Transistor "kits"

[paulca]

Why has nobody got this right yet?

Over the years I have picked up really useful "many values" kits for pF nF uF caps, diodes, zenners, schokty, resistors and so forth.

I did pick up a kit of transistors for £7, but it's literally ALL T92 NPN, PNP bipolars and 90% of circuits you need one for any of them would do.

I also found a set of 5 different darlingtons, but not a nice big kit of the 20 or 30 most useful transistors.

It seems to me there is or should be a market for a "Most useful transition" kit.

1 or 2 *values NPN BJTs
1 or 2 PNP BJTs
1 or 2 of the same but power transistors in To220
1 or 2 Darlingtons
1 or 2 N channel enhancements
1 or 2 P channel enhancements
1 or 2 N channel power mosfets
1 or 2 P channel power mosfets
1 or 2 N channel LL mosfets

The difficulty would be in picking which 1 or 2 variants. 

[Brianf]

Don't bother with kits.

Deciding on what to buy means that you will understand the specs of what you are buying. And it's not expensive. Just a couple of weeks ago I topped up my bipolar selection and bough 1,000 transistor for a total of £23.70 including VAT.

[Doctorandus_P]

I don't agree with Branf's remark.

I also once bought such a kit, and though all transistors are TO92, there is quite a lot of variance between the transistors. It's got the bog standard BC547 and 557, but also some Fets 2n7002 some darlingtons, some for higher currents and also some high voltage MPS types (I think 300V or so).

These kits are quite nice for a bunch of experiments on breadboards and for learning the differences between those parts. And because you've got 25 or so of each, it does not matter if you let the magic smoke escape out of several of them.

But if you have some project in mind, then go for some transistor that fits that project.

I also quite like the "SMD Books" which can have upto a few thousand of parts for EUR20 or so.

But do be aware of what you buy. Different sellers have different combinations in their sets. So gather the datasheets of the parts before you buy a kit. You will need the datasheets anyway. If a kit does not show which parts are in it, then just don't buy that kit and go forward to another seller.

[BillyO]

There are so many different kinds of transistors out there it might be hard to narrow it down to something that everyone would be interested in.

Write down your categories
Then write down the specs for each
Find a cheap transistor that meets them
Buy a bunch of each.

You may find the process time consuming only to find out that the rest of the world does not agree on your choices.

My best advice is to keep a small stock of maybe 6 to 8 simple dirt cheap small signal devices (2N2222, 2N2907, 2N3904, 2N3906, 2N7000, etc..) and then buy bigger/faster/higher voltage or other special quality ones when and if you need them.

[shapirus]

I did pick up a kit of transistors for £7, but it's literally ALL T92 NPN, PNP bipolars and 90% of circuits you need one for any of them would do.

Exactly. Unless you have very specific needs (high voltage, high frequency etc.), then pretty much any transistor of given polarity and of given power handling capability (aka package) will work.

From my (not particularly rich) experience the limiting factor for a general use (low voltage, low frequency) device is most frequently the package, while the rest of the parameters don't make any practical difference.

MOSFETs are a little different in this regard, but they too can be narrowed down to a few models which are good for general use.

[jasonRF]

I did pick up a kit of transistors for £7, but it's literally ALL T92 NPN, PNP bipolars and 90% of circuits you need one for any of them would do.

Exactly. Unless you have very specific needs (high voltage, high frequency etc.), then pretty much any transistor of given polarity and of given power handling capability (aka package) will work.

From my (not particularly rich) experience the limiting factor for a general use (low voltage, low frequency) device is most frequently the package, while the rest of the parameters don't make any practical difference.

In my experience (also limited) one parameter that really matters is the max dc current, which can vary a lot between to92 bipolars.  Bc549 is only 100 mA, 2n3904 is usually 200 mA, 2n4401 is 600 mA, and bc337 is 800 mA.  This does matter in some situations, for example when designing circuits to be robust against things like shorted outputs.  If you ignore this stuff you (or at least your transistors) can get burned. 

