breakpoint circuits to approximate the sine function with multiple linear segments. The XR2206 used the tanh function of BJT diffferential-pairs to approximate a sine. Gilbert took this idea much farther with his multi-tanh circuit concept. Metz did something very simple and clever using translinear circuitry invented by Gilbert, a consequence of creative minds in close proximity.

This brief historical survey of measurement instrument integration is but the first part of a much longer saga. In the coming parts, we'll transition from history to look at possibilities for future integration of analog circuitry for test instruments.

Related posts:

14 comments on “An Instrument on a Chip? A Look Back

  1. Dirceu
    May 22, 2013

       I wonder how the FPAA (field-programmable analog array) technology has advanced. It would be a way to build a medium quality home-made analog instrument? I think the higher integration would reduce thermal noise in the low voltage measurements.

  2. RedDerek
    May 22, 2013

    I remember the 2206 to the point that I built up a function generator using it as the core. Still have that function generator on the bench today.

  3. RedDerek
    May 22, 2013

    A while back on another forum, Michael Dunn mentioned of a nice integrated DMM to be picked up on eBay using the Auduino processor. This unit will figure out the device and tell you what it is. 3 terminal input, and it will figure out if there is a resistor, capacitor, inductor, diode or transistor. And then tell you about the orientation and basic parameters. The eBay number was 160943358553. Not a bad toy that I ended up buying two. For its functionality, it beats the typical DMM any day.

  4. D Feucht
    May 23, 2013

    The major IC companies such as Maxim or Fairchild have multi-GHz BJTs and equally fast (in their own way) JFETs and MOSFETs. In the 60s, a few companies put out dedicated transistor array ICs lacking the metal interconnect layer(s). The customer sent in an interconnect file (or drawing, actually) to complete the semi-custom IC design. In principle, this scheme for low-cost integration is appealing though it died out with the rubylith era and the advent of low-cost, wide-spread IC-level CAD. However, most analog circuit designers are not really good IC designers yet could interconnect transistors and some on-chip resistors. The resistor values are a real problem for versatility of design.

    Maybe the best process is to go back to the arrays, make them transistor-only with plenty of pins (or many array ICs, as in the past), and let the customer interconnect them off-chip. When the circuit is refined adequately, then an IC can be made.

    The performance difference between semi-discrete and monolithic circuits is mainly attributable to the additional parasitic elements of the discrete components. Thermal noise reduction is, apart from choosing low values of resistors to minimize Johnson noise, a matter of transistor characteristics: base spreading resistance, surface leakage and other process-related anomalies. These are problems for both discrete and integrated transistors. A big advantage of monolithic integration is the reduction in thermocouple effects on-chip, though Kovar leads to copper board traces still have some residual Seebeck effect.

  5. Vishal Prajapati
    May 23, 2013

    Sir DMMs are being sold here in India even at less than $3 pretty similar to what you have shown in the image. I have never been able to understand their economics. Don't know how they manage the retail price of $3 for whole product?

  6. Brad Albing
    May 23, 2013

    DF – you could go back to the arrays with a bunch of pinouts, but I'd be concerned that, even for preliminary testing of a concept, the parasitics as you've mentioned might be troublesome.

  7. Brad Albing
    May 23, 2013

    I did the same, but with the (now nonexistant) ICL8038.

  8. Brad Albing
    May 23, 2013

    I recall that blog. That was a nifty piece of test equipment, and a good example of integrating lots of analog functionality into a small amount of real estate.

  9. bjcoppa
    May 23, 2013

    Ahhh, those were the days…running manual multi-meter measurements and recording the data manually. As more applications are developed for mobile devices, I look forward to EMF sensing capability for smartphones that can automate electrical device testing and upload the data to a server. Well, heck if they have GPS and one can pay with a phone by scanning their credit card, the tech is probably at the lab level somewhere.

  10. amrutah
    May 25, 2013

    Dennis, Thanks for a interesting post.

      I have used and worked on IC's that have internal ADC's for measurement of chip parameters like internal voltages, current, power and so on to send the information to other devices or processor.  To an extent it can be used for testing (A DFT kind of device).

      I really dont understand the use of FG as an intrument on a chip?  Is it used for checking the behaviour of the analog by applying different signals?

  11. D Feucht
    May 26, 2013

    The reasons for integrating a function generator on a chip are the usual ones: to reduce parts count, cost, and size. FGs are versatile waveform sources and to be able to simplify their construction can reduce instrument prices, size, and power usage. Although waveform generators (WGs) – digitally generated waveform sources – are ascending, they cannot provide the frequency performance of a well-designed analog FG.

    As for the use of FG circuits as subsystems in larger ICs, this has not been explored much, to my knowledge. Any on-chip analog testing might better be perfomed by built-in WGs. To test circuits on-chip dynamically, however, is a topic at the frontiers of integration. Dynamic on-chip measurement would be useful for dynamic compnesation and adjustment of amplifier response, for instance, and would replace trim capacitors.

  12. Brad Albing
    May 28, 2013

    >>… dynamic compnesation and adjustment of amplifier response, for instance, and would replace trim capacitors. I can see applications for that for ICs that are doing dynamic cable compensation (coax or twisted pair), used for either video or high-speed data.

  13. WKetel
    June 20, 2013

    Vishal, I would seriously question the linearity and accuracy of a $3 DMM, since those are the things that add to the cost of production, as well as being more difficult to design. And it may be that they are the production-line “fallouts” from a maker of DMMs that normally sell for a bit more. Counterfeiting products does happen in more than one way. It would not be the first time that rejected items have been represented and old as the good product.

  14. WKetel
    June 20, 2013

    A function generator is used for many different kinds of testing and circuit development. Puting most of the circuit on a single chip allows the creation of a rugged and inexpensive piece of test equipment, as opposed to the earlier versions that  used quite a few discrete components to produce the various waveforms. 

    Evidently you don't do much real-world analog dvelopment work, although function generators can be very useful in the development of non-microcontroller digital development. NOT all of electronics is involved with software and controllers, after all.

Leave a Reply