An Instrument on a Chip? Some Emerging Instruments & the China Factor

Which measurement instruments have yet to be integrated? In my previous article, I did not include an emerging instrument: the waveform generator (WG), a partial successor to function generators (FGs). They digitally synthesize and process waveforms that are then converted to continuous (analog) form by a fast DAC.

Analog Devices Inc. makes the AD9833, applying direct digital synthesis (DDS) followed by a DAC. It outputs the classic three FG waveforms (triangle, sine, and square) at rather low amplitudes.

More WG ICs with more integration can be expected. To drive the usual 50Ω cable from the 50Ω source, an on-chip output amplifier needs to source and sink at least 100mA to produce the usual ±10V output. The power dissipation works against the TC tracking of the precision R-2R dividers of the DAC, though it should not be a problem for the DDS. Inclusion of the output amplifier is possible for a WG and would be much worse for a classic FG, where the four-decade range on frequency sweep for a given full-scale frequency corresponds to a voltage range within the triangle-wave generator of four decades. A 1V full-scale control of frequency has a zero-scale voltage of 100μV. Trans-chip thermals can interfere with such circuitry.

Now consider a complicated instrument: the oscilloscope, the mainstay of electronics benches. Pop the lid on a scope, and you will see nonmonolithic subsystems. The human-machine interface, or front panel, is of course a basic limitation. The entire waveform path of a scope is never built from either CRT or LCD technology, though active devices are integrated into the better LCD panels. For instance, input attenuators have sheet-metal shields — without which the volts/division knob would go down only to something like 0.1V/div. Any lower range would encounter too much radiated noise. The power supply also has large magnetic components.

Before the digital storage oscilloscope (DSO), analog scope integration was increasing. Vertical amplifiers were reduced in the 1970s to preamps, channel switches, and CRT-driver amplifiers with considerable off-chip support. Thermal distortion was compensated by external RC networks. Silicon integration of the high-bandwidth delay line remains a formidable challenge, and its frequency compensation is non-exponential. The horizontal axis was also capable of significant integration, with a trigger generator IC followed by a sweep generator IC and then a horizontal output amplifier that drove the CRT plates. An additional IC handled sweep control logic.

The functions contained in a DSO lend themselves to integration by the very nature of the scope — they are digital. The horizontal functions are replaced by digital circuitry. The trigger input waveform, routed from the vertical amplifier, is continuous and is turned into an edge by a fast comparator and a DAC trigger level. The sweep generator (or time-base) is replaced by fast digital counter-timer circuits, with count modulus and sampling rate set by a μC. One grand benefit of DSOs is that the display need not be written to in real-time.

Most of the analog in a DSO is in the analog front end — the analog circuits before the digitizer. Much of this circuitry can be and now is integrated. What remains is the attenuator, though it too shows possibilities for greater integration. Long ago, attenuators were completely passive networks. They were voltage dividers made up of precision resistors and capacitors (the capacitors for AC compensation) and housed in a sheet-metal shield (enclosure). The shields, intended to minimize interference from electrical fields, are hard to integrate, but they could be reduced to the point where all they enclose is an IC.

In the 1970s, H-P scope engineers began combining multi-path amplifiers with attenuators. That will be a topic for another article.

Not much circuitry is left in a DSO that is not in ICs. That makes it easier for new companies to bring DSOs to the market. Tektronix built a leading-edge IC fab facility in the 1970s, with bipolar junction transistors having fT values of a few gigahertz. That was leading-edge technology at the time — ahead of other semiconductor companies.

Eventually, it was not feasible for Tek to cloister all this technological capability only for scope manufacturing (though it put the company in the driver's seat competing with H-P in analog scopes). Now DSOs are being made in China. For its low-end DSO line, Agilent rebrands scopes from Rigol. It sells two-channel, 100MHz DSOs for $399. This would not be possible without extensive, nonproprietary analog integration. Two more rising stars in China making DSOs are Owon and Atten. Both are selling competitive measurement instruments.

Related posts:

15 comments on “An Instrument on a Chip? Some Emerging Instruments & the China Factor

  1. Scott Elder
    June 5, 2013

    I guess with scopes, the absolute accuracy of voltage is not as critical as with waveform generation.  I would think though with switched capacitor techniques, the thermal drift problems associated with resistors goes away.  So it seems reasonable that higher levels of integration should be possible.

  2. Scott Elder
    June 5, 2013

    Dennis — It is an interesting thing to ponder about who controls the value in such a product.  When the world was all about amplifiers and data converters, everybody had their own domain.  Semiconductor companies would make the pieces, equipment companies would find clever ways to put them together.

    But once the cleverness is moved to the chip, then one either makes the chips in their products or they compete solely on price.

    If you look at how successful fabless companies have become (i.e. Qualcomm) along with readily available semiconductor manufacturing and test, one begins to wonder how this will all shake out in the end.

    I think companies like Synopsys and Cadence believe the model will ultimately be:

    1.  Companies with hard/soft IP (drop in layouts), analog and digital

    2.  Companies with tools to assemble the IP into a database.

    3.  Companies that then manufacture the database and test the products.

    This is a substantially different picture than today where the IP in analog is designed assuming that it drops on a board and is useable for every possible application.  There is a lot of overkill in 8 pin op amps because of this reasoning (2V-40V, 0uA-20mA load, 1fF to 1nF load stable, etc.).


  3. Barry Harvey
    June 5, 2013

    Sure, as an engineer I'm lead to believe all integration is good.  Too bad none of the synthesized waveform generators can compete with the good old HP 8116A function generator, which used diodes to shape sinewaves out of triangle waves, and had better harmonic distortion at 10's of MHz than most digital synthesizers.  Really big output swings too, more than today's.

