Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - Roly

Quote from: EnzoYamaha makes good stuff.

Quote from: J M FaheyThe right side amp has doubled value (R4 430r vs R3 240r) because the other NFB resistor, R1 15r is in series with the other amp one , R2 15r , so it's also effectively doubled.

Both halves end up having the same gain, 240/15=16X but out of phase

Sorry, but I beg to differ.

I calculated U1 Av = x33.
LTSpice says x32.98.

I calculated U2 Av = x28.6.
LTSpice says x28.6.

Pin 2 of U2 is a virtual earth point.
Pin 3 of U1 is an actual earth point.
Therefore R1 and R2 are effectively in parallel.

LTSpice result;
attach: TDA2003 Canakit bridge.jpg
Every company that makes an IC has a datasheet for it, but not all datasheets are created equal.  Bookmark that datasheet site.  I use it all the time and find it pretty good (compared to some others).

I'd still be interested to know what peak-to-peak voltage you are getting across your 4 ohm load at the onset of clipping (i.e. is the amp weak or the speaker deaf?).
Quote from: galaxiex
My thought was to see if I could calculate and increase the gain of the P/A somewhat,

Increasing the gain isn't going to help.  You can already drive it into clipping and it isn't loud enough.  Increasing the gain will only lower the input level for a given output, it can't give you more output - that's limited by your supply voltage.

Quote from: galaxiex
Actually, If you wouldn't mind...
A complete in-depth analysis with a layman's description of what every part does and why it's there would be good too...

As it happens the TDA2003 datasheet has a table that explains what each component does and what effect changing its value has.

Taking your redraft, from the left-hand side;

This IC is basically a power op-amp, so the gain setting is the same;

Av = 1 + (R3 / R2)

1 + (240 / 15) = 17  i.e. a gain of x17

As with any op-amp the output of U1 will go to the value that makes the join of R2 and R3 identical to the input voltage on pin 1.

The drive for U2 is derived via R1 from the NFB attenuator on U1.  As noted above the signal at the join of R2 and R3 is identical to the input (but at much lower impedance).

U2 differs from U1 in that it is in an inverting configuration which means that the join of R1 and R4 are at virtual earth.  This in turn means that your R1 and R2 are effectively in parallel giving 15/2 = 7.5 ohms to ground, so the gain above must be modified accordingly;

1 + (240 / 7.5) = 33 or a gain of x33.

In the case of U2 the gain is;

Av = -(R4/R1)

-1 * (430/15) = -28.66666667 or an inverting gain of about x29.

This means that U1 will clip against the rails somewhat before U2, but that U2 will be driven with this clipped signal so it will appear to clip just before it gets to full rail-to-rail swing.

Since this is a "bridge" amp U1 and U2 drive opposite ends of the speaker load in opposite directions.  This allows twice the voltage swing of the supply, a considerable advantage in a vehicle where the supply is limited to 12V.  This means that overall we get twice that, ~24V peak-to-peak, across the load.

It also means that each amp "sees" half the total load impedance, i.e. 2 ohms with a 4 ohm speaker.

C8 and C3 are low frequency and high frequency bypass caps to tie the supply rail to ground for AC.  These should always be fitted as physically close to the amp supply pins with the shortest connections possible.  {you don't need to go to extremes, but they certainly shouldn't be back in the power supply or some other remote location.}

R5/C6 and R6/C7 are Zobel stability networks to prevent high frequency instability (and again should always be fitted physically close to their respective amplifiers).
Thanks, thought so.

Quote from: Vitrolinany how hed had 400VAC on all outlets in the listening room

AAAARRRRggggghhhh...   xP

One very lucky fella messing with stuff he obviously shouldn't.
Quite agree about the futility of a total rebuild, but I think (hope) that this was a throw-away line rather than a serious suggestion.

Quote from: J M FaheyYou are generalizing from a sample size of >1<


Quote from: Rolywhy would a modern amp from a reputable builder have such an obvious parasitic (if nothing else is wrong)?

