A long, long time ago, Jocko pointed us towards the famous Jung article "Realizing High Performance: Bandwidth Limitations", available here : http://www.elecdesign.com/Articles/ArticleID/7207/7207.html
This article is based on two articles by Barrie Gilbert, but links from the Jung article were broken. Articles are however available here :
Opamps myths : http://archive.chipcenter.com/analog/c007.htm
Spicing Up The Op-Amp : http://archive.chipcenter.com/analog/c011.htm
Are Op Amps Really Linear? : http://archive.chipcenter.com/analog/c014.htm
Something from way back but could be useful to some. Would it be only for your library
This article is based on two articles by Barrie Gilbert, but links from the Jung article were broken. Articles are however available here :
Opamps myths : http://archive.chipcenter.com/analog/c007.htm
Spicing Up The Op-Amp : http://archive.chipcenter.com/analog/c011.htm
Are Op Amps Really Linear? : http://archive.chipcenter.com/analog/c014.htm
Something from way back but could be useful to some. Would it be only for your library
Thanks for the links. Interesting reading. Where is Andy_C when you need him? (Over at Audio Circle..........last that I saw..............)
Anyway, reading them fully may give a clue as to why Iinsist on open-loop transimpedance amps for I/V stages.
Jocko
Anyway, reading them fully may give a clue as to why Iinsist on open-loop transimpedance amps for I/V stages.
Jocko
Articles are indeed great, sadly an interesting one about current feedback is MIA. 
OK. I've just one question pending. Reading those articles twice or thrice (I'm no EE engineer, I need that and more to understand a small part of them), one of the points stressed again and again is the low open-loop bandwith of opamps.
Jung shows that for the good old 5532, it is only 100hz. Unity gain bandwith is 10MHz, open loop gain is 100dB. Let's do the maths : 100dB = 20. Log 100 000. Thus 10M/100 000 = 100. Here are our 100Hz. Other opamps like the opa627 are not any better under this respect.
But today, an opamp like the THS4631 from TI seems to achieve a way higher bandwith. Unity gain bandwith is 325MHz, open loop gain is a typical 80dB. Open loop bandwith is thus (if I didn't messed up somewhere) 32.5KHz.
My question is (finally) : does this achieve specs sufficient to reconsider the use of opamps of I/V for transimpedance use (the TI datasheet goes for pages and pages about how great this opamp is for such applications) or not ?
Not looking for a complete answer, just a hint on what to look at would already be great
OK. I've just one question pending. Reading those articles twice or thrice (I'm no EE engineer, I need that and more to understand a small part of them), one of the points stressed again and again is the low open-loop bandwith of opamps.
Jung shows that for the good old 5532, it is only 100hz. Unity gain bandwith is 10MHz, open loop gain is 100dB. Let's do the maths : 100dB = 20. Log 100 000. Thus 10M/100 000 = 100. Here are our 100Hz. Other opamps like the opa627 are not any better under this respect.
But today, an opamp like the THS4631 from TI seems to achieve a way higher bandwith. Unity gain bandwith is 325MHz, open loop gain is a typical 80dB. Open loop bandwith is thus (if I didn't messed up somewhere) 32.5KHz.
My question is (finally) : does this achieve specs sufficient to reconsider the use of opamps of I/V for transimpedance use (the TI datasheet goes for pages and pages about how great this opamp is for such applications) or not ?
Not looking for a complete answer, just a hint on what to look at would already be great
00940 said:
Actually, Gilbert's articles don't claim that open loop bandwidth is important at all. Here's a quote from "Spicing Up The Op-Amp":
"The problem with this model, to my mind, is that it introduces two rather artificial parameters: the dc or zero-frequency (ZF) gain AO, and the open-loop corner frequency defined by wO = 1/TO. Neither of these are very important in practice: there will usually be an immeasurable impact on performance when replacing an OPA with an AO of, say, 100,000 with one having AO = 10,000,000, provided the gain-bandwidth product is the same for both OPAs."
In other words, open-loop bandwidth by itself is largely irrelevant. What's important is gain-bandwidth product. For example, the AD797 has a smoking gain-bandwidth product of 100 MHz and an open-loop bandwidth of about 5 Hz.
Many audiophiles will nonetheless claim that wide open-loop bandwidth is important, but there's no mathematical justification for it.
Thank you for making this point clear. Some things are (slowly) getting clear in those articles.
In the case occupying us, I/V conversion, Gilbert has this sentence (opamp myths article) : " As the OPA output integrates this "error voltage" of the full final value at its input, the error declines exponentially at a rate fully-determined by the unity-gain frequency, that is, on a time-constant of T1. During all this time the inverting node is far from being a "virtual ground!"
Following this, the "least" harmful conventionnal opamp would thus be an opamp with very high unity-gain frequency (=gain bandwith). Which is what you said (I'm a bit slow but it's late over here. I wish I were still in Boulder, CO, it made easier for me to follow the threads here).
Coming back to the THS4631 (I'm not obsessed with it, just have the datasheet under the eyes), the reason to use it (not that I really plan to) would thus be the wide gain bandwith it has.
Now I've to understand Jocko's simple I/V.
In the case occupying us, I/V conversion, Gilbert has this sentence (opamp myths article) : " As the OPA output integrates this "error voltage" of the full final value at its input, the error declines exponentially at a rate fully-determined by the unity-gain frequency, that is, on a time-constant of T1. During all this time the inverting node is far from being a "virtual ground!"
Following this, the "least" harmful conventionnal opamp would thus be an opamp with very high unity-gain frequency (=gain bandwith). Which is what you said (I'm a bit slow but it's late over here. I wish I were still in Boulder, CO, it made easier for me to follow the threads here).
Coming back to the THS4631 (I'm not obsessed with it, just have the datasheet under the eyes), the reason to use it (not that I really plan to) would thus be the wide gain bandwith it has.
Now I've to understand Jocko's simple I/V.
- Status
- Not open for further replies.