    I suspect the medium power/current pnp's that support the output stage are not available anymore, hence the drive to digital synthesis.

    We could argue about the analog quality of today's digital 'scopes relative to the Tek 'scopes of the later '80's, but that takes too long.

  4. bjcoppa
    June 5, 2013

    As noted at the end of the article regarding the production of these ICs, there was once a time when this technology was newer and fab costs were considerably less that it made sense to produce one's own chips. However, times have changed as fab costs have risen to over $3 billion for high-end Si fabs and traditional ICs and even more advanced ones are now increasingly outsourced to foundries in Asia.

  5. D Feucht
    June 5, 2013

    Scott – Our brains must be tuned to the same frequency.

    While information product (IP) integration by end-users to craft semi-custom ICs is an interesting ongoing growth path, I have an article in the queue for later this year suggesting that perhaps the integration expansion path for semiconductor companies is to start efforts to build things that have large-integration advantages. A concrete example is, of course, the market sector I am pursuing in this article.

    My view is that if a company has the technology to build X and X has some possibilities in the marketplace, then even if X does not come in an n -pin package, it is still an opportunity. There are, of course, caveats …

  6. D Feucht
    June 5, 2013

    Barry – Being steeped in the old Tek culture, I cannot agree more. All that is said or hoped for regarding the prospects for greater integration rests on the assumption that there will be engineers who are excellent at circuit design. I cannot say that the trend in the Trilateral world is toward greater engineering acumen. It will also take some time for the developing world (in which I now live) to produce the future Barry Gilberts, Bob Widlars, Howard Vollums and {your electronics hero here}. On the other hand, there still are some really good analog engineers around (some of them from Stanford U., even!).

    It is quite possible to improve on the reduction of power in the output amplifiers of former H-P pulse and function generators (especially PGs) given the same output specs. I'll not take that detailed subject up here, though it is worked out in my Instrumentation Notebook, a future book project. If this is of particular interest, get ahold of me by email and I'll scan that notebook page.

    Scopes are not up to the quality of the old ones. The people who designed in the past had a certain passion – a certain personal investment – in achieving perfection that is lost in a who-cares society. There is no real argument about this. I remember one guy at Tek who would occasionally be found in the early morning by arriving employess waking up from having fallen asleep at his desk. Nothing great was ever achieved by casual dabblers.

  7. D Feucht
    June 5, 2013

    There seems to be a different kind of Moore's Law at work here. I do not have the statistics but I wonder if the cost per transistor has hit bottom and is now coming up, especially when reduced MTTF at smaller scales is taken into account. Of course, the mitigating factor is the greater speed. But is speed everything? Our brains are marvelous “computers” running on millisecond processing elements. Perhaps new paths need to be explored conceptually about what to do with a million transistors.

  8. Brad Albing
    June 6, 2013

    @Dennis – >>…suggesting that perhaps the integration expansion path… is to start efforts to build things that have large-integration advantages . I look forward to this blog. When can we expect it?

  9. Brad Albing
    June 6, 2013

    @Barry >>We could argue about the analog quality of today's digital 'scopes relative to the Tek 'scopes of the later '80's, but that takes too long . Not so – we've got all the time and space you need right here. You write it and we'll make sure it sees the light of day.

  10. Brad Albing
    June 6, 2013

    @Dennis >>Our brains are marvelous “computers” running on millisecond processing elements . This seems familiar. Wasn't s/o else discussing the human brain's ability to do amazing computations thru some parallel electrochemical processes that we didn't quite understand? Thought I read that somewhere….

  11. D Feucht
    June 11, 2013

    My present series on an instrument on a chip is a prelude. Later, another series of articles will go into more detail, using the Z meter as an example. To my knowledge, no IC company has come out with a Z meter on a chip, yet it should be achievable.

  12. D Feucht
    June 11, 2013

    Brad – Brian Bailey has brought in this topic in his articles. The fact that the brain does so much more than digital computers (though digital computers can do some things the human brain cannot, excepting the brains of idiot-savants), tells me that we do not need multi-GHz computing to perform more powerful computing functions. We simply haven't figured out how to use the technology we have with computing concepts. It is new concepts, not higher performance of old parameters, that is the intriguing avenue of exploration to me.

    Here's an analogy. Barrie Gilbert discovered (or invented – he would agree that invention is discovery) translinear circuits in the 1960s. It was there to be discovered and required no improvement in the performance parameters of transistors (or any other circuit components) to achieve. Yet the discovery resulted in improvements in circuit performance. Something like this is bound to happen for computing, with the von Neumann (or Harvard architecture) bottleneck wrung as tight as it is nowadays, with multi-level caching in microprocessors.

  13. Brad Albing
    June 11, 2013

    That sounds got. I await the blog on the Z-meter eagerly.

  14. Brad Albing
    June 11, 2013

    @D Feucht – re [6/11/2013 7:21:06 PM] – that was it – it was the blogs by Brian that I was remembering that I was blanking out on. No wonder it all seemed familiar.

    It was and


  15. CameronRobertson
    October 1, 2018

    I resonate a lot with whatever D Feucht  has said here – I personally think that without the human brain, we would not have any of these diagrams or systems at all. Yes a computer can calculate in bigger numbers and faster, but without someone who has first programmed it to do so, they would just be a big hunk of nuts and bolts and components. At the end of the day, we have to remember who is the master and who is the servant in this arrangement…

Leave a Reply