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.

TLV9062: Is it difficult to design a band pass filter of gain x150 using TLV9062?

Part Number: TLV9062
Other Parts Discussed in Thread: DUAL-DIYAMP-EVM, , TINA-TI, TLV906X, OPA365

Hello guys,

I have similar question to the previous one.

One of my customers is evaluating  TLV9062 using Inverter Amplifier type (Figure 11. Inverting Amplifier Schematic) of TI OPAMP evaluation board, DUAL-DIYAMP-EVM.

https://www.ti.com/jp/lit/ug/sbou193/sbou193.pdf

They use the following parameters for the evaluation.
R1/6=Open, C4/7=47nF, R3/8=1kohm, C3/6=22pF, R2/7= 150kohm, R4/9=560kohm, R5/10=560kohm, C5/8=100nF, C1/2=100nF

Also each terminal voltages are below.
V+=2.2V, V-=GND=0V.

And REFA/B=2.2V supplied from a power supply separated from V+ power supply,

They took a gain-frequency characteristics in these conditions. As the result, the gain characteristics was lower about 5 dB than TINA-TI simulation result (about 43dB)
when 10kHz signal was input.

Their questions are the follows.

Q1. What do you think the cause? Is it caused by TLV9062 gain limitation at 10kHz?

Q2. Is it difficult to design a band pass filter of gain x150 using TLV9062?

Your reply would be much appreciated.

Best regards,

Kazuya.

  • Hi Kazuya,

    the open loop gain of TLV9062 at 10kHz is 60dB. With a closed loop gain of 44dB (151V/V) the loop gain -also called "linearizing gain reserve"- is 60dB - 44dB = 16dB. This should be sufficient to not cause a drop in the frequency response at 10kHz. But this gain reseve will decrease to less than 7dB at the higher frequencies:

    7dB is way too little gain reserve for a proper operation. I recommend a minimum gain reserve of 20dB at the highest signal frequency. A good circuit should have 40dB. And 60dB for a high precision circuit is not unusual.

    A good remedy when not having enough gain reserve is to split the gain evenly on two OPAmps in a row. Or to use a faster OPAmp.

    Also keep in mind that the speed performance in the datasheet of TLV9062 is defined at a supply voltage of 5V. At 2.2V the speed performance may suffer and you may see a degraded frequency response.

    Another cause for the observed drop in the frequency response could be a hidden source resistance of sine generator driving the input of your circuit. Many low frequency sine generators come with a source impedance of 600R. This will effectively reduce the gain of your circuit by 4dB.

    Or do you have a 22pF cap with a too high manufacturing tolerance?

    Kai

  • Hi Kai、

    Thank you very much for your reply.

    Could I ask you several questions as the below?

    Q1. You wrote that I recommend a minimum gain reserve of 20dB at the highest signal frequency. 
           Is this 20dB gain reserve to avoid affection of TLV9062 manufacturing gain variation?

    Q2. You wrote that a good remedy when not having enough gain reserve is to split the gain evenly on two OPAmps in a row.
           Does this mean that lowering the gain extends the region of the flat gain?
           How much higher frequency of flat gain region is needed  than the highest signal frequency?

    Q3. This is a question about TINA-TI. Is the power supply voltage reflected to TLV9062 characteristics
           (gain-frequency characteristics, output drive current, settling time etc) on TINA-TI?  
     

    Thank you again and best regards,

    Kazuya.

  • Hi Kazuya,

    Q1: No this has nothing to do with the manufacturing tolerances of OPAmp. This is simple "OPAmp physics". The negative feedback improves many parameters of a OPAmp circuit. The "closed-loop linearity error", for instance, becomes improved according to this relation:

    L_cl = L_ol / (1 + A_beta)

    where "L_cl" is the "closed-loop linearity error", "L_ol" is the "open-loop linearity error" (usually less than 5%) and "A_beta" is the "loop gain" ("open-loop gain" divided by "closed-loop gain"), also called "gain reserve"

    In this example an "A_beta" of 20dB would decrease the linearity error from 5% to about 0.5%. A "gain reserve" of 40dB, on the other hand, would result in 0.05%. A "gain reserve" of 40dB is adequate for most applications.

