Misterious Cap, noise issue on NOVANEX AUTOMATIC 6 amp

[razabri]

Alright - I have to say, all you guys are saying when you get into the issues of the circuit, thermal compensation, heatsink bias diodes and everything, is something I can't really wrap my head around, but I'm willing to take action if there's kind of secure way to do any of things mentioned and if it would be beneficial. I'm actually amazed how much particularities there can be with this sort of relatively simple circuit!

The simplest test is to listen to the output of the Volume control.   If the signal gets that far and is still free of the fizzy sound it means nothing in the preamp is creating the problem.  It shifts the blame to the power amp in one step.  If you do hear the problem in the preamp then check the signal at the JFET.  That will tell you if the fizz is at input of the preamp.   At that point it gets trickier to narrow down but it might be worth removing the JFET.

Thanks on the guideline, Rob, I have managed to probe the preamp as described and that fizz is not present, so I guess the issue generates later in the circuit. I have nice and clean signal through the volume pot's lug and it reacts well on low and high setting.

My absolute amateur guess would be that those main power amp transistors may cause this, as those components could have been stressed out - the amp has arrived to me with a burnt power fuse, connected between the transformer and power switch. Replacing those wouldn't be much trouble and replacement is something I can get easily and without waiting, not the original but 2N6488, that may fit well.

[razabri]

So I've been probing now through several spots, after the preamp stage, and the strange thing is that there's no fizz I usually get - finally I placed the probe at speaker out spot and there too the signal was clean... Beats me, as I've checked the speaker by connecting another and getting the same issue. Perhaps there's nothing wrong with the circuit and it's parts, but once placed together it gets that tiny dirt that's a bit annoying. I'm using simple probe, connecting the output to my other, clean amp, via 100nF cap, having the ground of the probe connected to the Novanex's ground. and the guitar is plugged in the Novanex, with setting that usually has that slight fizz...

[PRR]

...I actually did some simulations  playing around with the minimal bias on output transistors to at least reduce the crossover distortion. ...

It's a well-plowed field.

But in my bed I realized: this is a OLD, CHEAP amplifier, and IF the complaint is "crossover fuzz" (not clear), it had this so-called "problem" all its life. Play it loud, crossover does not matter. If you must play softer, today's world is full of cheap amps which do not have (much) crossover distortion. It is even a candidate for a chip transplant (though the half&half heatsink is awkward, and most of the medium audio power amplifier chips are on very tight supply this year).

[razabri]

I'd have to plunge a bit more into it to better know about the phenomena of crossover distortion/fuzz to know how it manifests itself, to be able to determine what sort of noise is it.

"Problem" might as well be something the amp had since it was made, no way of knowing (although I kind of doubt it, no matter of it being cheap) - it just bugs me, it's not that the amp distorts, there's just that slight dirt that follows the note when you get the pickups on full and amp on low and it dissapears when you mute the strings - vice versa, amp on high and guitar on low, there's no dirt, as the amp stays clean and it's uncapable of having a natural distortion, thanks to the set compression, so speaker can't distort.

What bugs me is that at the end of probing (monitoring through another amp), I had no issue with the signal and it seems to me that fizz is only present when the amp is given a task to actually push the speaker.

[PRR]

I was going to say "Chip It!!" but so many chips are on back-order from labor and shipping disruptions.

The LM1875 is one of the least-sexy power chips; both Mouser and DigiKey show stock in the hundreds as of this moment, with more coming within months. (Many chips are sold-out through 2023.)

The headline on the LM1875 says "20W...over 30W", but that is at high supply voltage. If you keep the Novanex 6 supply you will get nearly the same maximum output as before, with a little less distortion at middle power and potentially much less distortion near whisper level. The stock TI datasheet has enough details. You can even buy more-or-less complete LM1875 amp boards on the usual auction sites, though fake chips is a real problem. (99.44%-certain real LM1875 are $4-$7 at Mouser and DigiKey.)

[razabri]

To me LM1875 looks like a complete different animal, by just looking at it, but what do I know. Original chips in this guy are two 2N5492 and I mentioned before that I can get something like 2N6488 should I try and replace them, but I do have two or three other candidates that are maybe even cheaper in local store. Original says 50W power dissipation, the other one 70W... Perhaps I should go for less than the old ones...

Just realized now that by "Chip it!" you probably meant to replace the insides of the amp completely with something based on LM1875 - that might be a nice idea, to use it as a experimenting station, but I kind of like it as it is, with it's own character and feel, and all this talk was actually to try and make it as good as it can be, or as good as it were when it was new, but, as you said, maybe it's already there...

