FET amplifier information biasing voltage and circuits etc
 

Happened upon this , may be of use to some solid state builders out there.
The good old MPF102 2N3819 J201.
http://hawestv.com/amp_projects/fet_...fetpreamp1.htm

Re: FET amplifier information biasing voltage and circuits etc
 

It seems to completely fail to notice the huge tolerances on all the FET parameters.

Results will vary an awful lot on the characteristics of the actual device you have. This cn wipe out any confidence a novice constructor has been building. People tend to blame themselves.

David

Re: FET amplifier information biasing voltage and circuits etc
 

It seems to completely fail to notice the huge tolerances on all the FET parameters.

Results will vary an awful lot on the characteristics of the actual device you have. This cn wipe out any confidence a novice constructor has been building. People tend to blame themselves.



To be fair it does on the schematic page http://www.till.com/articles/GuitarPreamp/


" I should point out that FETs in general suffer from a serious lack of manufacturing consistancy. The FET VGS and IDSS, the parameters that determine the bias point, can be anywhere over a 5-to-1 range and still be within spec. That's pretty awful, but such is life. It's an engineering accomplishment to design a circuit that can function exactly the same over a wide variation of component parameters. But I can't guarantee that in this situation; there's not enough supply voltage headroom.

The upshot is that while this preamp circuit is designed to work with typical J201 FETs, it will not work with all of them and it would be a good idea to try a handful of FETs and throw out any that don't bias correctly. An easy test is to measure the voltage at the drain of Q1 and if it's between 5.0 and 7.0 volts things are fine. "

Re: FET amplifier information biasing voltage and circuits etc
 

Ah, that's good, and tells it like it is. Designing a sure-fire circuit with an FET in it that will work properly with any in-spec FET in it is one heck of an achievement.

Sometimes you see mention that the FET may require selection. The danger here is if you buy a number of FETs at the same time, they likely all came out of the same batch!

For a preamp with a guitar pick-up feeding it, the source impedance will be a better match for a typical bipolar transistor than for a typical FET, so the FET is going to be noisier than can be achieved with plain old bipolars. The same goes for magnetic record-playing cartridges.

It's often said that FETs 'have a valve-like characteristic' and this gives people a warm and fuzzy feeling. So they get used in places sometimes to make a marketing point, sometimes because constructors think they feel less culture shock. If the former, then if overload characteristics are important, as in overdriven guitar situations, then the benefit gets shot down by the variability and not all units will have the 'good' sound. So purchasers and users have a hard time finding a good unit, and manufacturers don't get much fun either. Reputations suffer, and the pricing of the good units sold has to carry the costs of making and rejecting the bad units somewhere along the line.

Working for a manufacturer there was a certain horror felt when a design needed an SOT (Select-On-Test) part, and you had lots in stores, all from a batch you discover to be unsuitable.

I came across a case where hundred-quid crystal oscillator units were averaging only one out of ten parts good. some months. PCB pads were getting worn out by the production line folk frantically changing oscillators until a noise problem stopped. A spec change to the oscillator fixed the issue. It pushed the problem back to the quartz manufacturer who changed the design and things were dramatically improved.

So, it's easy to fall into the "They're like valves" mindset, but the variability of FETs is nothing like the dependability of parameters you get with valves. So there is a much bigger learning curve than many people think.

David

Re: FET amplifier information biasing voltage and circuits etc
 

It gets much worse if you buy from dodgy Chinese suppliers, or in my case a 3rd party English component supplier supplying those over here- I bought some "2N3819" type devices that in reality weren't even FET's at all, they were bipolar NPN jobbies and showed up as such on a tester and curve tracer.
After that, I am extremely cautious what I buy and from whom. :-X

Re: FET amplifier information biasing voltage and circuits etc
 

Back around 1970 when I was running a university teaching lab, we encouraged students to actually build circuits and measure their performance. When FETs were a bit novel, I introduced a single-stage FET amplifier using a 2N3819. The prototypes had worked fine, but I have a painful memory that out of a class of 40 only around 20 worked as expected: the rest all needed the source bias resistor to be increased from 4k7 to 8k2, requiring some rapid on-the-spot creativity to maintain the all-important confidence of the class. It was interesting that, rather that being randomly distributed, the device characteristics fell into two definite groups of bias requirement.

