Why doesn't the vendor specify the input impedance or bias currents?

[MattHollands]

Hi,

TLDR; I pretty frequently find opamp data sheets without satisfactory detail about input impedance and/or bias currents. Why do vendors sometimes leave out this kind of critical information? Do they just not know it? Or do they assume that if you're a big enough customer you can ask them the question directly? Or they just assume it won't matter for most use cases? Any ideas?

Specifics:
I've been working with the AD8337 variable gain chip for some ultrasonic RX circuits. It consists of one uncommitted op-amp that is then fed into the variable gain stage. What is frustrating and confusing is that they don't specify the input impedance of that uncommitted op-amp, nor the bias currents, at all in the data sheet!

Because I am working with a single rail supply, I need to AC couple my input signal into the midrail which is 2.55V. So I decided to use two 100kΩ resistors to form a divider and then AC coupling through a 1nF capacitor. The value of these resistors is just a guess since I don't know the input impedance of the op-amp. When I actually come to build this circuit, I find that the bias voltage is about 2.8V instead of 2.55V due to the input impedance/bias currents of the op-amp.

After a bit more googling, I found a help topic stating that this uncommitted op-amp should match the behaviour of the AD8001, which has an input resistance of 10MΩ on the non inverting input and a bias current of up to 6µA. This can explain the 250mA discrepancy is voltage, since a bias current of 6µA would create a 300mV offset to a 50kΩ output resistance (2x 100kΩ resistors in parallel).

So yeah, going back to the TLDR, why don't they bother to include this information?

[Slh]

I find that often reading data sheets is about what isn't there as much as what is.  Ideally, one wants to say how great the part is without making the flaws obvious.

This is especially important for the first page where all the woolly claims are made but also applies to the tables. If there's something that you're not proud of then either don't put it in (if you can get away with it) or try to find a condition where it's not too bad and hide the bad bit in an iffy plot.

Hopefully most customers won't see the issue and, if they do, they might fix it somewhere else.

[PartialDischarge]

While the specs should be there, it clearly says that it is a current feedback opamp, that says it all. Also in all the test circuits the resistor values are very low which is consistent with their application .
If you use a large feedback resistor the bandwidth is going to be very low btw

[MattHollands]

I find that often reading data sheets is about what isn't there as much as what is.  Ideally, one wants to say how great the part is without making the flaws obvious.

That's true - often I will chose a part that has a good data sheet over a part that has a bad data sheet, even if the performance isn't critical - I don't know what they don't tell me and so I have to assume the worst.

While the specs should be there, it clearly says that it is a current feedback opamp, that says it all. Also in all the test circuits the resistor values are very low which is consistent with their application .
If you use a large feedback resistor the bandwidth is going to be very low btw

I wouldn't really agree that "that says it all". I still need to know the figure to make an educated design. If I was making a product, how can I guarantee performance if I don't know min/max/nom values?
I am using the 100Ω value recommended by the data sheet for feedback - luckily I am not having to stray too far from their recommended design, since they don't really give enough information to do so anyway.

[Someone]

If there's something that you're not proud of then either don't put it in (if you can get away with it) or try to find a condition where it's not too bad and hide the bad bit in an iffy plot.

Bias current worst case X pA! yay....  (at Vcm = 2.2V)

[PartialDischarge]

I wouldn't really agree that "that says it all". I still need to know the figure to make an educated design.

Again, the specs should be there. But if you had experience with CF opamps you'd know what to do and what not. No one uses >10kiloohm resistors in CFBs. Also did you know that CF opamps usually have a 1.5V output limit wrt to rail in their output voltage? Not exactly rail to rail so you need to design with that limitation.

[David Hess]

Input impedance is the change of input bias current for a change in input voltage and is irrelevant for most circuits.

Parts intended for use in low impedance high frequency circuits like video and RF might leave input bias current out assuming that the source will always be low impedance.

What annoys me is when datasheets give the input bias current but not the polarity, or give a single polarity without indicating that input bias current cancellation is in use so it could be either polarity.  The schematic for the part could be checked to resolve this but of course more often it is left out as well.

[MattHollands]

Again, the specs should be there. But if you had experience with CF opamps you'd know what to do and what not. No one uses >10kiloohm resistors in CFBs. Also did you know that CF opamps usually have a 1.5V output limit wrt to rail in their output voltage? Not exactly rail to rail so you need to design with that limitation.

Well, I didn't know that it was specifically due to it being a CFB, but I did know it because it's specified in the data sheet  (Vcom +/-1.3V) and so I was able to design around it.

Input impedance is the change of input bias current for a change in input voltage and is irrelevant for most circuits.

Yes of course, change in current due to change in voltage is how impedance is defined. But disagree that it's irrelevant in most circuits. How can you design a filter or a bias network without knowing how it is going to be loaded? Sure I can whack a buffer amplifier before it, but not good if I want to save space/power/money/noise.

[harerod]

...How can you design a filter or a bias network without knowing how it is going to be loaded? Sure I can whack a buffer amplifier before it, but not good if I want to save space/power/money/noise.

The standard application is for 50R impedance. The CFA's non-inverting impedance is of no significance in that case. Since this is ADI, just tell them about your special needs application and ask them for the specs. Please share their answer. 

[thm_w]

They show you how to AC couple the input on page 24 of the datasheet right?