LME49720: popcorn noise

Hi Jose,

After some research, I found this isn't a new problem: https://e2e.ti.com/support/audio/f/6/t/415907?LM4562-LME49720-low-frequency-noise

When prototyping I tested maybe 10 pieces of LME49720 with no issue, but, now we are getting actual board samples I am experiencing exactly this issue, and, at approximately the same rate; 10% to 20% have this popcorn noise. It's not flicker noise, its random, low periodicity, but comparatively large amplitude fluctuations.

It's absolutely unacceptable, and, moves these samples from 'low noise' into 'terrible' territory.

So much so, a cheap NE5532 or similar from NJR is better and more consistent - so I am on the verge of signing off a design change.

We are talking quantities of potentially millions for this product Jose, we cannot accept this.

I wish your ICs were as silent as your technical support...

Here are two plots from my APx555 - showing the noise from my prototype, this is including all contributions, from DAC to output stage. The first plot is a bad LME49720, the second a cheap-as-chips NJM2114. The situation is actually worse than it looks here, because this noise is impulsive, much like popping corn, and the FFT is integrating it.

I have other samples which are similar, or worse.

Hi Andy,

what do you expect from a chip which has no maximum specification of 10Hz noise?

This is not a mistake of the LME49720 but is typical for almost all such OPAmps. By the way, your NJM2114 does not have any maximum specifications referring to noise! So, don't expect too much from the NJM2114. It might show the same increased noise some day.

And even the NE5532/4 is showing huge manufacturing tolerances referring to noise, regardless of the manufacturer. I remember, many years ago we had to screen all of our NE5534 for the use in a professional audio application. There were three groups, one group containing the lowest noise devices for the use in high gain microphone amplifier stages, another group with noisier devices for high gain but less demanding amplifier stages and a third group for low gain amplifier stages. 5% of all screened NE5534 fell in the lowest noise group, 50...70% fell in the second group and about 30% fell in the third group. 10...30% of the devices, depending on the batch, couldn't be used at all and would be better placed in the dustbin.

So, if you want to get state-of-the-art low noise devices, then you should screen them.

Kai

Hi Kai,

thank you for your input.

To begin: If you want state of the art audio performance, you look at physics and materials technology, develop relevant technologies, and design such into components such as transformers, resistors and capacitors, making them in-house and/or working with large component manufacturers (such as Rubycon) to develop audio specific components.

It's my day-job to do that Kai. It's unusual, I know.

I do indeed select components, although, many of them are correct by design, and, for example, Rubycon 100% test every capacitor, as do our resistor manufacturer, so we don't have to do so much ourselves on that level.

However, when entering into an enterprise where the quantities are 500,000 pieces plus, it might conceivably become 'inconvenient' for a large, non-specialised OEM manufacturer to screen op-amps for noise. For example, my experience, and that of the earlier poster shows that maybe 15% of the LME49720 have this noise, that means we would have to analyse possibly 1.2 million pieces, or more, to get the 1 million we need.

This is per year Kai.

And, we are talking here not about expected noise, not fine-grass or 1/f, this is popcorn noise, which is impulsive in nature, and not what we would call a 'natural' textbook noise, it's what we would call 'excess noise' - something which occurs in an unexpected, non-ideal manner.

That's why it should be mentioned.

And, furthermore, the FFTs belie the real problem - pops and farts are not the same as stochastic noise, they are extremely annoying to human listeners.

I don't care about the NJM2114, it's a peanuts chip, and, it's not going into this product. But, regardless of datasheets, I have measured a ton of them on my Audio Precision APx555. None of them exhibit this problem. No op-amp I have ever measured does in this way.

As you say, I can specify any old device, and select for them. But, here, we are talking about something which is offered as a premium component - it costs 10x to 20x what a NJM2114 costs. I expect to pay what amounts to several million dollars for something better, not for something worse.

And, Kai, 15% of the LMEs are worse, by a country mile, than any op-amp I have used in my career, in the worst possible way - popping. Take a look at the previous poster's scope traces. I guess you wouldn't mind a bit of hiss in your headphones, but popcorn?

Furthermore, this kind of defect, and I use that word with purpose, is usually created at wafer processing - which implies low quality, bluntly, it's caused by wafer defects.

Anyway, the problem is solved for me, although I am still using excellent TI parts elsewhere, I am using ADI and LT parts in these specific locations.

And, as an edit, I do actually lay awake at night, wondering about people who are such flatlanders that believe an extra zero before the one in their THD figures is going to make their product somehow satisfying to a human listener.

I am raising this issue because it is a processing fault, pure and simple. The product is not as described. If you want low noise, do not use the LME49720.

A

Hi Kai,

Thanks for your comment. When I get the bit between my teeth, erm, well, I hope I didn't come over too strong.

I would love to be able to screen these parts and use them, the good ones are good, and the price is acceptable, (and it sounds OK too), its an ideal combination, that's why we intended to use it. But I can't see how it's possible in our case, especially for a human listener to check a million op-amps!

We would be accused of 'cruel and unusual' treatment!

Although I do already have a solution, it's a little expensive, so, I will have to look to another, find, the Middle Way.

By the way, readers of these posts might be interested in the story of the Chevrolet Corvair, and the book 'Unsafe At Any Speed' about how certain corporations would rather deal with litigation than fix the problem itself.

Ivan,

I received 4 more pieces from our partner today and all four of them have this popcorn issue - so we are looking at approximately 50% of samples overall so far. Below is a typical noisy LME49720, below that is a LM4562 obtained locally:

Well, further to this story. I received another batch of LME49720s, and ALL of them displayed this issue. Every one was noisy. To try to help, they have been submitted to the local vendor.

Overall, it seems TI have a problem with process QA.

The LME49720 is not an IC which can be recommended for audio applications.