How to measure the frequency response of TIA

[tongkin]

Greetings,

I would like to measure the frequency response of TIA using Frequency Response Analyzer. How to make a connection?

My connection is in the attachment, but it does not work. Normally, my design should have 114dB, but the measurement only show -28dB

Anyone help me please...

Thank you,

 

[LvW]

The "gain" of a TIA is a transresistance - that means: ratio of output voltage to input current - without feedback!

 

[keith1200rs]

You really need a high frequency current source. You could get an idea by using a voltage source and a small series resistor (eg 50 ohms) and then re-calculating or simulating the response with that test setup and comparing the two. It won't give you the frequency response as a TIA but it should give you a guide to whether you have enough bandwidth.

Keith

 

[LvW]

Keith,

I don't understand why you recommend a "small resistor (e.g. 50 ohms)" to mimic a current source. For my understanding, a current source can be simulated using a voltage source with a high source resistance.

 

[FvM]

The question is about a real measurent rather than simulation, if I understand right. In this case, the usable resistor frequency range is mainly limited by it's parallel capacitance. For 1 MOhm, it will be a few MHz at best. On the other hand, the "current source" resistor must be large compared to the TIA input impedance in the frequency range of interest to give correct results.

 

[LvW]

On the other hand, the "current source" resistor must be large compared to the TIA input impedance in the frequency range of interest to give correct results.

Yes, exactly! The flowing current must be dominated (that means: determined as much as possible) by the source resistance and not by the load (in this case: the low impedance inverting TIA input).

 

[keith1200rs]

As FvM suggests, you are limited with practical measurements and resistor values. While a 1M ohm might be a useful approximation of a current source, the stray capacitance will probably kill any high frequency response.

My suggestion is to measure the TIA with a small series resistor so it is being used as a voltage amplifier and compare that with a simulation of the same setup. It won't directly give you the bandwidth of the amplifier as a TIA, but it will tell you if the amplifier is behaving as expected.

Another way is to use fast optical source and a fast photodiode as a current source, but then you have a lot more components which could influence the reading.

You could design a voltage to current converter to connect to your signal generator.

Keith.

 

[LvW]

Keith, I see your point and - of course, I agree (in principle!).
But, on the other hand, I think you touch a problem which is connected with EACH real measurement (in particular, when high frequencioes come into the play): A good and exact measurement must find the best compromize between systematic errors (which occur in each measurement) and errors/uncertainties caused by parasitic resp. unknown influences.
For example, this is true also for measurement of the opamp output voltage - a load resistor (input of the measurement device) that is as high as possible is good only at first sight. Because of unavoidable load capacitances a value around some tens of kiloohms may be best.
And for the case under discussion I am sure that a source resistance of 50 ohms is not the best compromize.
And the same applies probably for a 1megohm resistor (because of capacitive influences).
But what about 20...100 kohms? I would guess this is large enough if compared with the TIA inverting input impedance (typical values in the range 10...40 ohms).
That's exactly what I had in mind writing my last posting.
(By the way: I am sure you do not need such a detailed answer, however, perhaps it is useful for some beginners who are not yet familiar with some measurement problems).

 

[keith1200rs]

You are right, 50 ohms is probably not the right value, particularly with 500k transimpedance. I cannot remember what values I have used in the past - something between 1k and 10k I think, but it depends on if you are making a 1GHz 1k ohm amp or a 10MHz, 10M ohm one. I have usually simulated the test setup first to try to pick a suitable test circuit.

Keith

 

[LvW]

How to measure the frequency response of TIA?
I would like to measure the frequency response of TIA using Frequency Response Analyzer. How to make a connection?

Hi Tongkin,

coming back to your original question: As an alternative to a current source at the inv. TIA input, of course, you can use also the non-inverting input and connect a voltage source. This may be more convinient than a current source.
Then, to get the result in V/A (transresistance) you only have to multiply the result (which is in V/V) by the inv. input resistance that must be known in this case.
But be aware of the offset voltage that may disturb the measurement. In this case, I suggest to use 100% DC feedback.
This can be accomplished with a large resistor (e.g. R=1 megohm) between output and inv. input - together with a large capacitor C (e.g 10uF) between inv. input and ground. The large time constant RC leads to a rather low cut-off frequency of the resulting lowpass in the feedback path. Thus, the measurement/simulation error goes continuosly down above w=1/RC.

 

[tongkin]

Thank you for every helps!!

Before I come to post my first post, I got my measurement as below:

Gain : 60 dBohms
BW : ~500KHz

But from my simulation, it should have:
Gain = 114 dBohms
BW : ~2.6MHz

If I connect "a large resistor (e.g. R=1 megohm) between output and inv. input - together with a large capacitor C (e.g 10uF) between inv. input and ground." as LvW suggest, could I able to get more accurate result?

Thank you very much