Objective Guide to Op-Amp Rolling - Part 1

[orchardaudio]

As per moderator, here is full article from the site:

Objective Guide to Op-Amp Rolling - Part 1
July 30, 2021

Op-Amp Test Board

This blog article/post is a deep dive into currently available 8-pin DIP dual op-amps, for those that like to swap op-amps in their designs and/or DIY projects.
The article will compare the performance of well know op-maps in two configurations;

1. Inverted 6dB gain
2. Unity gain buffer

The contenders are:

• Texas Instruments LME49720
• Texas Instruments OPA2134
• Texas Instruments LM4562
• Texas Instruments LM833
• New Japan Radio NJM2742
• Analog Devices LT1352

This list is not all inclusive of all available dual 8-pin devices, however the big hitters are all there. If you would like to see other devices tested send me a message.
Before getting into the measurements I had to design a circuit board which would allow me to easily swap out the op-amps (op-amp rolling). This is the schematic:

Op-Amp Test Board Schematic

The circuit board has decoupling capacitor C1,C3,C2, and C4 very close to the op-amp so that it can achieve it's best noise performance.
For the dual op-amps you can see one of the op-amps is configured for inverting 6dB gain with 10k and 4.99k resistors. The second op-amp is unity gain buffer. Both op-maps have the option to add either a 1k, 2k, or 10k load via the 2-pin jumper headers J1, J2, J3, J4, J5 and J6.
Lastly the board has four (4) mounting holes to make it easy to setup up.
Test Setup:
The circuit board is connected to a +/-12V linear power supply (Siglent / SPD3303X-E)
It is also connected to an Audio Precision APX-525 audio analyzer with analog performance package.
To start out I measure all op-amps and the audio-analyzer looped onto itself at 1kHz in the inverting 6dB configuration with 100k load (this load is internal to the audio analyzer). The bandwidth for these measurements is 22.4kHz.

THD+N Ratio vs Measured Level / Inverting 6dB Gain / 100k load / 22.4kHz bandwidth

Here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse. Here we can also see that the audio analyzer itself is the limiting factor for the measurements between 3.5 and 8V as all the measurements do not get any better in this area beyond what the audio analyzer is capable of achieving on it's own.
Next the same exact test but with op-amp in unity gain configuration.

THD+N Ratio vs Measured Level / Unity Gain / 100k load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Next I measure the intermodulation distortion (IMD) of the devices in the inverting 6dB gain configuration.

IMD (60Hz and 7kHz) vs Measured Level / 6dB Gain / 100k Load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Next I measure the intermodulation distortion (IMD) of the devices in the unity gain configuration.

IMD (60Hz and 7kHz) vs Measured Level / Unity Gain / 100k Load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Come back to the blog for the next set of test results of these devices.
Questions/Comments/Feedback all welcome.
Regards;
Leo

 

[solderdude]

Hi Leo,

Driving 100k is a walk in the park.

Some folks may be more interested in lower loads.
I think this is where the differences between most op-amps becomes most interesting.

Say a 2k load, maybe 600 ohm and 100 ohm or so.

 

[orchardaudio]

Hi Leo,

Driving 100k is a walk in the park.

Some folks may be more interested in lower loads.
I think this is where the differences between most op-amps become most interesting.

Say a 2k load, maybe 600 ohm and 100 ohm or so.

I know that is why there will be more measurements in the future. If you notice that board that I made lets me easily switch in 1k, 2k, and 10k loads.

Regards;
Leo

 

[solderdude]

Some folks drive headphones directly with opamps.

33 ohm and maybe 120 ohm may be fun (but requires even more measurements)

 

[levimax]

Hi,

For my last blog article I compared the performance of:

• Texas Instruments LME49720
• Texas Instruments OPA2134
• Texas Instruments LM4562
• Texas Instruments LM833
• New Japan Radio NJM2742
• Analog Devices LT1352

See here:
https://orchardaudio.com/blog/f/objective-guide-to-op-amp-rolling---part-1

Regards;
Leo

Nice work and very interesting. The problem I have with "rolling" op amps is that there can be subtle differences in the performance of the op amps beyond noise and distortion (such as how oscillation needs to be managed) so a good circuit design using one op amp may oscillate (or do something else undesirable) if another op amp is dropped in even if the other op amp has lower distortion. The performance of all these op amps are excellent and I doubt anyone could hear a difference caused by the distortion differences.

