Sometimes I don't understand part pricing. Take FJH1100, the $8 signal diode.

[JoeN]

Maybe the experts here can tell me what makes this part so special it can sell for $8?

http://www.mouser.com/ProductDetail/Fairchild-Semiconductor/FJH1100/
https://www.fairchildsemi.com/datasheets/FJ/FJH1100.pdf

The manufacturer's claim to fame is a 3pA reverse leakage.  But some other parts have the same spec for far less like the NPX BAV199 which claims the same level of leakage current.  These are five cents each, dirt cheap, in quantity.

http://www.mouser.com/ProductDetail/NXP-Semiconductors/BAV199215/ 
http://www.mouser.com/ds/2/302/BAV199-840428.pdf

How does Fairchild "get away" with pricing a small signal diode like this?  What am I misreading?  Thanks if you can make me understand.

[Kleinstein]

There is a big difference between 3 pA typical and 3 pA of maximum reverse current. The testing for low currents takes quite some time and thus adds costs. Also the plastic case is usually cheaper than glass - though not that much.
Though I still don't see a reason for $8. Sometimes it is just a last batch of a type that gets expensive.

[wraper]

The manufacturer's claim to fame is a 3pA reverse leakage.  But some other parts have the same spec for far less like the NPX BAV199 which claims the same level of leakage current.  These are five cents each, dirt cheap, in quantity.

Really? One has 3 / 10pA max leakage current spec, another 5nA, although at much higher test voltage but nonetheless.

[dmills]

Yep, and sometimes that **REALLY** matters, see also the LSK389 jfet, something else you only design in if you really need the performance.

Another possibility is that they are locked into making this thing by some contract with a defence supplier (An industry where updating an old design can take a years worth of really boring pointless paperwork), in this situation you do not want to change **anything** and $8 for a diode is completely insignificant if annoying.

I for one love companies that make the obscure, even if they charge for it, because sometimes you really need a 3pA diode or a 10G Ohm 1 % resistor or something else weird, and cost is not that important.

Regards, Dan.

[JoeN]

The manufacturer's claim to fame is a 3pA reverse leakage.  But some other parts have the same spec for far less like the NPX BAV199 which claims the same level of leakage current.  These are five cents each, dirt cheap, in quantity.

Really? One has 3 / 10pA max leakage current spec, another 5nA, although at much higher test voltage but nonetheless.

Well, the Fairchild part doesn't have a graph like NXP.  That 3pA "max" number is at 25°C only, you get to guess how badly it derates above 25°C because they won't tell you.  If it is anything like the NXP part, it derates quickly.  The only thing we "know" is that the 3pA reverse leakage number is "max" for Fairchild and "typical" for NXP at 5V and at 25°C.

[wraper]

Well, the Fairchild part doesn't have a graph like NXP. 

Because FJH1100.pdf is not a full datasheet.

[JoeN]

Well, the Fairchild part doesn't have a graph like NXP. 

Because FJH1100.pdf is not a full datasheet.

Fairchild seems to disagree with you.  I cannot find a more detailed datasheet.

Product page:  https://www.fairchildsemi.com/products/discretes/diodes-rectifiers/small-signal-diodes/FJH1100.html

"Download FJH1100 Datasheet"

https://www.fairchildsemi.com/datasheets/FJ/FJH1100.pdf

[wraper]

"Download FJH1100 Datasheet"

Often such "Download datasheet" button means downloading 5 page "datasheet" for very complicated IC, which is not much more useful than nothing at all. While real datasheet is something like 500+ pages which you can get only under NDA.

[orin]

The manufacturer's claim to fame is a 3pA reverse leakage.  But some other parts have the same spec for far less like the NPX BAV199 which claims the same level of leakage current.  These are five cents each, dirt cheap, in quantity.

Really? One has 3 / 10pA max leakage current spec, another 5nA, although at much higher test voltage but nonetheless.

"At a much higher test voltage..."

They are not comparable with the given datasheets.  You can't compare 5nA maximum at 75V reverse voltage with 10pA maximum at 15V reverse voltage.

However, given the price, if you are prepared to test them yourself, the BAV199 looks pretty good.

[Alex Nikitin]

Hmm,  I am familiar with both FJH1100 and BAV199 as we use both in the equipment we make at my workplace. There are differences and sometimes you have to pay a premium for the FJH1100 and tolerate it's problems, such as a strong light sensitivity. For a start, I can recommend to look at the BAV199 datasheet from ON Semi as that one contains a graph with actual measurements of a reverse current vs voltage  .

