National Semiconductor LM series & Ti LM series differences that affect circuit funct

[tomchr]

Binning can be automated.

Binning IS automated.

ICs are measured at wafer sort (i.e. before packaging) and again after packaging. Binning happens after wafer sort. This adds an additional step and now requires two separate product paths from that point. The "A" grade product will have to undergo final test under different criteria than the "B" grade product. Binning also generates multiple levels of inventory, which the manufacturer and distribution chain now has to manage. In high performance, high price ICs every penny counts. In general purpose, low price ICs, every fraction of a penny counts. If binning increases the total packaged cost of an IC by more than 1-2 cents per part, it is more cost effective to just make the high-grade part. That's why binning isn't as common anymore.

Some customers don't like multiple part numbers for the same chip either. This comes up with PLL/VCO chips. Chip A may cover, say 2000-2400 MHz. Chip B covers 2400-2800 MHz. If the customer manufactures gear that operates at, say, 2200 MHz in the US market and 2600 MHz in the EU market, they will rather pay more for one chip that covers the range of 2000-2800 MHz than buy the two separate chips. It's less expensive in the long run as the customer doesn't have to qualify two different ICs.

~Tom

 

[theaudiopath]

In answer to Mark Whitney the resonant frequency of a single FE 85 is 125 Hz
My line array coaxes a lower Fs with usable response down to about 85 Hz. It can best be described as a broad bandwidth hybrid reflex design. Still experimenting to see if the trade off with usable lower Fs is worth the increased distortion below 100 Hz
lf someone knows as an absolute fact that the practice of binning is carried out with the Ti LM series chips could they PLEASE present evidence to end this endless speculation.
Cheers Mark.

 

[Mark Whitney]

I was just trying to show you that there is another solution and its impact on your design. If you were more forthcoming with your design at the start it would be easier.

I did some sims with a LM3886 model at 85 Hz and is still see almost no phase shift. Maybe you can find another way of adjusting the offset instead of just blocking it.

I thought you had given up on your binning quest. Binning is the process of sorting and marking a production run into pre determined categories. These categories are documented in the datasheet. Without traceability it is pointless. There is some practical info on binning and cherry picking on the web. My dream chip has the spike circuitry removed, bonded to a large copper plate and has silver pins, isn't she beautiful.

From the clues you have given us so far are you building a Line array with some kind of horn construction? Will each driver have its own P2P chip-amp? Sounds like a nice project.

 

[KSTR]

lf someone knows as an absolute fact that the practice of binning is carried out with the Ti LM series chips could they PLEASE present evidence to end this endless speculation.

Which LM chips exactly? There are thousands as LM was the standard prefix for ANY chip National has designed in the "Linear Monolithic" regime in the last 40 years or so.

TI/NS does NOT bin LM1875 because there is only one grade available, did that never occur to you? Simple as that.

Further, if you're bothered about offset, provide a trim for it and all is fine.

Other than that, please believe what tomchr has to say, he KNOWS this stuff because it's his day job.

 

[tvrgeek]

Binning IS automated.

ICs are measured at wafer sort (i.e. before packaging) and again after packaging. Binning happens after wafer sort. This adds an additional step and now requires two separate product paths from that point. The "A" grade product will have to undergo final test under different criteria than the "B" grade product. Binning also generates multiple levels of inventory, which the manufacturer and distribution chain now has to manage. In high performance, high price ICs every penny counts. In general purpose, low price ICs, every fraction of a penny counts. If binning increases the total packaged cost of an IC by more than 1-2 cents per part, it is more cost effective to just make the high-grade part. That's why binning isn't as common anymore.

Some customers don't like multiple part numbers for the same chip either. This comes up with PLL/VCO chips. Chip A may cover, say 2000-2400 MHz. Chip B covers 2400-2800 MHz. If the customer manufactures gear that operates at, say, 2200 MHz in the US market and 2600 MHz in the EU market, they will rather pay more for one chip that covers the range of 2000-2800 MHz than buy the two separate chips. It's less expensive in the long run as the customer doesn't have to qualify two different ICs.

~Tom

I remember way back when I dealt with buying chips. The difference between a Z80 A and B was binning. Same die. Of course, in a processor that sells for dollars, it is easier to recover the cost over parts that sell for cents. Aren't we glad we don't have to keep separate MIL-SPEC or JAN numbers any more?

When I see simple transistors in A,B,C spec, usually sorted by gain, I assume this is binning. For low volume, it is easy to see the cost of multiple part tracking is not worth it, but for the users of millions of parts, if they are half a cent cheaper, it is.

The binning I worry about is the path from the trash bin to e-bay.

 

[ricortes]

I'd give an 'attaboy' to look at the testing for a batch of chips, any takers?

I used to write specifications for medical devices. Mostly you write specifications such that you have near zero manufacturing drop out. The sweat goes into making the product consistent enough that the consequences within those specs are inconsequential.

Here's the point: There will be a range of devices, even for good parts. Just a rule of thumb estimatation, three standard deviations mean approximately a 1 in a million device.

So if you have the testing data and it shows a 1% SD, the results are probably inconsequential for cherry picking. Not worth the time or effort to find that one in a million chip that performs ~1.5% better then average.

If on the other hand you get a SD of say 20%, you could conceivably get a range of 60% high to low for things like voltage and current capabilities.

w/o the testing data, things will remain anecdotal. I'm reasonably convinced that there is at least some deviation in the manufacturing process. Question is, how much?

 

[jackinnj]

I remember way back when I dealt with buying chips. The difference between a Z80 A and B was binning.

I still have a couple of those, plus some 6502's and 8255's

 

[tomchr]

I remember way back when I dealt with buying chips.

Keyword: "way back". Like I said, binning of ICs is largely a thing of the past. I outlined the reasons a few posts back.

When I see simple transistors in A,B,C spec, usually sorted by gain, I assume this is binning.

If this is backed up by a section in the data sheet explaining what the A,B, and C specs are and the specs are different, then the part is probably binned.

~Tom

 

[theaudiopath]

Hi Again
In answer to mark's question I intend to use 1 amp per 4 speaker line array. The cabinet is unremarkable
400x300x130mm series parallel connected with single neodym magnet added to the back of each driver.The tuned reflex system is remarkable. The design belongs to Adelaide Speakers but has been further modified by me. As Adelaide Speakers is a commercial concern I feel more than little akward sharing it with any third party. Part transmission line. Part BLH, part slot loaded reflex it offers bass extension and excursion control over broader frequency range than anything I have previously known.
Cheers Mark

 

[theaudiopath]

The motivation for the speakers and amp is maximum fidelity at minimum cost

 

[theaudiopath]

Hi Again
In answer to mark's question I intend to use 1 amp per 4 speaker line array. The cabinet is unremarkable
400x300x130mm series parallel connected with single neodym magnet added to the back of each driver.The tuned reflex system is remarkable. The design belongs to Adelaide Speakers but has been further modified by me. As Adelaide Speakers is a commercial concern I feel more than little akward sharing it with any third party. Part transmission line. Part BLH, part slot loaded reflex it offers bass extension and excursion control over broader frequency range than anything I have previously known.
Cheers Mark