It's hard to win a debate when someone thinks that two identical chips are different.
"Identical" chips can in fact be slightly different. Depending on production batch, criteria of acceptance, an so on. This is particularly the case with NE5534 : Many years ago, I have tested lots of this device (Signetics, Philips, TI...). But I am retired and now I do not own sufficient quantities of LMxxx from various manufacturers to make a serious study on this question.
The "sound" of one particular AOP is depending essentially on the good or bad adaptation in one particular circuit, simple thing such decoupling capacitors can affect very diversely the total net result...
Exactly.
Just like transistors, in circuits where hfe matters, the higher tends to produce better result/sound. And transistors from production batches are classified based on this hfe. With opamps, we know the "D" suffix (e.g. NJM2068DD) which is a tested batch conforming to certain noise criteria...
I have many LF353 from various manufacturers and the best sounding one came from Comset Semiconductors.
And THD is not the only thing that affect sound. There are so many other things such that usually it is impossible (even with expensive opamps) to find a single opamp that excel in every categories. One criterion that is very important for me but seems to be ignored by most people is CMRR.
Last edited:
Driving headphones is not an easy task for most opamps. Even the old classics 4556 can drive headphones better than many expensive opamps. LM6172 has a high current output and has little issue in other categories such that it will be hard to beat when driving headphone is the objective. Sure the ultra high slew rate require extreme care in routing and everything.
But when cost is no-object, we can use more than one chip to achieve the same objectives. For example, output drive capability can also be achieved by adding buffer to an opamp that excel at everything else but output drive.
Would it be fair to say that there would be greater variability between different manufacturer's implementations of the 5532 than a single sourced 4562/49720/49860? Run to run and wafer level yield is something everyone has to deal with.
The minimum and maximum difference in the characteristics specification is sufficient to make two chips of the same type to sound different!
When big manufacturers decided to move their plants to china and asian countries, nothing they can do control the minimum quality of the product and what will happen to these products (it can go out through the windows and roofs).
Here in my country you can buy at least 5 different LME49720 chips. Sometimes, the best one (the most expensive) is called "original" and the next best thing is "KW1", "KW2" and so on.
I have learnt (from transistors) that these low quality parts are not always fake (it's original with a caveat).
You should not listen to people who compares opamp based on these low quality parts (it is not a fair comparison, besides they dont understand that different opamp needs different treatment). The parts from asia is not supposed to be sold to US but I think ebay and the like is free for US citizens.
Johnego, I believe many IC design companies (called manufacturers though they don't manufacture) have lost control with the items being sold on the world market, claiming to be an original product. The use of various foreign wafer production sites through decades makes tracking very difficult.
My understanding from the past of IC manufacturing was the ICs would be tested, at least for basic parameters, on wafer-level such that the clearly deficient items would be discarded before being put in a housing.
Would you know if this is still the case? Or, is the problem the foreign wafer houses produce batches where the testing is much more lenient than for the official products?
I believe no large scale IC production has taken place in Europe for more than a decade. We simply don't know what's going on anymore.
My understanding from the past of IC manufacturing was the ICs would be tested, at least for basic parameters, on wafer-level such that the clearly deficient items would be discarded before being put in a housing.
Would you know if this is still the case? Or, is the problem the foreign wafer houses produce batches where the testing is much more lenient than for the official products?
I believe no large scale IC production has taken place in Europe for more than a decade. We simply don't know what's going on anymore.
Last edited:
I think that's a nice wishful thinking. It seems that only random sampling are being tested. LinearTech has ops that are 100% tested. But of course they cost $10 and are not suitable for audio. AMD has made PC processors in East Germany for a long time. When the State did not want to pay anymore, the fab was closed.
Last edited:
My modest knowledge goes back to when ICs cost considerably more than today and the photographic masks were an important way to protect against copying. For the ICs discussed in those days, 100% testing of essential parameters was foreseen (wafer-level). Today, the masks are no doubt replaced by a data-file and the production robots no doubt much cheaper. So, with a copy of the data-file, setting up a production of theoretically "original" products, to the extent the production gear is good enough and maintained well, may not be costly. But, then comes the yield and the screening as variable parameters. A complex situation.
Johnego, whatever is sold in your country is likely counterfeit. If you go to ti.com and look up the options, you will not see different grades of LME49720. As was stated above, be careful where you buy them.
As far as test flows, TI generally does 100% testing after packaging and many product are also tested at the wafer level. All reputable companies will use a combination of wafer-level die testing and 'final testing' to test all parameters with minimums and/or maximums on the datasheet; any deviations from this practice will likely be indicated with footnotes. Despite advances in semiconductor technology, there remain certain levels of 'defect density' on wafers even if the fabrication process itself were perfect (which it isn't). Customers are also demanding higher outgoing quality levels so skipping test steps is not a viable option if quality goals are to be met.
As far as test flows, TI generally does 100% testing after packaging and many product are also tested at the wafer level. All reputable companies will use a combination of wafer-level die testing and 'final testing' to test all parameters with minimums and/or maximums on the datasheet; any deviations from this practice will likely be indicated with footnotes. Despite advances in semiconductor technology, there remain certain levels of 'defect density' on wafers even if the fabrication process itself were perfect (which it isn't). Customers are also demanding higher outgoing quality levels so skipping test steps is not a viable option if quality goals are to be met.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.