I wouldn’t buy kits from places like amazon or aliexpress, since you are probably getting fakes with unknown specs.  I just keep a stock of a few standard ones that I buy in quantities large enough to get some discount from mouser/digikey (usually just 10 at a time, but have purchased a couple batches of 100). 

Jason

[Kleinstein]

There are 2 kind of kits:
1) A collection of Excess inventory parts, often quite a lot of the same type and not a good mixture. They may be cheap but it is a gamble. It sometimes works out well and sometimes a lot a junk.
2) Kits combined in a way to be usefull.  So a reasonable selection, though depending on the interest this may not fit what you need. The price is often relatively high.
  Ideally they came with datasheets - no longer important with the internet.

Unless one gets a good price the kits may not be worth it unless they come with a nice case or similar.  The standard transistors are relatively cheap, and if needed one can order something like 10 of the one one would need. Chances are one could get them cheaper than a Kit of the 2nd type. For the start chances are one can get away with rather few types, Like 2N3904+2N3906 + 2 somewhat higher power types (e.g. TO220). The best transistor "kit" I remember buying was a bag of 100 unlabled TO92 NPNs (turned out to be low noise and high gain) that was rather cheap for the time (AFAIR some $1,50).

After some projects one anyway tends to have a reasonable collection from ordering a few more than actually needed for projects, either because of MOQ, price steps or just to have a spare as it happens to damage parts in a DIY circuit.  Later on there is a tendency to design around the parts one has at hand and knows anyway. One than usually only needs a few types, not a full set.

[paulca]

The thing I try and avoid happening is scenarios like happened several times last week.

Wanted to make a buck converter but I didn't own a single P channel mosfet and I only had one inducter.

I wanted to test a power supply filter with a ferrite ring, didn't have any of those either.  Nor did I have any inductors to test that LC circuit.

Now the approach a lot of people are prescribing is to, at that point, sit down and research some values of inductors and P channel mosfets and go to a reputible supplier, such as "Farnell" and order a few values.  Great.  Inductors are under a penny, minimum order of 10 if you are lucky more like 100.  Ok.  Mosfets are anywhere from a few pence to a few pounds.  Pchannel TO220 probably £1.  That's all great.  However you get to checkout for your £5.20 of parts and the £9.00 handling fee for small orders arrives, then the VAT @20%, then the shipping at 4.99 and suddenly those inductors are costing you 20p each!

For the same amount of money via Amazon and AliExpress I got a box of 10 different values, 100 peices of axial inductors, 10 values - 50 pieces of wire wound toroidal inductors a bag of 3904Ns and a few others.  Oh and a bag of 100 ferrite rings.  (Nobody sells a useful Mosfet kit).

[jonpaul]

kits were popular in 1970s..1980s eg the TI FET kit, or National Semiconductor

some came with books or app notes.

We found little practical design use and still have some...

Nowadays with the internet and distribution like DK, Mouser, Farnell, RS, easy to make a small order of exactly what you want

Jon

[paulca]

Yes, you can put in a small order for whatever you want on Mouser.  Yep.  This is what happens in reality when you attempt that.

2x 2.048V 0.1% LM4040
2x 3.3V 0.1% LM4040
... voltage references.

[shapirus]

Not all transistor kits are completely useless. FWIW, here's an example. I bought a kit of TO-126 transistors on Ali: 15 values, 10 pcs each. $10, or thereabouts, shipped.

One of the parts is a thyristor, one is a triac (didn't check the breakdown voltages -- will do if I ever need them).

Most of the others are generic complementary (as much as their fake markings are close to reality, but the transistor tester shows that it may be quite close actually) PNP+NPN pairs such as B772/D882, and yes, most of them could have been combined into just two values, one for PNP, another for NPN.

One however is usefully different: marked as E13003, which is a high-voltage NPN transistor. The datasheets I found specifiy V(br)ceo at 400V and hFE at about 20. Guess what, the DY294 high-voltage tester measured V(br)ceo = ~540V and the T7-H transistor tester showed hFE =  21 for the actual transistors in this kit. Sounds good enough to me.

I already used one of the low-voltage transistors in this kit as the load element in my USB dummy load. Works just fine at 2A/4W (with an adequate heatsink). TO-126 package, let me remind you.

