Big test of 13 op-amps in inverting configuration with gain = -1 and non-inverting configuration with gain = +1

[pma]

Big test of 13 op-amps in inverting configuration with gain = -1

This post focuses on distortion as it is often not well specified for specific circuits. Distortion tests were done on each op-amp type, with an output of 1.9Vrms. I measured shunt feedback (inverting topology) with gain of -1,load 1kΩ. 13 types of op-amps were measured, as they are shown in the following chart. Distortion measured at 10kHz is added to the basic datasheet parameters.

I measured THD vs. frequency with sampling rate of 96 kHz and measuring bandwidth 45kHz approx., for all types. Op-amps were loaded by 1kΩ. In all graphs, one can see a comparison of system loopback distortion (green line) with the distortion after insertion of the op-amp test rig. There is a systematic error in all measurements below 100Hz. The rise of distortion below 100Hz is due to distortion in input coupling electrolytic capacitors of the E1DA Cosmos ADC and is always the same.

Op-amps used and results measured

1. OPA134
is an op-amp with JFET input stage. It has acceptable noise. THD vs. frequency plotted below:

We can see some rise of distortion above 1kHz compared to the loopback result.

2. OPA627
is an op-amp with JFET input stage. It has low noise for JFET op-amp and is very expensive.

High frequency distortion is very low and the lower values above 5kHz compared to the loopback are probably a result of lighter DAC load with the test rig than in case of a loopback, or a phase cancellation of distortion components.

3. OPA445
is an op-amp with JFET input stage. Its advantage is very high power supply voltage range up to +/-40V. It has higher noise.

Distortion starts to rise quite early above 500Hz.

4. TL071
is a general purpose, cheap op-amp with JFET input stage. It has higher noise.

Disappointing result of distortion vs. frequency.

5. MAE411
is an op-amp with JFET input stage, a clone of LF411, produced in former Czechoslovakia. It has higher noise and parameters seem to be same as for TL071. But the results are not same.

Though distortion is also rising with frequency, it is much lower than in case of TL071 and is in fact acceptable.

6. AD744
is an op-amp with JFET input stage. It was declared as an improved version of TL071/LF411 and produced by Analog Devices. It has higher noise.

The op-amp is better than TL071, but not much better than MAE411. Interesting.

7. LT1122
is an op-amp with JFET input stage. It is fast enough, but it has higher noise.

Distortion is well controlled and remains low.

8. OPA177
is an op-amp with BJT input stage and is rather supposed to be used in DC applications. It has low noise, high open loop gain, however very low GBW and slew rate.

This is a disaster and the part is unusable for audio, as expected from very low slew rate and GBW. High DC open loop gain does not help here.

9. LT1028
is an op-amp with BJT input stage. It has very low noise, high GBW and sufficient slew rate.

Very low distortion, perfect copy of the input signal.

10. AD4898
is an op-amp with BJT input stage. It has very low noise and moderate open loop gain. High GBW and slew rate.

Very low distortion. The lower values above 5kHz compared to the loopback are probably a result of lighter DAC load with the test rig than in case of a loopback, or a phase cancellation of distortion components.

11. LM6171
is an op-amp with special design of BJT input stage. It is extremely fast and has moderate to low open loop gain. It has higher noise.

Distortion is very low and well controlled.

12. AD844
is an op-amp with BJT input stage, however it has CFA (current feedback) circuit topology. A comparison with voltage feedback op-amps is not that easy. It is very fast and has low noise.

Distortion is almost independent on frequency, but is generally higher than distortion of best VFA op-amps.

13. WSH111
is a hybrid IC op-amp with BJT input stage, built around ancient uA709 core. It has additional input BJT pair, parallel feed forward HF signal path and new output stage. It is very fast, but has quite high voltage noise. It can be only used in the inverting circuit. It has moderate to low open loop gain.

Distortion is similar to AD744. Not perfect, but incredible improvement to core uA709.