I.e. I don't think this is generic to this model - it is because I don't think someone like Fender are knocking out marginally stable amplifiers that I posed the question.

What I'm wondering is if this newly introduced rolloff is only needed because the amp has lost a rolloff or HF bypass somewhere.

Quote from: J M FaheyTransistors vary in their characteristics

Which in turn invites the question, didn't Fender designers fully account for possible device spread?

I rather doubt that (given that they would have done a Monte Carlo and a min/min max/max on the design).  Given that these have been built in large numbers I would expect such a problem to emerge from production line testing - an accumulating pile of problem units in the corner of the factory.  {been there, done that  ::) }

Quote from: J M FaheySomething may have been replaced and compensation became marginal, or something similar.
or maybe the Zobel network was open ...

Exactly my own thoughts.  I'm suspicious that the other boot has yet to hit the floor.

Again I stress that, until evidence turns up to the contrary, I don't think this is a model generic problem, rather something specific to this particular amp has been overlooked.

Enzo goes to the heart of the matter;

Quote from: EnzoThat is why we want to always identify the actual problem and then fix that.

Is this an actual repair, or is it only wallpaper covering the actual fault?  This particular amp seems to be fixed, but is it actually repaired?  That is my residual concern.

As you say JMF, this amp has been built in large numbers, and if it's a generic problem of component spread then there should be other examples of parasitic instability in this model.  ATM we have no evidence that this is the case, so the obvious conclusion is that there is still something not quite right about this particular unit.

Probably too late now to follow up, but I have a feeling we know the what but not the why.
Ooow-arr!  That there's a parasitic oscillation all right, now we can see it.

Looks like you nailed it - well done.   :dbtu:

{Now my only question is, why would a modern amp from a reputable builder have such an obvious parasitic (if nothing else is wrong)?}
Hi gez239, welcome.

"Doc, I don't feel well - what could it be?"

Quote from: gez239it randomly stopped working

For a start, give us more detail about how it failed, for example was it cutting on and off while you were using it, or did it sometimes fail to turn on when you flipped the switch?

Does the pilot light come on?

If so, have you tried it with an external speaker and/or headphones?

When it was "randomly" stopping was the pilot light staying on, or did it reflect it working or not?

Since it was your car sound system, did it take a tumble at some point?

Please post answers to the above questions.

Quote from: gez239I'm sure it had something to do with me running it through that cheap car converter.

Maybe, maybe not.  Let's not jump the gun.  Here we work by first evaluating the symptoms, then making a possible diagnosis.  At this point it could be cancer, or it could be indigestion, or just about anything else in between.

{"randomly" suggests that it could be as simple as a dodgy connection, and is not consistent with damage due to running it from an inverter.}
Given a large enough heatsink there should be no stability problems with this circuit.

Current feedback was first tried with valve Hi-Fi amps back in the 50's but never really caught on because it increases the output impedance of the amplifier, and because of the infatuation of Hi-Fi-ists with damping factor the ideal was seen as an amplifier with zero output impedance, i.e. the speaker sees a dead short and therefore has maximum damping.

Solid state amps came along that go pretty close to that ideal but the Hi-Fi world had/has other objections to them, some well based, others only subjective fantasy.

So about the only place you find mixed voltage and current feedback these days is in solid state guitar amps where the current feedback gives a highish output impedance and more "valve sound".  Therefore I would say that any design like you have posted with mixed feedback was most likely aimed at guitar amp use.

I found the source of your circuit here;

...but unhelpfully it has no commentary on its intended application at all, but I think we can be pretty sure it wasn't Hi-Fi.

For a good overview of guitar amp design I suggest that you read;

Rod knows what he is talking about.   :dbtu:
Quote from: Vitrolinsomeone applied 380VAC to mains instead of 230

Just out of personal curiosity whereabouts are you?  Here in Aussie it's nominally 240V single phase and 415V phase-to-phase (and yagotta wonder how anyone could even manage to do that).

Quote from: DrGonz78it could have been someone who had a fault in the amp and just start shotgunning the thing to death.