[PRR]

> looks like a complete different animal,

Yes. The equivalent (and more) of all those eight transistors is in the one LM1875 chip.



You can even keep the input and feedback networks.

You need a heatsink, and the 1Ω+0.22u output network.

The LM1875 is nearly blowout-proof (especially at 1/4 nominal output), and crossover distortion should be inaudible.

(It should: it has like 5 times as many transistors inside, to correct things that Novanex didn't.)

Once that works well, _I_ would up-size those power supply caps. 235uFd effective? 1000uFd at least for 8Ω amplifiers (two 2000uFd series).

[Rob Strand]

My absolute amateur guess would be that those main power amp transistors may cause this, as those components could have been stressed out - the amp has arrived to me with a burnt power fuse, connected between the transformer and power switch. Replacing those wouldn't be much trouble and replacement is something I can get easily and without waiting, not the original but 2N6488, that may fit well.

It's more likely a design issue.  And all those weird technical details are very normal when designing amplifiers.   

But in my bed I realized: this is a OLD, CHEAP amplifier, and IF the complaint is "crossover fuzz" (not clear), it had this so-called "problem" all its life.

I guess that's the crux.   Highly likely to be the case but is it?  Is a faulty part or connection tipping it over the edge.   Maybe one side of the swing is totally screwed and it's not crossover distortion after all.

So I've been probing now through several spots, after the preamp stage, and the strange thing is that there's no fizz I usually get - finally I placed the probe at speaker out spot and there too the signal was clean... Beats me, as I've checked the speaker by connecting another and getting the same issue. Perhaps there's nothing wrong with the circuit and it's parts, but once placed together it gets that tiny dirt that's a bit annoying. I'm using simple probe, connecting the output to my other, clean amp, via 100nF cap, having the ground of the probe connected to the Novanex's ground. and the guitar is plugged in the Novanex, with setting that usually has that slight fizz...

Crazy result.  So when you did this test did you have the speaker connected?   If you did the audio probe test without a speaker it will change the result.   Also it might support the alternate fault hypothesis that there's a fault and the swing of the amp is stuffed-up in one direction.  All very easy to workout if you have an oscillator and an oscilloscope but requiring more experiments if you don't.
[The reason this is the case is under light loads the small driver transistors will do the output and the power transistors can remain completely off.]

If the amp design is at fault, I agree 100% with PRR's LM1875 suggestion.   Good performance, problem free, very easy to get going.    Modding the amp is a total redesign and there's so many things to go wrong.   Things that are hard for an inexperienced person to pickup and you don't really have the equipment to make sure it's all working.


FWIW,  another issue with the amp seems to be the 18k (R14) at the bottom of the  power amp.  The value creates an offset at the output which might promote crossover distortion issues.  IIRC, it needs to be reduced to around 5k but I found that value early on when I started looking at the power amp.

[razabri]

Actually no, I have disconnected the speaker while probing, to better hear what's happening - didn't realize that it matters in whole equation. Guess I'll have to try again with connected speaker to actually confirm the result.

Also, I may just take you on that 18k and reduce it's value and see what happens.

Going into that LM1875 thing may be an interesting take, but I'm not desperate to have it all "cleaned", as it's certainly usable the way it is and I kind of like keeping it with it's original parts if possible. Perhaps I'll do that excessive mod in the future when I have more experience, even though I could get a kit board for les than 4$ and with free shipping and just bypass the old power amp with it...

[razabri]

Crazy result.  So when you did this test did you have the speaker connected?   If you did the audio probe test without a speaker it will change the result.   Also it might support the alternate fault hypothesis that there's a fault and the swing of the amp is stuffed-up in one direction.  All very easy to workout if you have an oscillator and an oscilloscope but requiring more experiments if you don't.
[The reason this is the case is under light loads the small driver transistors will do the output and the power transistors can remain completely off.]

Ok, had probed it again with the speaker connected and I have located the rails that have distortion - passing the signal through these into a clean amp.

Have them marked on the sketch I made and then I tried to correctly locate them on the schematic.

[Rob Strand]

Actually no, I have disconnected the speaker while probing, to better hear what's happening - didn't realize that it matters in whole equation. Guess I'll have to try again with connected speaker to actually confirm the result.

Ah, OK that might explain it.

Also, I may just take you on that 18k and reduce it's value and see what happens.

