Newly discovered Neovibe failure mode

[R.G.]

One of our members had a problem with a 'vibe build, and it proved impervious to remote debugging. So I asked if he could mail it to me for some hands-on work. He did, and I did, and I have found a new problem. I'm documenting it here for the assistance of someone else who hits it when I'm not around.

His problem was that the vibe, although being competently built, had an ugly distortion on signals over some level. Having eliminated the obvious issues remotely, I suspected oscillation, which he was not equipped to cope with.

In short,  yep, it was oscillating at about 3MHz. Here are some observed symptoms.
- distortion above some fairly well defined audio level, like with humbuckers or fully-up single coils.
- distortion seems to start in the first three transistors, and is audible on transistors all the way down the audio path
- seems to quit this whenever my fat finger is on any component in the first three transistor section

The oscillation wiggled in frequency in time with the light pulsing. Aha! Power supply. I checked on the 78L15, and sure enough tiny fuzz on the +15V line. That vanished when I put a 0.1uF ceramic cap on the input-ground leads of the regulator. Some three terminal regulators are unstable if there is not good decoupling right at their input terminals.

Now it took more signal level to make it start, but start it did. C1 seemed loose, and touching it made the oscilation quit. New C1. No change.

Ah. Power supply impedance at the input preamp. I hand-held a 0.1uF ceramic cap across +15 and ground near Q2, and the oscillation vanished. Got it. Soldered the cap in, fired it back up and ... much better, but still oscillates some. Now I'm frustrated. The compensation cap (330pF) is the correct value; removing it has no effect.

I then took out my cutters and cut all three 2N5088 transistors out, replacing them with 2N5088's from my own stock. No oscillation, no matter what I do.

I think that what really fixed it was the change of brand of transistors. The builder had put in a new set of his own transistors, but obviously the were the same lot as the first three.

So -

1. Neovibes can oscillate under some fairly rare circumstances. I do not believe this is predictable based on transistor type number or manufacturer, so don't go flooding the forum with "what type number/brand/year/ vintage, etc etc is the good ones?" And please don't create one more internet rumor or myth that only X, Y, or Z transistors work in a vibe.
2. It can be the regulator. Fix with 0.1uF monolithic ceramic cap at the regulator pins.
3. It can be general power supply decoupling at the preamp. Fix by adding a 0.1uF monolithic ceramic  across the power/ground at the preamp.
4. It can be the characteristics of the three input transistors, and maybe only one of them. Fix by getting a different type number or different brand of the input transistors.

At one time this would have mystified me more than it does now. It is possible that counterfeiting parts has reached down into the sub-jellybean transistors. But don't overreact. This is rare. 99.999% of the problems are still parts placement, soldering, wiring, etc. This is after all the first of these I've seen since the first vibe boards back in the 90s, so it's notable primarily because it is rare.

I'll put spaces for the extra decoupling caps on the board if I ever update it again.

[Paul Perry (Frostwave)]

I don't think it's a counterfeit.
The phenomenon of the odd transistor taking off at RF is not all that uncommon, especially when you have those little NPNs that have a gain of almost a thousand.
Putting a ferrite bead on the emitter leg is said to stop it, but it is so rare it isn't worth doing until you know it is happening - and then a different transistor is easier.
I have had it happen two or three times in my products - very rare, and very puzzling. Naturally, it stops the instant you put a CRO probe near it.

[BDuguay]

I'm sure the owner thanked you for helping out and I too would like to thank you for being such a cool guy and for documenting this for our benefit.
I plan to build the Neovibe someday, soon I hope!
B.

[Mark Hammer]

I second Brian's motion.  Let it be known that what makes this forum the jewel it is, is the manner in which people sort out problems, do the legwork, and THEN publicly provide that knowledge for others in a manner that enables them to be more self-sufficient.  Kudos and thanks.

[yodude]

Here I am. You can read the debug thread here:
http://www.diystompboxes.com/smfforum/index.php?topic=69918.0
The thread currently has clips at effect out and internally with the audio probe. I'll update with clips of the corrected product as soon as I can.

I'm very grateful to RG not just for going way above and beyond on this pedal, but for all the great info he's put out there over the years. And that appreciation goes out to all the contributors to the forum, even those whose main contributions are coming up with super bugs for the heavy hitters to scratch their heads over.

