Advertisement

Blog

New Analog ICs — All You Have to Do Is Ask

I've been asking around (colleagues, online engineering groups, polls) for suggestions on what engineers want if they could dictate choices for new integrated analog ICs. I received some reasonable and interesting suggestions.

Dr. Tom Bach (Brighton, UK) had a rather specific request for a 16 input analog multiplexer that in some ways would be similar to the CMOS 74HC4067, MPC506, or DG506. The important difference: any of the inputs not selected (connected to the output) would instead be connected together and to a separate output pin.

Dr. Bach intended to connect that pin to a voltage source. We will note that any analog muxer is bidirectional. So perhaps he needed this function to confirm (via some sort of read-back circuitry) which input is selected and which is not selected. An interesting idea, in any case. In actual fabrication, this could be done by putting two 16-channel muxers plus suitable control circuitry on the chip.

Clarence Foytik (Houston, Texas) said he'd like a delta-sigma ADC with a PGA (programmable gain amplifier) front end. And he'd like it with higher gain setting resolution than what is available now. If it could be made with 7-bit control (i.e., 128 individual steps), that would be great. But he would settle for 64, 32, or at least 16 different levels of gain setting. And while we're at it, the amplifier should be unconditionally stable, repeatable, have an ENOB equivalent to the published number of bits, operate on a single supply, and draw hardly any current.

Joerg Schulze-Clewing (Sacramento, California) would like a programmable analog/mixed signal device. JSC mentioned the Cypress PSoC as a starting place, but with some additional features. His list is reproduced here:

  1. Fully programmable;
  2. More analog blocks. Lots more;
  3. Faster analog;
  4. Good (or software trimmable) offset specs;
  5. Fast current mode control features;
  6. Adjustable leading edge blanking; tens of ns granularity;
  7. Modest uC horsepower to implement “strange” loops;
  8. Complimentary beer and pretzels.

No one could argue with that feature list. Especially Homer Simpson. “Mmm… beer and pretzels.”

Ross Ramm (Melbourne, Australia) would like an IC that has most of the circuitry needed for a grid-tie micro-inverter. This would have all of the analog measurement circuitry, miscellaneous control I/O, and power-supply circuitry including the FET gate drivers. He adds that if isolation is needed for portions of this (entirely likely), he'd like to see that included as part of the IC. Then, wireless communication functionality would be good, too. Could be ZigBee or power line communication (PLC). He acknowledges that perhaps we would need to add external power FETs or IGBTs, a transformer, and some L-C filter components since maybe it's not practical to integrate all that.

Lastly, Chris Watts (Bath, UK) wants:

…a chip like the Lime Microsystems LMS6002D but covering lower frequencies and… optimized for narrower bandwidths. For those unfamiliar with Lime Microsystems, this is a transceiver front end (antenna to DAC/ADC) with a wide frequency range and programmable bandwidth. What I would like is coverage below the bottom of the LMS6002D range and with narrower bandwidths.

The part to which Chris refers, the Lime Microsystems LMS6002D, is a nifty part. This part is intended for the UHF band (plus 800MHz into the SHF band for those keeping score at home). This means the device is intended for cellphones and WiFi applications, but not for more general-purpose transceivers operating at lower frequencies. I expect Chris is looking for an IC suitable for a general purpose transceiver like those that armed forces, safety forces, and ham-radio operators would make use of. The likely frequencies of operation would be 1MHz to 300MHz.

Thanks for sending along your ideas. There's surely more ICs that you'd like to see implemented in silicon. Tell us what you have in mind.

Related posts:

28 comments on “New Analog ICs — All You Have to Do Is Ask

  1. Scott Elder
    August 16, 2013

    Brad, It would also be helpful if we knew WHY such an IC is important.  Certainly each one of those circuits could be done with a few parts.  So is the WHY because of board space?  Price? Not possible to build with several ICs because….?  Time to design?

    I'm wondering about the value proposition from the requestors perspective.

     

  2. Brad_Albing
    August 16, 2013

    @Scott – 'tis a very good point. I'll go back the the original posters and ask them.

  3. Steve Taranovich
    August 16, 2013

    Give me a high speed op amp in front of the ADC in a high IF receiver design on chip. One that can be truly transparent to a 14 or 16 bit dynamic range IF signal at bandwidths in excess of 100+ MHz (Ultimately GHz bandwidths would be ideal, but let's take it a step at a time) and a huge slew rate. In communications, we are trying to move the receiver as close to the antena as possible and eliminate those pesky down-conversion RF circuits (SDR-software defined radio is the goal). In many cases we don't want to use a transformer here since it will limit the lower frequencies and limit dynamic range.

  4. fasmicro
    August 18, 2013

    @Brad, you could explain this more >> There's surely more ICs that you'd like to see implemented in silicon. >> Do you mean they are implemented in other materials like GaAs or Ge? Or simply they are not done monolithicly. If they are ICs, assume, I think they are already in silicon. Perhaps, they are at board level; in that case, we have to specify clearly.

