Let us hope that he gets consistent useful results.
Then if needed we can order the small quantities of unobtanium we all need sometimes.
Even the Wright brothers had issues. So do not make fun of him, just tell him if he is doing an obviously wrong thing.
Then if needed we can order the small quantities of unobtanium we all need sometimes.
Even the Wright brothers had issues. So do not make fun of him, just tell him if he is doing an obviously wrong thing.
Mind blowing !
You have one more youtube subscriber !
Regards,
Tibi
PS: I really hope that some day we will be able to 3D print any transistor or ic.
Last edited by a moderator:
Erm, in short, no. For one I'm not interested in poisoning myself with silane or HF or phosphine any of the extremely toxic materials required for semiconductor processing(!) And home-made devices are going to be crude-as-you-like (don't expect super-beta transistors or low-noise!!!)
Its worth doing some parts research - lots of semiconductors you think are vanishing are actually available as SMT parts (with different part numbers, like the BC850/860 replacing the BC550/560).
[ BTW: I think you mean PNP or p-channel, not P-type? No semiconductor device can work without both n-type and p-type regions! ]
Lots of the parts you might think you need for an input section are not used much any more because better versions are present as the front-ends of opamps - people don't bother complex designs for front-ends when they can pop in a jelly-bean opamp and get reliable, cheaper and better performance.
Often the lack of availability of a particular sort of part is an indication there's a better circuit topology available using other parts (like class D replacing class B, or CAZ amps replacing low 1/f-noise JFET front-ends for precision intstrumentation).
I don’t care if everything is still available in SMD. You can’t physically separate the heat producing components from the board like you can with TO 126’s on aluminum heat sinks and 2 watt MOX resistors on standoffs. A cute little unreliable one inch fan blowing on the board directly is no substitute - and I see that more and more these days. The overall cooling scheme used in the Behringer iNukes is why they fail after a year or two of road use. Now they want to rely on air over board for everything and we wonder why electronics is going to hell in a hand basket.
And reliability is the main reason any DIY semiconductor isn’t going to be much good. On top of that, packaging is just as much of a bitch (with its own reliability concerns) and wire bonding is even less of an option for most of us. I’m not going to spend money on a surplus probably pretty ragged out bonder any more than buying $20k in new SMD inventory. I’m sure it’s great to experiment with things like that at University - they have resources for everything including disposal. I have a hard enough time getting rid of copper-contaminated ferric chloride.
And reliability is the main reason any DIY semiconductor isn’t going to be much good. On top of that, packaging is just as much of a bitch (with its own reliability concerns) and wire bonding is even less of an option for most of us. I’m not going to spend money on a surplus probably pretty ragged out bonder any more than buying $20k in new SMD inventory. I’m sure it’s great to experiment with things like that at University - they have resources for everything including disposal. I have a hard enough time getting rid of copper-contaminated ferric chloride.
You can think like that, or you can think like this:
Time to be a mammal, not a dinosaur - the world of electronics won't bend to your will, adapt! SMT is here, has been here for 30 years. As for thermal engineering there's multilayer boards, thermal vias, IMS, loads of ways to improve heat management, and SMT
pcbs can even be bonded on the underside with a heatsink/spreader.
As for hell in a handbasket we've never had computational power so cheap or power-miserly, never had such performant AD/DA converters, new techs like wide-bandgap materials are on the rise, even battery performance is better - embrace it all.
Time to be a mammal, not a dinosaur - the world of electronics won't bend to your will, adapt! SMT is here, has been here for 30 years. As for thermal engineering there's multilayer boards, thermal vias, IMS, loads of ways to improve heat management, and SMT
pcbs can even be bonded on the underside with a heatsink/spreader.
As for hell in a handbasket we've never had computational power so cheap or power-miserly, never had such performant AD/DA converters, new techs like wide-bandgap materials are on the rise, even battery performance is better - embrace it all.
If I had $100,000 coming in every year for the rest of my life maybe I would. But by the time I get enough in savings to do so iI’ll be drooling over the workbench and working from a wheelchair. Working with tweezers under a microscope just won’t be an option. I’m cutting it off much sooner than that, and I’ll be working mostly from inventory using parts I can see and can pick up with my fingers. Doing $500 board runs every month just so I can get advanced features won’t be an option either. Or buying a new computer and operating system. every year to keep it virus-safe. Then there is four racks full of mostly old-school PA amps that I’m not going to spend 50 grand to replace, and want to keep working indefinitely, or until I can’t move speaker cabs around anymore. I’ve already done my lifetime buys of the necessary transistors. Some of us “dinosaurs” are on the way out and we recognize that. And plan accordingly.