Cheers

Alex

[wraper]

FJH1100 and tolerate it's problems, such as a strong light sensitivity.

According to the datasheet it should not be light sensitive unless body coating is damaged.

[T3sl4co1l]

"At a much higher test voltage..."

They are not comparable with the given datasheets.  You can't compare 5nA maximum at 75V reverse voltage with 10pA maximum at 15V reverse voltage.

Sure you can!
Sort of.

Diode leakage varies relatively little with reverse voltage.  Maybe a factor of 10x between 10% and 100% of Vrrm.  That's a huge difference compared to the 500x between diodes!  (Which is reasonable for, say, the difference in leakage between 25C and 85C or more.  Leakage does vary strongly with temperature!)

"Sort of", meaning: They don't give a curve of reverse current versus voltage or temperature.  So it's unclear how this diode was actually designed: is it a high voltage diode like 1N4148, selected and tested for one parameter (current at a mere 15V)?  Or does it actually break down at low voltages, using a very different die?

However, given the price, if you are prepared to test them yourself, the BAV199 looks pretty good.

JFETs have long been used as low leakage diodes (for N-ch, use G for anode, tie D+S together as cathode).  The downside that the channel resistance is a big factor: it more or less adds in series with the diode junction; and the gate current limit is always tiny (a few mA).  They're often specified as well as the FJH1100 (~pA), yet cost mere pennies.

Supposedly, a lot of BJTs have excellent leakage characteristics (B-C junction; tie B to E), particularly small junction RF transistors like BFR92.  But again -- you have to test and select the parts yourself if you need guaranteed low leakage, not just "typical" performance.

Tim

[edavid]

JFETs have long been used as low leakage diodes (for N-ch, use G for anode, tie D+S together as cathode).  The downside that the channel resistance is a big factor: it more or less adds in series with the diode junction; and the gate current limit is always tiny (a few mA).  They're often specified as well as the FJH1100 (~pA), yet cost mere pennies.

Have you priced a PN4117 lately?  $2.29 each from Mouser.

(P.S. The FJH1100 is only $5.05 at Arrow.)

[T3sl4co1l]

Okay, not ALL JFETs cost pennies...

Tim

[David Hess]

Well, the Fairchild part doesn't have a graph like NXP.  That 3pA "max" number is at 25°C only, you get to guess how badly it derates above 25°C because they won't tell you.  If it is anything like the NXP part, it derates quickly.  The only thing we "know" is that the 3pA reverse leakage number is "max" for Fairchild and "typical" for NXP at 5V and at 25°C.

We also know that the BAV199 is 5nA maximum at 25C so it is potentially 3 orders of magnitude worse than the FJH1100 maximum of 3pA at 25C.  No leakage versus temperature graph is needed because PN junction leakage roughly doubles every 10C. (1)

As Kleinstein points out, the difference is testing.  Time spent on the test fixture costs money and low leakage current tests take lots of time.

A good example of this is the National Semiconductor $0.83 LMC6081 and $5.76 LMC6001.  They are the same part but the former is 10fA typical and the later is 25fA maximum at 25C.  These articles suggest that the testing time at this level is a full minute which is consistent with the roughly 10 cents per second cost that I remember and the price difference between the LMC6081 and LMC6001:

http://electronicdesign.com/test-amp-measurement/whats-all-teflon-stuff-anyhow
http://electronicdesign.com/test-amp-measurement/whats-all-femtoampere-stuff-anyhow

Many common small signal transistors have leakage in the picoamp range but their specification will be something like 25nA, 50nA, or 100nA maximum because that is what the ATE (automatic test equipment) can quickly measure and for their intended use, that level of guaranteed leakage is all that is required.  At 10 cents per second of test time, not much time is available to test a 6 cent 2N3904.

(1) Back when I was working on low leakage integrators that had to operate at high temperatures, I ended up using bipolar super beta 308 type operational amplifiers like the LM11 (50pA maximum at 125C) and graded CMOS amplifiers like the LMC6081 (10fA typical at 25C and 4pA maximum at 85C) instead of other CMOS and JFET amplifiers because starting from say 10pA, the later devices had too much leakage at high temperatures and tests found the rough doubling of leakage every 10C rule was pretty close.  Analog Devices and Burr-Brown had JFET input electrometer amplifiers which would have worked but they were more expensive than grading LMC6081s.