Conclusion
There are considerable differences in distortion of op-amps even in the simplest circuits. It is a myth that op-amps make no difference. Parts like TL071 should be avoided in audio. It can be also seen that high open loop gain and strong feedback is not everything and that the op-amp must have sufficient slew rate and GBW to be able to keep very low distortion at higher frequencies. As shown in examples of ADA4898 and LM6171, despite the low OLG is HF distortion low.

Follow-up
Measurements with gain +1 (voltage follower) were added in post #5 and comparisons with measurements in gain -1 are shown.

 

[DVDdoug]

It is a myth that op-amps make no difference.

I'd probably avoid the OPA177 but the others should be better than human hearing in low-gain applications... I doubt they make an audible difference... I'm pretty sure I can't hear 1% distortion so the OPA177 is probably OK but I wouldn't choose it. Hopefully, nobody claims no measurable difference.

In a high-gain mic preamp, of course I'd want very low noise.

 

[Salt]

Big test of 13 op-amps in inverting configuration with gain = -1

This post focuses on distortion as it is often not well specified for specific circuits. Distortion tests were done on each op-amp type, with an output of 1.9Vrms. I measured shunt feedback (inverting topology) with gain of -1,load 1kΩ. 13 types of op-amps were measured, as they are shown in the following chart. Distortion measured at 10kHz is added to the basic datasheet parameters.

View attachment 446722
I measured THD vs. frequency with sampling rate of 96 kHz and measuring bandwidth 45kHz approx., for all types. Op-amps were loaded by 1kΩ. In all graphs, one can see a comparison of system loopback distortion (green line) with the distortion after insertion of the op-amp test rig. There is a systematic error in all measurements below 100Hz. The rise of distortion below 100Hz is due to distortion in input coupling electrolytic capacitors of the E1DA Cosmos ADC and is always the same.

Op-amps used and results measured

1. OPA134
is an op-amp with JFET input stage. It has acceptable noise. THD vs. frequency plotted below:
View attachment 446725
We can see some rise of distortion above 1kHz compared to the loopback result.

2. OPA627
is an op-amp with JFET input stage. It has low noise for JFET op-amp and is very expensive.
View attachment 446726
High frequency distortion is very low and the lower values above 5kHz compared to the loopback are probably a result of lighter DAC load with the test rig than in case of a loopback, or a phase cancellation of distortion components.

3. OPA445
is an op-amp with JFET input stage. Its advantage is very high power supply voltage range up to +/-40V. It has higher noise.
View attachment 446727
Distortion starts to rise quite early above 500Hz.

4. TL071
is a general purpose, cheap op-amp with JFET input stage. It has higher noise.
View attachment 446728
Disappointing result of distortion vs. frequency.

5. MAE411
is an op-amp with JFET input stage, a clone of LF411, produced in former Czechoslovakia. It has higher noise and parameters seem to be same as for TL071. But the results are not same.
View attachment 446733
Though distortion is also rising with frequency, it is much lower than in case of TL071 and is in fact acceptable.

6. AD744
is an op-amp with JFET input stage. It was declared as an improved version of TL071/LF411 and produced by Analog Devices. It has higher noise.
View attachment 446737
The op-amp is better than TL071, but not much better than MAE411. Interesting.

7. LT1122
is an op-amp with JFET input stage. It is fast enough, but it has higher noise.
View attachment 446738
Distortion is well controlled and remains low.

8. OPA177
is an op-amp with BJT input stage and is rather supposed to be used in DC applications. It has low noise, high open loop gain, however very low GBW and slew rate.
View attachment 446739
This is a disaster and the part is unusable for audio, as expected from very low slew rate and GBW. High DC open loop gain does not help here.

9. LT1028
is an op-amp with BJT input stage. It has very low noise, high GBW and sufficient slew rate.
View attachment 446740
Very low distortion, perfect copy of the input signal.