Fake parts? Maybe. Very likely. So what? Even so, I consider them totally usable as general purpose parts for hobby use, and some, like the mentioned E13003, are very likely going to do their less-than-general-purpose high-voltage job as well.

[donlisms]

My first thought reading this thread was "Because of many parameters."  In very general and approximate terms, suppose you divide each interesting parameter into three ranges: low, medium, high.  Then if there's two parameters, 3^2 is nine combinations (parts) to cover all that.  If four parameters in five ranges... 625 parts.

With other components, maybe you typically have the value and "sufficient" voltage or power rating or whatever - much easier.

I know this is not an answer for those who prefer thinking in specifics, but it's... kind of a concept, eh?

[james_s]

The kits of resistors and capacitors can be useful, but for transistors I usually find that if a 2N3904/3906 won't do what I need then I probably need to order something more specific. If there is a particular kind of circuit you tend to play with then just buy a selection of transistors used in that sort of circuit.

[shapirus]

if a 2N3904/3906 won't do what I need

...for example?

Ok, high voltage is clearly one of such cases. High frequencies may require specific parts also. But what else?

[paulca]

if a 2N3904/3906 won't do what I need

...for example?

Ok, high voltage is clearly one of such cases. High frequencies may require specific parts also. But what else?

It doesn't matter what it is.  It will just mean you don't have it.

The last instances of this I had where:
Needs extremely high gain, so a darlignton.
Needs to be a Mosfet
Needs to be a P Mosfet

Then you start on "logic levels" and the perils of using transistors with MCUs.

EDIT:  If all of the above occured at the same time, there might be a Farnell order in there, but it would take me a month to pick them.  However they don't happen at the same time and a full supplier order is out of the question for 1 or 2 items.  So you buy a bag off Amazon, next day deliver for £7.... each time.  Yes, in the end you end up spending £60 on possibly suspect parts, when if you'd ordered it all together you would have got Farnel sourced parts for the same price.  But when you want to make a circuit and you are missing one part, next day and £7 for 10pcs sounds great.

However, in the tutorials, they will happily use all of the above and show the scope trace that they feel is relevant.  For example a capacitive  multiplier power supply filter.  Looks great in the tutorials until you attempt it and your rather noisy 6V turns into a lovely smooth 3V and the transistor starts to glow red.

[paulca]

Transistors in particuar NEVER give tutorial answers when you "wing" it.

The last one I set up my 5V output ended up 2.5V.  It took me a while and a burnt finger tip to realise the other 2.5V were dropping across the transistor.  Load resistor too small, too much current on the gate?  not enough current on the gate?

I swear I have more burn marks on my fingers than I have working transistor circuits.

And I know what you are all going to say, "Do the sums first!".  ... and yes, I have done that exercise with an NPN circuit, even before I had a scope and I choose my resistors and circuit as close to the numbers as I could.... and got results I was expecting.  It's just that sometimes you have the "text book" numbers for the ideal circuit in your head, you plough ahead with "How ya doing" values and a slightly modified, but "close enough" circuit and ... completely random results and burnt fingers.

Sometimes it frustrates me.  Why do people tolerate the idiosyncrasies of these components so much?  In software we just don't do that.  We get the smart people to wrap it in something that affords its design purpose.  Want a transistor to work as a switch, wrap it in an IC and make it work like a switch then.  Stop fucking around with all the devil in the detail for every single instance.  How did electronics get so far as it has?

Nobody is going to explain to a newbie how write a async FIFO buffer and teach the idiosyncrasies of the hardware access of that.  We give them a BufferedReader and a FileInputStream.  Why don't these abstractions already exist in electronics?  Why can't you buy a simplified "switch" component that actually behaves as a layman would expect?

Rant... Like, for example.  If a "switch" says it will be ON at 3.1V volts, then it's "ON", like "ON" as in how your grandmother would describe it.  Not in some grey beard curved approximation covered by 6 test circuits and 14 equations "on" that depends on temperature and the colour of the capacitor beside it.  I mean WTF?  Why is that hard?