It helps but with the 330 ohm in there the biasing is probably be too far off for the 18k to make a noticeable difference.  The change shouldn't hurt.

Ok, had probed it again with the speaker connected and I have located the rails that have distortion - passing the signal through these into a clean amp.

When crossover distortion is present it is all normal to get a distorted signal on those points.  I don't recommend putting the audio probe on the bases of Q7 and Q8 (version 0.3 schematic), the amplifier could oscillate and burn out the speaker.

So at this point I'm reasonably convinced the amp has crossover distortion.   You could do a test with no input signal and the speaker connected.   Measured the voltages across the base and emitter of Q7 and Q8 (version 0.3 schematic) ie. across the 470 ohm resistors (R21, R22).  Also measure the voltage across the 330 ohm resistor (R19).

The problem is, going forward past this point is difficult.   Without having emitter resistors on the output transistors (see examples in the links reply #58) any mods attempting to remove the crossover distortion could cause the amplifier to overheat and if left in that state too long that could cause Q7 and/or Q8 to fail.

[Rob Strand]

Something occurred to me which probably isn't helping.

Normally we think a diode drop and transistor VBE drop as around 0.65V (0.6V to 0.7V say).   The circuit around R18 puts one diode drop across R18 (2x diode - 1xVbe), then the voltage drop across R19 is about double that so about two diode drops 1.3V.  Normally two diode drops across R19 is pretty safe.

However! R17 is a high value and that causes a low diode current and that causes the diode drop to be small, perhaps only 0.52V.   
https://2n3904blog.com/1n4148-diode-forward-biased-i-v-curve/

The transistor Q5 is operating over 2mA so the VBE drop is more like 0.66V.   So that leaves 2*0.52 - 0.66 = 0.38V across R18.     That means the voltage drop across R19 is only (330/150)*0.38V = 0.84V.

In those details it says the actual voltage across R19 is somewhat less than what you work out by eye-balling the circuit.   It's somewhat less than a safe value, so it's way off.  That's going to make the crossover distortion even worse than it needs to be.

So at this point I'm reasonably convinced the amp has crossover distortion.   You could do a test with no input signal and the speaker connected.   Measured the voltages across the base and emitter of Q7 and Q8 (version 0.3 schematic) ie. across the 470 ohm resistors (R21, R22).  Also measure the voltage across the 330 ohm resistor (R19).

It would be good to confirm the low voltage.   You could also measure the voltage across the 150 ohm R18.

If we can confirm that then we might be able to increase R19.  I have a preference to replace R19 with a resistor in series with a diode.  The resistor value would need to be tweaked.    Also I'm not sure how confident you are about trying to attach a 1N4148 diode to the heatsink - that would help prevent the amp getting hot.  You cannot have the diode leads shorting to the heatsink.

[razabri]

Thanks Rob, for your time and input - much appreciated.

I have made readings on those points:
Q7 - 0.2v (base and emitter)
Q8 - 0.65v (base and emitter)
R21 - no readings with no load, low voltage when the signal passes through and decreases with the signal fading
R22 - 0.52v
R18 - 0.38v
R19 - 0.83v

I am pretty confident to apply any changes, once I know what I'm doing that is. So I'd probably have to have a visual reference on that 1N4848 + heatsink idea, and no problem with changing R19 in series with a diode, I'd just need to know the value and to double check the diode orientation.

[razabri]

Something occurred to me which probably isn't helping.

Normally we think a diode drop and transistor VBE drop as around 0.65V (0.6V to 0.7V say).   The circuit around R18 puts one diode drop across R18 (2x diode - 1xVbe), then the voltage drop across R19 is about double that so about two diode drops 1.3V.  Normally two diode drops across R19 is pretty safe.

However! R17 is a high value and that causes a low diode current and that causes the diode drop to be small, perhaps only 0.52V.   
https://2n3904blog.com/1n4148-diode-forward-biased-i-v-curve/

The transistor Q5 is operating over 2mA so the VBE drop is more like 0.66V.   So that leaves 2*0.52 - 0.66 = 0.38V across R18.     That means the voltage drop across R19 is only (330/150)*0.38V = 0.84V.

In those details it says the actual voltage across R19 is somewhat less than what you work out by eye-balling the circuit.   It's somewhat less than a safe value, so it's way off.  That's going to make the crossover distortion even worse than it needs to be.

It would be good to confirm the low voltage.   You could also measure the voltage across the 150 ohm R18.

