The tyranny of CJ22

[iceowl]

There is a story here involving my building an MOTM VCO 300, working with Paul Schrieber when I failed twice, and then succeeded. Maybe this will help. Here it is.

I buy transistors in lots of 100 from Mouser or Digikey. For devices like 3904s it hardly justifies the postage to buy any less. I build as many modules as I can. I don't want to buy any ready-made. I'm not a pro-musician. This is a hobby for me, and part of the hobby is the building. The journey is the reward, and the joy of modular synthesizers is they never need to be "finished". It can go on for as long as you and your bank account can take it.

Along the way I have built modules from many different designers. I've acquired a bunch of gear from Paul Schreiber/Synthtech. Many of you know Paul, so you will not be surprised to learn that the moment he discovered I was having a problem with my assembly of his VCO, he got on the phone to me - called me directly (TWICE) and spent an hour each time helping me debug the problem.

Which is to say that the first time I assembled the VCO, it didn't work. Now, assembling a module "kit" isn't anywhere near the same as designing a module, nor does it involve a deep understanding of the circuit design, though the latter is useful in debugging. But as we all know, debugging is typically a part of the module build process, and when you have to debug, having the designer at hand is a huge plus, and I had Paul on the phone to me running me through DC operating points, and levels, and theory. A rather massive advantage.

Yet, after a month of debugging, my first VCO did not work. Failing everything we could think of, Paul sent me a new board, and some key components, basically for the price of postage. I ordered a new set of parts from Mouser/Digikey, and built the second board. And it failed to function in exactly the same mode as the first.

Paul spent another hour on the phone with me - called me directly out of the blue and we went through the thing, nearly component by component. We thought we had an answer in the age/functionality of a particular somewhat obsolete IC used in the design. In fact, Paul had sent me some of the ones he had on his bench (they are still in transit in the US Mail). Meanwhile, I bought 10s of the things from Mouser/Digikey/local supply. I now have several hundred LM319s. I only needed one functional one.

Turns out that was not the problem.

Out of sheer desperation, after nearly 3 months of debugging, I purchased a device by BK Precision called a 510 semiconductor checker. This is a monstrously expensive device (read $500) that performs a function Paul assures me is available on a $50 Radio Shack DVM. He knows this because he holds the patent on the transistor checking circuitry they use.

That device showed up last evening. One feature of the device, the marketing blather promises, is that it works perfectly well on devices in circuit. I proceeded to use it on the discrete transistors in the design of the VCO 300 that were planted on my second VCO300 board. My objective was to determine if any of the transistors were somehow "rotten".

The 2n4403s all checked out. The 2n3904s also checked out - but with a weird anomaly for some reason the "checker" wanted to tell me the pinout on the device was ECB instead of EBC.

Immediately, many of you who have read this far emit an "ah ha!" of recognition. Being naive, I did not.


Now rarely when I insert a part into a board, do I look beyond the first couple of identifying numbers on the part. My 2n3904s are indeed 2n3904s. The datasheets I have for them specify a pinout of emitter-base-collector when looking at the "face". All the datasheets. ST. On. Ti. Etc. These datasheets come from the suppliers I send money to - Mouser and Digikey.

Yet I had this little expensive device (that I paid 10x too much for, according to Paul) telling me the pinout on the 2n3904 on my board was ECB, and not EBC, as all the datasheets promise.

I went to my transistor box and pulled out a couple 2n3904s and tested them in isolation. The tests all showed the pinout was indeed EBC. So then one imagines that somehow the attending circuitry on the board fools the checker into misidentifying the pinout. Of course, that would invalidate the entire reason for the price of molten evil they exact for their foul offering, because if it cannot distinguish the parts in circuit - why would any rational, breathing being expend so much filthy lucre for their spawn, especially during a world wide recession when there are children starving on the streets of Beverly Hills?

But then, because I was at my wit's end - some muse of electrical engineering caused me to actually care about the OTHER numbers on the transistor's face. The FULL name of the transistors in the holes on the VCO 300 board was: 2n3904-CJ22

The full name of the transistors in my box of 3904s was: 2n3904-J04. And because I buy these things in lots of 100 - and because I have several hundred, bought on different occasions, there are lots of both "types" in my box of transistors.

