Why are Instrumentation amps so expensive?

I don't think Joerg is missing the point but *you*. An Instrumentation amp has the high CMRR even with an unbalanced source. A bridge *is* unbalanced exept in the calibration point. Any ripple on the supply will result in a common mode ripple on the outputs. A diffamp will have an unequal impedance on both inputs for these signals and maybe 20 to 40dB less attenuation for them. It will also have more noise due to the resistors in the input. It will attenuate high impedance sources and give non-linear readings because of this. You can imagine an IA as a diff-amp with two additional high quality opamps on the input, even when now IAs are done with only two opamps and a pair of transistors. You can *not* adjust the gain of a diff amp with a single pot, that is why audio mixers always use IAs or discrete equivalents for mic inputs, SSM2017, THAT1510/2, INA163. Actually a lot of the INA range are instrumentation amps. Get a data sheet and read up a bit on them to understand the difference.

--
ciao Ban
Apricale, Italy

Va bene, magari non mi sono espresso bene visto che lei ha capito l'inverso di quello che stavo dicendo :-)

A quanti mi pari della tua risposta, lei sarebbe abbastanza d'accordo con me che un IA =E9 superiore ad un amplificatore differenziale. Sarebbe anch=E9 d'accordo con me ch=E9 ci sono dei gamme di prodotti (tipo la gamma IPA dello BB) che contiene ambedue i dispositivi, pero i IA costano di piu.

La mia domanda =E9 semplicamente **perch=E9** i IA dovra costare di piu? Le risposte ovvio non quadrano.

Solitamente quando un componente di tipi "migliore" ha una specifica superiore, si va a finire che quello migliore viene utilizzato per tutti l'applicazione sullo principio che maggiore volume di un singolo compnente si abatte le coste molto di piu di non qualch=E9 semplificazione dello circuito integrato. Quindi si suppone che c=E9 qualch=E9 motivo perch=E9 lo IA costano **molto** di piu di produrre di non il diff amp (di precisione, calibrato a laser, si intende).

Non =E9 una domanda "polemica". Sono curioso perch=E9 il classico vantaggio che lo IA abbia rispetto ad un diff amp con due buffers davanti, =E9 che l'abbinamento dei resistenze sia meno critico per un determinato CMRR. Ovvio che i questione di offsset, stability ecc. sono uguale. Allora perch=E9 i diff amp di precisione sono presente nella gamma a costi inferiore? Perch=E9 non un IA universale, facendo maggiore quantit=E0 di un singolo componente si rischia che costa meno dello diff amp.

Perch=E9 lo chiede? Perch=E9 sono convinto che esiste qualch=E9 ragione preciso, e vorrei sapere qual'=E9!

ise

es

ps

of

17,

So to sum up you are saying that there is no technical reason why an IA costs so much more to produce than a precision diff amp that it justifies the precense of both in a product range. You are saying that the reasons are purley commercial:

1) Nobody has tried taking the plunge and marketed a universal IA at diff amp prices, with a view to eliminating the IA costs.

2) There is a commercial advantage as the diff amp respects the true cost of the device, and most applications will try and use this if at all possible (eg. 200K input diff amps on 2K bridges), but they can cream extra profits from applications that must have true IA's or just want the possible spec.

Joerg wrote:

Neither, I have just been selecting an IA for a project, and it got me thinking about the issues and, frankly, I do not understand why the IA's should be more expensive than other precision laser trimmed devices.

Nor do I understand why there is not modern alternative such as a switched capacitor IA replacement that can eliminate offset drift problems.

I am just curious, as experience teaches me that there is a reason for everything and understanding those reasons is essential to improving your skills.

I can understand why a titanium bike frame costs more than a steel one.

I can understand why a laser trimmed device costs more than a untrimmed one.

I can understand why an IC that requires a different process, a larger chip, or different packaging can cost more than another.

I can understand that a chip with larger volumes costs less.

I cannot understand why a chip that uses the same process, chip size, and packging can cost so much more to produce that it is the case to make a seperate low cost device (the precision diff amp). Why not a "once size fits all" IA at diff amp costs?

The only issue I can think off where the diff amp may be superior to the IA is in speed/phase. But that is hardly an issue for most IA apps.

So I am convinced that there is something I do not know, but nobody (yet) has offered me an answer, except for "commercial reasons".

In article , Roger wrote: [...]

Try google with "flying capacitor". This method of making a dif-amp has been used for years. It has the advantage of a huge CMRR and the disadvantage of a sampling nature. If you make one with a relay you can handle 100s of volts of common mode.

--
--
kensmith@rahul.net   forging knowledge

Hello Roger,

It depends on what you'd want in an IA. If, for example, you like them to include the feedback network or be gain programmable then there is of course extra cost in the manufacturing process. Just like you'd pay a lot more per pound of food for a TV dinner versus buying the individual groceries. Call it the "convenience factor" which isn't free.