10. AD4898
is an op-amp with BJT input stage. It has very low noise and moderate open loop gain. High GBW and slew rate.
View attachment 446741
Very low distortion. The lower values above 5kHz compared to the loopback are probably a result of lighter DAC load with the test rig than in case of a loopback, or a phase cancellation of distortion components.

11. LM6171
is an op-amp with special design of BJT input stage. It is extremely fast and has moderate to low open loop gain. It has higher noise.
View attachment 446742
Distortion is very low and well controlled.

12. AD844
is an op-amp with BJT input stage, however it has CFA (current feedback) circuit topology. A comparison with voltage feedback op-amps is not that easy. It is very fast and has low noise.
View attachment 446743
Distortion is almost independent on frequency, but is generally higher than distortion of best VFA op-amps.

13. WSH111
is a hybrid IC op-amp with BJT input stage, built around ancient uA709 core. It has additional input BJT pair, parallel feed forward HF signal path and new output stage. It is very fast, but has quite high voltage noise. It can be only used in the inverting circuit. It has moderate to low open loop gain.
View attachment 446745
Distortion is similar to AD744. Not perfect, but incredible improvement to core uA709.

Conclusion
There are considerable differences in distortion of op-amps even in the simplest circuits. It is a myth that op-amps make no difference. Parts like TL071 should be avoided in audio. It can be also seen that high open loop gain and strong feedback is not everything and that the op-amp must have sufficient slew rate and GBW to be able to keep very low distortion at higher frequencies. As shown in examples of ADA4898 and LM6171, despite the low OLG is HF distortion low.

Once again a lot of effort, Pavel, Respect!
I very appreciate Your findings about 'what can be measured to be a difference' beyond 'what could be audible'.
Thanks!

 

[solderdude]

That's why it is important to select the right op-amp for a job at hand.
This is the responsibility of the designer.

Unfortunately, for large companies, sometimes the purchasing department opts for a slightly cheaper part that works about just as well.
As can be seen in this test various op-amps perform very similar (in the test conditions present).
Another reason for manufacturers to select a certain op-amp can be availability (in large quantities) or liaisons with certain manufacturers they source their components from.

And for the numbers people... note how well all of these op-amps perform well at 1kHz... this also happens to be the frequency a lot of gear is specified at.

 

[pma]

Some more measurements -

Measurements with gain = +1 (voltage follower) and comparison to results with gain = -1 (inverter)

1. OPA134
Distortion as a follower is almost same as an inverter, only slightly higher above 10kHz.

2. OPA627
Up to 3kHz, distortion as a follower is same as an inverter, above 4kHz the follower distortion is higher.

3. OPA445
Distortion as a follower is considerably lower than an inverter above 500Hz.

4. TL071
Distortion as a follower is considerably lower than an inverter above 100Hz.

5. MAE411
Distortion as a follower is almost identical as an inverter.

6. AD744
Distortion as a follower is almost identical as an inverter.

7. LT1122
Distortion as a follower is almost same as an inverter, only slightly higher above 10kHz.

8. OPA177
Distortion as a follower is considerably lower than an inverter.

9. LT1028
Distortion as a follower is almost identical as an inverter.

10. AD4898
Distortion as a follower is almost same as an inverter, however higher above 10kHz.

11. LM6171
Distortion as a follower is almost identical as an inverter.

12. AD844
CFA (current feedback) circuit topology and it is not intended to be used as a simple voltage follower, so it was not measured.

13. WSH111
is a hybrid IC op-amp with BJT input stage, built around ancient uA709 core. It has additional input BJT pair, parallel feed forward HF signal path and it is not intended to be used as a simple voltage follower, so it was not measured.

Some of the op-amps (MAE411, AD744, LT1028, LM6171) have very consistent distortion behaviour if we compare follower and inverter results. Some others have very inconsistent behaviour (OPA445, TL071, OPA177). A proper design choice is always needed. Reckless op-amp swapping by users is a nonsense.

