Lost in BJTs

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: lisa, luketeaford, Kent, Joe.

Post Reply
BWBKc6VyUr
Learning to Wiggle
Posts: 44
Joined: Thu Sep 26, 2019 1:25 am
Location: UK

Lost in BJTs

Post by BWBKc6VyUr » Fri Jun 26, 2020 1:59 am

I'm sure this has been discussed before, but I can't find a clear synthesis anywhere so I'll start a new thread. Feel free to point me to an older thread if you know of one.

What I remember from my student days about BJTs is : provide a base-emitter current that breaks a well-defined threshold -> get a proportional, higher collector-emitter current. (I'm happy to be corrected if this is inaccurate but it isn't my main objective for starting this thread.) I read on Wikipedia that "Many types of transistors are made to standardized specifications by multiple manufacturers."

I'm assuming those "standardised specs" define specific part numbers and how they differ. I'm trying to learn which parts are a good starting point, or even a definite go-to, for particular synth/audio circuits use-cases. (For example, and again correct me if you disagree, but it is my understanding that 2N3904/6 are usually cheap and great for turning LEDs on/off, but not so great for audio signal chains.)

I realise my question is somewhat generic, maybe too much so, but could someone give a rundown of which BJTs are commonly used in audio or synth circuits, and to perform which function?

User avatar
EATyourGUITAR
has no life
Posts: 4770
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Re: Lost in BJTs

Post by EATyourGUITAR » Fri Jun 26, 2020 3:41 am

the question is a good question. in 2020, silicon NPN or PNP transistors are cheap and good. for switching an LED on, there are many ways to do it. a transistor is one possible way to switch an LED on. a BJT provides current gain which is used by a lot of people with resistors to achieve voltage gain. the device performs current gain but the circuit provides voltage gain for the typical NPN audio amplifier. I will spare you the most of the theory for now or I will write you a book. so if you want an LED to turn on from some arduino for example, you can hook up a transistor to the GPIO output pin of the arduino. you can put the LED on the transisor. but you could also drive the LED from the arduiono directly because the current consumption for the LED is maybe 3mA but it does not exceed the arduino output maximum rating or 10mA. the transistor would be unnecessary. if for example you wanted to use some crazy huge 3w LED, you will need more than 10mA. to control it from an arduino, you will need a transistor that can pass enough current to power a 3W. you will also need to work backwards from the 3W to get the collector current to get the base current from the beta. for now lets say beta = Hfe = current gain. a transistor with a beta of 200 will have 200mA through the collector 1mA into the base, and 201mA through the emitter. but if the beta is 10 then 500mA out is 50mA at the base so you will not have enough current coming from the arduino to drive the transistor to turn on this 1W LED that needs 500mA. you can however put a transistor on a transistor. transistors all the way down. a 2N3904 switches the small currents 100mA or less. the 2N3055 switches larger currents.

using a transistor to switch a load on or off has different requirements compared to the design of a perfectly linear amplifier. I have a big problem with that book definition they gave you. it was entirely bullshit. the word proportional implies linear. it implies a perfect scalar, multiplication if you prefer to call it that. a transistor has a linear region between shutoff and saturation. this is called the "linear region" for a reason. too little or too much current into the base, you will be out of the sweet spot. effectively having a linear amplifier depends on how you use the transistor. there are differences in different part numbers but there are also differences within a single part number even within a single vendor even within a single batch. for applications that require matches transistors, designers will spec laser trimmed matched transistors and matched resistors. this is much more practical than sitting at the table with a bag of 1000 transistors writing down numbers and taping them to cards.

so yes there are many transistors and transistor based designs. I probably have one of the largest collections of old transistor books and circuits here. most people have moved on. books from 1967 are so cheap you can buy them all.

the focus today is CMOS for switching, opamps for everything else. the opamps are made from transistors however they are not exclusively made from BJT. they can have BJT, CMOS, JFET inputs and outputs or any combination thereof. linearity of an opamp is much easier to manage. linearity from discrete transistor designs requires an engineer. before computers, there were formulas and models to get the best out of these circuits. hybrid pi model is one for example. Ebers-moll is another. there are about 9 models last time I checked.
WWW.EATYOURGUITAR.COM <---- MY DIY STUFF

loki
Common Wiggler
Posts: 98
Joined: Fri Apr 25, 2014 4:20 pm

Re: Lost in BJTs

Post by loki » Fri Jun 26, 2020 12:25 pm

Download the: National Semiconductor - Discrete Databook 1978 and the 1982 National Transistors Databook.
You can find them at the internet archive and many other places. There is also the: National Semiconductor Fet Databook 1977

They contain descriptions of the processes used for their entire line of transistors and die photos of them. You will notice that some processes are associated with multiple part numbers. Process control was poor enough in the early days that parts were tested and binned according to the distribution of their properties. Process control is much better now and wafer of 2N3904s probably yields at better than 99%.

