Tweet
Hello guys,
Since some of the topics discussed cases of fake components/chips, I decided to do a few basic experiments on the actual parameters of some operational amplifiers, compared to the declared catalog data.In this case, I was specifically interested in the frequency bandwidth of the amplifiers.
For this purpose, I used a modest unipolar pulse-induction metal detector, in which I used different amplifiers, keeping the same gain, the same offset, he same damping point of the coil, etc. I received the data from the ADC of the microcontroller (PIC16F1787), through sequential sampling. I tested 3 operational amplifiers - NE5532, TL082 and TL072. The downloaded data is on the DATA sheet from the attached excel file.
After sorting the data, I was surprised to see a drastic difference in the characteristics of the TL082 compared to the TL072, even though according to catalog data they are almost identical. Apparently (at least to me), the NE5532 and TL072 amplifiers met their specifications, but the TL082 behaved quite strangely.
To assess how big the difference is, I first tried to simulate the coil damping process (with MOSFET limitation included) by deriving computational data from the output of an ideal op amp. The calculations used the actual values of the components (inductance and resistance of the coil, damping resistor, amplifier gain, etc.) that participated in the experiment of downloading data from the ADC. This is on the "IDEAL AMPLIFIER PROCESS" page of the attached file.
I expected that the "long tails" of the obtained experimental data are due to the inclusion of a second time constant in the process (tau_coil+tau_op), which changes the general exponential character of the decay curve. For this I set different time constants for "tau_op" to find a real amplifier that would behave like the ideal one, so that I could then make comparisons with the amplifiers used in the experiment. The calculations are on the page "Calculations with two taus". The result was an amplifier with GBW = 207MHz. This, as a result, is again strange, because logically the bandwidth should not be limited (the higher the GBW, the more accurately the input decay curve will be followed). Obviously, the computational model used is not unipolar, so the data obtained are unreliable.
For this I approached from the other side - the dependence of GBW on the slope of the characteristic in the "most active" part. To isolate the signal amplitude, ADC normalization, gain, etc, I use logarithmic ratios, which I apply/normalize to amplifiers with verified/reliable data. This is done on the "GBW" page. In the end, the following results were obtained -> GBW is equal to:
GBW = arctg(ln((S2-S1)/(t2-t1)))*120 - for operational amplifiers with bipolar transistors at the input;
GBW = arctg(ln((S2-S1)/(t2-t1)))*60 - for operational amplifiers with field effect transistors at the input.
To see if this is a reliable result, I also tried with LM358 - an extremely weak amplifier, but still the detector worked with it. Due to the really weak characteristics of the chip, the "active part" was shifted to the left, but to be consistent with the other experiments, the same interval of 12 us was used. The obtained GBW for it is 0.755 MHz (according to the catalog it is 0.7 MHz).
The calculated GBW of TL082 is equal to about 1 MHz, which is actually not TL082, but TL062.
Why did I write you so many unimportant things - because I can't answer one really important question:
Is it possible that the cheapest possible chip in the world, bought from an authorized dealer (COMET/FARNELL), manufactured by a world-renowned manufacturer - Texas Instruments, could be a fake?
Since some of the topics discussed cases of fake components/chips, I decided to do a few basic experiments on the actual parameters of some operational amplifiers, compared to the declared catalog data.In this case, I was specifically interested in the frequency bandwidth of the amplifiers.
For this purpose, I used a modest unipolar pulse-induction metal detector, in which I used different amplifiers, keeping the same gain, the same offset, he same damping point of the coil, etc. I received the data from the ADC of the microcontroller (PIC16F1787), through sequential sampling. I tested 3 operational amplifiers - NE5532, TL082 and TL072. The downloaded data is on the DATA sheet from the attached excel file.
After sorting the data, I was surprised to see a drastic difference in the characteristics of the TL082 compared to the TL072, even though according to catalog data they are almost identical. Apparently (at least to me), the NE5532 and TL072 amplifiers met their specifications, but the TL082 behaved quite strangely.
To assess how big the difference is, I first tried to simulate the coil damping process (with MOSFET limitation included) by deriving computational data from the output of an ideal op amp. The calculations used the actual values of the components (inductance and resistance of the coil, damping resistor, amplifier gain, etc.) that participated in the experiment of downloading data from the ADC. This is on the "IDEAL AMPLIFIER PROCESS" page of the attached file.
I expected that the "long tails" of the obtained experimental data are due to the inclusion of a second time constant in the process (tau_coil+tau_op), which changes the general exponential character of the decay curve. For this I set different time constants for "tau_op" to find a real amplifier that would behave like the ideal one, so that I could then make comparisons with the amplifiers used in the experiment. The calculations are on the page "Calculations with two taus". The result was an amplifier with GBW = 207MHz. This, as a result, is again strange, because logically the bandwidth should not be limited (the higher the GBW, the more accurately the input decay curve will be followed). Obviously, the computational model used is not unipolar, so the data obtained are unreliable.
For this I approached from the other side - the dependence of GBW on the slope of the characteristic in the "most active" part. To isolate the signal amplitude, ADC normalization, gain, etc, I use logarithmic ratios, which I apply/normalize to amplifiers with verified/reliable data. This is done on the "GBW" page. In the end, the following results were obtained -> GBW is equal to:
GBW = arctg(ln((S2-S1)/(t2-t1)))*120 - for operational amplifiers with bipolar transistors at the input;
GBW = arctg(ln((S2-S1)/(t2-t1)))*60 - for operational amplifiers with field effect transistors at the input.
To see if this is a reliable result, I also tried with LM358 - an extremely weak amplifier, but still the detector worked with it. Due to the really weak characteristics of the chip, the "active part" was shifted to the left, but to be consistent with the other experiments, the same interval of 12 us was used. The obtained GBW for it is 0.755 MHz (according to the catalog it is 0.7 MHz).
The calculated GBW of TL082 is equal to about 1 MHz, which is actually not TL082, but TL062.
Why did I write you so many unimportant things - because I can't answer one really important question:
Is it possible that the cheapest possible chip in the world, bought from an authorized dealer (COMET/FARNELL), manufactured by a world-renowned manufacturer - Texas Instruments, could be a fake?
Comment