Rock on.

[amz-fx]

I then took out my cutters and cut all three 2N5088 transistors out, replacing them with 2N5088's from my own stock. No oscillation, no matter what I do.

Interesting...  these were new Fairchild 2N5088 transistors in it originally?  You replaced them with some old 1985 vintage devices you had in your stash?

Fairchild apparently has changed how they make bipolar transistors in the last few years...  the new method uses a different process that allows them to pack more transistors per square inch (therefore, saving money for the MBAs)...  the new process transistors are more noisy (definitely) and maybe less stable (conjecture). I hear the same has happened at ON Semi who also makes this device.

Is there a date code on them that you could check?

regards, Jack

[R.G.]

Interesting...  these were new Fairchild 2N5088 transistors in it originally?  You replaced them with some old 1985 vintage devices you had in your stash?

Right. Only NOS vintage 2N5088s will give you the smooth, creamy tones we want. Modern 2N5088's won't do that because they don't make them like that any more...

<sarcasm>Right.</sarcasm>
What a lot of nonsense. That's why I went into the tirade I did about not getting nutso about the transistors. Defective transistors are always possible, and counterfeit electronics even more so. Many manufacturers will not accept semiconductors for manufacture that cannot show a provenance trail that an expensive painting would be proud of for just this reason.

Fairchild apparently has changed how they make bipolar transistors in the last few years...  the new method uses a different process that allows them to pack more transistors per square inch (therefore, saving money for the MBAs)...  the new process transistors are more noisy (definitely) and maybe less stable (conjecture). I hear the same has happened at ON Semi who also makes this device.

Could be. I always assume that whomever makes a type number will have made it by using their then-current semiconductor process to turn out stuff that is within the datasheet for the type. There is no Standards Guardian of the Type Numbers, at least for non-military stuff. Very much like Bill Clinton's sworn testimony, a 2N5088 is whatever its maker says it is - and that includes the unspoken caveat "at the moment I say it..."

Most people don't realize what making semiconductors today entails. Companies that are serious about it will build a new fab(rication) facility for a few billion (!) dollars, then tweak it into production. It is so expensive to set up a process line and keep it running that all semiconductors that can conceivably be run in that Process (capitalization intended) will be run so.

So what has happened in all makers of sub-jellybean stuff like transistors is that the big makers like Fairchild and On Semi will make everything on the then-current Fab and Process. They can't not do so. Most big makers simply quit making low cost stuff. The MBAs have figured out that there is so little margin in making transistors that they would rather shut down the Fab than go into the transistor market. Fairchild and On Semi are definitely mavericks in this regard. Maybe they have old sub-fabs they're squeezing. Maybe the change has been that they can't keep the old sub-fabs going. The diode, transistor, analog, low- and medium-scale logic chips are referred to in the industry as "the dog-food market".

Generally, transistors are either made by subverting the existing Fabs, or running them on old equipment. The old equipment is usually sold or run by a smaller company that accepts the lower rate of return, or thinks they can squeeze the market somehow. This last is what Germanium Power Devices does (or did - I haven't called them in a few years). The Fab hand-me-down market is where a lot of transistors come from.

That being the case, yes, Fairchild probably has changed the way they make 2N5088's in the last few years. I'm not surprised at all.

Is there a date code on them that you could check?

Sorry. I did exhume the bodies from the trash can and check. I don't think they're Fairchild. I'm pretty familiar with Fairchild's markings and I couldn't pervert what was there into the normal Fairchild markings.

At this point, I think the thing to do is just note "Huh. Bum transistors. That doesn't happen often."

It's the old adage - once is an accident, twice is a coincidence, three times is enemy action.

[amz-fx]

The difference in performance of old process Fairchild transistors and the new process was discovered by the DIY microphone guys who were doing the Scott Dorsey mod to the Octava mic, which is an even more noise critical application than guitar pedals. The new fab 2N5087s were much more noisy than the old process devices made pre-2005.  Scott reported this info in RECORDING magazine, Aug 2008 issue, page 6, in case anyone want to get more info. Apparently the old process devices are still available from Digi-Key by special order.

Does this have any impact on the Neo-vibe?  I don't know, but I think it is a legitimate matter to investigate, or at least be aware of it.