^^^^^^ That .... sort of 🙂
That many components are being phased out (do you hear, through hole? 🙂 ) does not mean they dissappear instantly, a few in hobby quantities will always be available, one place or another.
The horror stories we read here daily apply to very specific parts, some mythical ultra low noise Japanese transistor not made since the 80's, some very specific power transistor, lateral MosFet, etc; but in many cases most are replaceable by reasonably close equivalents (sadly without the Mojo number), at least close enough to be functional, and in many cases indistinguishable by ear, unless one "listens with his eyes".
In my particular case, I am into Guitar amps, 99% of what I make needs some TO92 general purpose transistor (think BC547/557) or higher voltage equivalent, think 2N5401/5551 , some TO220 mid power transistor, think TIP31/32C , some TO247 power bipolar or vertical Mosfet 😱 , some general purpose Op Amp, think TL072, all of which will be available in TH packages, even if by some second-third-fourth tier supplier.
At least for the foreseeable future.
That many components are being phased out (do you hear, through hole? 🙂 ) does not mean they dissappear instantly, a few in hobby quantities will always be available, one place or another.
The horror stories we read here daily apply to very specific parts, some mythical ultra low noise Japanese transistor not made since the 80's, some very specific power transistor, lateral MosFet, etc; but in many cases most are replaceable by reasonably close equivalents (sadly without the Mojo number), at least close enough to be functional, and in many cases indistinguishable by ear, unless one "listens with his eyes".
In my particular case, I am into Guitar amps, 99% of what I make needs some TO92 general purpose transistor (think BC547/557) or higher voltage equivalent, think 2N5401/5551 , some TO220 mid power transistor, think TIP31/32C , some TO247 power bipolar or vertical Mosfet 😱 , some general purpose Op Amp, think TL072, all of which will be available in TH packages, even if by some second-third-fourth tier supplier.
At least for the foreseeable future.
Yeah, yeah, jellybeans in hobby quantities will be available indefinitely. But there are times a TIP31 just won’t cut the mustard and you NEED a 2SC4973 (to use in place of maybe 500 similar types, all NLA). Those and many others have long since been secured back when I saw the writing on the wall. And the MJL2119x, back when they were under two bucks apiece (even longer ago). I’d hate to do it again at today’s prices. Or tomorrow’s. I tend to use a little more than “hobby” quantities and the budget is fixing to go down from 12 grand a year down to 12 hundred. The advance build-up was necessary and I chose to stay with the tech that I could still work with in 20 years.
The BC550/560 pair is one of the shittiest BJTs ever made. No matter how high the beta is, these are crap. Even OnSemi / Fairchild ones. Are anyone who still use them for audio?
Mark, for example, the others thought of the Toshiba 2SC2240/SA970 pair. These are really missed by many. 2SK246/2SJ103, 2SK30A these was real HQ devices, not the ordinary BC550.
Last edited:
I mean the new age of audio is to build/design stuff that doesn't rely on specific components or suffer the fact that you might have to redesign the circuit. Make your equipment/designs resilient to component differences and you won't have any issues.
As far as the BC550/560 being crap, well, they're fine sand, the 850/860 series are better, and the BCM847/857 matched pairs are even better. Put them in enough NFB and you'll never know they're there. But if I can pull a zero-NFB transimpedance amplifier with the BCM847/857 with less than 0.01% THD, you should be able to as well.
Last edited:
Sam, I am a retired research engineer who worked at Motorola for 41 years. I have been following your transistor and IC making exploits ever since I found some of your early stuff in a random Youtube search. I say that you are a true leader in garage physics, far above the dreams of most.
I returned to school in the late 90's to get my MSEE degree, and took several classes in IC design. We used Magic on UNIX for layout and the school got the chips fabbed through MOSIS under a government (NSF) research grant program. I'm not sure if that program still exists, but it does offer some shortcuts.
Our chips were made on an old 2 micron CMOS process in 1999. All had to be laid out on identical sized die using a common 40 pin pad ring, and each student got 5 packages 40 pin ceramic packaged parts ready for testing.
My chip used about 1/4 of the die and contained 5 different CMOS opamps, all sharing the same schematic, but having different sized transistors, often just changing the size relationship between the N-channel and P-channel transistors. All 5 opamps did "function" and two worked pretty well having a high frequency roll off in the hundreds of KHz and maybe 60 dB of open loop gain.
Of the 20 some students in the class, few (3 or 4) had any function in their chips, despite promising simulations from the data extracted by the layout software.
