This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OPA2210: Is there a version in production or planned with lower optimum noise resistance?

Part Number: OPA2210
Other Parts Discussed in Thread: OPA211, OPA189, LMP7731

Hi,

The OPA2210 has exceptionally low noise temperature (optimum noise resistance calculated by dividing the input voltage noise by the input current noise and noise temperature calculated by determining at which temperature an ohmic resistor having that resistance value generates the same voltage and current noise as the amplifier input), being about 160k @ 1Hz and about 70k above 1kHz.

Furthermore, the optimum noise resistance is quite stable over frequency (similar 1/f corner frequencies for voltage and current noise).

Unfortunately, the optimum noise resistance is about 5kohm (rising to about 5.8kohm in the flicker region), which is quite high for many applications.

Is there a similar OpAmp from TI with a lower optimum noise resistance but similarly low noise temperature (especially at 0.1 to 10Hz)?

If not, is an according amplifier planned?

If both would not be the case, may I suggest this as a new product?
If you would increase the current of the input differential stage by a factor of about 6 to 7, then the optimum noise resistance would decrease by the same factor and the noise voltage would be like 2.5 times lower while the noise current would be 2.5 times higher. Maybe it would be advisable to increase the geometry of the input transistors (like take what is now in the OPA2210 and just put 3 additional input transistor pairs aside to the existing one, quadrupling the emitter area) in order to keep the flicker corner frequency about the same (I assume flicker noise might not scale as linearly with collector current, but if it would, the extra emitter area might not even be necessary).

With these measures, you could bring the input voltage noise down to values only achieved by the famous LT1128/1028 while even beating it with respect to flicker current noise to quite an extent (the LT1028 has about 66k noise temperature above 1kHz at about 800 ohms noise resistance, pretty much the same as my above proposed amplifier, but it has about 372k at about 127 ohms noise resistance at 1Hz, which would be clearly beaten by the presumed 160k at about 930 ohms noise resistance of an OPA2210 modified to about 6.25 times of input stage current).
If then there was also a slightly decompensated version (like the LT1028 is with respect to its 1128 brother), there would finally be a new king of low noise in the low frequency low source impedance area.

Even if there would be just an unity gain stable version, I would soo love to buy it and I am sure many others would too.

Many thanks in advance for considering my suggestion (I still hope that you are just about to announce that very product tomorrow).

Best Regards,

Gerd

  • Hi Gerd

    Thank you for your suggestion. As you pointed out there are other factors beside pumping current to get lower 1/f. Geometry being one of the factors, it can be prohibitive for some devices depending on target applications, markets and of course cost. Yes, the LT1028 is a respectable part but the OPA2210 was intended to be a Super Beta input thereby targeting lower total noise. In other words current noise, which in my opinion is often overlooked, at least with bipolar input op amps. 

    We have other options as well for low 1/f. One example is the LMP7731 which, I believe is underrated. We also have the OPA211 and its audio version called OPA111. There is also a different category for low 1/f as you probably know. Zero Drift amplifiers like the OPA189 (100nV) can be very useful and are often sought for exactly that purpose in applications like seismic, displacement measurements, gas exploration and even high end instrumentation including medical. What is the application of interest for you? 

    I do appreciate you sharing your candid feedback with us and I'll be sure to convey your suggestions to our design team. 

    Please do let me know if you find any of the parts I mentioned helpful to you. 

    Soufiane 

  • Hi Soufiane,

    Thank you very much for your answer.

    I know that the Super Beta transistors are good for low current noise, but as well as enjoying the lower current noise one can also increase the collector current and then enjoy the lower voltage noise (or a compromise between both).

    Unfortunately, the other amplifiers you suggested also aren't much better with respect to the voltage noise. The OPA2210 really has the lowest noise temperature by far, it's just the noise matching to lower source impedances that is the problematic thing. It's a bit like having the most powerful engine but lacking a few gears in the transmission box in order to not only be able to drive fast, but to be able to climb the steepest mountains.

    There are quite some interesting autozero amplifiers, only they also lack the last bit of enthusiasm with their developers in longing to achieve extremely low noise voltages.
    The OPA189 has no input current noise spectrum in the datasheet, but from the 1kHz value it seems that its effective noise resistance is quite high (some 31kohm at 1kHz).
    Give the OPA189 an input stage with 27-fold transistor geometry/stage current, then you have an amplifier with 1nV/Hz^0.5 and still less than 1pA/Hz^0.5.

    I know, silicon area matters, but that would be a hell of an amplifier.

    BTW, what is the autozero frequency of the OPA189? I can't find it in the datasheet.

    One more question regarding the OPA2210:

    Is the current noise spectrum in the datasheet showing the differential (balanced) current noise or the common mode (unbalances) current noise?

