It’s Raining Germanium!

I just bought 500 germanium transistors. Yes, as a matter of fact, I am insane. Why do you ask?

I just bought 500 germanium transistors.
(Yes, as a matter of fact, I am insane. Why do you ask?)

UPDATE: My Dunlop Fuzz Face Mini review is live at Premier Guitar. Audio clips included!

How and when did it get so frickin’ easy to procure great-sounding germanium transistors?

I’ve been building stompboxes for four years or so. I used to consume article after article detailing the sheer horror of dealing with germanium. Sure, those old-school transistors sound great, I’d read, and they’re necessary for vintage distortion circuits. But they’re unstable. They’re expensive. They’re hard to find. You have to sort through dozens to find the few good ones. And once you do, you must spend countless hours matching and biasing them for optimal sound.

I believed everything I read — until I finally admitted to myself that I seldom encountered any of those problems.

(If you don’t know much about germanium transistors and why they’re cool, here’s my manifesto.)

I used to buy germanium transistors from Small Bear and other parts sites, and was always happy with the results, even though I had to pay eight or ten bucks per transistor. (Small Bear even does the matching for you, offering sets of transistors suitable for various vintage fuzz circuits.) However, it was a little tough finding NPN (negative-ground) germanium transistors. In fact, Small Bear once rejected my order of a dozen or so NPNs because they were so scarce. (To his credit, Small Bear’s Steve Daniels explained that he restricted sales so that everyone who wanted to build a couple of great DIY fuzzes would have the opportunity.)

The workaround is to build pesky positive-ground pedals, or jigger with the schematic in order to use PNP (positive-ground) transistors in negative-ground circuits. (This site’s Fiendmaster project is an example of the latter workaround.)

Then some odd things happened:

The new Dunlop Germanium Fuzz Face Mini has brand-new germanium transistors. It sounds great and only costs $99.

The new Dunlop Germanium Fuzz Face Mini has brand-new germanium transistors. It sounds great and only costs $99..

First, they’ve started making germanium transistors again, though I haven’t managed to figure out exactly who “they” are. BYOC has been selling fuzz kits with new transistors from New Jersey Semiconductor, who have an exceedingly uninformative website. The guys from Mammoth told me their germanium transistors are new-production, though they wouldn’t or couldn’t tell me more than that. Dunlop is reportedly making new germanium transistors for their Fuzz Faces. (I just wrote a review of the $99 Dunlop Germanium Fuzz Face Mini for Premier Guitar. It sounds fabulous, and I’ll link to my review when it goes live.) In each case, these transistors work great, and NPNs seem as plentiful as PNPs. [UPDATE: Nope, I was misinformed. Dunlop says they are using strictly NOS transistors.]

Meanwhile, a friend shared a couple of online sources for cheap germanium transistors, but only after extracting a promise not to share the info. (Sorry! I know how obnoxious that is.) But with a little snooping, you can find them too. As with most electronic parts, they’re cheaper in bulk. I was buying PNP AC-128s for two bucks a pop, and used those for all the pedals in Fuzz Detective, my mondo germanium fuzz comparison video. After that, I ordered 100 NPN AC-187s for a buck each. It was such a good deal that, on a lark, I asked how much 500 would cost. Answer: a little over $200. That’s not much more than the price of four Boss DS-1s! And that’s how I acquired the 500 transistors pictured above.

But get this: They’re not new, but NOS (new-old stock). So how the hell can once company sell me seemingly unlimited NPN transistors for forty cents each, while Small Bear declined my offer to buy the equivalent for ten bucks each?

I don't care what the hFE meter says — this three-transistor fuzz sounds AMAZING with hot AC-187s!

I don’t care what the “experts” says — this three-transistor fuzz sounds amazing with super-hot AC-187s!

All those articles I’d read all stressed how essential it is to measure transistors, using only the ones whose gain levels fall within specific ranges. I bought a relatively expensive digital multimeter with an hFE function (hFE is the measurement unit for transistor gain) and dutifully test every transistor. But over time, I’ve found that a) the vast majority of transistors I test are just fine, and b) the “rules” about optimal hFE levels for classic fuzz circuits often aren’t true. Or rather, transistors whose gain levels fall outside the specified ranges can sound great.