Note to the naysayers:
Someone who worked in a high tech production facility, or even a well stocked research lab at a college can "find" or get "samples" of all sorts of chemicals and material not easily found on Amazon or Ebay. Did I make rocket powered paper airplanes with LN2, and dissolve a cockroach in HF, yes after freezing it in it's tracks. The HF ate a hole in the concrete floor too before we could neutralize it. A friend made a gold plated mouse skeleton too.
Last edited:
Even if I could *get* the chemicals, it might be a stretch to set up even a 2um processing facility in my shop. Sort of like trying to construct a mnemonic memory circuit using stone knives and bear skins. I’d need a bit more controlled an environment than a concrete floor and window AC unit. I suppose if people can sound tens of thousands on room treatments for their listening rooms, you *could* set up a clean room in your house if you wanted to badly enough. Just not where I’m going to put my money - I’m trying to get *away* from semiconductor processing (had my fill).
Some of my “friends” back in high school could get all sorts of chemicals you supposedly needed a license to acquire. I’m not sure which acid it was that ate a hole to China in the parking lot, but I bet the hole is still there. About the size of a quarter, but straight to China. Never could find the bottom with a 10 foot piece of rebar. It all sort of ended when one of us (not me) tried making plastic explosive. One step required curing in a freezer. Got unstable and that was the end of the freezer. And the end of the experiments.
Some of my “friends” back in high school could get all sorts of chemicals you supposedly needed a license to acquire. I’m not sure which acid it was that ate a hole to China in the parking lot, but I bet the hole is still there. About the size of a quarter, but straight to China. Never could find the bottom with a 10 foot piece of rebar. It all sort of ended when one of us (not me) tried making plastic explosive. One step required curing in a freezer. Got unstable and that was the end of the freezer. And the end of the experiments.
My DIY rocket experiments ended when my brother attempted to get more match heads into the cylinder by compressing it in a vice. Nearly blew himself up. By then it was possible to get Estes rocket engines at the hobby shop.....and attach our own warheads.
The hole in the sidewalk came from a DIY thermite reaction, the only one out of a dozen or so attempts that actually started, the others just burned. No China, it was only an inch or two deep.
My DIY vacuum tube in a Mason jar experiments produced many different glow lamps, because I could never get, or keep a good enough vacuum with an AC system evacuation pump or automotive engine vacuum. Diffusion pumps, the needed hivac plumbing, and the roughing pump were far beyond my budget or motivation level. A few did manage some ugly distorted sound, but no real gain. A few years later the quartz crystal manufacturing operation at the plant where I worked was shut down, and all of the equipment was scrapped or sold. There was some serious hivac stuff in that lot. They had a real clean room too.
What's a "clean room?" We had a clean room where thin film microelectronics were assembled, including wirebonding chip on ceramic modules, but all wafer fab was done in Arizona or Texas. It was not uncommon to see an operator eating while running a wirebonder. I wonder what the Special Sauce from a burger does to long term product reliability. The thick film micro assembly room was even worse. You did need the white coat and hat to get into the thin film substrate manufacturing operations though.
As the thin film manufacturing operations were being moved to the other side of the planet, the research guys were experimenting with thin film transistor matrixes, but the results were not so great. This was in the late 80's or early 90's. I'm guessing that pretty much all of the equipment we used at that time can be found on the surplus market today if you dug deep enough.
People that fit see-through windows on their hard drives have a simple way to make a temporary clean-room. They do it in the bathroom after running the shower for a while, flushing much dust from the air (and bathrooms have less dust to start with).
Hard drives dust particle size is much bigger than the traces on ICs
And showers have moisture in the air.
And showers have moisture in the air.
Guys, you are absolutely amazing. Do any of you have the foggiest idea of the cleanliness (physical and chemical) and precision (not only for photolithography) required to obtain a stable and reproducible NMOS transistor, as claimed by our DIYer?
I find the image of a fan blowing water vapour from a boiling jar in a tube, for wet oxidation, apocalyptic 😀. In particular thinking of a modest university lab that still uses mass flow controllers to dose the oxygen and water vapour, and zoned temperature control for the oxidation tube.
Really guys, get real 😀.
I find the image of a fan blowing water vapour from a boiling jar in a tube, for wet oxidation, apocalyptic 😀. In particular thinking of a modest university lab that still uses mass flow controllers to dose the oxygen and water vapour, and zoned temperature control for the oxidation tube.
Really guys, get real 😀.
I wonder if they should start a forum for the blowin' on the wind theories...