Martin

Re: FET amplifier information biasing voltage and circuits etc
 

So if these FET,s have / had variations (the old styles from years ago) then what of the surface mount types of today then, has any improvement been made ?.
This leads me to think that it may be possible to salvage SMD FET's from todays old thrown away SMD equipment , however there may be a problem with ident whats what !!!.

Re: FET amplifier information biasing voltage and circuits etc
 

The 2N3819, MPF102, BF245, BF256 and 2N4416(A) are all process 50 JFETs. One classic way to get around the spread in parameters for process 50 JFETs is to buy either the 2N5484, 2N5485 or 2N5486 as these are all process 50 but they are graded across the range of process 50. The 2N5484 has the lowest Idss and 2N5486 has the highest.

It's difficult to find the 2N5484, 2N5485 or 2N5486 today though. They are sometimes available in SMD. The BF256 is/was available in A, B and C grades.

I think the BF256B version can still be purchased in TO-92 as there was a recent manufacturing re-run of these parts by Onsemi. Farnell have them for 38p each and have 17,000 in stock with more to follow. The Idss range for the BF256B is typically 6mA to 13mA. I think process 50 JFETs can have an Idss range from 1 to 20mA.

Re: FET amplifier information biasing voltage and circuits etc
 

Jeremy

I've seen mention of process 50 and other numbers with regards to FETs on this forum. Please could you give a brief explanation of the meaning of these process numbers.

John

Re: FET amplifier information biasing voltage and circuits etc
 

Hi John

Sadly, I'm not really an expert on this stuff, but if you go back 40 or 50 years then there were quite a few JFET manufacturing processes that were established across the industry. The same applies to BJT transistors. I think this allows a manufacturer to make a process 50 JFET from an agreed JFET process recipe book.

The two images below are from an old Nat Semi databook. The first one shows all the devices they make using process 50.

The well-known 2N3819, MPF102, BF245, BF256 and 2N4416(A) are included here and the parts with an asterisk are the preferred parts. You can see an image of the internal die.

The second image shows process 92. The common process 92 JFETs are the U309, J309, U310 and J310

You can see the die is very different between the two processes. In the past I've compared the die for process 50 across several manufacturers and the die pattern can differ slightly in terms of the shape and the number of interlocking fingers. So, I'm not sure what the rules are for each process.

I'm not sure why there are so many part numbers within process 50, but some of it can be explained in terms of device packaging and pinout differences. Inside the process should be the same.

I guess in the case of the 2N5484, 2N5485 and 2N5486 the parts get assigned a part number depending on where they fall within the accepted range of process 50. One batch might all end up as 2N5484 for example. However, I'm really just guessing. Maybe someone else can describe it all better.

Re: FET amplifier information biasing voltage and circuits etc
 

Process 51 is another popular JFET process, and this process is used to make JFETs that are suited for use as switches or choppers. You can see there are some familiar part numbers in the image below.

The J111, J112 and J113 span the parameter range of process 51. These parts are still available today.

The die shape is very different for process 51 and I guess this is done like this to make the JFET more suited to switching applications.

Re: FET amplifier information biasing voltage and circuits etc
 

Must admit, I have always had a weak spot for the U309/310 JFETs, they were my go-to devices for grounded-gate HF/VHF amplifiers in the past.

It's not always appreciated that JFETs are kinda symmetrical; the source and drain can be interchanged. Indeed, experimenting with such interchanging while using a pulsed noise-source[1] can sometimes get you a significant improvement in the balance between noise-figure and gain in a particular circuit.

Best gain and best noise-figure rarely, if ever, coincide.


[1] usually a diode valve run with the heater at a low voltage so emission is 'thermally limited'.

Re: FET amplifier information biasing voltage and circuits etc
 

The two images below are from an old Nat Semi databook. The first one shows all the devices they make using process 50.

Thanks Jeremy, that info. is most interesting, which data book is it in?
I couldn't find it in my small collection of those Nat Seminconductor books I have.

The uploaded photo's are not easy to read.