If we can confirm that then we might be able to increase R19.  I have a preference to replace R19 with a resistor in series with a diode.  The resistor value would need to be tweaked.    Also I'm not sure how confident you are about trying to attach a 1N4148 diode to the heatsink - that would help prevent the amp getting hot.  You cannot have the diode leads shorting to the heatsink.

All the suggestions seem realistic to me, and as I can see, voltages are close to your calculations - guess one step would be to decrease R17 and the other to increase R19, and sure, I can place a diode in there too. Still I am not sure about what it means to attach a diode to the heatsink...

Other thought I had, and again it's completely amateur guess, is that there could be possibility that in putting together the circuit some parts could have been unintentionally swapped, perhaps Q5 and Q7 - this is complete guess as amp probably wouldn't work if so, as those differ, PNP and NPN, I'm just looking what's on that path where noise occurs.

[duck_arse]

an ordinary silicon diode's forward voltage changes with temperature. an amplifier output stage can go into thermal runaway, where a device gets hot and allows more current to flow, making it hotter and allowing more current etc, until it melts down and goes China Syndrome. to prevent this, a diode/transistor junction is placed in thermal contact with the heatsink and therefore the output devices (this is why two heatsinks is a bad idea, they don't thermal couple), and is used to throttle the current allowed to the output transistors - so more heat equals less current.

that's the general gist, anyway.

[razabri]

an ordinary silicon diode's forward voltage changes with temperature. an amplifier output stage can go into thermal runaway, where a device gets hot and allows more current to flow, making it hotter and allowing more current etc, until it melts down and goes China Syndrome. to prevent this, a diode/transistor junction is placed in thermal contact with the heatsink and therefore the output devices (this is why two heatsinks is a bad idea, they don't thermal couple), and is used to throttle the current allowed to the output transistors - so more heat equals less current.

that's the general gist, anyway.

Very well explained, thanks, even though my knowledge is not so great so I'm probably not able to apply this concept on my own, or to determine if this sort of mod would be beneficial... I'm unable to imagine how to add the diode to this particular amp's circuit, and not to cause any damage (Rob did warn about the diode shorting out).

[Rob Strand]

I have made readings on those points:
Q7 - 0.2v (base and emitter)
Q8 - 0.65v (base and emitter)
R21 - no readings with no load, low voltage when the signal passes through and decreases with the signal fading
R22 - 0.52v
R18 - 0.38v
R19 - 0.83v

Ok so it's as we expected.  The 0.52V on R22 vs zero on R21 is related to the DC offset at the output.

Very well explained, thanks, even though my knowledge is not so great so I'm probably not able to apply this concept on my own, or to determine if this sort of mod would be beneficial... I'm unable to imagine how to add the diode to this particular amp's circuit, and not to cause any damage (Rob did warn about the diode shorting out).

There's little doubt the amp has too much crossover distortion.

The biggest dillema is how to fix it.  As it stands the whole design is stripped back.   What that does is each solution to fix the crossover has more risk than we would like.   There's a risk something will get fried making a lot of mods.  More the concern to me is making mods causes the amplifier to oscillate.  It's very difficult for someone without power amp design experience or equipment to deal with this.  Mods like this need quite a bit of testing to verify all is OK.  It could also blow something up like the output transistors or speaker.

Without a sensing diode on the heatsink and without emitter resistors on the output transistors is certainly limits how far we can push things.   Normally you need at least one of these to have any confidence things won't become thermally unstable.

Instead of making a heap of mods perhaps we should try the least number of changes to edge it over the line.   While the mods will help I can't guarantee they will be enough.

Call this phase 1 mod.

The idea is to edge-up the bias.  Which will reduce some of the crossover distortion.
Currently the voltage across R19 is only 0.83V which is very low and quite a bit
under the safe level.  Changing R19 to two diodes will increase the bias to 1.3V.
This is still safe level since there are three junctions (Q7, Q8, Q9).
We are still under-biased just not as bad as before.
[No need to put the diodes on the heatsink.]

The two added diodes can be placed where R19 went.  You can join them in air.
The both should be orientated to point towards Q6, Q8.   One the PCB
they should have the same direction as the other two diodes (D2 and D3).




When you power the amp up you need to,
- be prepared to turn off the amp if something bangs,  smells like
  something is getting ho  t, or if you hear any strange a sizzling sound
  (which could mean oscillations)
- Measure the voltage across added two diodes.  It should be about 1.3V
- Measure the voltage across R21 and R22.
- Check Q7, Q8, Q9, Q10 are not getting hot.

Play the amp at a loud level for 5 to 10 mins.
Keep checking Check Q7, Q8, Q9, Q10 are not getting extremely hot
They should get hotter than idle but not cooking.

Stop playing
- Measure the voltage across added two diodes.
  It should be about 1.3V maybe a little less than before.
- Measure the voltage across R21 and R22.
- Make sure Q7, Q8, Q9, Q10 cool down.

[Rob Strand]

The aim isn't to do this now, it's just FYI,  but here's some ways use a diode to sense the heatsink temperature.

The key point here is the diodes have pads on the PCB and the act of soldering them in holds
the diode stable.   You don't have this so you will need to add some silicone (adhesive) to hold them.
You need to keep in mind you have to solder wires to the bottom and you don't won't the
diode and wire rotating where it could short against a track.   You also don't want the
diode leads to touch the heatsink.

In order to drill holes in the heatsink you would need to disassemble power transistors and heatsink.
Something could get damaged.

We can probably get away with one diode on one sink.  We will deal with that later if we need it.

In these examples, Peavey use a specai dual diode.

Fender:


Peavey:


Leach:

[razabri]

Call this phase 1 mod.

The idea is to edge-up the bias.  Which will reduce some of the crossover distortion.
Currently the voltage across R19 is only 0.83V which is very low and quite a bit
under the safe level.  Changing R19 to two diodes will increase the bias to 1.3V.
This is still safe level since there are three junctions (Q7, Q8, Q9).
We are still under-biased just not as bad as before.
[No need to put the diodes on the heatsink.]

The two added diodes can be placed where R19 went.  You can join them in air.
The both should be orientated to point towards Q6, Q8.   One the PCB
they should have the same direction as the other two diodes (D2 and D3).




When you power the amp up you need to,
- be prepared to turn off the amp if something bangs,  smells like
  something is getting ho  t, or if you hear any strange a sizzling sound
  (which could mean oscillations)
- Measure the voltage across added two diodes.  It should be about 1.3V
- Measure the voltage across R21 and R22.
- Check Q7, Q8, Q9, Q10 are not getting hot.

Play the amp at a loud level for 5 to 10 mins.
Keep checking Check Q7, Q8, Q9, Q10 are not getting extremely hot
They should get hotter than idle but not cooking.

Stop playing
- Measure the voltage across added two diodes.
  It should be about 1.3V maybe a little less than before.
- Measure the voltage across R21 and R22.
- Make sure Q7, Q8, Q9, Q10 cool down.

Hey Rob, thanks for returning to this thread - I'm obviously helpless without your lead. I'll do this mod exactly as you specified and get back with the numbers as soon as I'm able.

Thing about heatsink diode has intrigued me and it is something I would be able to do, if it comes to be something we may consider.

Thanks, again!

[Rob Strand]

Hey Rob, thanks for returning to this thread - I'm obviously helpless without your lead. I'll do this mod exactly as you specified and get back with the numbers as soon as I'm able.

Thing about heatsink diode has intrigued me and it is something I would be able to do, if it comes to be something we may consider.

See how things turn out with the two diodes.

To go any further than that we will need to add a resistor in series with those two diodes (perhaps around 150 to 220 ohms).  However, I would be *very* hesitant about going to this step without either the sensing diode or the "emitter" resistors. 

The 2xdiodes + resistor will very likely fix the crossover distortion and a slightly higher value resistor will sound cleaner. BUT it comes at a very high risk of melting down the amp.   Even with the diode or "emitter" resistor mods we would need to tune that resistor to the absolute minimum.

As far as heatsink diode vs emitter resistors I'm pretty sure the heatsink diode is going to be a lot easier to do.  It doesn't quite add the same level of insurance against catastrophes.

A skimped version of adding "emitter" resistors is to add a single resistor between where the Q8 emitter and Q10 collector join and the output terminal.     Like R25 in this circuit (but without R22),

https://1.bp.blogspot.com/_YpJ002RhDIU/TOariOVWrYI/AAAAAAAAAJA/3-9S7R4uQcs/s1600/Quasi+circuit.JPG

However due to the ordering or the tracks that mod is quite messy to implement.

A much easier mod, with slightly poorer audio performance, is to add the equivalent of R24.  (Ideally we want both R24 and R25.)

The added resistor(s) need to be large (say 2W) so you have to find a place for them where they won't get damaged or short on something.