User avatar
uniqview
Learning to Wiggle
Posts: 16
Joined: Mon May 18, 2020 3:18 pm
Location: Silicon Valley, California
Contact:

Re: Lost in BJTs

Post by uniqview » Fri Jun 26, 2020 11:48 pm

I too am lost in BJTs :love:. My SDIY work alas has had many phases of flip-flopping. Some years have been lost due to thinking along the lines of digital sofware, or using DSPs, FPGAs, CMOS, and TTL memory and logic.

Mostly, I prefer analog, and right now I'm hard at work on SDIY based on a really weird (but incredibly useful) single-supply voltage quad opamp, the LM3900. Mixed in with this will be a number of BJT circuits I developed.

I have a blog which mixes in some of this, but more of the posts are actually about my original BJT development work, including discrete opamps, comparators, and various other really useful circuits. In case useful in your BJT quest, here is one link to start from: EMS Development Blog.

I made a number of discoveries doing development with BJT discretes. First, virtually any amount of performance you want can be had, and with parts that literally cost pennies. Second, a number of key integrated circuit specifications are easily exceeded by using more power, more wire, more components, etc. Of course, some specifications, not all. Fabricating a number of transistors on a single silicon die definitely has immutable advantages. And technical improvements from various process nodes over time have enabled tremendous performance.

At this point in time, I'll just say that I'm generally a non-fan of CMOS and other lower-voltage opamp devices. In any case I most definitely do not like surface mount components whatsoever -- for SDIY.

I personally find many of the trends in VLSI and SOCs really kind of boring, to be honest :sarcasm:. Now, of course, for mobile phones that have 64-bit supercomputers in them, tablets, or laptop computers: that's a different story. Late-modern silicon tech makes all that possible. I also use a number Raspberry PI 4B computers as hosts for Software Defined Radios,

But the use of continuous time signals with analog for creative sound generation and processing has a very great appeal.

You'd requested some notes on specific devices. I started with a wider selection in early 2000s, and this got whittled down as availability declined. It's still declining too. If you do find some devices you like working with, you might consider buying a batch. Notes follow.

Aside from the general purpose 2N3904 and 2N3906, I find the 2N5088 and 2N5087 devices very favorable for linear signal processing because they offer high-beta at lower collector currents, often usable as matched pairs without actually matching them. The 2N5961 are now harder to find, but they are very good for extremely low noise figure gain blocks. A good complementary pair for medium-power switching is the 2N2222A and 2N2907A; alas they are more likely available now in TO-92 plastic form with worse thermal resistance as the PN2222A and PN2907A. Another device I like a lot, for extremely fast switching circuits (think discrete ECL logic) is the 2N2369A, which has a derived part the PN2369A; again, both are essentially antiques now :bang:. Today, it's not commonly recognized that one can operate BJTs on much higher voltages than typical ICs. This situation is advantageous for Class A ultra-low distortion audio signal processing circuits: making the required signal swing a small fraction of that available!

Bipolar Junction Transistors are still really cool. There are number of things these devices can do quite directly. Among them are: provide accurate switches in inverse mode; oscillate reliably and with good temperature stability in inverse mode; create accurate linear sweeps; form simple comparators that allow direct signal interface; provide super clean wideband phase-splitting for differential circuits; AC gain is cheap and easy; ... the list is infinite :hyper:.

Enjoy your quest!
Applying the LM3900 for amazing things in SDIY 8-)

BWBKc6VyUr
Learning to Wiggle
Posts: 44
Joined: Thu Sep 26, 2019 1:25 am
Location: UK

Re: Lost in BJTs

Post by BWBKc6VyUr » Sat Jun 27, 2020 10:45 am

Thanks for your answers, they are mostly less synth-centric than what I was hoping for, but full of valuable points. I know my question itself is fairly generic.

