You Can’t Build A Roland TR-808 Because You Don’t Have Faulty Transistors

That headline sounds suspect, but it is the most succinct way to explain why the Roland TR-808 drum machine has a very distinct, and difficult to replicate noise circuit. The drum machine was borne of a hack. As the Secret Life of Synthesizers explains, it was a rejected part picked up and characterized by Roland which delivers this unique auditory thumbprint.

Pictured above is the 2SC828-R, and you can still get this part. But it won’t function the same as the parts found in the original 808. The little dab of paint on the top of the transistor indicates that it was a very special subset of those rejected parts (the 2SC828-RNZ). A big batch of rejects were sold to Roland back in the 1970’s — which they then thinned out in a mysterious testing process. What was left went into the noise circuit that gave the 808 its magical sizzle. When the parts ran out, production ended as newer processes didn’t produce the same superbly flawed parts.

This is an incredible story that was highlighted in 808, a documentary premiered at SXSW back in 2015. The film is currently streaming on Amazon Prime (and to rent everywhere else) and is certainly worth your time just to grasp how seminal this drum machine has been in hip hop and several other music genres.

For modern product developers, betting your production on a batch of reject parts is just batty. But it was a very different time with a lot fewer components on the market. What worked, worked. You do have to wonder how you stumble upon the correct trait in an obscure batch of reject parts? Looks like we’ll be adding Ikutar Kakehashi’s book I Believe in Music: Life Experiences and Thoughts on the Future of Electronic Music by the Founder of the Roland Corporation to our reading list.

[via EMSL]

55 thoughts on “You Can’t Build A Roland TR-808 Because You Don’t Have Faulty Transistors

  1. A diode, when reverse biased, will generate noise. Some are better at it than others. But so long as you get good noise, all should be well. “Distinctiveness” would suggest a diode isn’t generating random noise.

    A 1N21 microwave diode was !ong used for adjusting VHF and UHF front ends. But you don’t need that for audio noise. PAIA Electronics used 2N2712s or something like that, but I’m not sure if it was chosen for noise or other factors. It was “obscure” but very cheap, they used it all over in their electronic music equipment. I followed along, but used some cheap transistor I had around.


    1. I have used zener diode noise generators for RF noise and dual transistors for RNG noise but they don’t all produce the same noise
      Mainly in the low end were I think this transistor shined

      It would be nice to have an expert crack one open to see why it’s failed but that’s a long shot as the 808 is a legend

    2. I recall that when I put together a PAiA Gnome microsynth back in the golden age, the “noise” transistor had a blob of white paint on it. I later read that at PAiA, they had manually sorted, tested and selected those marked transistors for their noise characteristics.

      It’s simple to rig up a test jig to slightly reverse-bias transistor junctions, and couple to a mic preamp to hear the noise.

      1. Transistors are hit or miss, the noise is just a byproduct of the semiconductor fabrication and how it works electrically. Zeners seem to be a much larger, beefier source, I’m guessing due to the larger boundary layer of the PN junction

        1. Like the “warm feel” of tube amplifiers.

          Given enough time, money, bits and megahertz, I’m pretty sure you could digitally recreate the tube amp sound so that it would be indistinguishable from the real thing in double blind testing.

          1. Bob Carver did that exactly. He did it with several amplifiers but is perhaps most famous for producing the awesome Silver Seven tube amp and then designing and selling a transistor-based version that provably sounded exactly(enough) the same for far less.

          2. Already done years ago and I own one. I hate tube sims, a 50$ tube amp sounds better, so when the first digital tube sim had me fooled I just had to buy it. The Kemper even reproduces and stores existing amps by ‘profiling’ them accurately (you just put a mic in front of the real amp and connect it to the Kemper).

          3. Tubes age, getting weaker as they are used. If a tube amp is perfectly duplicated in solid state and both are used for a year, they won’t sound the same any more. Some tube aficionados will then claim that result as proof that tube amps are superior and can’t be duplicated.