Bets regards, Jack

[R.G.]

Yeah, in spite of what an effect-centric world might think, guitar pedals are not all that noise sensitive.

That issue seems to be Fairchild specific. A bigger issue is what process is used by whomever provided whatever transistor one happens to be using. Some of the Taiwanese providers are excellent.

In this particular instance, I was getting gain-phase oscillation that could not be easily removed by the 330pF single-pole compensation cap on the collector of Q2. Doubling the cap didn't remove it, oddly enough. I actually think that one of the three transistors had either extra high gain or extra high Ccb, or both. Could have been a marginal BVceo, I guess. But now that I'm working only with equipment I can buy myself and don't have access to a corporate-funded semiconductor testing lab, I really have no good way of nailing it down any further. I suppose I could have cherry-picked the devices to figure out which one it was, but even then I could not have burrowed deeply enough into the silicon to figure out exactly why.

Ah, well. Someday, a surplus wafer-prober will show up on ebay... 

[Skreddy]

1. Wouldn't paralleling a small ceramic cap (say 50pf) with R4 (negative feedback) also be able clamp down any rf in the input stage even if the transistors themselves were responsible for it?

2. I also prefer older transistors in audio paths. 

[Eb7+9]

I actually think that one of the three transistors had either extra high gain or extra high Ccb, or both. Could have been a marginal BVceo, I guess.

you're dreaming ... only thing that can cause parasitic oscillation in that circuit is an improperly filtered PSU and poor grounding ... an extra high Ccb would dull oscillations if anything - high-Beta devices don't cause problems in this circuit because the signal path is roughly unity gain throughout ...

[R.G.]

you're dreaming ... only thing that can cause parasitic oscillation in that circuit is an improperly filtered PSU and poor grounding ... an extra high Ccb would dull oscillations if anything - high-Beta devices don't cause problems in this circuit because the signal path is roughly unity gain throughout ...

Oh, hi, J.C. Glad  you could drop by.

Unfortunately, I believe that you're dreaming that I'm dreaming. Here's why:
1. Unity gain is no preventer of oscillation. One reason power MOSFETs need gate stoppers is that they are prone to oscillation, perhaps especially in the source follower setup. Do a little reading. If that's too much trouble, let me know and I'll find some references for you.
2. Take a look at the LFO in the ... um, univibe. (note correct use of ellipses!) That's a darlington emitter follower, no? So howzat oscillating if oscillation is impossible in unity gain setups?

Take your time. I'll tell you in a minute.

3. The first three transistors are NOT in a unity gain setup. They form what amounts to a discrete opamp for AC signals, with two, count them two feedback loops, one of which is neutered down to almost DC by a bypass cap, that being the one from emitter of Q2 to the base of Q1. The AC signal loop is the feedback from R8 and R9 to the emitter of Q1. Let's count inversions. From base of Q1 to collector of Q1 is one invert. From Q1c and Q2b to Q2c is two inverts. From Q2c/Q3b to Q3C is three inverts. Only the signal comes off Q3e, which is only two inverts. That would be positive feedback if we went back to Q1b. But it goes instead to Q1e, which is ... one invert! Yay! We have negative feedback and Mother Nature is happy.

But what's the open loop gain? It's the forward gain through Q1 and Q2 to Q2c, and that's then buffered by Q3 for feedback purposes. I can whip out the simulator and do an example circuit for an approximate gain faster than I can do the math these days, but I promise that the forward voltage gain is high; it's high enough that R8/R9 cause the correct forward AC gain, so we know the G = (alpha*beta)/(1+alpha*beta) approximation is satisfied (alpha and beta being their control systems meanings, not transistor gains). Think about it, work a little on it from the hybrid-pi model standpoint, and I'm sure you'll see it.

And I did promise to tell you how oscillators with unity or less voltage gain oscillate.

The voltage gain doesn't matter. It's the POWER gain that actually determines whether a circuit can oscillate. Audio-only people are so accustomed to thinking only in terms of voltage that it's hard to see, but the 'vibe LFO is a good example. It has about unity voltage gain, but a huge current gain. The current gain is enough that the phase-shifted signal through the impedances back to the transistor bases convert the excess current gain into enough voltage gain to keep it going. 'Member my post about the sensitivity of that point to loading? That's why it's sensitive - too much loading can kill the necessary power conversion back to the input to keep oscillation happening.