Take my mountain of AC-187s, for example. The AC-187 is a loud NPN transistor whose hFE values check in between 300 and 500, a gain level widely considered too hot for Rangemasters, Fuzz Faces, Tone Benders, and their derivatives. I don’t know enough to challenge that statement on technical grounds, but I will on musical grounds: The stuff I’m building sounds frickin’ awesome. Germanium transistors are so dynamic that rolling back the input level (either from the guitar’s volume, or via a pre-gain pot) produces sounds equivalent to transistors with lower hFE values. I find I can get great vintage tones with the more gain on tap if needed. So I call b.s. on the conventional wisdom here.

(I suspect those “optimal” values came about from measuring the transistors in particularly great-sounding vintage pedals. The pedals in question probably do sound great — but they’d also sound great with many other values.)

So where do these new-production germanium transistors come from? And why is it suddenly so easy to find good, cheap old ones? Can anyone give me a clue? If you do, I’ll mail you an envelope full of enough cheap-ass rare, expensive germanium transistors to build a couple of bitchin’ fuzz pedals.

28 comments to It’s Raining Germanium!

  • Ben

    THOMASNET lists 57 germanium transistor suppliers, but I wouldn’t know how to verify whether any of them are causing the sudden availability: http://www.thomasnet.com/products/germanium-transistors-87352209-1.html

  • Oinkus

    If Dunlop is making them for Fuzz Faces there might be some other explanation besides China ? Remember if people want to buy things someone will supply whatever is in demand. I will go back to searching now but I am pretty sure I already will get the one in the Fiendmaster I just ordered .

  • bear

    You drive me a little crazy every time you tease your source, Joe. I’m pretty sure my Google skills still haven’t turned up your source yet. I have found a good candidate for under $2 a pop, though.

  • smgear

    great score!! I’m guessing more germ-based projects coming up.. )

  • Oinkus

    First one that shows up is on Amazon AMS and they go for 4 to 13 bucks each ,I haven’t found any bulk deals yet. The BYOC Octave Fuzz I built sounds great too just don’t really have a use for it.

  • Jermaine Eyum

    Interesting situation. Whoever is making them perceives a growing market. China’s production of raw Ge is 10x Russia and US combined. Making transistors requires expensive equipment, less so than CPUs so probably existing fabrication line was turned back on. I’d be really surprised if Dunlop or any pedal mfr actually was making them.

    What I’d like to see is a spec sheet that is circa 2013 rather than 1958. If they can control specs better now I’d like to see parts with tighter specs. Unusual market for DIY which will accept almost anything without complaint.

    Where’s Colombo when you need him?

  • Shizmab Abaye

    You can get lower stage gain when using higher hFE in a grounded-emitter design just by reducing the collector resistor. Make sure you don’t exceed the collector current rating though.

      • Shizmab Abaye

        To give a more concrete example, supposing you are designing a stage with target hFE of 150 but you have hFE = 300. The original design has a 12k collector resistor. Use a 6k resistor instead. Supposing a 9V supply, the max current that would ever go through the collector would be 9/6000 = 0.15 mA.

        Or you can leave everything as it was and use your volume knob. I’m not trying to tell you what to do (lie).

  • Shizmab Abaye

    (cough) let me see, I should have offset everything by my BAC (what can I say, I’m on vacation)…

    Instead of 6k, I’m sure a 5.6k would be fine, and 9/6000 = 1.5 mA.

  • bear

    (hFE issues are now screwing with me on the silicon side. I just got an order of 50 2N3904’s hoping to get some in the 200 range and maybe some 100-ish. Most of them are in the 360-375 range, only a couple barely below 350. Drag.)

  • Perhaps there are parallels with what happened with tubes. Western manufacturing stopped and prices for NOS started rising so companies like Groove Tubes found places in Russia and China that could still make tubes.