Cheers :beer:

Re: FET amplifier information biasing voltage and circuits etc
 

It's from the 1977 National Semiconductor FET Databook.

It does still seem to be available as a pdf online, but the pdf quality may vary.

See below for process 52. The J201 was the original device used in the guitar project and you can see it listed on the screenshot below as a process 52 device.

The FET databook has quite a few app notes at the end. There's some good stuff about noise in chapter 6 including app notes about cartridges and noise. Also, some noise comparisons for various process types as in the image below.

Re: FET amplifier information biasing voltage and circuits etc
 

Thanks for that!
I have a .pdf now and must say it is a damn good book (as those National books always are- the audio one is superb).
I would love to own a hardcopy, but can't find one yet.
Appreciate your help.

Re: FET amplifier information biasing voltage and circuits etc
 

Too much information can be just as bad as too little LOL, i have never heard of these FET processes before so looks like it warrants some reading, is this something to do with the etch masks in the manufacturing?.
https://www.researchgate.net/figure/...fig1_340701863

But do the SMD FET,s of today still have such wide variations ?.

Re: FET amplifier information biasing voltage and circuits etc
 

You'll find the 2N4391, 2N4392, 2N4391 group in that listing. They are a graded set of high Idss devices. 2N4391 ha the highest Idss grade (and lowest RDSon) then the higher JEDEC numbers are reducing Idss (Increasing RDSon)

HP used the 2N4391 as i855-0020, its preferred analogue switch FET.

Marconi used, I think it was 2N4392 or it may have been 2N4391 as a high bias current HF amplifier in some of their high performance HF receivers. I first came across it in the H2900 'Sosin's folly' star of the famously withdrawn receiver comparison article sosin wrote.

The FET channel is long and is folded in a serpentine under the gate fingers. the fine-ness of the fingers shows how short the channel is, and the total length of all the gate fingers shows you how wide the channel is. Process 51 are wide, short channel FETs, folded to fit on an economical area of silicon.

There used to be a sort of industry rule of thumb of so many dollars per square millimetre, plus packaging. Can't remember the number, but it held for quite some time and even worked for specialist devices like digital camera sensors!

David

Re: FET amplifier information biasing voltage and circuits etc
 

This is also something I have recently observed, or rather observed large differences between batches of the same component type. So much so that a fresh supply of the same type from the same brand were so different as to render an analog switch circuit inoperable.

It's probable that devices from the same wafer have reasonably close characteristics as would have been suggested by the investigation and subsequent testing by the young friend who was tripped up by this feature, there was definitely two very different discreet levels of bias required to switch the device, such that the circuit needed to be redesigned.

Greg.

Re: FET amplifier information biasing voltage and circuits etc
 

This is also something I have recently observed, or rather observed large differences between batches of the same component type. So much so that a fresh supply of the same type from the same brand were so different as to render an analog switch circuit inoperable.

I found similar, the only easy answer was to match the usable FET's on a curve tracer. As long as they are not fakes, like I mentioned in a post above.

Having worked for Siliconix for knocking on for a decade it would appear that "process xx" corresponds to "geometry" in the Siliconix FET catalogues of yore.

Back then FETs were tested in wafer form at "die sort" where the entire wafer (an enormous 3" diameter) was tested for function and various parameters.

I looked up the 2n3819 and it appears that it was sourced from two geometries: NH and NRL which would explain why different batches might have different spreads of IDSS etc.

Siliconix do not make small signal FETs anymore following the closure of the Santa Clara fab.

As to the question of the surface mount part differing from the through hole part, they would use the same die sorted on the same parameters so no improvent there then.

Further details of "process xx" are available in an ONSEMI/On Semi/Fairchild application note AN-6609 from the dear dead days of 1977.

https://www.onsemi.com/pub/collateral/an-6609.pdf

Then again if you were prepared to pay for it, Siliconix would create a special device spec with tighter test conditions.

It saddens me to see various devices carrying such numbers asked about on various boards when there's no means of figuring out what the baseline device was now the factory has been closed for more than 30 years.

So if you see a device with "WNxxx" and the Siliconix logo, it came from Swansea.