Advertisement
Advertisement

Tucson_Mike

's profile
image
Principal, MAS Amplifier Consulting

After 8years of analog IC design at IBM and StorageTek, joined Comlinear in 1985 where we originated the Current Feedback Amplifier (CFA). Eventually taking the Apps Manager role, I wrote many of the original CFA app notes for the industry. Moving to Tucson in 1995 to take the strategic marketing manager role with BurrBrown’s high speed amplifier group, set the product development roadmap and launched numerous industry leading products like the OPA2677 and OPA698. With the TI acquisition in 2001, moved over into more of training and ADC interface solutions role – writing and presenting a large part of the high speed tech days technical material. Joining Intersil in 2009, continued the high speed ADC interface solutions role (introducing the ISL55210 FDA) along with developing the op amp online design tools and leading the DSL/PLC line driver apps support. Returned to TI-Tucson in 2013 taking a leadership role in the high speed amplifier roadmap. Those products included the THS4541 and THS4551 FDA's and the more recent OPA837 and OPA838 VFA op amps.With 32 years in the high speed signal path area, have introduced over 80 products, written over 30 app notes and published over 50 articles and/or conference presentations. Currently focused more on being a contributing author to EDN and selective consulting assignments across the component industry.


Tucson_Mike

's contributions
  • 01.13.2018
  • Active-filter design tools shootout
  • You are of course correct, and I have done a lot of that in the past to validate some of these issues, but if you don't start out close (which was kind of this first level discussion going on here), going to board level will not be getting closer except by chance. The most recent bench level work on MFB filter validations would be in the THS4551 data sheet - in the data sheet, those all look like sims (and that was all wrapped up in getting the model correct for what we were seeing the bench) But many were tested on the bench for stability mainly. That one has a bit of phase margin issues tied up with the reactive open loop output impedance.
  • 12.08.2017
  • Testing op amp tools for their active filter design accuracy and dynamic range
  • Thanks guys, I just noticed I needed to respond - sa_penguin, your point is obviously where I was headed, and after considerable sim work and shape extractions that now has posted on EDN as part 2 - if you have any questions on that stuff let me know. I also have those excel data files if you are interested. And then uana42log, thanks to Mark Fortunato for pointing that out to me. We all have different backgrounds and it turns out omniscience on the history of every common technique is surprisingly difficult. I tend to attribute everything to Dr. Budak and Dr.Huelsman, since those where the folks I studied in the 70's. But in fact we are all standing on the shoulders of legions of early pioneers from the 60's and 70's in this area.
  • 06.14.2013
  • Op amp input current noise
  • There it is, something I have wondered about for years. Sounds like we need to split our data sheet In numbers into common mode and differential mode to give good predictive match over all external circuits. Great work, thanks Sergio.
  • 06.16.2013
  • Signal distortion from high-K ceramic capacitors
  • There is probably a valid point to the cap linearity discussion here, but there are many layers to SKF dynamic range. Though not stated, the test filter was set to 1kHz Q = .7 or a butterworth. That is a pretty benign stage from an internal noise gain peaking standpoint. I do wonder about all the curves above the 1kHz cutoff frequency as I would assume that is data beyond the band of interest and kind of into a collapsing loop gain region for the total circuit. The key point to add seems in this example is the noise floor is totally determined by the relatively large resistor noise terms vs. the awesomely low 2nV and 3pA terms on the op amp itself. That seems to the core issue is that to lower that noise you really want much lower resistors but then that pushes you into very high valued caps at these lower cutoff frequencies and hence the desire for the HighK stuff. I normally suggest folks use C0G for the tempco but they kind of stop about 12nF in standard cost. Great article pointing out yet another trap in getting to high dynamic range. I might have simply driven a RC circuit with with an audio precision with the different cap types and measured distortion. A non-linear load will always generate distortion from a resistor source in just that simple circuit and can in fact be used to extract the polynomial coefficients for an impedance model(at least when i was younger I could do that - another article for you?).
  • 06.09.2013
  • Wideband fully differential amplifier noise improved using active match
  • Hey Hooman, how you doing over there at the Kifer road TI design center? Good questions, I was thinking of these myself as I was working through this stuff. Kind of comes down to targeting a minimum return loss through the highest frequency of interest. I could imagine a new data sheet plot of say a target 20dB number and an x-axis of gain and y-axis of max frequency to stay above 20dB return loss using the active match approach. The 2nd question comes down to simulated vs. measured input impedance. I did a little bit of that on an example circuit (in an app note on the Intersil website) where I got nearly perfect agreement through about 400Mhz and then the board tilted down probably due to unmodelled parasitic C on the board. I was surprised how well the sim macromodel was picking this active input impedance up - Jian Wang did that model and did a better job than I even imagined. In the early days of looking at this (Aug. 2011), he was running ADS sims for me with the full internal cadence model and we were getting great match to bench. As an added caution, this circuit (and the other transformer one I like so well) obviously depend on the source impedance being about what you expect to be flat to the outputs. If the source is perhaps a narrowband match, that can still work but I would be adding a modest BP on the output side to strip off any weirdness out of the band where the source is the impedance you are designing for. The ISL55210 is internally compensated to handle any source Z and stay stable, but not necessarily flat over frequency.
  • 04.09.2013
  • Simulating op amp noise
  • I do a lot of noise work also of course and the Art Kay material is drawn from years of BurrBrown content. Normally you do need two seperate noise currents at the inputs - particularly to include CFA type devices in tools/design work. I did not look too closely but I think the curve above is focused on total noise at the noninverting input including kTR, en and InR terms RSS together. The inverting side has its noise as well, and an interesting question is if the there is no Ios spec, those elements should maybe be scaled down in impedance. I am interested in the correlated current noise question also. I think LTC made some headway spec wise there, but it has been awhile since I looked. I am actually doing spot noise extraction in the lab right now on various things (3 config to isolate the 3 terms of CFA) and have wondered how to approach/model a correlated term. I am assuming it increases the apparent output spot noise, but how to get that term extracted?
  • 06.16.2010
  • Op-amp Spice macro-models article from Intersil
  • Steve T. asked me to reply here. For Intersil, if an amplifier has a model, that will be listed under the documents and/or resources tab of the product info page. These are detailed .txt files for import as Spice subcircuits. Also, the iSim PE tool includes the most current set of op amp model files, but there they are compressed to use fewer nodes and model listing is not visible, but normally the same as the .txt file. The online design tools generate circuit blocks that then port into iSim PE for further editing and combination with other blocks.
  • 04.15.2013
  • Wideband matched input Z using active match and a new type of amplifier (Part two of two)
  • Good catch, I of course enter the axis labelling later and should have entered V/rootHz These were simulations here, and I actually found the sim model is just bit high on the FDA voltage noise, the app note that I just finished for the demo board for this has measured data. I fired up my old spot noise measurement system and re-took that data there. Need to fix that sim model when I get a chance. We itereated quite a lot at the intro of the ISL55210 to make sure that 0.85nV differential spot noise internally was correct and feel pretty confident on that one.