    By the same factor other parameters of OPAmp circuit are improved as well. Think of "closed-loop output impedance" or "closed-loop gain stability as a function of temperature", for instance, to call only a few of them.   

    Q2: Yes. With a standard slope of "open-loop gain" of 20dB/decade a 20dB "gain reserve" translates to a headroom in bandwidth of factor 10. But this is only valid for constant gain applications without any filtering components in the feedback loop. Because of this I prefer to think in terms of gain reserve.

    Q3: I don't see any change in the frequency response when changing the supply voltage in the TINA-TI simulation. So I think the supply voltage dependency is not modelled here.

    Kai

  • Hi Kai,

    Thank you very much for your reply and I'm very sorry to be late my response.

    Could you ask you a few additional questions because the customer and I'd like to know about the gain reserve you said?

    Q1. Does "closed-loop linearity error" mean closed-loop gain linearity error?

    Q2. You said that the gain reserve should be 20dB at least. Also the gain reserve is 7dB in the customer case.
          Could you please tell me how to calculate the maximum gain which they can take when 20dB gain reserve is taken?

    Q3. Could you please tell me any document which explains why 20dB gain reserve is needed at least?

    Could you please tell me your reply?

    Thank you again and best regards,

    Kazuya. 

  • Hello Kazuya, 

    Q1. Does "closed-loop linearity error" mean closed-loop gain linearity error?

    Yes, I consider the two terms the same.

    Q2. You said that the gain reserve should be 20dB at least. Also the gain reserve is 7dB in the customer case.
          Could you please tell me how to calculate the maximum gain which they can take when 20dB gain reserve is taken?

    You can use TINA-TI for this, or calculate by hand. If the customer requires gain of 150V/V, you have to select an amplifier with more BW or distribute the gain across multiple TLV906x devices to maintain the 20dB gain reserve. Usually, it is most common to frontload the gain to the first amplifier stage to reduce noise, but you could always distribute the gain equally across two stages SQRT(150) = 12.25V/V,( see below). 

    See This App Note for further details on Cascading Amplifiers  

    Q3. Could you please tell me any document which explains why 20dB gain reserve is needed at least?

    Give the Analog Engineer's Calculator a try and select the category "Gain Error vs Bandwidth". You can change around the parameters to calculate the minimum GBW required to realize a desired Gain error. This will help demonstrate how Gain and BW influence linearity. 

    Please let me know if you have any questions,

    Best,

    Jacob

  • Hello Jacob,

    Thank you very much for your reply.

    I installed TI Analog Engineer Calculator and tried to use the Gain Error vs Bandwidth calculator.
    I have a few questions about this calculator as the below. Could you please give me your reply?

    Q1.  Non-inverting circuit is shown in the display of the calculator. Is the Gain Error vs Bandwidth calculator applied to Non-inverting circuit only?
            Or can the calculator be applied to Inverting circuit too? The customer uses a inverting circuit for Bandpass filter which shown at top of this tread?

    Q2. How to display the 3 colored (Red, Green and Blue) column which located at right side bottom?

    Q3. The customer designed the inverting circuit has 150V/V(43.5dB). But the actual measured gain was about 110V/V(40.8dB).
           In this case, the gain error was about 40V/V and the error rate is about 26%(40V/V divided by 150V/V).
           When I input 10kHz to Signal Frequency column, 150V/V to Closed Loop Gain and 26% to Acceptable Error.
           As the result, the minimum gain bandwidth product was 1.65MHz. This number was lower than TLV9062(GBW=10MHz).

           The customer and I'd like to know why the gain reserve is needed 20dB at least.
           Is there any documents which explains about it?

    Thank you again and best regards,

    Kazuya.
         

  • Hi Kazuya,

    Q1.  Non-inverting circuit is shown in the display of the calculator. Is the Gain Error vs Bandwidth calculator applied to Non-inverting circuit only?
            Or can the calculator be applied to Inverting circuit too? The customer uses a inverting circuit for Bandpass filter which shown at top of this tread?