"But I want my transistor in the trans-conductance zone". 
"Fine, then it's not a switch.  Call it a variable resistor, define it's interface, make it work like a variable resistor, wrap it, sell it in it's millions."

I think I know the defacto reason.  Cost.  I even know the above "ideal" devices do exist and they do work.  Ideal diodes, ideal transistors, "digital" analogue switches, remote controlled variable resistors etc. etc.  It's just that they cost 10 times more money than doing the sums, doing the design and using basic passives/descrites.  It also helps maintain the "Ivory castles" protection which many/most industries surround themselves with as a moot or learning cliff.  (Software and IT do it too).  At least "we" all speak languages other than Latin!

EDIT: Sorry I think I cross linked my two threads here.  I've had my coffee now I'll be fine!

[shapirus]

Needs extremely high gain, so a darlignton.

Can be made from 2x discrete transistors, as well as the Sziklai pair, which can sometimes be preferred.

Needs to be a Mosfet
Needs to be a P Mosfet

Aliexpress is your friend here, as long as you're ok with general-purpose transistors for use in low-voltage circuits and don't absolutely need the very best values of Rds(on). Yes the part numbers are not to be trusted, they are always fake, at least for the brand name transistors such as IR. However, it doesn't make them useless: for example, it's easy to find ones with Rds(on) at or below 10 mOhm which are fine to use as switches in low-voltage circuits. Read the reviews, those where people post actual measurements. If you know gate capacitance (which is a function of the die size) and Rds(on), you can estimate the boundaries of the transistor's applicability for your own purposes.

Buy a YR-1035 (aka TR-1035) to measure your DS milliohms at different GS charge voltages. Expensive? Not so, really, but if you think it is, build a CC source and measure Rds via the voltage drop across DS.

Know your transistors.

Then you start on "logic levels" and the perils of using transistors with MCUs.

These are far from general purpose parts, so not to be expected to be in the kits anyway. But, you can always use a workaround. Buy or build a step-up dc-dc converter to power the gate driving circuitry. Get a bunch of really handy isolated 5V->12V converters, power them from the MCU's 5V rail, or, if all you have is 3V, there are 3V->12V devices, and you're no longer limited to the logic level voltage.
Necessity is the mother of invention.

Try to design your circuits in a way that they don't need very special parts and that any part could be replaced by something generic instead of trying to stock up on fancy stuff.

But do maintain a shopping list so you're ready to order what you need once it's long enough to justify the P&P costs.

Looks great in the tutorials until you attempt it and your rather noisy 6V turns into a lovely smooth 3V and the transistor starts to glow red.

Some tutorials mention thermal management and the fact that real components aren't ideal. Some don't and assume that it's the reader's responsibility (and you should always assume it is). Don't trust anyone. Verify. Simulate.

[paulca]

Aliexpress is your friend here, as long as you're ok with general-purpose transistors for use in low-voltage circuits and don't absolutely need the very best values of Rds(on). Yes the part numbers are not to be trusted, they are always fake, at least for the brand name transistors such as IR.

I watched a few videos on "fake components" from AliExpress.  It was interesting to see how economics works differently there.  I think it was Great Scott that found most of the "lower end" transistors over delivered while the high end ones under delivered.  I would see that it's cheaper to fake/clone/refurb one model in the middle of the range and just relabel it all the other numbers in the range.

[shapirus]

I watched a few videos on "fake components" from AliExpress.  It was interesting to see how economics works differently there.  I think it was Great Scott that found most of the "lower end" transistors over delivered while the high end ones under delivered.  I would see that it's cheaper to fake/clone/refurb one model in the middle of the range and just relabel it all the other numbers in the range.

Yes that's basically what they do.

These products are unsuitable if you need a specific genuine part (for which be prepared to pay accordingly), but very acceptable if you just need to get a handful of generic, cheap, but still quite decent, even if fake, parts for general purpose use.

[jasonRF]

Sometimes it frustrates me.  Why do people tolerate the idiosyncrasies of these components so much?  In software we just don't do that.  We get the smart people to wrap it in something that affords its design purpose.  Want a transistor to work as a switch, wrap it in an IC and make it work like a switch then.  Stop fucking around with all the devil in the detail for every single instance.  How did electronics get so far as it has?