Could there possibly be a difference between J04 and CJ22 devices? EVen the data sheets say the digits are nothing more than a manufacture date code.

Because I had nothing else life-affirming to do last evening, I rummaged through my pile of transistors and managed to find, like needles in a haystack, a couple bare 2n3904-CJ22s amid nundreds of 2n3904-J04s

And, as you now expect having read this far - the pinout on the CJ22 devices checked out, on my expensive checker - to be Emitter-Collector-Base, where the ones labeled J04 were Emitter-Base-Collector.

I pulled the 2n3904-CJ22s off the VCO board, replaced them with their J04 cousins, and a three month debugging nightmare ended. The VCO works. The Semiconductor Checker earned its expensive keep. Now I am impoverished and oscillating.

This is truly a Pyrrhic victory. How come I didn't catch the different pinout, which is probably obvious to those not in North America? How come I ordered the parts from Mouser and downloaded the datasheet and it assured me the pinout was ECB? How, for the love of God, in the twenty first century could I spend 3 months of my synth building time, two whole hours of Paul's time, and hundreds of dollars - nay - THOUSANDS of dollars, including all the times I replaced parts that didn't need replacing, buying an analog oscilloscope because I thought there was an off-chance my trust worthly Fluke scopemeter somehow could not withstand the virulent scrutiny of sensitive analog design - because I swapped out all the ICs and expensive MAT02 transistors many times, never presuming that I could be betrayed by the reptilian treachery I endured having been offered a theoretically reliable, rock solid, sanctified datasheet that spewed nothing but technical effluent, tearing valuable hours from my ever shrinking lifespan and tarnishing forever my faith in internet commerce. The pig fu**ers should be fried in their own liver fat.

Let this be a lesson to all of us.

[davemoog]

I know what it feels like, yup, had a similar problem with 2N5910 for ARP VCO's. I bought a batch from a reliable supplier and every piece that we both had was NPN not PNP...When stuffing a board I always check base orientation with a simple multi-meter check but missed / didn't expect the wrong polarity. This took way too long to figure out. ('') Ended up using a very cheap, available and later substitute that ARP recommended - the 2N4125

[frijitz]

Wow, what a nightmare! I don't see the CJ22 designator either at Digi or Mouser. Do you know what PN you ordered?

It's puzzling, though, that the problem didn't pop out at you (or Paul) during analysis/troubleshooting. I'm not seeing offhand how switching C and B in a circuit couldn't lead to unreasonable voltage levels.

Glad you tracked it down, though. Sounds like a reall killer.



Ian

[iceowl]

[Bodega]

That's the problem with DIY - you do spend a lot of time tearing your own hair out, especially at first.

It's worth it, I'd say. But still, lots of hair-tearing.

When you're trying to actually write a song, there's nothing more distracting than a filter that won't self-oscillate when it should be, for example.

[iceowl]

Ok, I have done my research and I see the fault is entirely - utterly - absolutely - completely my own.

Mea culpa, as they say in Latin.

Here's the mouser part number 512-2N3904CTA

If you check out that part and read the not-so-fine-print, you will see it clearly states, "Collector middle pin" in the "description", which I did not read. I only looked at the part number, the cost, and the delivery mechanism when I ordered.

Now I will slurk out the back door in utter shame. My public apologies to Paul Schreiber. This sort of thing is why DIY is a time-intensive business for designers like him. I need to stop being right about things, and start checking more.

[frijitz]

Wow, I never even knew there was such a thing, at least for the 2N3904.

Troubleshooting a VCO can be very frustrating, since the circuit is a loop, and something amiss anywhere in the loop can kill the oscillations. I had a frustrating couple of days getting my new TZ VCO circuit to run at all. It has two separate discharge circuits that were interacting some with each other.

A good general principle for troubleshooting is to isolate individual sections of the circuit to test them separately. If you had tried the osc with the sync tranny pulled it would have run, no?



Ian

[iceowl]

[frijitz]