    Yes, you should be able to utilize the gain error vs BW calculator for inverting configuration

    Q2. How to display the 3 colored (Red, Green and Blue) column which located at right side bottom?

    The other three colors derive from the other options in the panel: " AOL Error Considerations" and "Resistor Tolerance"

    Note, both options are designed around the non-inverting schematic, but similar tests can be conducted in TINA-TI.

    Q3. The customer designed the inverting circuit has 150V/V(43.5dB). But the actual measured gain was about 110V/V(40.8dB).
           In this case, the gain error was about 40V/V and the error rate is about 26%(40V/V divided by 150V/V).
           When I input 10kHz to Signal Frequency column, 150V/V to Closed Loop Gain and 26% to Acceptable Error.
           As the result, the minimum gain bandwidth product was 1.65MHz. This number was lower than TLV9062(GBW=10MHz).

           The customer and I'd like to know why the gain reserve is needed 20dB at least.
           Is there any documents which explains about it?

    Being off by such a large factor(26%) makes me believe other issues may be at play.

    Firstly, resistor tolerance can easily cause the gain to be different than expected. Do you know the resistor tolerance for the customer circuit? 

    Secondly, the filtering capacitors may be causing the gain to roll off prematurely. 

    Can the customer test the gain across the filter passband to help me understand where the error is sourcing from?

    I do not believe we have any specific TI document (that I know of) which references using a minimum of 20dB gain reserve. This value depends on the application and accuracy requirements for the intended system, therefore it is difficult to generalize a guideline. 

    The customer is welcome to design more or less gain reserve into their application as they see fit. Ultimately, more gain reserve helps many factors of op-amp performance as Kai mentioned in his previous post. Additionally, many characteristics change over process and temperature, so it doesn't hurt to design in some amount of guard band. 

    This Application Note covers how loop gain influences the accuracy of op-amps: Operational amplifier gain stability, Part 2: DC gain-error analysis

    The mathematical derivations in the app note prove how less loop gain creates increased gain error in amplifiers. 

    Thank you,

    Jacob

  • Hi Jacob,

    Thank you very much for your strong supports.

    >Being off by such a large factor(26%) makes me believe other issues may be at play.

    Yes. I thought that. But they are evaluating TLV9062 using the type B board (Inverting amplifier) of DUAL-DIYAMP-EVM.
    https://www.ti.com/lit/ug/sbou193/sbou193.pdf.
    Also they use 1% accuracy resistors. 

    So I think their measurement environment is not bad.

    >Can the customer test the gain across the filter passband to help me understand where the error is sourcing from?

    I have their measurement data. But they don't want to disclosed the data on E2E.
    So I will send you friendship request to share the data with you. could you please accept it?

    After your approval, I will share the data on friendship site.

    Thank you again and best regards,

    Kazuya.  

  • Hi Jacob,

    I just sent you the data by e-mail I could know your address.

    Could you please take a look and please give me your comment by e-mail?

    Thank you and best regards,

    Kazuya. 

  • Hi Kazuya,

    please don't feel offended, but the mistake here is that the customer uses a way too slow OPAmp in this application. Even if the error has another cause, it makes no sense at all to amplify 10kHz with a gain of 150V/V with a 10MHz OPAmp. The circuit will never properly work but will suffer from distortion and temperature drifts. 

    The usual way is to split the gain evenly onto two amplifiers or to take a much faster OPAmp which may not be a good idea for a beginner.

    Kai

  • Hello Kazuya-san, 

    Jacob is currently out of office, so I will continue the support here. 
    I am not able to see the data that you have shared on Jacob's email, however if you would like to follow Kai's suggestion of a high speed op amp here is the op amp portfolio on ti.com: https://www.ti.com/amplifier-circuit/op-amps/high-speed/overview.html

    All the best,
    Carolina

  • Hi Kazuya,

    in the following simulation I have compared three scenarios, the customer's circuit (VF1), the circuit with an ideal OPAmp (VF2) instead of the TLV9062 and a circuit with two TLV9062 (VF3) with the gain evenly distributed over the two OPAmps:

    kazuya_tlv9062_2.TSC

    It can clearly be seen that the "single OPAmp circuit" drastically differs from the circuit with the ideal OPAmp. This is the consequence of not having enough gain reserve.