 

[melomane13]

Why NE5532 isn't in the list?

 

[pma]

Why NE5532 isn't in the list?

Because dual op-amp cannot be directly inserted into DIP-8 socket in a test rig for single op-amps. The socket pins have different signals. Single version NE5534 is not stable as a follower and inverter, so it could not be tested.

NE5532 has already been tested in the test of dual op-amps.

 

[solderdude]

Can you post system loopback ?

 

[DonH56]

Nice work!

Is 1 k-ohm a reasonable load for all of these op-amps? I do not know, curious, knowing 10 k-ohm is a typical lower value for most audio inputs (sans the unbuffered class D amp modules). I calculate 1.9 Vrms is about 2.7 Vpk so only about 2.7 mA (not accounting for any capacitance and slew requirements) thus does not seem unreasonable, but I know a number of op-amps have low quiescent bias and rated output current.

 

[MAB]

Nice work!

Is 1 k-ohm a reasonable load for all of these op-amps? I do not know, curious, knowing 10 k-ohm is a typical lower value for most audio inputs (sans the unbuffered class D amp modules).

I've noticed some headphone amps with 1k Ohm input impedance. For instance:

Drop + THX AAA™ ONE Linear Amplifier Details | Audiophile | Amps | Headphone Amps

Discover 490 Community discussions, tips, and reviews on Drop + THX AAA™ ONE Linear Amplifier from the Audiophile enthusiast community on Drop.

drop.com

Should be worst-case, I hope!

I calculate 1.9 Vrms is about 2.7 Vpk so only about 2.7 mA (not accounting for any capacitance and slew requirements) thus does not seem unreasonable, but I know a number of op-amps have low quiescent bias and rated output current.

 

[wwenze]

Kind of refreshing to know that OPA627 and LT1028 (the two best op-amps among "audiophiles" 20 years ago) do measure the best, even if inaudible.

Meanwhile the signal goes into OPA1612 in the E1DA

What will happen if you put a OPA1611 into this test (or the OPA1612 in the dual-opamp test?) Will we see perfect matching with E1DA, or will we see the impact of the test circuit?

 

[pma]

Now, please let me show you what happens if we measure op-amps with a mediocre interface - Focusrite 2i2. TL071 vs. OPA627 were tested in Gain +1 (follower) configuration, loaded with 10kohm. We now get identical distortion plots, that are a result of very poor 2i2 higher frequency distortion qualities. And, also test music samples analyzed by Deltawave are almost identical. That is the reason why Amir's test of op-amps placed into mediocre amplifiers say that "there is no difference". The differences are masked by poor DUT devices. I have shown in another threads that op-amp swapping in Topping D10s DAC and PRE-TC10 preamp (EQ bypassed) makes clear measurable differences. Possible audibility or inaudibility is to be tested in well prepared and controlled DBT tests, it should not be a subject of mere speculations.

 

[solderdude]

That is the reason why Amir's test of op-amps placed into mediocre amplifiers say that "there is no difference". The differences are masked by poor DUT devices. I have shown in another threads that op-amp swapping in Topping D10s DAC and PRE-TC10 preamp (EQ bypassed) makes clear measurable differences. Possible audibility or inaudibility is to be tested in well prepared and controlled DBT tests, it should not be a subject of mere speculations.

Yep, this has been made clear throughout those threads. The amps performance was limited by the class-D output stage not by the compared op-amps which both are 'better performing' than the output stage.
A question that still remains (for me) is why there is no (significant) difference in the P4 test ..
Is that perhaps because the swapped op-amp in the design is not the only op-amp the signal passes through and the swap-able op-amps are both 'better' in performance than the one signals possibly may pass through ?
My eye spies some SOIC-8 SMD op-amps in there that may or may not be 'active' when the tone circuit is 'bypassed'.

It is also clear that different op-amps measure different. The audibilty question remains as that depends on multiple conditions.