Market share has a very large impact on the cost of a chip. Design and mask costs usually run into the millions. This needs to be amortized, something the accountants at companies such as AD will certainly insist on. As will the SEC and the IRS. Has to be done "by the book", been there many times. Then if the sales numbers don't pan out this amortization charge increases.

What looks like "creaming extra profits" might actually just be recouping up-front costs.

As to taking the plunge you are right. They would need to do that. Trying to skim just the glitz applications leads into a corner, except that Sales & Marketing in many companies does not understand this. Folks like me will never take another look and move on to competing devices or make do with a lesser chip plus some "tricks of the trade". Clamping, firmware compensation or whatever it takes. For example, there is a reason why many outdoor/indoor digital thermometer do not use typical RF chips even though they'd be a glove fit. Or why many of my designs do not contain a micro controller although these would have been perfect for them.

--
Regards, Joerg

http://www.analogconsultants.com

Hello Roger,

Why do I now crave a nice glass of Barbera? :-))

Had one yesterday. Ok, I had two glasses because it was so good.

Hope I understood this right. A better part can indeed become "universally" adopted. But only if it is marketed at a competitive and aggressive price right from the beginning. That is often not the case. As long as the inferior part plus any added external components come to a grand total of one penny less, then the "better" part will not be used.

--
Regards, Joerg

http://www.analogconsultants.com

Have you been doing your homework? I happen to live right in the middle of Barbera country (little village in the Asti province) :-)

Well yes, sorry but the Italian was for Ban, who seemed to have got the idea that I thought IA's were no better than diffs. What I said to him was pretty much what I resumed to you in English ;-)

But to sum everything up, nobody has come up with a technical reason as to why an IA priced at a diff amp price could not absorb both markets with a single component at a diff amp price. Hence I think will are all converging on the conclusion that it is all down to "commercial policy".

As for my thoughts about grabbing a voltage with a cap and feeding it normal & contrarywise to a single ended SAR to get offset nulling, well, I am going to try this with discretes! Any ideas for the best analog switches to use? I suppose the two factors are On resistance vs cap charging time and operating range of the switch vs CM range desired. Anything else?

Hello Roger,

It was just the Italian language that did it. Reminded me of some good old times in and around Siena. Italian always sounds like music to foreigners.

Well, maybe there really is no technical reason such as process costs. But I am not sure. You could ask the folks at AD, they should know.

What voltage range? My favorite used to be the SD5400 quad switch. Nicely matched but it seems the ball has been dropped on it.

For higher voltage stuff look at Supertex:

Matching and temperature tracking is paramount here. Else the polarity reversal trick wouldn't work. That almost demands that the FETs be all on one chip.

--
Regards, Joerg

http://www.analogconsultants.com

Charge injection.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

12ohms @ 180V signal....thats not bad for a non vertical FET!

Mind you, nice as it is to know that 200V is available, it would be a bit exaggerated for testing the principles.

Digikey have Pericom devices at interesting prices, 2 bucks for the PS392. This has a more modest voltage range but would non the less be ideal with +/-5V for grounded thermocouples and in bridge applications could go in single supply off the bridge supply rail.

The snag with this is there are not enougth switches to allow me to connect the cap arse over apex, I will need a chip each side. Still, for the experimentation.....

Elaboration?

Hello Roger,

When I used their muxes in ultrasound machine designs they didn't quite go down to 12ohms but sure were nice chips.

Either you have to distributed them so Rdson and CI differences cancel out or you could try the low tech way with reed relays.

--
Regards, Joerg

http://www.analogconsultants.com

Hello Roger,

This is the amount of gate drive that makes it into the path via the gate capacitance. Depending on polarity it adds or subtracts charge from the signal path.

There is a nice way to avoid this and I have done it a few times: A diode quad driven via a toroid. You could also buy that from companies such as Mini-Circuits. I think they even have an app note on it. But then it gets expensive.

--
Regards, Joerg

http://www.analogconsultants.com

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Yes, charge injection *is* the main issue here, as Speff rightly remarked. Another problem will be that any cheap high resolution ADC will be of the delta-sigma kind and thus have an integrating behaviour, so your cap will have to be buffered. Have a look at the AD7718. This converter works with switched caps internally AFAIK. There are a lot of sampling high-speed ADCs with flying caps, but their resolution is low, and they even use possibly very good compensation schemes. It would also be smarter to feed the bridge with a voltage of changing polarity and not turn that cap upside down, as then the charge injection is attenuated before it reaches the input. Indeed this is done since ages for sensitive sensors.

What relates to your original question, I can appreciate your polite Italian explanation. And I agree with your further logical conclusion, even if your original English question was not immediately understandable for me.

--
ciao Ban
Apricale, Italy