        2. Don’t know who said that ” no one left dancefloor only because tb303 line was not done with tb303″( more less – I red only translation). Same goes for any other gear. But there is a difference between getting acid sound and real tb303 line. Same goes for tb303. You get close enough with samples and VST (check d16 group – heard they did very good job) and for 95% applications it’s fine or even better. But in some productions close enough is not enough.

      1. while the new compact series using digital modeling (they have a library where they simulate at the component level really interesting) us synthheads what hardware analog goodness. Hardware analog synths are popular again and no longer rare vintage artifacts. This is why I’m pumped for the Behringer RD-808 , it will sound close enough and be analog.

  2. Like the 6581 (aka “SID”) had that “fourth voice” (actually a way to use the basic noise as a voltage offset to output 4 bit samples through volume control without much affecting the three “real voices”) … something that REMAKES don’t cover, because the specific glitch in the original SID seems to still be unknown in nature.
    Have a look here – the osci displays 3 real voices plus the “hacked” sampled fourth voice:

    I am pretty sure that a lot of “distinctive characteristics” in old-times-sound-machines come from such glitches. And no, modern remakes using samples and sophisticated “it should work this way, really it should” designs quite often do NOT get it.

  3. Replicate it like Queen’s little amp was replicated. Carefully take that transistor out of one or more of those drum machines, test it every way possible then design a new transistor that’s made to duplicate its special performance.

    That’s what was done with the amp’s crappy transformer and speaker. The transistors were analyzed then batches of vintage transistors were found and individually tested to (rotten)cherry pick just the ones that performed precisely as poorly as the originals.

    1. It’s not just the transistors, but the capacitors, the audio transformers, and perhaps even a few substandard decoupling capacitors. In other words, you’d have to cherry pick everything. Don’t forget that parts age. So what you have today may not sound exactly like the one that was used to make the recording.

      Perhaps music is actually supposed to be ephemeral. We can play the same sheet of music, but it will never sound the same way twice.

  4. Oh, the 808!

    First, chasing “the” sound is for suckers. The 808 was built inexpensively, for the bottom end of the market, for bands that were too cheap to hire a drummer. It was not built for golden-ear studio types, and the QC was done by many different humans on an assembly line. The result: cross-machine variation is huge. There was no single sound.

    And don’t get me started on the temperature dependence of the circuitry in an 808! The same machine will sound different six months from now.

    But second. If you’ve ever heard an 808 first-hand, it doesn’t sound at all like an 808. F’rinstance, it has a crappy anemic bass drum. (“But, but, the Beastie Boys!”) Nope, the trick with an 808 is to record it so hot that it saturates the tape that you’re recording it on. The resulting compression and overdrive/distortion is what makes the 808 sound like a monster, and not just a decaying single sine wave with a little punch up front. At least half of the 808’s sound is in the studio rather than in the box, IMO.

    So then the question is, can you get an 808-enough sound that it will work for your audience? The modern remakes, as the OP suggests, will get you there, funny transistors or not. And if you’re not going for pure nostalgia, you should be able to do better than the 808 anyway.

    Unless you’re really after “the” handclap. Then only psychiatry can help you.

    1. Im no music geek, but reminds me of when you hear a 303 it doesn’t sound like you think it does till you realize people run them thro guitar distortion pedals and compressors and the like.

        1. Want to sound like the 80s-90s? Produce like the 80s-90s. :)

          I used to have this old stereo Revox reel-to-reel, with tube amps. I was a kid, so I was a) using used tapes and b) re-using them myself. It sounded horrible in 1992 in a way that would sound awesome in 2018. If you know what I mean. People are strange.

        1. The ‘noise’ seems to have been uniquely weighted, with exactly the spectrum they felt worked best for the application – also working ‘as is’ requiring less overhead to implement.