     

  5. D Feucht
    August 19, 2013

    I'll second Clarence Foytik's recommendation. Ever since the aging TI ADS7871 (or others in that series), which has a 12 or 14 bit ADC preceded by PGA and AMUX, nothing seems to have appeared to replace or expand upon it. This is a no-brainer for DAS system integration, yet nobody has done it that I am aware of. There are various n -bit sigma-delta (or delta-sigma) converters on the market, where n is a large integer, but they lack the front-end stages that complete a DAQ. Applications with multiple thermocouples, for instance, require more than 14 bits of resolution and in a DAS product, also need the gain and channel switching.

    It's time for some new DAS-on-a-chip ICs!

  6. dates
    August 19, 2013

    Dennis

    There are some new DAS devices. ADS1248, LMP90100, LMP90077 series.

    Also some AFE (no ADC) for specific end applications:

    LMP91002 – H2S and CO

    LMP91050 / 51 for NDIR

    LMP91200 for pH

     

  7. SunitaT
    August 20, 2013

    According to a recent report by Strategy Analytics (Boston), 92 percent of the chips used in Electric Vehicles and Hybrid Electric Vehicles are analog ICs. Between 2011 and 2018, the value of analog ICs for EVs and HEVs is expected to grow at a compound annual growth rate of 23 percent, well-above the projected CAGR of 7.2 percent for standard automotive analog and linear products, according to Strategy Analytics.

  8. fasmicro
    August 20, 2013

    It is evident everyone sees the quality in the old analog players like ADI, TI and co via their datasheets. Making quality datasheet is a realy comparative advantage as it builds credibility for companies. 

    On >>It's time for some new DAS-on-a-chip ICs! >> Except the interface, I think we have that already. All DAS are powered by special ICs that are largely DAS-on-a chip.

  9. fasmicro
    August 20, 2013

    >> 92 percent of the chips used in Electric Vehicles and Hybrid Electric Vehicles ar

    That is true. When EETimes had the 2012 tour of the VOLT by GM, they visited only analog IC makers. That was an indication that the innovation came from those companies. Nevertheless, do not overlook the amount of digital ICs inside these units.

  10. Brad_Albing
    August 21, 2013

    @fasmicro – again, I'll check with the folks who originally made these suggestions (work in progress). A general comment here is that the devices would be done in Si and done  monolithicly. Why? Because that was my intended starting point in the discussion. But do they have to be? No. Ultimately, my intent was to get a discussion going regarding what we need fabricated in Si, GaAs, GaN, SiC, or even Ge.

    Some of the functionality already exists, so in that case, the intent is simply to shrink the physical size and reduce the cost compared to existing ICs or to board solutions.

  11. Brad_Albing
    August 21, 2013

    @D Feucht – I agree with you on the need for more versions of a delta-sigma combined with a programmable gain stage and a mux. I know there are potential customers who would want such a device or family of devices for a DAS. This sometimes shows up as an AFE in some IC companies product lines, but we need more complex (feature-rich) versions to be made.

  12. Brad_Albing
    August 21, 2013

    @Steve – don't cross off the idea of a transformer. I know in some apps you want to go right down to DC. But the transformer is helpful a lot of the time. I sure you recall this excellent article we published a few months ago by Michael Steffes:

    Measuring & Modeling Wideband Baluns for Application to ADC Input Stages

  13. Brad_Albing
    August 21, 2013

    @fasmicro – well sure, there is some digital content in EVs, but the analog parts are the important ones.

  14. Steve Taranovich
    August 21, 2013

    You are absolutely correct—-Michael Steffes is the best source of high speed amplifier and data converter architectures that I know—-transformers absolutely have their place in some designs and even add performance enhancements in many cases. Steffes was a mentor of mine at Burr-Brown/TI for many years and still is a mentor and source of some great tech articles on EDN and PA

  15. kendallcp
    August 21, 2013

    >> have an ENOB equivalent to the published number of bits

    Or perhaps that should be

    “not claimed to have more distinct output levels than 2^(the claimed ENOB number)”

    A single ENOB number is not a characteristic performance metric for a converter.  It's the result of a simple calculation that takes the results of one experiment as its input.  However, this one point is arguably a better indicator of the 'performance' of an ADC (or ADC plus front end) than the number of published bits.

    The noise-shaped front end of a delta-sigma ADC converter doesn't have any meaningful “number of bits” figure.  It just has a noise density versus frequency, and a linearity profile that's a complicated (though hopefully small) function of the (continuous) input signal.

    The del-sig ADC in my employer's top-end SoC – or rather, the virtual component that you design with in the GUI, the one with the pins and the datasheet – can be programmed to have any resolution that you like from 8 to 20 bits – or maybe it's 6 to 21 in the laterst version, I can't remember OTTOMH.  How many bits you can have is a function of the sample rate you want (and one or two other factors).  It's actually a matter of some annoyance to me because I would prefer that the underlying promitive actually delivered every 'bit' that comes out of the machinery, which could be up to 24, whatever the output rate.  We limit the size of the output word because that's what people are familiar with.  They expect an ADC with a 48 dB SNR to be an 8 bit ADC and get confused if you give them 24 joggling bits but still only 48 (actually closer to 50 now because no output quantization noise) dB.  Perversely, restricting the output word width makes the performance worse than it needs to be.  In other words, if you have a front end which yields an ENOB calculation of exactly 8 bits, then if you quantize the output to 8 bits, the resulting converter will have an ENOB of less than 8 bits…

    So, i'd say (and maybe I should blog this) – forget the number of bits.  All our processors' data buses have more bits than you can shake a stick at and bits are not a limitation any more.  Just scale so it's MSB aligned, and process way more bits that you think your system SNR demands.  Then you won't artificially be adding extra noise in the name of not confusing yourself…  This has worked really well in the high quality audio world, and IMHO industrial applications should start trying it!