Like Doc I've been assuming (there's that word again) that ogeecheeman has been lumbered with the outcome of applied idiocy, but perhaps there is more to the history and a clue there.  "Replaced the bridge"  "But why?" is a very fair question.  Perhaps there was mention of an original fault or symptoms prior to the bridge replacement.  {but then I'm sure we have all had clients who swear blind that "it just accelerated into the brick wall all by itself!"  Yeah.  Riiight.}

Normally we are dealing with a single component failure and maybe some associated components that got killed by the primary failure.  With over-voltage or reversed supply things are different because all sorts of things all over the place could have been taken out, and multiple faults (like in a new build) don't just add, they multiply the difficulty.

Quote from: DrGonz78"you can always say NO to a repair"

As it happens I had one of these only yesterday.  A Deawoo Royale EG-1 PCM baby grand piano (I kid you not).  I'm apparently only the latest of a line of techs who have had a look at this, and I spent most of the afternoon probing its innards with a CRO on the free list for a local community group.  As far as I can tell it has a "CPU" that consists of a couple of Field Programmable Gate Arrays (for speed I presume) and one of them doesn't seem to be doing what it should, talking to PROM but not I/O, no signs of keyboard scanning or the like.

The builder, Deawoo Electronics, was spun off when GM bought their car operation in the late 90's and apparently closed not long after, so it's 20 years an orphan.  The thing doesn't even have a serial number much less other ID and wasn't built in large numbers (which is hardly surprising given the limited market for an electronic baby grand that would take up half the average living room).  It's basically a $200 keyboard in a very fancy baby grand case and apparently originally sold for around five grand!

They bought it second hand for $250 "going", but it obviously wasn't, and it now seems to have consumed technician time worth at least twice that.  No documentation on the net, no sign of any spare parts supplier, and apparently dead custom silicon.  Up close the electronics are what I'd call a pretty average build, certainly not "quality".

Taken together I had to tell them I think it's a lost cause (and I'm a sucker for a bird with a broken wing) but they are clinging to a faint hope that it can be rescued somehow and don't understand that when Roly tells them it's hopeless that is the most optimistic diagnosis they are likely to get.

However this is an amp, we have a circuit, common parts, and ogeecheeman is obviously capable, no beginner, so I think that we can still collectively pull this fat out of the fire (and besides, I'm dying of curiosity to find out how this story ends - I hate unsolved mysteries).

Until we get the voltages asked for above I'm still of the view that the problem is somewhere around the diff-pair, that the VAS and downstream is only doing what it is being told to do by the diff-pair, but the "why" still plumb evades me.  I cobbled up the diff-pair and VAS in LTSpice but no lightbulb moment there, it only confirmed what we already know and started me wondering if the problem is off-circuit, e.g. a blown trace.
Thanks Enzo.  I have to admit to being baffled by this one.

Quote from: EnzoI am left to wonder at the health of R49 and R46

I think Enzo means R49 1k5 and R76 3k3 (at least on the cct extract I posted), the bootstrap resistors.

Confirm +ve rail on R49,
measure voltage on the bootstrap cap C62 (the join of R49 and R76),
and the half-rail
Apply Ohms Law.

Also; I asked you to check the voltages down the diff pair tail, D12, R66 and R50, same idea as above.  Please post both sets of voltages.

I'm wondering if a track somewhere has been opened by excessive current because I remember a situation where a few mm of track between two donuts had been so neatly blown off the board it looked okay to casual inspection, needed a bright light an lens to confirm.

As a sidebar, given the damage to the muting FET Q9 I'd be giving D14 a close look too.  If enough current passed to explode the FET I'd be surprised if D14 wasn't damaged too (not that I think this is the cause of your basic problem).
"????" indeed.  This one has certainly got me scratching my head.  ATM you are the front runner for my Dead Transistor Award for the most puzzling fault this year.   ;)

Quote from: ogeecheemanthe only damaged components I initially found were  jfet transistor Q9 which literally exploded and R35- 4.7 ohm emitter resistor

I think that we can put R35 to one side for the moment.  You have replaced it, and its associated output transistor Q12, D29, and I presume you have checked R74, but with -31V on the collector of the VAS it would seem that the output stage is only doing what it is being asked to by the VAS, Q14, which in turn is only doing what the diff amp is asking it to.