    It seems to me that the biggest problem with germanium transistors and with NOS NPN transistors in particular, is leakage (and contrary to some opinions leakage does not contribute to the unique sound of germanium and is never desirable). All the NOS NPN germanium I have seen leak like a sieve. Leakage current completely screws up the DC bias of circuits like the Fuzz Face. Also simple transistor testers (such as those found on multimeters) cannot tell the difference between collector current and leakage current so they show transistors with high leakage as having high gain. I wonder if that is some of what you are seeing Joe?

    • joe

      Hey Terry — thanks for the info.

      Since you seem to know what you’re talking about, let me pick your brain: Does inserting a transistor backward in an hFE tester provide any sort of leakage measurement? Typically, a germanium transistor that measures around 350 might display 25 or so when tester the wrong way around.

      I don’t doubt your statement that these components leak like sieves — but what does that really mean sonically? Almost all the germanium transistors I’m getting these days sound fabulous in fuzz circuits. And the more of them I build, the more I suspicious I become of much tech advice that the experts post on various DIY forums. More specifically, I don’t doubt their claims on a technical level, and probably wouldn’t even if I was smart enough to fully understand them. But it just doesn’t seem all that relevant to how the circuits sound. In particular, statements about optimal hFE levels for various vintage circuits seem way too narrow. If I’m using leaky transistors with the “wrong” values, why does the stuff I build sound so frickin’ good?

      There’s something of a cultural divide here between scientifically ignorant players (like me) and folks coming into the DIY process from an engineering perspective. I don’t doubt their science skills — just their musical expertise.

      Oh crap — now I’ve probably gone and picked another fight. ;)

  • Hi Joe,

    Well I guess if by inserting a transistor in a multimeter Hfe tester reversed you mean with the base the same and the emitter and collector reversed then it will just try and operate as a low efficiency transistor and you will probably see a low ‘Hfe’. You would be better off testing the right way round with the base lead disconnected.

    As far as I know the majority of multimeter Hfe testers have a resistor to supply base current and a resistor connected to the collector which goes to the supply (9V). The emitter of the transistor goes to the other supply rail. This set-up is either one way up or the other depending on whether the test is for PNP or NPN. The meter reads the voltage drop across the collector resistor which is the product of the collector current and the resistor value. The resistor values are chosen such that if leakage current is very small then the true gain of the transistor will be displayed as a ratio of the base current to the collector current. Any leakage current will simply be added to the collector current flowing as a result of the base current. If the leakage current is large the meter shows a false value for gain.

    I’m sure you must be familiar with R.G. Keen’s web page The technology of the Fuzz Face. He shows a way to test transistors there and to check leakage. Essentially his test circuit mirrors what is inside those multimeter that have a transistor test socket, but places it outside the meter and allows you to disconnect the base and measure the leakage current that then flows between the emitter and the collector with the base floating. I use a PEAK DCA 55 digital tester that makes a separate measurement for leakage and then compensates the gain measurement for leakage.

    As an example I have a pair of AC176 transistors supplied to me for a Fuzz Face. If I measure these on my multimeter I get Hfe’s of 406 and 419 (a suspiciously high value for these transistors). In my PEAK tester I get 94 with 0.6mA leakage and 72 with a leakage of 0.64mA. I tried to use these in a Fuzz Face and although I got some signal it did not sound very good. The Fuzz Face did have a trimmer, but I could not set a sensible bias point at the collector of the second transistor. I got a bias adjust range of from 9.25 (the battery voltage) to about 6V. I would expect to be able to set bias to at least somewhere near 3 to 4.5V (relative to signal ground).

    Incidentally when measuring germanium transistors – they are incredibly temperature sensitve, just holding them between the fingers to plug them in to a tester is enough to warm them and it then takes a couple of minutes for the readings to settle down.

    Since then I have for the moment given up on having a negative ground germanium Fuzz Face, put sockets in the build, and fitted a low gain silicon (70) in the first position and a high gain (900) in the second, so I don’t disagree with you that higher gains can sound good. This thing now sounds phenomenal! I think all the emphasis on particular gains is partly to do with getting a sensible DC bias in a circuit with all fixed resistors and partly to do with obtaining a particular range of fuzz sounds with good interaction with the guitar controls.