    2. Sorry but this is not true. The 808 was very expensive at the time (and still is) and aimed at prof bands to be used for demo’s. It was a commercial failure because it didn’t sound like real drums, which the Linn LM-1 did.

      An 808 has a very distinguishable sound, even clean. Post processing it makes it cut through a mix better, punchier etc, yet it’s a pretty “finished” sound. Unlike a 909 which sounds very dull without any processing, imho.
      If you really ever heard a clean 808 you’d know, and could easily spot it in any track.

      My 808 and it’s still one of my favourite drum machines. Probably the last one to leave my place. I’ve never noticed it suffering from temperature changes at all. When you hook it up to a sub woofer, it becomes a real neighbour killer.

      1. I had to look it up, but it retailed for just less than half of a Linn Drum when it was new. Roland was not aiming at the top of the market. That said, I thought it was hundreds of dollars, not $1,200. So that’s still a fair chunk of change.

        Sound was/is unique, for sure. But you don’t find the bass drum too clean/sine-wavy? If that’s what you’re going for, there you go. :)

        That all said, I haven’t had hands on an 808 in like 15 years, and I’ve only built the snare and cowbell circuits myself. So for musical judgement, I’ll defer!

  5. The mythology and nostalgia surrounding these products is what they are all about and this is a great example of it, the search for authenticity inspires a lot of people to go a long way and spend a lot of money. If you want the exact feel and sound then it’s true that you need the exact technically flawed hardware. Ironically tho I supect the original users would have dropped them and forgotten about them fast if they were given some of the modern analog gear around today. Artists are a fickle bunch and trying to rationalise it kinda misses the point

  6. Getting the “best” sound out of an instrument as much about how the instrument inspires the musician versus pure technical performance. You can make technical improvements/samples/tweaks/new materials and parts which are absolutely, unequivocally better than older instruments. They sound better, work in a wider range of conditions, have more consistent performance, etc. For most instruments every single technical metric can be made better with newer stuff.

    But the inspiration part you really can’t do anything about. Creative inspiration comes from a lot of different things, which is why guitarists play shitty old guitars or go to old recording studios where famous bands recorded tracks…or seek out 808’s. It’s as much about the story as it is the gear, and that’s all there is with the 808. No question it was a shitty synth, but it hit the market at just the right time and ended up getting used by a bunch of famous people, so the legacy will live on.

    1. Agreed. Back around 1995ish, my sister got a free used full size upright piano to practice on. The piano could only be described as hideous sounding. Like the bottom octave+ strings were literally caked with crud. They had the most ridiculous dead thump of a sound. It was comical.

      And then my musician friend came over. He was big into mixing, synths, and all kind of interesting music. I thought he’d be appalled.

      His eyes lit up and he instantly played the most rad hip-hop style bass line I’ve ever heard. Like crazy slap bass. I still have it somewhere on an old DAT tape.

      That’s a real musician.

  7. Fascinating subject, but good lord, does that Synthesizers blog need an editor. The whole article was written like a high school book report. Is the author paid by the word, or by the length of his run-on sentences? Hard to get through.

  8. The headline reminds me of all the nay sayers who said I couldn’t replicate an original Minimoog Model D by designing my own PCBs from the Moog repair schematics. Having done it successfully, with indistinguishable sound, I can confidently say that no matter what strange nuance an original design has (in the case of the Model D it was weird stuff like “capacitors” made from intentional trace designs), it *can* be replicated with modern techniques and tools, the only question is how much time and money you want to throw at it. The reason is *physics* haven’t changed.

  9. Consider the recent invention in the video realm of the “Deep Fake” and similarly the use of GAN neural networks for generating image clarity out of what amounts to algorithm. Shoot, now that I think of it google has been doing this already on the audio world with NSynth .

    What I’m getting at is rubles to robots if took a neural net generated 808-like cpu and shoehorned it into an empty 808 shell… maybe slap an old vintage front 242 sticker on it, you could not only fool anyone other than someone with direct day to day experience with the real deal, you most likely could get turn an easy profit as well .