What I said about how BJTs operate was more a vague recollection than a definition, thanks for the refresher on linearity and non-linearity it rings a bell.
uniqview wrote:
Fri Jun 26, 2020 11:48 pm
I started with a wider selection in early 2000s, and this got whittled down as availability declined. It's still declining too.
EATyourGUITAR wrote:
Fri Jun 26, 2020 3:41 am
the focus today is CMOS for switching, opamps for everything else.
This corroborates what I'm finding out in my readings. It's actually one of the reasons I'm trying to better understand them : they are ubiquitous in historical circuits that enshrine interesting functional ideas but deserve a good dusting off. But it's not always easy to understand the historical reasons for putting a BJT somewhere, or why a specific part was used. (Side note -- I'm comfortable interpreting opamp designs but FETs are yet another can of worms (to me) that I also intend to open soon.)

As for SOCs or plain MCUs -- though they use MOS as their building blocks, they are an altogether different topic... (FWIW I do understand the specifics of driving an LED directy with a GPIO; I just mentioned the LED example to demonstrate which kind of info I'm after.)
loki wrote:
Fri Jun 26, 2020 12:25 pm
Download the: National Semiconductor - Discrete Databook 1978 and the 1982 National Transistors Databook.
You can find them at the internet archive and many other places. There is also the: National Semiconductor Fet Databook 1977
I've only had a quick look yet but they are very interesting, e.g. Section 4: Selection Guides (1982 National Transistors Databook, p122). I need to go over that stuff properly.

Each answer mentions 2Ns, why are those most common?
And what about other references, e.g. the many BC5xx ? Why/when would you use them in a synth circuit ?

User avatar
belltones
Learning to Wiggle
Posts: 28
Joined: Wed Aug 16, 2017 10:27 pm
Location: Philadelphia
Contact:

Re: Lost in BJTs

Post by belltones » Sat Jun 27, 2020 11:22 am

My perspective is a little different because my head is 80% in servicing designs already built and 20% in designing, but I do have one simple thing to mention which was implied in some other people's answers, but which I can maybe help untangle from the more in-depth and specific discussions I saw above. When trying to select a transistor for a certain application (be it a new design, or a substitution in a design that already exists) the most important thing to do is first figure out what "job" the transistor is doing, or needs to do, in the circuit. IE is it switching, amplifying a voltage (eg an audio-frequency signal), or supplying/amplifying current for a signal whose voltage amplitude will not be changed? Your long-term goal should be to familiarize yourself with which specifications/parameters are most important for different types of applications and how their values affect a given transistor's performance at that task. One thing that helps is to note which transistors are used in other applications you know to be similar, and look at their datasheets. See what trends you can notice. I learned a lot of transistor theory which was really abstract and didn't help me much with actual parts selection until I started looking at the datasheets and learning what the *actual numbers are.*

Another thing to mention which I think people often miss, which is true for semiconductors in general, is that a lot of the qualities that make certain parts "better" (as well as certain board layout concerns and other design concerns, but nevermind that for now) only matter at high frequencies (multiple megahertz) and actually don't really matter in analog, audio-frequency applications. People sometimes conflate "high frequency" in the way that people use it with regards to audio with the way that electrical engineers use it in general, but they're actually a factor of at least a thousand apart, so don't let concerns about "high frequency performance" lead you astray.

User avatar
EATyourGUITAR
has no life
Posts: 4770
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Re: Lost in BJTs

Post by EATyourGUITAR » Sat Jun 27, 2020 11:50 am

There is a book on designing differential amplifiers from BJT if you are interested. Not an easy thing. Not a cheap book.
WWW.EATYOURGUITAR.COM <---- MY DIY STUFF

User avatar
uniqview
Learning to Wiggle
Posts: 16
Joined: Mon May 18, 2020 3:18 pm
Location: Silicon Valley, California
Contact:

Re: Lost in BJTs

Post by uniqview » Sat Jun 27, 2020 1:19 pm

EATyourGUITAR wrote:
Sat Jun 27, 2020 11:50 am
There is a book on designing differential amplifiers from BJT if you are interested. Not an easy thing. Not a cheap book.
So exciting to see such interest, here on BJTs :hyper:. I thought I was completely alone in the world!

