"The Crocodile"

[Mark Tillotson]

Paralleling opamps can increase the current, or if the load is the same it reduces the loading on each opamp, which can improve THD (if a single opamp would be overloaded in the same position). There's no inherent change in THD in paralleling, but noise can be definitely reduced.


Inverting and non-inverting have different noise-gains, so are not the same from a compensation perspective.

 

[Shaq888]

Thanks, that explain a lot!
I have a few questions more:
- Does parallel opamp only increase current output or does it improve THD too?
- As i see the inverting and non-inverting configuration using different stability compensation, i think both type of compensation will work with both opamp configuration, just need to adjust value a bit, is that right?

Well, Mark already gave you pretty good answer. The only thing that here will be no visible noise reduction since it is mostly defined by the front end.
As for compensation of inverting vs non-inverting - the type of compensation i used on the very beginning may not be ideal for non inverting configuration. So may be worth of optimizing it a bit more than just adjusting values.

Does this mean the 3rd harmonic of the LM6172 57db down, can you explain where the 83db come from?
Thanks!

Apparently there was "the worst case LM distortion" at level 57dB - 83dB = -26dB. If you show where you are quoting from I may be able to tell more from a context.

 

[proid]

Well, Mark already gave you pretty good answer. The only thing that here will be no visible noise reduction since it is mostly defined by the front end.
As for compensation of inverting vs non-inverting - the type of compensation i used on the very beginning may not be ideal for non inverting configuration. So may be worth of optimizing it a bit more than just adjusting values.






Apparently there was "the worst case LM distortion" at level 57dB - 83dB = -26dB. If you show where you are quoting from I may be able to tell more from a context.

It's from this post in the first page

Buffer distortions will be suppressed by the gain depth the first opamp has at a particular frequency in a particular gain configuration. The lower GBW opamp you use, the more it matters.
You can estimate the worst possible case distortion for LM6172 for my case, assuming that it all comes from the LM's side. For example OPA has around 63dB gain @ 60kHz, minus 6dB for the noise gain, so there is around 57dB of the loop gain left. It means that LM's 3rd harmonic is at least 83dB down.



Yep, there are two channels on the PCB. You can see it from the schematic too.
It is 50x50mm. Was the aim to fit into Seeedstudio's smaller PCB size to get a good price.



I run it on +/9V. Hard to tell what it would be inside a box. Can have a look with an infrared camera what is going on while it is on the bench.



You can also make a switch to change the supply voltage, for a high and low impedance loads.

 

[Shaq888]

Oh, that...
I was the other way around. The third harmonic was at around -130dBc, external loop gain is around ~57dB, so the buffer third harmonic estimated somewhere at -83dBc, if no other factors involved.

 

[DNi]

As to the stability and driving capability of the final version: I used one channel (w/o 3uH coil at the output) to drive an LS 3/5a loudspeaker, and listened to the music for a few hours at an average SPL of about 80dB@1m (measured, DIN A).
No signs of instability, and my fingers tell me that the temperature of the op-amps may have been slightly above 40 deg. C.

Not bad for a 7x5 cm headphone amplifier!

Regards,
Braca

 

[Shaq888]

Hey DNi!

Kind of reminds me Self's crazy amp on multiple 5532 (or 5534?)
I guess SMPS may be a limiting factor while driving speakers.

BTW, I recently added OPA2156 measurements on that forum. It can deliver slightly more current than TLVx172 and quite a bit more linear even with a heavy loading.

 

[DNi]

The massive parallel amp of D. Self was built with 5532's.
I've built a version with 8xLM4562 a couple of years ago, but not to drive loudspeakers!
With "Crocodile 2", I measured SPL peaks in excess of 100dB, so I guess the SMPS met its limits at each of them.
Pity that OPA2156 is not compatible with the current PCB as the input IC - there is no OPA1156.
I could not enlarge your new screenshots in the other forum, now I must be a member in order to do so. The old attachments can still be enlarged.

Regards,
Braca

 

[Mark Johnson]

If the rails are +/- 18V then the output sinewave is +/- 15V peak, which equals 10.6V RMS. Thus the power delivered to the 8R load is (10.6 ^ 2) / 8 = 14 watts RMS.

If we assume that the NE5532 is approximately a Class B amplifier whose "efficiency" (Power delivered to the load)/(Power drawn from the supply) equals 50%, then the power dissipated inside the array of NE5532s equals the power delivered to the load: 14 watts RMS.