  16. Brad_Albing
    August 21, 2013

    @KCP – yes! Do blog on this.

  17. Dnoosh
    August 22, 2013

    How about a good LVDT front end. The choices out there are either expensive, hard to use, not very accurate (requiring calibration). Also for a front end an AD625 can be used for the first stage of gain, with the second stage in the ADC not having a gain of more than two. This would double the gains in the ADC.

  18. Brad_Albing
    August 22, 2013

    @Dnoosh – I had completely forgotten about about LVDTs. Good suggestion for an IC to interface with them and reduce the cost.

  19. WKetel
    August 22, 2013

    There has been an LVDT interface IC around for quite a few years. I don't recall the part number, but we used then in Diesel fuel injection system testing machines at delphi. They worked well and were very linear.

  20. Brad_Albing
    August 23, 2013

    @WKetel – if we could get a little more functionality added to the LVDT IC, that would probably make it more useful. That's the real key to marketing success – figure out what to combine with and existing IC's functionality, add it in, end up with a larger IC and a smaller PC board, and make a million bucks. Easy, right?

  21. WKetel
    August 24, 2013

    It seems that the real problem is feature creep: Keep on adding functions until something becomes either unuseable or so unrliable that nobody will use it. Then somebody else comes out with the plain version, only more reliable, and takes their place. Sometimes more bloat dose not equate to more value.

  22. Brad_Albing
    August 26, 2013

    It's a valid point. I've been thru that situation and gone back to the simpler part.

  23. Brad_Albing
    August 26, 2013

    @Scott – additional comments from Clarence Foytik “In a lot of our designs, board space, single supply [operation], and low power is important. There is a realization that the best possible analog front end is not designed into these ICs, but they integrate very well with our design strategy.”

  24. Brad_Albing
    August 26, 2013

    @Scott – more answers – from Joerg Schulze-Clewing: “… there are many cases where you need both analog and digital. I had cases where switchers could not have properly done the required job without some sort of processor in the loop. Would be great to combine that but it only works well if fast stuff is included. In a switcher that is the current mode control. Average current mode is often simply not possible, can't always compromise there. The other market is where you have to change system behavior on-the-fly.

    Board space is the usual challenge. I just completed another multi-stage switcher design and despite going to 0402 the board is full. Like a Tokyo subway at rush hour. If we'd had mixed-signal programmable ICs it would have been much easier.”

  25. Brad_Albing
    August 26, 2013

    @Scott – one more — Chris Watts said that regarding a device along the lines of the Lime Microsystems LMS6002D (but lower frequency), “… the application I was thinking of was low cost hand held spectrum analyser. The chip I mentioned will do the job just fine above 350MHz. A S/A that stops at 350MHz would be a hard sell. There are portable S/As on the market but they are pretty big. Not hand held DVM size. The process should be Si for cost and there is no need for the more exotic stuff at the lower frequencies. To hit the amateur market and get volume the instrument would have to be pretty cheap.”

  26. jkvasan
    August 29, 2013

    Brad,

    My wish may seem odd but I would like to share it any way.

    My IC has the following specs:

    1. Input voltage range 0 to 10 V or more.

    2. It is a ADC with 24 bit resolution.

    3. can work on power supply of 10 V.

    4. SPI outputs may be 5 V or 3.3 V compatible to suit mcus avl today.

    5. Internal Amp – Gain : 50, 100, 150 and so on.

    Most 24 bit ADCs today have an input voltage range of +- 500 mV. This limits the smallest step measured to nV. Instead, it the voltage can be amplified and then measured at a higher voltage level, dividing it into 24 V reolution could give a better SNR (my assumption, may be wrong).

    I am not sure if this makes a monster out of noise amplification which later becomes difficult to handle.

    Your views, please?

  27. Brad_Albing
    August 29, 2013

    @JK – the 0 to 10V would make it useful for industrial applications. Companies like Rockwell and Honeywell would like that. But with that input range (and the corresponding 10V supply voltage), it would of course need to be fabricated on a medium voltage process. That will make it cost a bit more. Noise might be an issue, but you could probably work around that.

    I like the features you've described, but my guess is that not enough companies would buy such a part in high enough volume to make it worthwhile for the IC manufacturers to build such a part. Just my 2 cents worth….

  28. jkvasan
    August 30, 2013

    @Brad,

    We need these chip companies to be profitable to get us more chips and innovations, I agree. However, small volumes from several buyers could pool into a large volumes. Only difference would be logistics cost, which the small buyer would always be willing to pay.

Leave a Reply

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