    You know… now that I’m thinking about it maybe I should have kept that all to myself…. These are not the droids you are looking for!

  10. Big detail nobody in this post is addressing… though the story of the defective-component-noise-generator may be true, the noise generator is only used in some of the 808’s sounds. It isn’t used at all in the kick drum (“twin-T” ringing oscillator with a trigger click at the front). It’s 1/3 of the snare sound (snare is noise+two ringing oscillators for the “tone” portion). It’s not used in any of the cymbal sounds (all of the cymbals originate from highly filtered, six atonally tuned square-wave oscillators, because the oscillator “hash” has a much more authentic metallic tonality than pure noise). The noise generator is most significantly heard in the hand clap sound.

  11. 1. what is the noise that the transistor made?

    if it was a distortion you could simply over drive a low gain transistor that is not made for high input or something

    2. how was the fault made?

    if the fault was made because of a lack of clean room (manufacturing in dirty conditions) then yes the fault could be re created today because the chinese like to cut corners so get a chinese manufacturer to not use a clean room (manufacture in dirty environment).

  12. Bent Circuit?

    The article title “You Can’t Build A Roland TR-808 Because You Don’t Have Faulty Transistors” would seem like a personal challenge. Has anyone tried damaging an appropriate transistor to see if the characteristics could be made similar.

    The answer might be as simple as, applying a specified over-voltage/over-current or heating them.

  13. I remember a paper from years ago about stressing solid state memory with temperature extremes. The conclusion was it allowed for the ID of bad blocks ahead of time and actually increased the longevity of the remaining sectors. I was thinking something similar to BrendaEM. Push the upper envelope of their operating specs then cryotreat them with a can of air duster while in operation. This process could be automated by mating the components in question with a thermistor that in turn triggers a solenoid that actuates an upside down can of duster aimed at the part. Or even something as simple as alternating between a toaster and the freezer for multiple cycles may do the trick.

  14. Mention is not made of exactly what the problem was with the transistor. This may not come into play, however, one can change the gain of a silicon transistor by reverse biasing the base-emitter junction. The higher the current used and the time over which one does this effect the resulting gain. This is an “old trick” used by radio guys; when restoring radios that had Ge transistors, with much lower gain than new Si devices, they would do this to make a replacement part.

  15. I read some of the posts here that have doubts about that story. That was some years ago and I also was in doubt. But if you really want to know it, I can only encourage you to hear samples of real TR-808s (web is full of it). And THEN: Listen close to any replicas to make sure you hear no difference. That’s what I did and I am sorry for it but: None of them really sounds like the real thing. So I came to the conclusion that this story must be true. Last year I bought a TR-606. Same principle applies here: A selected transistor forms the noise. Service Manual states: ‘2SC9445(P), Selected (NOISE)’. If you’re not a purist and don’t want exactly that sound you probably might use any other type. It’s as simple as that. But then, why did you land on this page, to begin with? Let’s not forget: Noise can be never be really observed in detail by the use of a spectral analyzer. It would roughly get you the levels of measured frequency components but it won’t tell you about the density and pattern structure of the it. For me it was more revealing to watch and interpret the waveform. This showed me: You can’t just take any noise and try to match it’s spectral image to that of the sample. It’s sad but: Not even Roland managed to remodel it good enough with their ‘ACB’ technology.

  16. The simplest way to replace the “faulty transistors”, for me, is taking a sample (at least ten seconds long) from the output of the white noise amplifier circuit on the original TR 808 PCB board;
    In this way a little PCB with a uController both with a small 16-bit converter, playing a loop of that sample, can perfectly replace the “faulty transistor” in any replica of the original analog circuitry of the TR 808 …
    I got a real 808, and it sound really wonderful, so I’m sure that nothing can sound like the “real thing”, except a perfect replica of the original TR 808 analog circuitry, even with a noise sample at the place of the “faulty transistor” …

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.