Diagnosis of a blown transistor

[leigh]

Please see the attached schematic. This noise generator is working, but not well. First transistor stage, Q28, creates nice smooth white noise. However, the output of the circuit has an audibly grainy, spitty quality.

Power into the circuit measures as spec'd, 12 volts.
On the collector of the 2nd transistor, Q29, I am measuring only about 1 VDC.

That's a sign of a damaged transistor, right? It shouldn't be so low, should be up closer to 11 volts?

 

[samgraysound]

Could be Q29 or C94, or a wiring error, or...

Do you have a multimeter? you could do diode checks on the transistor and check for shorts from C-E and across C94.

Did you build this? Is the transistor oriented correctly?

 

[JohnRoberts]

Reverse biasing a base emitter junction to cause it to zener is not a very pure noise source. That schematic says it is selected, probably for the quality of that noise...

the second transistor is just a very high voltage gain stage, amplifying the output of the first. The collector voltage of that second transistor probably depends on the current gain of that device, you could tweak the collector voltage higher by making R358 larger than 150k.

JR

 

[leigh]

Could be Q29 or C94, or a wiring error, or...

Do you have a multimeter? you could do diode checks on the transistor and check for shorts from C-E and across C94.

Did you build this? Is the transistor oriented correctly?

I didn't build this - it's in an Arturia Minibrute synth. And unfortunately for easy troubleshooting and repair, it's all SMT parts. I recall it used to have a decent-sounding noise, and I have the chassis opened up to fix something else, so I'm taking a look at what went wrong with the noise while I'm in there. (As it turns out, apparently lots of other Minibrute users have reported the same issue, so it doesn't seem to be an isolated case.)

Q92 has 0 VDC on the emitter, and the collector is steady at ~1V, so C-E isn't shorted.
The other side of C94 has 0 VDC, so C94 isn't shorted.

Q29 has ~0.6 Vbe , so that seems normal.

I've checked the resistances of R356 and R358, - cold, but in-circuit - so due to the caps the multimeter keeps changing its reading, but it levels out around their spec'd values.

So given all that, does the low collector voltage of Q29 make that part the likely culprit?

 

[leigh]

Reverse biasing a base emitter junction to cause it to zener is not a very pure noise source. That schematic says it is selected, probably for the quality of that noise...

That may be that it's not a very good "pure" noise source, but it's commonly used in synth circuits. In any case, there's a definitive, audible change in the noise between what comes out of Q28 and Q29 - like Q29 is either clipping, or "gating" (only letting the loudest peaks through) because it's mis-biased. Hard to tell which one from looking at the (analog) scope here. I'll see if I can capture an audio clip.

 

[JohnRoberts]

That may be that it's not a very good "pure" noise source, but it's commonly used in synth circuits. In any case, there's a definitive, audible change in the noise between what comes out of Q28 and Q29 - like Q29 is either clipping, or "gating" (only letting the loudest peaks through) because it's mis-biased. Hard to tell which one from looking at the (analog) scope here. I'll see if I can capture an audio clip.

Q29 is a high gain stage so it is expected to sound different. 

I already suggested the resistor value to tweak to bias it up higher for more symmetrical clipping.

JR

 

[leigh]

OK, I went in and recorded clips of the noise, both before and after the Q29 gain stage.

Attached is a screenshot comparing the two waveforms. You'll note that in the computer, the "before" clip has been gained up by 28dB to match the peak level of the "after" clip, and that the "after" clip has been Inverted for easier visual comparison.

Since I can only attach one image per post, I'll add more in a second post.

 

[leigh]

The clips in the previous post were 8 seconds long each. More telling, perhaps, is the zoomed-in view. These 2 clips capture 20ms each (to be clear, the recordings were made at 2 separate times, so the two clips are not correlated in time).

With the increased resolution you can see how after Q29, the signal is close to 0V much of the time, except for the more occasional peaks. So maybe that's just poor transistor biasing...?

John I hear your suggestion to try increasing R358 to be larger than 150k. Worth a try I guess. But still I'm wondering what changed in this circuit over time. I did find a recording from a few years back, when the synth was new, and there was a bit of Minibrute noise on that recording. The noise definitely did not sound so "spitty" back then.

 

[PRR]

> That's a sign of a damaged transistor, right?

No. In fact it seems to be WAD. BC847C has hFE near 500. The 150k base bias reflects as 150k/500 or acts-like about 300 Ohms in the collector-emitter path. With 10k above this gives 11V*(300/10,300) or 0.35V of drop in 150k. Add 0.65V for Vbe, it is spot-on the 1V you observe.

I would not bias a random signal quite that low without many hours of observation. It can handle only 1V peak signal. The broke-junction source is very weak so that may be ample? And biasing collector low gives double gain compared to centered. But I would not go as far as this seems to be biased.

I have _NO_ idea how/why it has apparently changed over time.

If the 150k won't come out (easy), can you tack-solder a 3k across the 10k? That should bring collector up a volt or so. And also increase loading on the broke-junction which *may* reduce output.

The alternative may be to torch-off enough SMD parts so you can jumper-out to a bit of perf-board with a similar but tinkerable noise generator.

 

[leigh]

> That's a sign of a damaged transistor, right?

No. In fact it seems to be WAD. BC847C has hFE near 500. The 150k base bias reflects as 150k/500 or acts-like about 300 Ohms in the collector-emitter path. With 10k above this gives 11V*(300/10,300) or 0.35V of drop in 150k. Add 0.65V for Vbe, it is spot-on the 1V you observe.

Thank you for running the numbers on this. I got lost in my own calculations. I follow yours, except for the first step where you calculate the 300 Ohm equivalence of the collector-emitter path. But good to know that in any case, the 1V is not a sign of transistor damage. (And, measuring in circuit, it would seem that the collector and bias resistors have not been damaged either.)

I would not bias a random signal quite that low without many hours of observation. It can handle only 1V peak signal. The broke-junction source is very weak so that may be ample? And biasing collector low gives double gain compared to centered. But I would not go as far as this seems to be biased.

The noise source is indeed very weak, about 40mVpp at the emitter of Q28.

The alternative may be to torch-off enough SMD parts so you can jumper-out to a bit of perf-board with a similar but tinkerable noise generator.

Yeah, I'm about at that point!

 

[PRR]

> except for the first step where you calculate the 300 Ohm

Whatever current flows into Base-Emitter is multiplied by hFE to make current in Collector-Emitter.

Assume Vbe is negligible.

V(150k) is now equal to V(c-e).

Same voltage, 500X the current..... C-E acts-like 1/500 of 150k. Which is 300r.

Now we have roughly a 10k collector resistor and an equivalent 300r "in transistor". The transistor must drop 300/10300 of 12V, or 0.35V.

Now un-do the Vbe=zero simplification, add the 0.35V to the ~~0.65V of a Silicon Vbe, we expect very nearly 1V at collector.

Whatever the circuit does, is in series with 10k (and 47k+7V), so the drop in the 22r is teeny. Less than a tenth volt.

 

[clintrubber]

it's in an Arturia Minibrute synth.

FWIW, this thread a remarkable coincidence... just pulled out mine last week and pretty much enjoyed the noise-generator on the MiniBrute. Wasn't aware of any issues. Was in fact about to build an alike circuit for a friend who has the MicroBrute, which doesn't have a noise-source.

Rambling aside, you'll most likely already be aware of Yves Usson, the person that is kind of the 'medium' to Arturia of this indeed common topology that they used for the MiniBrute (and in fact for his own modules as well, see link below).
He's pretty well reachable, in case it'd be relevant to have a chat with him about this:
http://yusynth.net/index_en.php?&arg=7

http://yusynth.net/Modular/Commun/NOISE/NoiseSH-sch.jpg

 

[clintrubber]

(As it turns out, apparently lots of other Minibrute users have reported the same issue, so it doesn't seem to be an isolated case.)

Wasn't aware of this so far, but thanks for bringing it up. Possibly I start to hear 'things' as well on my MiniBrute now...  :
As a bit inelegant & obvious workaround: external noise-source injected in the external audio input.
But I'm with you, would like to understand as well why this common circuit could change behaviour over time...

 

[JohnRoberts]

Wasn't aware of this so far, but thanks for bringing it up. Possibly I start to hear 'things' as well on my MiniBrute now...  :
As a bit inelegant & obvious workaround: external noise-source injected in the external audio input.
But I'm with you, would like to understand as well why this common circuit could change behaviour over time...

I am not expert in semiconductor physics but recall a little about device zenering (actually base-emitter junction) from low noise design research.

You may notice backwards clamp diodes across low noise input transistor base-emitter junctions in mic preamps to prevent the input devices from ever zenering caused by input transients reversing the base-emitter junction (like turning on phantom power). It takes about 7V to zener, so a diode clamp makes sure that never happens.

Reportedly if low noise devices are allowed to zener they become noisy. Undesirable for high gain stages thus the clamps. Perhaps on topic about junction behavior changing over time, reportedly noisy junctions that have been damaged by zenering can be rehabilitated by self-annealing (operating with forward current for significant time).

While this does not explain why reverse biased base-emitter junctions might change over time, they discuss how the forward direction junction noise performance can heal after injury from reverse bias. This suggests to me that at least some characteristics of junction performance can be plastic (changeable).

JR

PS: IIRC I read about this in my old copy of Low Noise Design by Motchenbacher and Fitchen... (several decades ago). I still have my copy so may dig it out and dust it off later. 

 

[clintrubber]

PS: IIRC I read about this in my old copy of Low Noise Design by Motchenbacher and Fitchen... (several decades ago). I still have my copy so may dig it out and dust it off later.

Great book indeed! (the Noise Bible)
FWIW, the 1993-version within armlength reach at my desk at work and an older personal copy at home.

Remains puzzling... if this topology is flawed then it wouldn't be widely used for this kind of synth-application... reported problems for other synths would have been more widespread...  Possibly the dimensioning then, as suggested here above?

Don't know what Moog for instance uses... I do recall that for at least one type (The Rogue?) they used a dedicated IC, so not the simple 'upside-BJT'.

Best regards

 

[JohnRoberts]

Great book indeed! (the Noise Bible)
FWIW, the 1993-version within armlength reach at my desk at work and an older personal copy at home.

Remains puzzling... if this topology is flawed then it wouldn't be widely used for this kind of synth-application... reported problems for other synths would have been more widespread...  Possibly the dimensioning then, as suggested here above?

This is a pretty mature  "stupid transistor trick", cheap and simple. I looked at it on the bench to make a noise source a long time ago and dismissed it as not very desirable.  By the 70s they were already making psuedo noise sources from CMOS ripple counters. Trouble with simple digital noise sources is you can hear them repeat, of course they have gotten a lot better in the decades since I last looked into making one.

JR

Don't know what Moog for instance uses... I do recall that for at least one type (The Rogue?) they used a dedicated IC, so not the simple 'upside-BJT'.

Best regards

 

[clintrubber]

Hello,

Was curious again which one Moog - The Rogue used and indeed, it uses such a pseudo-random noise generator, the MM5837. That's using a 17-bit shift register.

Spending more of such cheap stages (& applying the commonly known right combination of taps) will improve randomness, which may or may not be relevant to the application.

Obvious benefit of using such CMOS noise-sources for production is the skipping of component selection.

 

[leigh]

FWIW, this thread a remarkable coincidence... just pulled out mine last week and pretty much enjoyed the noise-generator on the MiniBrute. Wasn't aware of any issues. Was in fact about to build an alike circuit for a friend who has the MicroBrute, which doesn't have a noise-source.

Rambling aside, you'll most likely already be aware of Yves Usson...

Yeah, it used to sound pretty good! I don't know what happened...

Thanks for the heads-up on Yves' site. I had checked out that circuit before, and yes it's identical to what's in the Minibrute, except for the transistor part number (BC847C in Minibrute, BC547C in the Yves original) and of course that all the Minibrute parts are SMT rather than through-hole.

I put together audio examples, but mp3s aren't an allowed file type for attachments on this forum, so here's a link to the annotated clips:

https://fishboytech.tumblr.com/post/189759022143/three-short-arturia-minibrute-noise-samples-in

 

[leigh]

Wasn't aware of this so far, but thanks for bringing it up. Possibly I start to hear 'things' as well on my MiniBrute now...  :
As a bit inelegant & obvious workaround: external noise-source injected in the external audio input.
But I'm with you, would like to understand as well why this common circuit could change behaviour over time...

Well, you'll be able to hear if yours sounds as bad as the audio examples I posted! Hopefully yours doesn't...

Here's an example of other users talking about the same issue:
https://forum.arturia.com/index.php?topic=86245

One person notes in there that Arturia customer support offered to sell them a replacement digital board, because the noise issue is "supposedly a fault in that board". The Minibrute has 2 boards, which they call the analog and the digital boards (the latter handling MIDI and other microprocessor-related stuff, although it does also have all the analog i/o jacks on board). The noise generator is on the analog board though, not the digital, so I am highly suspect of that diagnosis.

 

[leigh]

Assume Vbe is negligible...

That's quite a helpful trick! Ignore Vbe at first, and then account for it later. Thank you for writing out this explanation, it helped me see how your analysis worked.

In order for this estimation (of the circuit's equivalent 300R "in transistor") to work, we also need to assume that the supply voltage is high enough to bias the transistor into its linear region, correct?