The exploded switch-on mute FET Q9 is interesting though.  This would have blown up when the nominal +40V was actually -40V and there is a direct uninhibited path from ground through the source-gate of Q9, and diode D14.  The G-S junction would have allowed a high current to pass from ground to the inverted supply rail via D14.  I would therefore expect D14 to be cactus as well, but that doesn't explain the negative voltages around the diff pair which seem to be coming from the -21V half rail via the NFB resistor R62 27k.

The voltages around the VAS suggest that it is turned on and pulling its collector down (and the rest of the OP stage is responding accordingly), but it doesn't explain why the diff pair is doing that.

The -24V at R56 and D18 will be reflecting the standing negative voltage on the half-rail.  This and the -0.8V at R51 and D20 are about what I would expect in the current situation - a result not a cause.

Quote from: ogeecheemanfrom the power amp side of R48 nothing but wildly high negative dc voltage.

I think I can see what is happening, just not why, so I'll give you my take in the hope that it produces a lightbulb moment for you.

From ground via R44 27k the -10V at the base of Q19 will be due to the E-B reverse zenering out to the -17V on the common emitters, which in turn will be coming via Q20 E-B and the NFB resistor R62 to the -21V on the half rail.  That's the what, but I just can't see why ATM.

I know that you have replaced the diff pair Q19 and Q20 but my intuition is that there is still something wrong around here, and had you not already replaced Q19 that would be my prime suspect (followed by Q20).

{come on Brains Trust - I'm grasping at straws here   :-\ }
Quote from: Vitrolinit was said to be stable down to 8 ohms

"Stable" in the sense that the magic smoke won't come out of the output transistors.   :lmao:

There are a couple of things that you need to keep a close eye on when designing an output stage.  Just for example the venerable 2N3055 will indeed pass a maximum of 15 amps, but the obscure gotcha is that the device current gain falls down a hole above about 5 amps, so in practical terms this generally means a load no lower than 8 ohms and a power limit of about 50 watts out per pair.

The second gotcha is under-estimating the amount of heatsink required to keep the transistor chip itself below its maximum temperature under all operating conditions.  An output stage that works fine on the bench at a 25ºC ambient might be in trouble on a 40ºC ambient summer night under stage lights.

The 2N3055 is rated at 115 watts and a maximum case temperature of 200ºC, however it can only handle 115 watts up to a case temperature of 25ºC and has to be derated above that, to about 60W at Tcase = 100ºC, and no watts at all at 200ºC.

The thermal resistance of the chip to case is 1.52ºC/W and with a really big heatsink giving 0.5ºC/W to ambient, at 25ºC ambient and 50 watts out you have another 50 watts of waste heat to get rid of.  With a total chip-to-ambient thermal resistance of 1.52+0.5 ~= 2ºC/W that's 25ºC ambient plus 2ºC/W * 50W or 100ºC + 25º = 125ºC chip, and the device has to be derated to around 50 watts rather than the 115 watts in the datasheet.  In an ambient of 40ºC it's down to only 35 watts.

Every now and then you come across a "100W into 4 ohm" amp circuit using a pair of 2N3055's that has been "designed" using only the optimum headline device specs, and yes, this can be made to work provided you don't mind bolting an air conditioner onto the back of your amp - otherwise it's a in-joke.   8|
Quote from: bluebooWith the device plugged in but not switched on : 0.3 mA
Or do you mean switched on but both volumes on 0?

We want the voltage, not the current, and yes, turned on and turned right down so the amp is active but idle.

Quote from: bluebooIt doesn't have a Send / Return only has Input, Footswitch, Pre-Amp Out, Power Amp In and Headphones

Here we can treat Pre Out/Main In as the same as Fx Send/Fx Return, so plug a known good lead between Pre Out and Main In and see if that makes any difference to your symptoms.

Have you checked Q8, the bias transistor?