where is pin 1?

I was using a 20 pin, break-off right angle header once, the kind with 0.1" spaced, 0.025" square posts. Other than the right angle bend, there would be no way to tell which pin was 1. It didn't even occur to me to check the data sheet to see where they expected pin 1 to be.

So, of course, when they started making the boards, they contacted me to tell me the documentation I provided was wrong, having the connector being bent over the board, rather than over the edge. I tried to explain to them that pin 1 is which ever pin I define it as, but they insisted they had to go by the data sheet. I forget what I had to do, but I finally got them to just install the connectors the *obvious* right way and get on with it. I may have had to write an ECN on the connector data sheet or something similar.

On 2023-08-23 22:43, John Larkin wrote:>

On Monday, we received two batches of test chips in QFN-64 packages. The pins are numbered clockwise from what is normally Pin 64. :(

It wasn't an accident--apparently it's an internal convention of the chip design house. It would make perfect sense if they were selling vacuum tubes.

Cheers

Phil Hobbs

At least tubes had keys. You can load an IC wrong two or four ways. Why are so many things symmetric?

I read the data sheet as showing a rectangular indent in the centre of the bottom. Looking at the bottom, if that's in landscape mode and there are pins at roughly 11 and 1 o'clock, then pin 1 is at roughly 8 o'clock.

I must be reading the drawing for the "Horizontal 3 Line Choke" on page

4 wrong then. The upper-left picture shows what appears to be an inverted triangle of registration pips in the plastic (the apex is at the "6 o'clock" position, in between pins 5 and 6). Pin 1 is the next pin clockwise (roughly between 8 o'clock and 9 o'clock).

Likewise, the 4-pin one has the apex of those registration pips between pins 1 and 2.

The vertical ones look entirely symmetrical though, so figuring out which corner (from the bottom) is the correct "upper left" is probably more difficult...

Well, it is wrong. It's impossible to wind as shown.

People handling real parts, the 3-phase version, can see no way to find pin 1. It's entirely 3-way rotational symmetric.

That is not correct. There appears to be a rectangular slot that creates a two way symmetric reference line. Combine that with the 3 way symmetry of the coils and separators, and there is only one orientation that will put all axes in the correct alignment. Also, the leads on the part only allow three rotations, of which only one orients the rectangular cut out correctly.

Where's the problem?

Yeah, the stylized wire wrap is wrong (one end has to come from between the core and the base plate). But I wasn't talking about those views, but rather the full bottom view in the upper left (with the dimensions / pin arc spacing / etc).

So the features shown on the "bottom view" don't actually exist?

Actually, it doesn't. these. These are just common modes chokes on a single core with two or three coils of equal inductance, and they are all wound in the same sense. It doesn't make any difference which way around you mount them, as anybody who understood inductors would know. Explaining this to people who don't understand inductors gets tedious, which is presumably what John Larkin is bitching about.

The only non-3-way-symmetric feature on the ottom is the part number sticker. It's not useful for production people to find pin 1.

Oh, the rectangle in the drawing is the sticker then? I misinterpreted it as a hole for that central pillar to be snapped into ... :/

They're supposed to be idiot-proof, but with EMC, physical placement of starts and finishes on the physical body can have unexpected benefits/pitfalls.

There are only a rare few EMC choke designs that are suitable for automated fab, so you're often at the mercy of visual inspection of manual windings, for 'internal' lead dressing or winding direction, though phasing may be accurate and measurable leakage terms by the book.

RL

The CM choke works at any rotation, as I noted and Sloman ignored. The only problem is not getting pestered by production and QC, namely "where is pin 1?" We'll have to note in the production documentation IT DOESN'T MATTER. We deliberately didn't include a pin1 dot on the PCB silk.

Sloman is never pestered by production or QC. His only mission is life is to tell everyone else how stupid they are.

The leakage inductance is tiny on that choke, like 15uH on one winding with the other two shorted. CM chokes tend to use very hi-mu core material, which means they saturate at very low levels of non-balanced current. One can be fooled by distributor searches into thinking these are really good multi-winding conventional inductors.

I guess they might be good transformers. Hey, maybe I can use a cool Coilcraft planar transformer instead of that monster.

No joy. The Coilcrafts are microhenries and I want millihenries.

The issue was a 6-amp-RMS several-KHz ground loop between two three-phase instruments. The dreadful CM choke mostly fixes that.

If that is a label, the drawing is dorked. If you zoom in on the orthogonal view of the bottom, the "label" would seem to be a depression. Maybe they put the label in a slot?

Scarcely. That was what my post was about.

I did say that this was what you were bitching about.

So why didn't you just do that rather than bitching about it here.

Congratulations.

Not recently. It used to be a big part of my life. They didn't exactly pester me - they sought help and I could be relied to deliver it, not always quite ad fast as they would have liked because I frequently had other things to do, but pretty quickly.

Well, I see it as more as pointing out when people have got stuff wrong. You don't have to be stupid to do that but you do feel stupid when it is pointed out, though usually relieved. Nobody gets everything right, though some do pretty well.

That isn't "the leakage inductance" though it can get pretty close. You are relying on the currents induced in the other two winding to cancel the mutual inductance, and their winding resistance limits those currents. You can characterise the parts more accurately if you try harder.

are really good multi-winding conventional inductors.

If it has got enough inductance. Big high-permeability cores deliver a lot of that.

yeah... I can't think of a cure for that.

Do you get the same grief for non-polar capacitors?

They have a better datasheet on their website.

page 7)

The 18uH leakage inductance is free and works om DM currents. It's also the 'oick-up antenna' for nuisance local EMI re-radiation in just the exact wrong place - in the filter hardware itself - often justifying local physical screening in hardware $ $ $ .

This unneccessary and (in this case)totally unwarranted) sniping at Sloman is childish.

RL

I had three high-current ferrite beads in the design to kill fast EMI, but took them out after I measured the leakage of the 3-leg CM choke.

A toroid isn't supposed to leak h-field, but the leakage inductance is from field that does NOT confine itself to the core material. The three clusters of windings will have very different and very nasty currents; each is a sort of dipole electromagnet. The simulated waveforms are ghastly, and I'm plugging in boards on 1.6" centers.

CM chokes are interesting, in good and bad ways.

A shield is a possibility. I should at least include some holes for spacers. At my frequencies, a modest aluminum sheet is a good e and h shield. Even a sheet of 1 oz copper is pretty good, so the shield could be a PC board.

He managed to agree with me and call me stupid at the same time. I usually ignore him but you've got to admire creativity at that level.