[jaromir]

At 10 cents per second of test time

I'm curious - is this somehow typical rate per test second? What about more complicated silicon devices, like MCUs?
Sometimes I wonder how on earth could they sell MCU for less than 50c per piece (and much cheaper in volume), when just the functional tests have to take "very long". I assume they load FLASH with simple program to exercise all the peripherals, that has to take a while (hundreds of miliseconds) at much more convoluted testers.

[orolo]

Have you priced a PN4117 lately?  $2.29 each from Mouser.

(P.S. The FJH1100 is only $5.05 at Arrow.)

MMBF4117 is only 0.128 euro for a hundred pieces at mouser. Another caution with JFETs as diodes is the exponential temperature-leakage dependence. MMBF4117 goes from 10pA at 25ºC to 25nA at 150ºC.

[rstofer]

At 10 cents per second of test time

I'm curious - is this somehow typical rate per test second? What about more complicated silicon devices, like MCUs?
Sometimes I wonder how on earth could they sell MCU for less than 50c per piece (and much cheaper in volume), when just the functional tests have to take "very long". I assume they load FLASH with simple program to exercise all the peripherals, that has to take a while (hundreds of miliseconds) at much more convoluted testers.

All of which is why they invented JTAG!  This was a way to test complex devices like MCUs.

[David Hess]

At 10 cents per second of test time

I'm curious - is this somehow typical rate per test second? What about more complicated silicon devices, like MCUs?

This would only be for simple analog devices like diodes, transistors, and operational amplifiers.

Sometimes I wonder how on earth could they sell MCU for less than 50c per piece (and much cheaper in volume), when just the functional tests have to take "very long". I assume they load FLASH with simple program to exercise all the peripherals, that has to take a while (hundreds of miliseconds) at much more convoluted testers.

BIST (built in self test) is a major feature of complex logic devices an especially things like processors and memory.  There is no way to supply all of the test data patterns through the pins in a timely manner so internal structures to generate and verify them are included.

Part of this process includes activating redundant circuits like spare rows of memory to replace faulty ones.  I read that one of the things that led to low yields resulting in high costs on the AMD K6-3 processor was lack of enough spare L3 (update: L2) cache rows.

I wonder how they test OTP (one time programmable) devices though.

[jaromir]

All of which is why they invented JTAG!  This was a way to test complex devices like MCUs.

But still, this takes time, which is very valuable in manufacturing. And not all MCUs do have JTAG interface.

BIST (built in self test) is a major feature of complex logic devices an especially things like processors and memory.  There is no way to supply all of the test data patterns through the pins in a timely manner so internal structures to generate and verify them are included.

That makes sense, thanks. The BIST is used along with specific program in FLASH to exercise it, which is erased after tests?

[Kleinstein]

The costs for testing also depends on the type of testing done. More simple tests for large quantity devices like the 2N3906 can use more special, less capable tests. This still makes the low current tests rather expensive as they need time and high quality test equipment and those super low leakage diodes are not high quantity.

Anyway there is not that much logic behind some of the semiconductor prices - some are just expensive because they have no reason to sell them for less. Also quantity and the factories that are used can be a big factor.

[edavid]

I wonder how they test OTP (one time programmable) devices though.

"OTP" EPROM based devices can be tested and erased before they are encapsulated.

True OTP devices have spare cells for testing, but they can't be completely tested, so invariably have higher field failure rates than other devices.
Good reason to avoid them if possible.

[David Hess]

BIST (built in self test) is a major feature of complex logic devices an especially things like processors and memory.  There is no way to supply all of the test data patterns through the pins in a timely manner so internal structures to generate and verify them are included.

That makes sense, thanks. The BIST is used along with specific program in FLASH to exercise it, which is erased after tests?

I do not know the details but that would work.  Or a separate ROM or state machine is included to run the test vectors using the BIST structures.

I remember discussions about this problem in trade magazines in connection with the increased size of DRAM chips.  For a while BIST was a major new feature in cutting edge integrated logic chips but I assume it is too common now to bother advertising.

[Hypernova]

I wonder how they test OTP (one time programmable) devices though.

Many OTP are just EEPROM without a local erase circuit. I have also heard from someone where they once were developing with some OTP MCUs. They just go to specialists that decap and window the chips and erase them with UV during development.

[EEVblog]

This is what I like to see, a nerd fight over diode specs