    In any case, is the not shown one lower or higher and if higher, how much?

    Regards
    Gerd

     

  • Hi Gerd

    The chopping frequency of the OPA189 is around 140kHz. For current noise (really noise in general) it's taken at the non inverting terminal so I can't call it balanced or unbalanced. Now, going back to your suggestions, choppers have higher input bias currents inherently and yes one could consider pumping more current and bootstrapping the input to minimize Ib and in but I think I would rather see that done on a "linear" op amp. I do appreciate the informative discussion and I have provided your feedback to our design community. By the way, what did you think of the OPA211?  

  • Hi Soufiane,

    I think you misunderstood me. The OPA189 has a much lower input noise current to input noise voltage ratio (at least at 1kHz) than the other mentioned amplifiers. So it would not needed to reduce Inoise but rather to reduce Vnoise. Like I said, think of 27 paralleled OPA189.

    Regarding the OPA211, while it has a lower wide band voltage noise, it has a higher low frequency voltage noise than the OPA2210 and a much higher current noise (mor than 10 times the OPA2210 current noise at 1Hz and still more than 4 times at 1kHz).

    Regarding the noise current being taken at the non inverting terminal, this would be the unbalanced case (like having the output shorted to the inverting input or the gain setting divider from the output to the inverting input having a very low impedance (like 999 ohms and 1 ohm for a gain of +1000), where any noise current of the inverting input would not matter.

    If one then assumes that the input stages are similar for the inverting and the non inverting input (like it usually is with operational amplifiers), one would assume that the noise current at the inverting input is about the same (in magnitude), and the total noise caused by input current noise would be sqrt(2) times higher in a balanced configutation (like the above example but with the effective source impedance of the gain setting divider being the same as the source impedance of the signal fed to the non inverting amplifier).

    Now in reality, the noise currents of both inputs might be correlated to some extent. Like if the tail current of the input differential pair is not noiseless, its current noise appears correlated at both inputs (divided by 2 (split between both input transistors) and divided by the current gain of the input transistors. Also input bias current compensation (if implemented in the circuit) is usually derived from a common current mirror, so the noise of the input bias compensation current would also be highly correlated.
    The correlated part of the input noise current would contribute to the total noise with the unbalanced amplifier, while it would not contribute at all with the balanced amplifier (same story as with input bias current in amplifiers with balanced source impedances).

    I just now realize that this fact seems to be completely ignored in the data sheets. Is this the case in all TI datasheets?

    So are the input noise currents in the data sheets the currents including the correlated part? Like this is what I get with one of the input terminals driven with virtually zero source impedance? If this would be the case, one might actually be able to lower the noise contributed by the amplifier even more if the input noise current is the dominant contributor.

    Just for comparison, look at the datasheet of the LT6018 (EDIT:I had initially typed AD6018, sorry for the confusion /EDIT) (sorry, I don't have a TI part datasheet at hand where balanced and unbalanced noise are depicted).
    In graph 14 (lower left chart on page 7, the graph numbers are printed with very small font) one can see that at low frequencies the balanced current noise is about 1/6.75 of the unbalanced current noise (8.3 vs 56pA at 1Hz).

    If you'd have a source resistance of 120 ohms, you would get 1.414nV thermal noise from the source resistance, 6.72nV due to the unbalanced input noise current times the source resistance (56pA*120ohm) and about 1.6nV due to the amplifiers input voltage noise, making a total input noise of 7.05nV. 

    If you'd add a 120 ohm balancing resistor to the non inverting input, you would get 1.414nV thermal noise from the source resistance, another 1.414nV thermal noise from the balancing resistor, 2nV due to the unbalanced input noise current times the source resistance (8.3pA*240ohm, because the balanced noise current is presumed to flow into one of the input terminals and out of the other one, so causing noise voltage as of the sum of both resistors) and about 1.6nV due to the amplifiers input voltage noise, making a total input noise of 3.25nV, being less than half the noise of the unbalanced amplifier.

    So once again, the values in the datasheets are the actual values measured with the negative input current noise not having any effect (neither adding noise nor compensating for any correlated noise current)?

    Regards,
    Gerd

  • Hi Gerd

    I think you mean perhaps the AD8016. For us, in precision devices we do not spec balanced and unbalanced. At least I have never seen that in 20 years dealing with precision amplifiers. Now, as far as correlation, I have yet to come across correlated noise sources which exhibit the same peak and valley at the same time (thinking 0.1 to 10Hz specifically) but I wouldn't say impossible. Since you do mention the OPA211, it does have input bias current cancellation which effectively increases the current noise by factor of square 2, so roughly 40%. I need to look at some of our high speed data sheets to see how they spec current noise. 

    Have a nice weekend. 