EATyourGUITAR, this is the book from 1963 that I have on differential amplifiers:
PSX_20200627_110133.jpg
Do you have a different one? I'd certainly enjoy learning from another perspective.
You do not have the required permissions to view the files attached to this post.
Last edited by uniqview on Sat Jun 27, 2020 2:55 pm, edited 1 time in total.
Applying the LM3900 for amazing things in SDIY 8-)

User avatar
belltones
Learning to Wiggle
Posts: 28
Joined: Wed Aug 16, 2017 10:27 pm
Location: Philadelphia
Contact:

Re: Lost in BJTs

Post by belltones » Sat Jun 27, 2020 1:59 pm

BJT fan club!

User avatar
Repeater
Phantom Powers
Posts: 765
Joined: Sun Oct 03, 2010 3:14 am
Location: Rochester, NY

Re: Lost in BJTs

Post by Repeater » Sat Jun 27, 2020 2:29 pm

I use 2N3904 and 2N3906 often for discrete oscillators, current source/sinks and in house-keeping stuff like LED drivers, buffers, etc. Haven't ever really considered using other BJTs because these get the job done for me.

Another author I'd recommend that doesn't get a lot of attention is Delton Horn. He wrote a lot about discrete designs in his books about synth DIY. Some are more technical than others, but you can always supplement with a book like The Art of Electronics.

User avatar
uniqview
Learning to Wiggle
Posts: 16
Joined: Mon May 18, 2020 3:18 pm
Location: Silicon Valley, California
Contact:

Re: Lost in BJTs

Post by uniqview » Sat Jun 27, 2020 2:44 pm

BWBKc6VyUr wrote:
Sat Jun 27, 2020 10:45 am
Each answer mentions 2Ns, why are those most common?
And what about other references, e.g. the many BC5xx ? Why/when would you use them in a synth circuit ?
Type registration for American electronic components began under the Joint Electron Tube Engineering Council (JETEC), in 1944. Some other trade organizations later emerged as part of this effort, including the Radio Electronics & Television Manufacturers Association (REMTA) and the National Electrical Manufacturers Association (NEMA). RETMA later became the Electronic Industries Association (EIA). In 1958, as some of the first new silicon semiconductors arrived, a new combined trade organization emerged for electronic component type registration: the Joint Electron Devices Engineering Council (JEDEC). The JEDEC semiconductor series numbering runs like this:
  • 1NXXXX diodes and rectifier
  • 2NXXXX transistors and 3-terminal semiconductor devices
  • 3NXXXX semiconductor devices having 4 electrodes, generally thryisters, multi-emitter transistors, or specialized MOS transistors, etc.
Prior to this system, there were also devices registered 2NYY and 2NYYY, earlier than under the 2NXXXX system of 1958. Some interesting history can be found in Andrew Wylie's retrospective web post. Another fascinating source documenting early semiconductors is the Transistor Museum.

The BC5xx and many other simlar and related parts originated in a British component registration system, which later became pan-European. This system is called PRO-ELECTRON. I'm much less versed in particulars for these devices but I have managed to collect a nice Mullard transistor applications manual from that time. However Wikipedia describes this history, and how the type registration system works: PRO-ELECTRON.

Update: I was looking around for more historical transistor information, and in this process, I found a rather nice general transistor explainer. This section Transistor & Diode Part Numbering Codes however explains nearly all the part registration schemes now in use, including Japanese parts.

The answer to your question as to why a particular device would appear frequently is simple: ubiquity of application within in a particular market trading area. Example: I'm in Silicon Valley, California, and can literally walk to where many of these devices were first made. 1N and 2N series are just much more available in this geographic area. Whereas, in Europe, BC5xxx and many other devices are likely more available.

Why/when to use in a synth circuit is a much harder question to answer: it depends. These are circuit design questions, and sometimes the answers are ambiguous; in come cases, more then one device could fulfill the same role. But in other cases, a particular design depends on a device having a specific useful property. An example might high-β under all currents and temperatures, which could be important for DC accuracy, high-gain, or low-noise. A different circuit might require very low Vce|SAT, to work effectively as a switch, or to more efficiently drive a load with low power dissipation. The list is endless ... it depends on what is more salient to a particular design problem at hand. After doing this a while though, a "set of rules" can start to emerge, to help guide one through a circuit design problem.