Guys with RF experience are well aware that any kind of tuned circuit on an emitter can make a single transistor oscillate.

Yell if you need help deciphering my Techoese into English. I'll be glad to help if I can.

[Eb7+9]

burp, nice dose of technotopia ... the newer transistors likely have lower parasitics by virtue of their small size, so smaller Ccb even ... regardless, the problem you're having lies in the interaction between your NFB based 7815 regulator that, on the one hand, isn't well supported by large enough caps AND the fact that it's feeding a single-ended circuit - which more easily allows a circulation of unwanted signal readily through the supply loop than an op-amp circuit would ... these regulators are known to have oscillation issues, not too well documented in the app. notes, a big reason why lots of audio guys avoid them ... with a low draw bulb (Steve Daniel carries the ones I recommend) a 7815 is passable as long as it's well filtered and the grounding BETWEEN BOTH SIDES OF THE CIRCUIT is full star all the way to the regulator ground pin, and there's another condition that I'll leave you to find with regard to the 78XX guys ...

I get a fair bit of emails concerning a grounding problem you have on your board or in your wiring - it manifests itself as a hum that creeps into the signal path when the bulb goes brite - part of it is poor filtering around your 7815 as you're discovering and also manifests into the audio path through a shared impedance ... there's another inter-coupling path between the two sides I'll say that much, but it ties in with the regulator filtering issue - together they invite problems ... it's your puppy to figure out ...

hope this helps ...

[R.G.]

burp ... the newer transistors likely have lower parasitics by virtue of their small size - so smaller Ccb even

Smaller is usually better for transistors, but that wasn't my issue. Why don't you and Jack go run that one down?

Sorry to hear about the indigestion. Those cheapo burritos for lunch are not really good for you, you know.

... the problem you're having, aside from technotopia,

Thanks for your concern, but I've enjoyed all the analysis I want to for today. Let's hear about your problems, OK?

In the words of every therapist everywhere, "And how did that make you feel?" 

is the interaction between the NFB based 78xx series regulator that isn't well supported by large enough caps AND the fact that it's feeding a single-ended circuit which allows a circulation of spurious signal more readily through a poorly bypassed regulator feed - known to have oiscillation issues (that's why lots of designers avoid them in specific situations) ... if you use a super low-draw bulb for the vibe (talk to Steve Daniels) the 7815 is ok as long as it's well filtered fore and aft ...

Well, JC, as I mentioned when I first posted this, that is in fact the first place I looked. And no amount of decoupling or tinkering would make it stop. It did change, reflecting the differences in power line impedance all right, but if that had been my problem, I'd have fixed there, right? Oops. You're in Canada, right? So I should have said "Eh?"

Or would you like to suggest that I was using high impedance solder, or the wrong temperature iron??  Holding the iron the wrong way? 

I get a fair bit of emails concerning a grounding problem you have on your board and/or in your wiring

I find that odd. I've never had an email concerning a grounding problem on that board or wiring that was not traced down to off-board wiring (like, no ground lead to a jack ... ) (note correct use of ellipses) or to someone wiring the PITA power entry wrong.

I find it odd to the point of questioning your, well, let's call it imagination. It does sorta seem like someone sometime would have asked me about it, since I did the board and I have been supporting it to the point of personally debugging and/or rebuilding them to get people going for about a decade now. Why exactly do you imagine that people would only email you about this issue and not me? Not ever?

which manifests itself as a hum that creeps into the signal path when the bulb goes brite ... these guys are often using a high current bulb ...

Yep. That's no mystery. Put too low an AC source on it, heavily load the unregulated side with a high current bulb and you run the regulator out of headroom. It makes a nice hum. Instead of grounding mysteries and misbegotten regulators, how about we think about ripple hum? Or did you even stop to scope it and see if it was 60 Hz or 120 Hz?

part of the problem seems to be the grounding

You know, a very large numbers of these boards have been built over the years and it doesn't seem to affect them all. In fact, I've never seen ONE that can be traced to poor on-board grounding. So it's facts time, fella. If you are going to toss around assertions like that, be prepared to back them up. I think you're making this up, and what you need to do is to talk me out of that by telling me *exactly* what is wrong with the grounding, and why. 

But before you get into that, I will tell you that a large amount of time went into the grounding on that board. The first protos were where I implemented a lot of theoretical info about designing ground paths to exclude hum and buzz from the audio path. I will be very interested in your theories of how the grounding is wrong. Please be specific, not hand-waving. Which trace is doing what?

No, it's not because the moon is just past full. 

and part of it is the lack of proper filtering around the reg ...

Once again. Write it down fella. And while you're at it, write down *exactly* what filtering is needed to make it not have this problem. Here's a hint - those little notes about decoupling on the datasheet are not the last work in power supply design. 

they aren't solid regulators once you start pulling a few tens of mA -

OK. Define your terms. "They aren't solid regulators... " (note grammatically correct use of ellipses) means what? DC voltage wanders? Ripple rejection suffers? They get soft and squish like jelly beans? What?

Handwaving allusions and innuendo were not part of the circuit courses I took. The guys who tried that got Fs. They all opted out of EE and went into Civil Engineering to design simple stuff like roads and bridges (yeah; think about that next time you're driving). But they got to drink more beer than we did because the tests were much, much easier.   

in the Vibe the 7815 well filtered is passable as long as you use a low-draw bulb ...

Gosh! You mean it's possible to overload a power supply?!? OHMIGOD. You mean it's possible to pull so much current from a rectified/filtered supply that the ripple will break through ANY regulator by violating it's headroom requirements. Oh, say not so! 

Or wait! Did you mean that there is an esoteric regulator somewhere that's better than the dirty, old, sick, weak 7800 series? Is there a solid-gold regulator that's better than passable? Maybe sold by hifi tweako stores for $100 a regulator? Izzat where you're going?

Then there's that little issue of the original 'vibe. It had a half wave rectifier and half the size of filter caps I put on the Neovibe.  And no regulator at all. So how is it that every one of those isn't a hum monster? That's right, just step a little closer and blame it on the PCB layout. Tee-hee. I know a few things about the original layout too. 

I had the same problem in the 360+ for many years until I switched to a low feedback way of regulating power to class-A audio circuits ...

I do sympathize with you and I hope you get over that soon. But are you absolutely 100% sure that was what your problem was? 

[R.G.]

Nice try JC. Edit your original post instead of responding. Of course the original text, all of it, is in my post as quotes.

That trick of yours is one reason I try to get your posts in quotes. You will and have revised history to suit you.

But it isn't that simple. What exactly makes you think that just dropping in to toss spears and then duck out is OK? If you're going to play like you're a techie, you're sometimes going to have to explain yourself in some detail. So here's the deal - unless you can spell out exactly what you're talking about and support it by logical, instead of handwaving debate, you're making it up. In fact, I take your not doing so as an admission that you are in fact making it up.

burp, nice dose of technotopia ... the newer transistors likely have lower parasitics by virtue of their small size, so smaller Ccb even ... regardless, the problem you're having lies in the interaction between your NFB based 7815 regulator that, on the one hand, isn't well supported by large enough caps AND the fact that it's feeding a single-ended circuit - which more easily allows a circulation of unwanted signal readily through the supply loop than an op-amp circuit would ... these regulators are known to have oscillation issues, not too well documented in the app. notes, a big reason why lots of audio guys avoid them ... with a low draw bulb (Steve Daniel carries the ones I recommend) a 7815 is passable as long as it's well filtered and the grounding BETWEEN BOTH SIDES OF THE CIRCUIT is full star all the way to the regulator ground pin, and there's another condition that I'll leave you to find with regard to the 78XX guys ...

I get a fair bit of emails concerning a grounding problem you have on your board or in your wiring - it manifests itself as a hum that creeps into the signal path when the bulb goes brite - part of it is poor filtering around your 7815 as you're discovering and also manifests into the audio path through a shared impedance ... there's another inter-coupling path between the two sides I'll say that much, but it ties in with the regulator filtering issue - together they invite problems ... it's your puppy to figure out ...

hope this helps ...

[amz-fx]

Why don't you and Jack go run that one down?

I don't have any interest in your squabble so don't try to drag me into it.

-Jack

[R.G.]

I don't have any interest in your squabble so don't try to drag me into it.

No intention to do so.

Simply prying JCs fingers off one of his spears.

[RedHouse]

Wow! ...I can feel the "love" in this thread!   

Sorry, I couldn't help it, but just to chime-in on the 78xx power thing, I too have noticed tendancies to cause/promote/contribute weirdness like oscillations. I too use the input and output caps on the regulators (and diode protection) and sometimes a smaller one like 1000pF on the output works well.

Not relevant to this issue, but on my 'vibe boards I actually use two regulators to power the boards, one powers the audio path, and another the bulb-driver/oscillator circuit. I keep them off-board and they use separate ground paths.

I too have had some recent experience with some funky Fairchilds (not 'vibe circuits) I have been working on some gear with a few stages of preamp level gain and I was getting some bad behaviour from them, I took som etime to troubleshoot them and got discusted and just replaced them with some Central Semiconductor 2N5088's and the problem went away. Damn strange when a known-working board can show other behavour with questionable parts.

I also noticed the Fairchilds are often measuring way up in the 700+ hfe range which I feel IMHO is too high, I'd expect a 2N5088 to be commonly somewhere in the 350-500 range.

[Paul Perry (Frostwave)]

Brad has some good points there.
One of the 'problems' with 78XX and 79XX regulators, is that they LOOK so simple - once you realise their pinouts differ! - that very few people ever bother to read the fine print of the application notes.
If they do, they will find comments on length of wires, and ranges of filter capacitors. It's quite possible to build an oscillator, when you think you are building a power supply. And so far as manufacturers go, I always prefer National brand regulators. Certainly, there are plenty of other good brands - but National have never let me down, and I have done some pretty rough things to them.

[R.G.]

Wow! ...I can feel the "love" in this thread!   

Sorry, I couldn't help it, but just to chime-in on the 78xx power thing, I too have noticed tendancies to cause/promote/contribute weirdness like oscillations. I too use the input and output caps on the regulators (and diode protection) and sometimes a smaller one like 1000pF on the output works well.

There is a variation in the performance of same-numbered parts from different vendors. Fabs and processes are too expensive to change for one or two parts, so each maker uses what their fab/process turns out to make parts that meet - by however slim the margin - the datasheet limits for the part they sell. This generally means that each maker's part is different inside, and only meets the specified limits of performance, not necessarily anything else. So yes, I believe that for some variants of three-terminal regulators, there may be issues. That's one reason I went to decoupling caps for one of the debugging procedures. It changed the behaviour a bit, but did not fix the oscillation itself, which is consistent with the transistor oddity.

Three terminal regulators were at one time source-specific items Three Initial Corporation's power supply designs. A specific vendor's 7805 would be qualified, and not necessarily all vendors' versions. Until it had passed component quality verification, it was not approved for use even if it had the right type number stamped on it.

As for feeling the love, I do love a rational technical debate. I respond very well to finding out something new, even if it means I'm dead wrong. I don't respond well at all to mysterious allusions of shadowy issues that can't be substantiated.    If you told me I was completely full of some spurious and unsavory substance and could present objective facts to show it, after I went through the facts I'd be happy to say that you were right and I was wrong. I've done that before and will no doubt do it again. Electronics is very much fact-based, or should be.

Not relevant to this issue, but on my 'vibe boards I actually use two regulators to power the boards, one powers the audio path, and another the bulb-driver/oscillator circuit. I keep them off-board and they use separate ground paths.

That's not a bad way to do it, and is extra-cautious.
It's funny. I wish I had saved the raging debate from usenet back when I first released the vibe clone board. I was castigated up and down for putting in a three terminal regulator instead of the two-section R-C filter from the original. The accusation was that it could not sound as good because the regulator was different, not original.

I too have had some recent experience with some funky Fairchilds (not 'vibe circuits) I have been working on some gear with a few stages of preamp level gain and I was getting some bad behaviour from them, I took som etime to troubleshoot them and got discusted and just replaced them with some Central Semiconductor 2N5088's and the problem went away. Damn strange when a known-working board can show other behavour with questionable parts.

I also noticed the Fairchilds are often measuring way up in the 700+ hfe range which I feel IMHO is too high, I'd expect a 2N5088 to be commonly somewhere in the 350-500 range.

Unfortunately, it's not all that strange. Different-brand parts can be a cause of different behaviors. It's one of those things that can easily bite you.