Nobody is going to explain to a newbie how write a async FIFO buffer and teach the idiosyncrasies of the hardware access of that.  We give them a BufferedReader and a FileInputStream.  Why don't these abstractions already exist in electronics?  Why can't you buy a simplified "switch" component that actually behaves as a layman would expect?

You can buy a switch component - search your favorite supplier for switch ICs and there will be hundreds or thousands to choose from.   


jason

[shapirus]

You can buy a switch component - search your favorite supplier for switch ICs and there will be hundreds or thousands to choose from.   

Yeah, in fact it's actually very much like software development: you can have things at any level of abstraction you want: from single transistors to ICs, modules and finished products.

[james_s]

if a 2N3904/3906 won't do what I need

...for example?

Ok, high voltage is clearly one of such cases. High frequencies may require specific parts also. But what else?

High voltage is one I've needed often, MPSA42 is my go-to. High current is another scenario, the 2N3904 is only good for 200mA max, it's not uncommon to want to control something that draws a bit more than that. Darlingtons, not used much these days but handy to have a few of. I also keep some N and P mosfets around. A box of scrap PCBs that you can harvest parts from as needed is something every hobbyist should have.

[james_s]

Sometimes it frustrates me.  Why do people tolerate the idiosyncrasies of these components so much?  In software we just don't do that.  We get the smart people to wrap it in something that affords its design purpose.  Want a transistor to work as a switch, wrap it in an IC and make it work like a switch then.  Stop fucking around with all the devil in the detail for every single instance.  How did electronics get so far as it has?

Nobody is going to explain to a newbie how write a async FIFO buffer and teach the idiosyncrasies of the hardware access of that.  We give them a BufferedReader and a FileInputStream.  Why don't these abstractions already exist in electronics?  Why can't you buy a simplified "switch" component that actually behaves as a layman would expect?

Rant... Like, for example.  If a "switch" says it will be ON at 3.1V volts, then it's "ON", like "ON" as in how your grandmother would describe it.  Not in some grey beard curved approximation covered by 6 test circuits and 14 equations "on" that depends on temperature and the colour of the capacitor beside it.  I mean WTF?  Why is that hard?

"But I want my transistor in the trans-conductance zone". 
"Fine, then it's not a switch.  Call it a variable resistor, define it's interface, make it work like a variable resistor, wrap it, sell it in it's millions."

I think I know the defacto reason.  Cost.  I even know the above "ideal" devices do exist and they do work.  Ideal diodes, ideal transistors, "digital" analogue switches, remote controlled variable resistors etc. etc.  It's just that they cost 10 times more money than doing the sums, doing the design and using basic passives/descrites.  It also helps maintain the "Ivory castles" protection which many/most industries surround themselves with as a moot or learning cliff.  (Software and IT do it too).  At least "we" all speak languages other than Latin!

EDIT: Sorry I think I cross linked my two threads here.  I've had my coffee now I'll be fine!

Because that's the way electronics work, it has always been like that, these components exist in the analog realm, you have to know what you're doing to work with them effectively, that's why there is occasionally debate on whether software engineers should be called engineers at all. Software is binary, somebody has already done all the hard work in the analog realm for you, the hardware is debugged and ready to go, it just works.

A switch just means a transistor is used to switch something, it's either on or off, two states, the voltage drop is not defined there, everything has voltage drop, even a mechanical switch. Those "ideal" ICs you mention have their own quirks and caveats, and they cost more because they took effort to design, test and market, and they are more specialized so less versatile than stuff like single transistors.

I think the only reason you find software easier is that you've already been through the learning process. Personally I find software hard, I can write code but I'm not very good at it, algorithms are the part I have the most trouble with. Building stuff with transistors is much easier for me, I understand how they work, they feel intuitive, I have a better sense of their limitations than I do with coding techniques. It's not anything to do with ivory castles, it's just a fact that engineering, whether software or hardware is complicated, there is a lot to know, it's the reason there are four year degree programs dedicated to one specific discipline of engineering.