If you can do this - make chips and transistors - on a small scale, this will do to the chip industry what the laser printer did to the typesetting business.
I have people saying additive manufacturing is the future of injection molding, I tell them that 144 cavity PET preforms are cycling 10 seconds or so, that means 864 preforms a minute, how many 3D printers needed to achieve that production?
That generally ends the discussion.
I had somebody saying he worked on ship main engines, I asked how he fitted the rings on the pistons, and the pistons in the bores. Those are typically 120 cm, 4 feet in diameter. Weight is in proportion, you need hoists.
Now I don't see him much, he goes away when I go to get my car seen to at the workshop.
If you can do this - make chips and transistors - on a small scale, this will do to the chip industry what the laser printer did to the typesetting business.
I have people saying additive manufacturing is the future of injection molding, I tell them that 144 cavity PET preforms are cycling 10 seconds or so, that means 864 preforms a minute, how many 3D printers needed to achieve that production?
That generally ends the discussion.
I had somebody saying he worked on ship main engines, I asked how he fitted the rings on the pistons, and the pistons in the bores. Those are typically 120 cm, 4 feet in diameter. Weight is in proportion, you need hoists.
Now I don't see him much, he goes away when I go to get my car seen to at the workshop.
Last edited:
I do have an idea of what's needed to produce todays IC's on huge wafers with excellent yields and uniformity in often a "million a month" run rate.
Nobody claimed that these were production quality parts, or that the yield was anything above minimal (stated as 1 good out of 15, 2 more partially usable).
We here on these forums think nothing of building our own stereo systems. Some of us even design them from scratch, not simply build from someone else's schematic. How many average people even know (or believe) that this is possible.
Some people build their own PC's. Some even claim to be a "computer designer" if they can buy some boards off the web, pop in a CPU chip and some memory. Some of us have even designed our own "computers" going back to 1975 (an 8 bit system for making simple music). The average person does not know that this is possible.
Does this make a DIY transistor array completely impossible for someone to DIY? Given the pieces of equipment and processes outlined on Sam's web site, I would have to say no, it is not impossible. It happened.
Does this mean the DIY CPU chips made in a garage lab are just around the corner, just fire up the CPU printer? No chance. These parts were made on a 10 um process. That's roughly 1000 times larger than a modern CPU chip, and the poor yield limits the transistor count to maybe a few flip flops.
The fact that someone took the time and effort to do this twice with several preliminary experiments up front to find a process worth trying shows that not all of today's youth is wasted on playing video games or throwing Twitter bombs back and forth.
Nobody claimed that these were production quality parts, or that the yield was anything above minimal (stated as 1 good out of 15, 2 more partially usable).
We here on these forums think nothing of building our own stereo systems. Some of us even design them from scratch, not simply build from someone else's schematic. How many average people even know (or believe) that this is possible.
Some people build their own PC's. Some even claim to be a "computer designer" if they can buy some boards off the web, pop in a CPU chip and some memory. Some of us have even designed our own "computers" going back to 1975 (an 8 bit system for making simple music). The average person does not know that this is possible.
Does this make a DIY transistor array completely impossible for someone to DIY? Given the pieces of equipment and processes outlined on Sam's web site, I would have to say no, it is not impossible. It happened.
Does this mean the DIY CPU chips made in a garage lab are just around the corner, just fire up the CPU printer? No chance. These parts were made on a 10 um process. That's roughly 1000 times larger than a modern CPU chip, and the poor yield limits the transistor count to maybe a few flip flops.
The fact that someone took the time and effort to do this twice with several preliminary experiments up front to find a process worth trying shows that not all of today's youth is wasted on playing video games or throwing Twitter bombs back and forth.
Agreed...but if the purpose was to make parts no longer available, there is a long way to go yet.
And some circuits can be modified or scrapped, the book value of 30 year old equipment is zero.
So the effort is good, but the results are not consistent enough to get predictable parts.
Parts per million is a long way away. And not needed, as he is doing it on a very small scale.
Also I have noticed a tendency here not to use the equivalent Japanese and European series parts in place of the American parts.
Sometimes - not always - it can be done, and your equipment is usable again.
And some circuits can be modified or scrapped, the book value of 30 year old equipment is zero.
So the effort is good, but the results are not consistent enough to get predictable parts.
Parts per million is a long way away. And not needed, as he is doing it on a very small scale.
Also I have noticed a tendency here not to use the equivalent Japanese and European series parts in place of the American parts.
Sometimes - not always - it can be done, and your equipment is usable again.
Last edited: