Related to another sample&hold thread I was wondering about the breakdown voltage of RF diodes. For example, the SMS7621 that John Larkin suggested looks really good at only 0.25pF capacitance. But diodes like this are rated 2V abs max and between the signal and the sampling pulses that really ain't much.
Or is that rating more for DC? What if you exceed the abs max and, for example, there'd be 3-4V for about 250 psec at regular sampling intervals?
How many uA is low? Says 10uA at 2V in the datasheet but that's it. During the pulse I don't care (as long as it won't go phut) but at up to
1V I need the leakage to be in the low hundreds on nA.
The 0201 version of the SMS7621 is listed at 0.18pF which would be really sweet. Of course, if you sneeze before getting it soldered on it's gone.
Thanks, that is good to know.
I was also considering the HSMS-2860 which is rated at 4V but nobody carries the 0.25pF SOT-343 version. Unfortunately it's like with RF transistors, you don't get much choice to try various flavors because they have many different footprints.
I don't know. The graph that I have (done by some intern) is a linear scale, and I can't resolve uA very well.
Hey, I'll send you some and you can measure them.
I did my old 5 GHz sampler with some HP SOT-23 dual schottky, 0.5 pF or some such. You don't need a super low-c diode to sample at a few GHz bandwidth.
50 ohms * 1 pF = 50 ps.
--
John Larkin Highland Technology, Inc
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
Since you device folks abdicated responsibility for datasheets to the marketing department, their value as engineering documents has declined drastically overall.
If you have to stick to the info in the datasheet exclusively, there are a lot of things that you can't design at all.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Jumping the impedance would create a problem for me because that causes multiple reflections back and forth, messes up the desired signals. Dropping it via resistive divider would work but then I'd easily lose
15dB of SNR. A balun could work but 4:1 or higher are rare up there.
It sure would be. But right now I can only simulate and pretty much have to nail the design that way. Done it before, except this is a much different domain where cost and stuff matters.
Precisely. The only alternative path is with companies that have excellent FAE or email support. Linear Technology, for example. Their chips are sometimes painfully expensive but I have obtained non-datasheet data that came directly from their IC designers and was very reliable. Plus they can then deliver production quantities :-)
Data from the designer doesn't mean it is data in the production test program.
Chip "behavior" is more or less fixed in the design, but nuances with parameters need to be tested. Leakage for instance is effected by the starting material, which in turn depends on the epi reactor and/or vendor. Design engineering only sees the initial product. They don't see how the chip is being built years later.
If you are just making a few devices and have the time to test the parameter in house, it is likely to be OK, with the exception of using the part in a manner than is destructive, such as breaking down junctions.
But if you are making a product that sells in the millions, you would be better off asking for a custom test flow. Sometimes when you see parts that got the "paint and re-mark" treatment, they could be from a custom test flow. Medical and military products generally.
Leakage tests are notoriously slow since they require the ATE to settle down. Remember, the DUT board has all sorts of relays on it, there is cabling to the test head SMU, etc. So routinely the leakage in the part is far lower than the datasheet. They just set a high limit so that the test time is reduced in production test. But you also could get leaky parts. I've seen it happen. You make a million of something, just about every limit will be hit for real in testing.
Note that marketing does not write the electrical specifications. The product manager, trying to be cost competitive, can mandate some parameter is not to be tested. That is one reason why you see typical values without limits on a datasheet. Often these typical limits can be tested in a custom flow. It depends how the company arrives at the typical values. Some companies just bench a few parts are deem them typical values. Others write two versions of the test program. One for characterization, and the other for production test. The characterization program will spend the time to get the low level high precision data. So the test for the parameter in question has already been written, but not present in the production test program. That means the test could be added to the flow.
If you bring money, just about anything can be tested. If you want cheap, you are on your own if the parameter isn't tested. Or you test it yourself if your quantity is small.
It is hard to find documented, but typically most chips are designed for a
10 year lifetime. So even then, the parameters of the original product may not be the same after a decade. You see this mostly in logic chips where the VT shifts with use. The worst case spice file should reflect this shift based on continuous clocking over ten years.
That's ok. Tektronix built tons of successful product that way, using parts in unorthodox ways. Like the pre-blackened neon lamps. That way they were pre-aged and would not significantly change further for decades.
Yep, like for rad hard certs. That is something I'll leave up to the client because it is a business decision.
I'd be happy with a typical graph. That wouldn't be too hard for them to provide. They could at least figure that a diode that is marketed as an RF detector will also be used in pulsed applications, UWB, and so on. Without knowing anything about leakage you can't even begin the design.
Thanks for all the info "from the trenches". Most people who'll read your post never really get into a fab or IC design place. I do, but only once a year on average. But I know many of the trade-off, some of the chips I was involved in only have to last 10h and then you can really step on it with some parameters.
There seems to be no consistency in how any of these chip companies are run regarding test and production. ISO9000 and similar nonsense did nothing to solve this. The only good thing out of TSMC and Global taking over the world is there are a lot less variables in fabrication.
In the old Tek IC center, there is/was a book of tricks. It had internal papers of design knowledge of the group. It would be nice if the old bipolar tricks were published, but I suspect the book will never see the light of day.
Regarding Tek oddball designs, you kind of assumed they would be around until you junked your instrument. Given the turmoil, I suspect Tek couldn't fix much of anything. There is/was a an ex-employee that was fixing old gear. I've long lost his name and address, but he was of some use regarding my curve tracer.
I got to go into Santa Clara 1 years ago. I was working at a now defunct company that was using Intel as a fab interface. [A really long story.] Lots of trade secrets out in the open, though you had to know what to look for. For instance, the masks had pellicles, which we didn't have.
Regarding those paint and re-mark chips, some places test the parts to meet the new grade, then paint re-mark and ship. A good flow is to test again after the new marking. If the part leave the factory, you don't trust it.
TSMC doesn't own the world. Yet. There's lots of other foundries with cutting edge processes. For example, I really like X-Fab's XH035 high voltage process.
That is sad. Although much of it will probably be outdated. Tek should have hung on to some of the old guys, should have built an assisted living place for them and their spouses. I am sure it could have avoided such embarrassments like the TDS-220 scopes.
At least in the old days their service manuals were so good that one can repair tyhe gear. I've always been able to keep the 7704A going and that must now be over 30 years old.
Some parts weren't built in a service-friendly way. For example, it is quite common to fry the mixer in spectrum analyzers when doing pulse-echo stuff. Comes with the game. Yet when that happens on a 7000 series analyzer plug-in module game's over. It's essentially bricked.
Until people become desperate. Once they find out that someone has a small stock of unobtanium they come banging on the door.
I visited Dean, a long time ago. He had a barn full of Tek gear, subassemblies, and parts. I bought a couple of custom linear ICs, for a sampling scope, from him for a couple of dollars. Cool old guy.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.