    Soufiane 

  • Hi Soufiane,

    I meant the LT6018. Sorry for the confusion, but since AD has bought LT, I occasionally mix up the part names. Take a look at the LT6018 datasheet, that's where you can actually see how much correlated noise currents can matter.

    If with the OPA211 the cancellation currents are derived from the same current reference, it is highly probable that at least this part of the input current noise is correlated. Does the OPA2210 use the same kind of cancellation and is that's noise contribution also about half the noise power?

    If you want to measure it directly (and actually see two traces you can compare), you'd have to set up a circuit with high ource resistances and measure the voltages at both inputs with some JFET amplifier.

    But there is a much easier way. Set up an OP amplifier as an inverting amplifier with a high ohmic feedback divider, so the output noise is heavily dominated by the input current noise. Then you insert a resistor between the non inverting input and ground and gradually increase the resistance value. If the common mode current noise is significantly dominating over the differential current noise (like it is with the above mentioned LT6018), then increasing the resistance will actually reduce the noise (you will have to do this in the lab, as it seems that your models don't take common mode noise into account).

    If you look at all the calculation examples in your datasheets and application notes (at least the ones I looked at), increasing any resistance always leads to increased noise.

    Have a nice weekend too.

    Regards,
    Gerd

  • Hi Soufiane,

    Did you take a look at the LT6018 datasheet?

    And were you able to find something about unbalanced or common mode input current noise with TI parts? Maybe there is something in the older Burr Brown documents.

    Regards,
    Gerd

  • Hi Gerd

    I was yes and I was rather intrigued. I have asked a few folks around but they couldn't think of anything either.

    Last time, you mentioned the bias cancellation. The increase in the current noise is for the uncanceled Ib of course (square root of 2). 

  • Hi Soufiane,

    Soufiane Bendaoud said:
    I was yes and I was rather intrigued. I have asked a few folks around but they couldn't think of anything either.

    I think I don't understand this line. Does the "I was yes" mean that you have found something about it in TI or Burr Brown documents? But then the "but they couldn't think of anything either" wouldn't match to the "yes" part.

    Or was the "I was yes and I was rather intrigued." about having looked at the 6018 datasheet and the balanced and unbalanced input current noise?

    Soufiane Bendaoud said:
    Last time, you mentioned the bias cancellation. The increase in the current noise is for the uncanceled Ib of course (square root of 2). 

    I have problems understanding this also.

    Did you mean that the increase is relative to how the noise would be if the Ib was not canceled? Which is how I understood it anyway.

    I'm sorry, for having trouble understanding what you meant. English is not my first language, so sometimes I have to ask things twice.

    Regards,
    Gerd

  • Hi Gerd

    I'm sorry. I meant "Yes, I did look at the LT6018". I asked a few of the engineers who have been around TI/BB longer than I have to see if they know of this method (balanced vs. unbalanced) but they hadn't seen that on our precision products. 

    For the last statement your understanding is correct. Your English is just fine but to be honest I'm having a hard time picturing correlated noise sources even if one uses the same reference. To me, correlated would be the. same peak to valley over the same time which is a somewhat counter intuitive since noise is a random phenomenon. If you like you can write me at Soufiane.Bendaoud@ti.com to share more details about your application and interest. Better yet, if you send me your phone number, I'll call you at your convenience and we can discuss the subject at length to hopefully come up with a solution for you. 

    Soufiane 

    1-408 702 0311

  • Hi Soufiane,

    With correlated I mean that the noise (or part of it) is the same on both inputs.

    If you have input bias current cancelation, you usually have a PNP based current mirror with two outputs (collectors), one for each input terminal, providing the current that flows into the base of the NPN input transistor.

    The current "feeding" that current mirror will have some noise (if you can build noiseless current sources, you have to sell me some), which will be mirrored to both input terminals. Now you have at each input terminal the current noise of the noisy base current of the according input transistor plus the current noise of the bias current cancelling current mirror output.

    The two base noise currents will be rather uncorrelated (except for some noise caused by the tail current source) while the cancelation currents will have at least some correlated noise stemming from the current source feeding the current mirror.

    While you can derive the cancelation currents from the same reference current that you derive the tail current from, which might eneble you to cancel out the noise of that reference current, you can not use the same current mirror for the bias cancelation current and the tail current, because one comes from VCC and one goes towards VEE. There has to be at least one additional current mirror involved, which adds some own noise, which makes the cancelation currents at least partly uncorrelated from the tail current. So at least some part of the correlated noise current coming from the tail current will not be canceled by the noise of the cancelation currents. This means that there will be some residual input noise current that is following the same peaks and valleys on both inputs.

    All noise sources are uncorrelated by definition, but if one noise source impacts two nodes, the noise at these nodes will no longer be completely uncorrelated.

    Regards,

    Gerd