    When you put leaky transistors in a circuit it is a bit like having an extra resistor (with a value that can vary a lot depending on the leakage) wired between the emitter and the collector. This does not seem desirable to me. In bias critical circuits like the Fuzz Face this can completely upset the DC bias conditions and this is probably one of the reasons why the original Fuzz Faces were so variable. Modern clones of the Fuzz Face often replace the 8.2K collector resistor for the second transistor with a 47K trimmer. This gives a very wide range of adjustment to bring the DC bias to somewhere sensible and probably allows these pedals to run with transistors that are quite leaky.

    Sorry about the long reply…

  • Further – I tried lifting the base connection with my multimeter Hfe tester and as far as I can see the number then displayed is proportional to leakage but it is not a direct measure of leakage in mA or uA.

  • joe

    Terry, thanks SO much for so much info. I owe you, man! :beer:

    Yeah, and seeing a lot of measurements around 400hFE. However, I’m finding they sound great in just about every circuit I try! I wonder why your Fuzz Face didn’t sound very good.

  • Well the AC176’s I have are spec’d at an Hfe of between 50 to 180. So for me to see an Hfe of 400 odd on my multimeter tester means they are very leaky, which I know to be the case because of my measurements with my PEAK tester. Your AC187’s on the other hand are specified at 100 to 500, so you may be seeing a true Hfe and many of your transistors may be low leakage. Just measuring Hfe like that on a multimeter you are only getting half the story. I really recommend you build RG Keens test rig and see what your leakages are. Although that would be more for my benefit than yours :smirk: I would just like to know what leakages you are getting. As long as you can build pedals that sound great with the transistors you have everything is cool, right?

    The Fuzz Face, and similar fuzz circuits, are designed as saturating amplifiers. It has enough gain for the guitar signal to push the output signal swing up to the supply rails to produce clipping. Nobody now knows if the original designer had an ideal operating point in mind when they designed it, but it is reasonable to expect the collector of the second transistor to be biased somewhere near half the supply (9/2 = 4.5V) which (in a negative ground Fuzz face) allows the output signal plenty of room to swing up to +9V and down to 0V (actually not quite to 0V, there is a small voltage drop across the transistor). Biased to 4.5V a Fuzz Face tends to produce square waves. You can get away with an output bias that is away from the centre point where you get some asymmetry in the clipping, but go too far and you start to get ‘spitting’ as notes die away and the input signal falls below the bias threshold.

    Excessive leakage in the transistors upsets the D.C. conditions in the circuit so that the bias point can be too close to the supply rails, there is no room for signal swing, and you get little to no signal with a lot of spitting.

  • joe

    Yup, I’m going to add that tester project to my to do list. :)

  • You can get a rough indication of transistor leakage using the resistance ranges on a multimeter.

    Switch to the 20K range. For a PNP transistor connect the red lead to the emitter and the black lead to the collector (assuming the red lead is positive and the black lead is negative for resistance measurements, most meters are arranged this way). Reverse that for an NPN. The transistor base lead should not be connected to anything.

    If the meter shows the resistance is too large to measure on the 20K range then your transistor has very low leakage. If you see only a few 1,000 ohms then your transistor is quite leaky. This is a crude method but it can be used to sort low leakage parts from high. I have some PNP transistors that show over range with this method and better than 0.00mA on my PEAK DCA 55.

  • I have been running some circuit simulations to try and answer the question – how much do transistor gain and leakage variations matter to the audio performance of the Fuzz Face circuit.

    I’m going to make the assumption that the ideal bias point at the collector of the second transistor is close to 4.5V and that the ideal clipped output is a square wave with a 1:1 mark space ratio. The input signal for these sims is 100mV peak to peak. All the component values are those for a standard germanium version of the circuit.

    So starting off with both transistors with the same Hfe = 85 and NO leakage the bias point is 4.602V and the mark space is 1.17

    With 100uA of leakage for the first transistor the bias point is 5.842V and the mark space is 1.21

    With no leakage and the second transistor Hfe increased to 130 the bias point is 4.559V and the mark to space is 1.18

    This is obviously a very limited set of data, but it may give some idea of how the circuit is affected by transistor gain variations and by leakage current.
    I think you might find it difficult to distinguish between the first set of conditions and the third. The second should be distinguishable from the other two in an A to B listening test.