 

[JohnYang1997]

Nice work and very interesting. The problem I have with "rolling" op amps is that there can be subtle differences in the performance of the op amps beyond noise and distortion (such as how oscillation needs to be managed) so a good circuit design using one op amp may oscillate (or do something else undesirable) if another op amp is dropped in even if the other op amp has lower distortion. The performance of all these op amps are excellent and I doubt anyone could hear a difference caused by the distortion differences.

The idea is for the end user: Do not roll opamps. They may have similar 8pin packages, they don't really have the same features.

For designers, there are many more stuff than this. Samuel Groner has made this massive compilation that's incredibly detailed. http://www.nanovolt.ch/resources/ic_opamps/

 

[Count Arthur]

I thought that the LM4562 and LME49720 were in fact the same thing; their specs are identical: https://www.ti.com/product/LM4562, https://www.ti.com/product/LME49720

 

[orchardaudio]

Some folks drive headphones directly with opamps.

33 ohm and maybe 120 ohm may be fun (but requires even more measurements)

That requires high current op-amps which none of these are. For example, I use OPA1622s in parallel for my headphone driver on the PecanPi DAC., however, this part is SMT and won't be tested as part of these tests.

 

[solderdude]

Still you would not believe how many headphone output stages use opamps like these.

 

[orchardaudio]

Still you would not believe how many headphone output stages use opamps like these.

They most likely have a resistor in series with the headphone output to limit the current to something the op-amp can handle.

 

[solderdude]

Most do not but some add resistors between 10 and 68 ohm hardly enough to prevent current clipping when the output is shorted.
Just check out 'Chu Moy' (or this one)

 

[jtwrace]

https://headwizememorial.wordpress.com/

 

[tomchr]

Most do not but some add resistors between 10 and 68 ohm hardly enough to prevent current clipping when the output is shorted.

That also conveniently prevents the amp from spot welding the headphone plug inside the headphone jack if the phones are plugged/unplugged while the music is playing. The Spot Welding Amp is less of an issue with opamps as they tend to be current limited, but beefy discrete stages can easily provide enough current.

Tom

 

[tomchr]

Actually what we see is that the measurement is noise limited at amplitudes up to 800 mV and limited by the residual THD of the AP from 800 mV and up for all but two test cases.

It would be really helpful if you could measure the opamps in the same sequence so one can compare blue vs blue, for example.

To get a true measurement of the THD, I suggest looking at the test circuit in the LM4562 (= LME49720) data sheet. I also suggest optimizing the circuit for low noise so you can tease the THD out.

Tom

 

[orchardaudio]

It would be really helpful if you could measure the opamps in the same sequence so one can compare blue vs blue, for example.

I will keep this in mind for the next set of tests.

 

[LTig]

As per moderator, here is full article from the site:

Objective Guide to Op-Amp Rolling - Part 1
July 30, 2021

Op-Amp Test Board

This blog article/post is a deep dive into currently available 8-pin DIP dual op-amps, for those that like to swap op-amps in their designs and/or DIY projects.
The article will compare the performance of well know op-maps in two configurations;

1. Inverted 6dB gain
2. Unity gain buffer

The contenders are:

• Texas Instruments LME49720
• Texas Instruments OPA2134
• Texas Instruments LM4562
• Texas Instruments LM833
• New Japan Radio NJM2742
• Analog Devices LT1352

This list is not all inclusive of all available dual 8-pin devices, however the big hitters are all there. If you would like to see other devices tested send me a message.
Before getting into the measurements I had to design a circuit board which would allow me to easily swap out the op-amps (op-amp rolling). This is the schematic:

Op-Amp Test Board Schematic

The circuit board has decoupling capacitor C1,C3,C2, and C4 very close to the op-amp so that it can achieve it's best noise performance.
For the dual op-amps you can see one of the op-amps is configured for inverting 6dB gain with 10k and 4.99k resistors. The second op-amp is unity gain buffer. Both op-maps have the option to add either a 1k, 2k, or 10k load via the 2-pin jumper headers J1, J2, J3, J4, J5 and J6.
Lastly the board has four (4) mounting holes to make it easy to setup up.
Test Setup:
The circuit board is connected to a +/-12V linear power supply (Siglent / SPD3303X-E)
It is also connected to an Audio Precision APX-525 audio analyzer with analog performance package.
To start out I measure all op-amps and the audio-analyzer looped onto itself at 1kHz in the inverting 6dB configuration with 100k load (this load is internal to the audio analyzer). The bandwidth for these measurements is 22.4kHz.

THD+N Ratio vs Measured Level / Inverting 6dB Gain / 100k load / 22.4kHz bandwidth

Here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse. Here we can also see that the audio analyzer itself is the limiting factor for the measurements between 3.5 and 8V as all the measurements do not get any better in this area beyond what the audio analyzer is capable of achieving on it's own.
Next the same exact test but with op-amp in unity gain configuration.

THD+N Ratio vs Measured Level / Unity Gain / 100k load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Next I measure the intermodulation distortion (IMD) of the devices in the inverting 6dB gain configuration.

IMD (60Hz and 7kHz) vs Measured Level / 6dB Gain / 100k Load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Next I measure the intermodulation distortion (IMD) of the devices in the unity gain configuration.

IMD (60Hz and 7kHz) vs Measured Level / Unity Gain / 100k Load / 22.4kHz bandwidth

Again here we see that the LM4562, LM833, LM49720 and OPA2134 all have the same level of performance with the NJM2742 having the worst performance of the bunch and the LT1352 being just slightly worse.
Come back to the blog for the next set of test results of these devices.
Questions/Comments/Feedback all welcome.

A few of these opamps (LM4562, OIPA 2134) and many more (LM741, NE5532/5534, AD797, OP27, OP270, OP275, TL072, TL052, OPA604, OPA627) have been tested by Douglas Self in inverting and non-inverting configuration, see chapter 4 of his book Small Signal Audio Design. Several of them failed in the non-inverting configuration due to common mode distortion (or failed with higher noise than specified due to input current cancelling). Since many circuits use opamps in non-inverting configuration it makes more sense to show test results in this configuration.

 

[orchardaudio]

A few of these opamps (LM4562, OIPA 2134) and many more (LM741, NE5532/5534, AD797, OP27, OP270, OP275, TL072, TL052, OPA604, OPA627) have been tested by Douglas Self in inverting and non-inverting configuration, see chapter 4 of his book Small Signal Audio Design. Several of them failed in the non-inverting configuration due to common mode distortion (or failed with higher noise than specified due to input current cancelling). Since many circuits use opamps in non-inverting configuration it makes more sense to show test results in this configuration.

The unity gain buffer tests are exactly this. As the opamp is used in the non-inverting configuration. For the next set of tests, I will add more realistic loads to the outputs that will start to show the difference between the opamps much more.

 

[levimax]

With all the requests I will put in mine for a couple more Op amps to add to the test.... I like NE5532/5534 as a reliable and cheap high performing option and the $0.50 TL02 as the "old school" reference. If not no worries. Thank you.

 

[LTig]

The unity gain buffer tests are exactly this. As the opamp is used in the non-inverting configuration.

Oops. I missed it in the schematics and didn't notice that these are all dual opamps, one channel in -6 dB inverting and one channel in unity gain non-inverting mode.

For the next set of tests, I will add more realistic loads to the outputs that will start to show the difference between the opamps much more.

You should use both a realistic load and the maximum load as specified or rated by the manufacturer. It makes no sense to use a higher load than allowed.

 

[sq225917]

1612? Gotta be worth a go, on an adaptor