The output current of the array of NE5532s is sqrt(14 watts / 8R) = 1.323 amps RMS. If there are 64 chips in the array (thus 128 individual opamps), each opamp is called upon to deliver 10.3 mA RMS, which is barely within the datasheet specs.

And if there are 64 chips delivering an aggregate total of 14 watts, each chip delivers 0.22 watts. Since power dissipated equals power delivered (efficiency = 50%), each chip dissipates 0.22 watts. Theta JA is 85 degrees C per watt for the plastic DIP, so the junction temperature is (85 * 0.22) = 19 degrees C higher than the ambient air. Which seems reasonable.

The numbers are worse for 4 ohm loads, of course.

 

[Shaq888]

I could not enlarge your new screenshots in the other forum, now I must be a member in order to do so. The old attachments can still be enlarged.

Hmmm... strange, should not be any different. I guess you could try registering. It will ask you a simple question, like Ohms law, to make sure that you are not a bot

I guess I could post those measurements somewhere here as well... I'm just being lazy since it is quite a truckload of images.

Pity that OPA2156 is not compatible with the current PCB as the input IC - there is no OPA1156.

It is actually not to hard to solder it under a bit of an angle with some pins bent up. Not as aesthetically pleasing but it works. The second opamp need to be terminated of course.

 

[JensH]

BTW, I recently added OPA2156 measurements on that forum. It can deliver slightly more current than TLVx172 and quite a bit more linear even with a heavy loading.

Where can those measurements be found ?

 

[Shaq888]

I posted them here - Пидсылювач "Крокодил" или "цiкавi дослiди"
It's mostly in Russian and Ukrainian, may be Google translate can help?
Please note, all those spectra show absolute values. To get to dBc you need to subtract a main tone level.

 

[JensH]

Thanks
Google translate helps a lot!

 

[Shaq888]

I suspect that opa1656 would be quite close or a bit better. They look similar except 1656 has slightly extended region with gain roll of >6dB/oct, which may be helpful in some instances.
1656 also does not have R2R input range.

 

[DNi]

Yes, OPA1656 appears to be a good choice for the input stage.
I also thought on OPA828 because of its lower DC offset voltage, but it costs 2.5 more.

I now finished the amplifier, including the protection circuit.
I noticed that after a first switch-on the output relay stays open indefinitely. Switching the unit off, and immediately afterwards on again seems to reset the protection circuit, and the relay closes.
Comparing the PCB with the protection schematics, i noticed that the connection EN between the collectors of Q1, Q2, etc. and the EN pin on LMR64010 does not exist on the PCB, which is probably OK anyway.

Regards,
Braca

 

[Shaq888]

If 1656 is anything like 2156, and i belive it is, than it's godd for the input and the output stage
Offset usually not a big deal with such a small gain.

Oops, apparently I missed something May be a wire link may help.

 

[Shaq888]

In fact, 2156 was so linear even driving 50 ohms, that with so many in parallel it does not really need this nested feedback configuduration. I didn't notice any traces of thermal induced distortions either.

 

[DNi]

Thank you, Sergey.

I'll test the protection circuit and report back.

I've already have some experience with OPA2156 in a Wien-Bridge oscillator, so it'll probably be the op-amp to use in my next project - I intend to build one more amplifier on your PCB, but for a different purpose.
For testing MC headamps I need signals below 1mV with a very low THD, and the idea is to use your amp driving a 50 Ohm attenuator. I might as well build both channels and connect their outputs in parallel.
For that application I do not need the protection circuit.

Regards,
Braca

 

[Shaq888]

Or you can just grab any half decent opamp and make an inverting stage with a gain of -0.001. Pretty much anything will be linear at these magnitudes. Or non inverting with a similar divider in front of it.

 

[DNi]

Well, as they say: "If all you have is a hammer, everything looks like a nail".
I've got a couple of HP355 attenuators on my bench, and thought of a way to give them something to do .

Re. the protection circuit, it seems that a direct connection to the EN pin does not solve the problem completely because by pulling the collectors of Q1, Q2, etc. hard to +U, the relay closes at start-up, but the circuit does not react any more to a DC potential at IN_A & IN_B.
By connecting the two points with a resistor (3K at the moment), the protection remains armed after the switch-on and trips if there is a problem at the amp output. Shorting this resistor temporarily rearms the protection circuit.
A small-size TH resistor fits nicely between C102 and the collector of Q1.

Regards,
Braca

 

[Shaq888]

Finally found some free time to put those spectra from opa2156 distortion measurement into a document. There are mistakes for sure, so please open with caution

Microsoft OneDrive - Access files anywhere. Create docs with free Office Online.