To help some with this question though I did recover from my transistor-focused blog my itemization from 2017 on roles for specific transistors and diodes. e.g. the capabilities each device seemed fitted well for in synthesizer applications: Semiconductor Parts Lists

The actual things (a subset of ...) these devices were particularly good at (or at least: It worked!) are posted in various articles in the blog.
Last edited by uniqview on Sat Jun 27, 2020 7:22 pm, edited 2 times in total.
Applying the LM3900 for amazing things in SDIY 8-)

User avatar
EATyourGUITAR
has no life
Posts: 4770
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Re: Lost in BJTs

Post by EATyourGUITAR » Sat Jun 27, 2020 5:39 pm

uniqview wrote:
Sat Jun 27, 2020 1:19 pm

EATyourGUITAR, this is the book from 1963 that I have on differential amplifiers:
PSX_20200627_110133.jpg

Do you have a different one? I'd certainly enjoy learning from another perspective.
yes that is the one. these are also good. you may not want a book that only does DC discrete amplifiers. the use for those is pretty non-existent.


WWW.EATYOURGUITAR.COM <---- MY DIY STUFF

KSS
Super Deluxe Wiggler
Posts: 2410
Joined: Mon Jan 25, 2016 7:28 am

Re: Lost in BJTs

Post by KSS » Sun Jun 28, 2020 1:12 am

Seconding what loki and belltones wrote. Both really cut to the meat and heart of the matter.

Will add that although audio may be slow in an overall sense, some parts really do need high speed. Like the reset of a saw core oscillator. Even PAIAs old 4720 could be greatly improved with the use of a transistor designed for RF switching. Discretes doing logic duty or combined with TTl, CMOS are another example where the throughput may be slow, but certain parts -transitions- need to be fast.

And will tie it together by saying beware of magic part number thinking. Read loki's comment along with belltones again and you'll come to the conclusion that certain 'magic' part numbers of a given designer had less to do with some magic quality than a host of other factors. Very, very few synth circuits require the original transistor to work the same to a level essentially undetectable by human ears.

User avatar
uniqview
Learning to Wiggle
Posts: 16
Joined: Mon May 18, 2020 3:18 pm
Location: Silicon Valley, California
Contact:

Re: Lost in BJTs

Post by uniqview » Thu Jul 02, 2020 7:45 pm

Repeater wrote:
Sat Jun 27, 2020 2:29 pm
Another author I'd recommend that doesn't get a lot of attention is Delton Horn. He wrote a lot about discrete designs in his books about synth DIY. Some are more technical than others, but you can always supplement with a book like The Art of Electronics.
Have to second that. And what's annoying, is Before Marriage, back in late-70s to early 80s, I know that I had Delton Horn's books! But as the sands of time weave through the glass, I cannot find those books anymore.

One item I was lucky enough to collect in my technical books library in the last decade is REFERENCE DATA for RADIO ENGINEERS, 4ed. It's from 1958, but wow it is one of the best EE references I have ever seen. The Fundamentals of Networks chapter alone is very worthwhile. This reference book seems still very relevant, its definitely complete, but also compact. Of course, some things have faded, like tables of frequency allocations. However, the full-color pull-out frequency spectrum sheet has not changed -- wavelengths are wavelengths! And yes, we use many fewer vacuum tubes today, and while this reference does cover junction transistors, the planar transistor had not quite happened by 1958.
Applying the LM3900 for amazing things in SDIY 8-)

KSS
Super Deluxe Wiggler
Posts: 2410
Joined: Mon Jan 25, 2016 7:28 am

Re: Lost in BJTs

Post by KSS » Thu Jul 02, 2020 8:08 pm

uniqview wrote:
Thu Jul 02, 2020 7:45 pm
Have to second that. And what's annoying, is Before Marriage, back in late-70s to early 80s, I know that I had Delton Horn's books! But as the sands of time weave through the glass, I cannot find those books anymore.
abebooks.com is your friend. Hundreds of Delton Horn books very inexpensive. Add "synthesizer" to his name in author and you better act fast!. Of his synth books, there are only a few copies for sale reasonably priced, with many others absurdly listed at ridiculous prices.

Post Reply

Return to “Music Tech DIY”