    The "dual OPAmp circuit", on the other hand, comes the ideal performance much closer. To simpliy the design and to not alter the bandpass curve, the first OPAmp U1 is only providing gain while the second OPAmp U3 is providing the bandpass curve. C5 is used to allow the necessary AC coupling and C2 stabilizes U1 by adding some phase lead compensation.

    And in the simulation below the much faster OPA365, as recommended by Caro, is compared with the ideal OPAmp:

    kazuya_tlv9062_3.TSC

    The OPA365 is a 50MHz OPAmp and has a 14dB higher gain reserve compared to the TLV9062. As it can be seen from the simulations, the "dual OPAmp circuit" with the TLV9062 is even a little bit better than the "single OPAmp circuit" with the OPA365, but consumes more components and board space, of course.

    It can clearly be seen that having enough gain reserve is the key factor in this application.

    Kai

  • Hi Kai, 

    Thank you very much for your strong supports.

    Could I ask you similar questions to my previous one?

    Q1. The customer designed the inverting circuit has 150V/V(43.5dB). But the actual measured peak gain was about 110V/V(40.8dB).
    Is the cause of this peak gain difference between TINA-TI and actual data the small gain reserve?

    Q2. TLV9062 open loop gain is about 10dB higher than 43.5dB at 12kHz. Is the open loop gain not sufficient for getting the peak gain which simulated by TINA-TI because gain reserve needs 20dB at least?

    Q3. You told me that 20dB gain reserve has nothing to do with the manufacturing tolerances of OPAmp and it is simple "OPAmp physics".
    Is there any reference or document about 20dB gain reserve? The customer wants to get it.

    Thank you again and best regards,

    Kazuya.

  • Hi Kazuya,

    the reference I have for the minimum gain reserve is a very detailed databook from 1988 from a competitor. So I cannot post it here. In this databook a minimum gain reserve of 40dB is mentioned for a standard application.

    There are applications where the precision requirements are lower than in a standard application. For instance, if an amplifier shall only increase the signal somewhat in front of a comparator. Or toy circuits. Also, when using HF-OPAmps in many applications distortion and stability don't play the major role. Think of low budget video circuits. In all these cases a minimum gain reserve of 20dB is recommended.

    And in high precision applications even 40dB gain reserve may be too little and 60dB gain reserve is chosen.

    I think you will find these recommendations in other textbooks on operational amplifiers as well.

    Kai

  • Hi Kai,

    Thank you very much for your strong supports and I'm sorry to ask you many questions.

    I tried to find a document which explains the gain margin but I couldn't find any document today.
    I will continue to find the document.

    Could I confirm you a few points as the follows?

    You told me that 20dB gain reserve has nothing to do with the manufacturing tolerances of OPAmp.
    I think OPAMP open loop gain is changed a little by the manufacturing tolerances.

    Q1. Is my understanding not correct? 

    Q2. Does the manufacturing tolerances not affect to the gain of their circuit?

    Thank you and best regards,

    Kazuya.  

     

  • Hi Kazuya,

    manufacturing tolerances of open loop gain only affect the gain of their circuit, if the gain reserve is way too low, which is the case here. This is just one reason why the gain reserve should be chosen high enough: If the gain reserve is high enough, the manufacturing tolerances of open loop gain will not play any relevant role and the gain of circuit will be as programmed by the feedback components.

    There's a simple method to find out whether the manufacturing tolerances of open loop gain is responsible for the wrong gain of circuit: Just take some fresh TLV9062, insert them in the same circuit and check the gain again. Buy the TLV9062 at different stores to be sure that the TLV9062 will show some statistical variation of open loop gain.

    I send you a "friendship request" and give you a photo of the reference which recommends a gain reserve of 40dB.

    Kai

  • Hi Kai,

    Thank you very much for your strong supports.

    Best regards,
    Kazuya.