 

[AnalogSteph]

Is 1 k-ohm a reasonable load for all of these op-amps?

Definitely not for the TL071, no. This one was designed for relatively high-level, high-impedance work way back in the 1970s. (In this context its noise level isn't so bad either.) In a mixer you might see them inverting with 22k feedback resistors, driving 10k or higher loads. In a speaker balanced input stage handling more modest levels (like +4 dBu or less) you might see a 10k/10k/10k/10k circuit followed by a 5-10k pot.

This is why it was one of the parts exhibiting lower distortion as a follower. The inverting amplifier has a noise gain of 2 and as such only has half as much GBW available to keep output stage distortion in check. This effect was entirely masking the common-mode distortion we should normally be seeing in a follower with this part, though the signal amplitude here may be too small to clearly show it in general. Samuel Groner used something like +20 dBu if memory serves.

 

[pma]

Definitely not for the TL071, no. This one was designed for relatively high-level, high-impedance work way back in the 1970s. (In this context its noise level isn't so bad either.) In a mixer you might see them inverting with 22k feedback resistors, driving 10k or higher loads. In a speaker balanced input stage handling more modest levels (like +4 dBu or less) you might see a 10k/10k/10k/10k circuit followed by a 5-10k pot.

When TL071 is loaded with 10k in a follower configuration, its behaviour is definitely acceptable. Below a comparison with OPA627 in a follower configuration, 10k load, 1.9V output.

Test circuit is more complex. Behind 10k load resistor, there is a SE/balanced converter with 2xADA4898 and very high input impedance. This rises level for the E1DA ADC and symmetrizes the signal, which results in suppression of LF distortion of the input coupling electrolytic capacitors in the ADC (see lower rise of distortion below 100Hz).

1k load asks only 2mA from the op-amp at 2V. But, as we can see, it is still a lot for output stage designs like TL071 has. I agree, TL071 is cheap (I bought 50pcs at $0.25/pc recently for non-audio purposes), but my goal is to seek for best parameters not cost related. And the parts that are not that much related on circuit used. Like OPA627. Audio is a hobby to me, not the means of my income.

It is complicated. Everything is application circuit related and simple deductions from tests with high noise gains (as shown in datasheets) are quite inaccurate. During real circuit testing, I always find something unpredictable. Again, simplified conclusions are difficult.

 

[pma]

In a speaker balanced input stage handling more modest levels (like +4 dBu or less) you might see a 10k/10k/10k/10k circuit followed by a 5-10k pot.

https://pmacura.cz/ULTIM2-sch2.jpg

 

[pma]

Gain +2, op-amp (DUT) + BUF634, 1kohm load

More measurements, now with gain = +2, op-amp under test is followed by BUF634 high speed unity gain buffer (inside FB loop), output load is 1kohm. Several op-amps are tested and as a reference overlay there is always the result of OPA627 + BUF634.

1. TL071

2. ADA4898

3. LT1122

4. LT1028

Measurement bandwidth is always 90kHz with 192kHz sampling. Distortion below 100Hz is not shown because of considerable distortion of E1DA Cosmos input coupling electrolytic capacitors. This is a SE circuit test, so the cancellation of capacitor distortion by balanced signal connection does not occur here.
Please note that the usage of BUF634 buffer effectively erases distortion due to op-amp output stage loading.
Output level is 4Vrms.

 

[Salt]

Not audible ... not?

 

[pma]

Such low distortion is not audible and the BUF634 buffer helps a low cost TL071 to get very good results. I did use such circuit for audio since 2002, and for measuring purposes, DC coupled, since about 1995.

 

[MaxwellsEq]

It is complicated. Everything is application circuit related and simple deductions from tests with high noise gains (as shown in datasheets) are quite inaccurate. During real circuit testing, I always find something unpredictable

This is a strong, backed by measurements, argument why people should never randomly swap op-amps.