  • Really misinformed article in a lot of ways.

    1. Dunlop is using NOS PNP trannies in their Ge fuzzes. They most likely come from New Jersey SemiConductor, but to tell you more about that relationship, I would have to kill you. They swore me to secrecy.

    2. To my knowledge, nobody is making usable new Ge transistors. There are a few companies making legacy parts for airplanes and military vehicles, but those are the two leg power transistors.

    3. DSI in Europe, though most likely made in China, also makes new Germanium transistors, but they are 15 hfe on average, and thus too low for a perfect fuzz face.

    4. I have contacted ALL the semi conductor peeps in China, and none of them could deliver me new Ge transistors with the hfe and leakage we need to make great fuzzes.

    5. While there are a plethora of both NPN and PNP Ge transistors on ebay and other online markets, you still need to sift through 10 transistors to find 3 good ones (DCA55 Atlas makes this quicker).

    6. Thanks to all the hype created by article like this one (and myriad others), the demand for ge transistors is so high that all the pedal builders bought the best ge stuff, and we are all left to glean what we can from what is left. They are out there, but after you get a useable pair, there will be no room left to sell said pedal for a profit. In fact, you might as well buy a Dunlop.

    7. Reading this article sent me on a 6 month search for the best NPN germanium transistors out there. Minifux on Ebay has OC139 that are pricey. And, only 3 in 10 will be good, but they are the best I have found. PNP are easier to find in the correct hfe and leakage as NPN seem to be hfe high= leakage high. My local electronics place had some great high gain ge pnp philcos, and I bought them all before you all got them.

    Germanium does sound better in a fuzz face in my opinion, though it depends on your playing style and technique. I have found that NOS lower gain silicon transistors can sound REALLY CLOSE if you get ones in the right hfe and leakage. I have a pair now that sounds just like Ge in that it cleans up well with volume knob, and is very wooly. That said, the lower gain means less sustain and that the fuzz is not fully saturated. There are trade offs.

    No offense to Joe, but just do more research next time. This article gave me false hope. On the bright side, I now know more about germanium transistors then I ever wanted to.

  • Kenny

    Hi guys, I’ve built Fuzz Faces with AC125s, AC128s, MU-25s, and with Si-2N5078s and Si-2N3906s for Q2. A germanium transistor that tests over 250 for hfe on a regular tester like on a cheap DMM is certainly leaky, and will sound crappy in any distortion circuit, especially the FF. The gain number isn’t that important, most Ges test around 100 (unless they are leaky). In my experience almost all Tungsram AC125s are unuseable (they go off the scale, ie, over 1000 on my very good quality beta tester), all the others listed above sound great. Make the collector load for Q2 adjustable for bias point (say a 25k trimpot) and you will not be disappointed after tweaking. It’s also worthwhile to make Q2’s collector load adjustable with an external pot for more nuances. I think the reason so many guitarists don’t like “fuzz” is because they’ve never tried a good one. I don’t like pedals generally, but a good FF is eminently useable especially for lead work.

  • Kenny

    Guys, that’s MP-25, not MU-25. ooops

    • joe

      Thanks for the info, Kenny!

      I’ve read stats like these before, and just have to add that I’ve made many fine-sounding pedals using germanium transistors that lie outside those supposedly optimal ranges. (Ultra-low-gain transistors are especially effective as Q2 in a Fuzz Face.) My advice is, be aware of such guidelines, but listen and judge for yourself.

Leave a Reply

  

  

  

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

:1up: :alert: :ban: :beer: :borg: :coffee: :cuckoo: :cuss: :finger: :goombah: :stupid: :megaman: :mad: :pity: :noshake: :oogle: :pacman: :pill: :poison: :poop: :rant: :satansmoking: :shake: :shiftyeyes: :shroom: :sick: :smirk: :spammer: :stfu: :thumbdown: :thumbup: :turtle: :what: :whatever:

Click to upload a JPG