View Full Version : Amp15-PS postmortem - blown TA3020 and resistors
As I described here (http://www.41hz.com/forums/showthread.php?3421-PS4-regulators-in-Amp15-what-to-leave-out&p=31666#post31666), I blew the TA3020 chip in a modified Amp-15PS as I plugged in a laptop source. In this thread I will describe the damage and hopefully figure out the cause. I will also describe the modifications to the amp.
I plugged in a Macbook Pro as a source. Bam! There were small explosions, fire and smoke. There is damage to the TA3020, and to the four largish 0.1 Ohm surface-mount resistors on the bottom of the PCB. Here's the chip: http://imgur.com/ETyWNl.jpg (http://imgur.com/ETyWN.jpg)
The pins were twisted and burned, blown completely off the chip, and remained stuck in the chip socket. The blown pins are pins 38-41, and pin 43. These are NC / VNN / VBOOT1 / OCS1LN / OCS1HP. (Funny -- a non-connected pin blew.) Pin descriptions:
NC - Not connected (bonded) internally. To minimize coupling between pins, tie these pins to AGND (pin34).
VNN - Negative supply voltage.
VBOOT1 - Bootstrapped voltage to supply drive to gate of high-side FET, Channel 1
OCS1LN - Over Current Sense inputs, Channel 1 low-side
OCS1HP - Over Current Sense inputs, Channel 1 high-side
Then the largish flat green SMD resistors on the bottom are oddly shriveled. No other damage is visible, and the board looks good.
Re. flux residue - The board was IMO not clean enough, in retrospect.
To answer Scratchy's question (http://www.41hz.com/forums/showthread.php?3421-PS4-regulators-in-Amp15-what-to-leave-out&p=31672#post31672), the AGND-PGND choke looks 100% intact. (I'll measure once I've finished documenting.)
That what you get for using a Macbook. The chip is coded by Microsoft :D
Hahaaah! Yes, Saturnus is right!:D
No seriously, this seems like the output fets have entered some state of short circuit which has blown the current sensing resistors.
The blown pins do not necessarily describe the exact cause of death, it's like a fire cracker doesn't always blow the same way. Inside such a chip there's layers of all sorts of materials that conduct and when something inside blows it can cause short circuits to any other layer, also NC pins....
All nice, but although most know it is not wise to change cabling on working equipment, people will now want to know about the input protection on the amps.
Does one need to implement it if not on the amp? Like some add a DC-protection on the output?
Because the input spike probably killed this amp, and most do not want this to happen.
Also, this prompts me to look into the (simple, some diodes) input protection on my amp-9-basic.
Now there are 2 diodes (1N4148) on a line, making it swing to about max. 6Vpp. (TAA4100A: -0.5 upto 5.5V)
On most other Tripath chips the input range is 4Vpp.
I think I might change it to a LED to have a visual aid to see if it trips = clip indicator.
Maybe put something like that on every amp to have a basic protection?
Will have a look at the schematics of other amps, because had not considered the input protection on the amps.
it is possible that since the VNN pin is beside the NC pin (if I understand correctly) and since VNN is a high voltage pin (with respect to gnd) and can provide high current, when the pin blew, the hot molten metal also melted the NC pin that's why it burned also. that's just my theory. :)
Diode clamping, not a bad idea! I have also endorsed this method for firewire ports. In the company I work numerous ports used to blow up, but now I have protected most of them with diode arrays. I've done tests with this on the scope and it works like a charm!
Now you might think, how CAN this work on an audio input when you clamp between ground and 5V?? Simple, the inputs on the chip are 2.5V biased, so the signal can nicely swing between +2.5 and -2.5V!
With Firewire it's the same story, these signal pins are 2V biased... On most Texas Instruments TSB43xxx series Firewire controllers the max. voltage on the inputs is 3.6V and on the newer ones 3.9V, but I've seen both types blown up... This controller uses 3.3Vdc power supply, so it's very simple to clamp between ground and 3.3V...
I'm not sure a LED is fast enough, normally Schottky diodes are chosen for such an application...
I like to use a package with two diodes, usually sot23. This makes it quite easy to implement.
Clamping diodes are on the amp-9-basic, but no protection at all it seems on most of the other amps.
So it seems like a recommended, permanent mod then for those. Why not done already ?!
I mean, really a pity when you've built your beautiful amp-15, 2, TP etc. working, and some bloke pulls a plug by accident (or on purpose): bye bye amp!
On a stereo amp, 4 LEDs cost about 1 Euro. Just do it, I'd say.
If you clamp between gnd+5V, then I think the diodes will start conducting at 2.5+0.7 = 3.2V?
Well, in any case a lot better than NO protection at all.
On average, LED's have a t-on of about 5-30ns, so the audio range is peanuts for them.
(just like normal, cheap diode-rectifier bridges can take 100Hz easily and are low-cost)
Fast LED's are only needed on data-lines with transmission rates of 288+Mbit/s.
We don't have those at home (yet).
Blown NC pin on the TA3020 (or any chip for that matter):
If one or two pins are blown out (=with force), then indeed, usually the pin in-between or adjacent joins in....
I can't find the thread right now, but does this amp have the pull-down resistors on the Mosfet gates?
Is is possible that a static charge when plugging in the input cable could have turned on the FETS?
No on resistors: http://www.41hz.com/forums/content.php?353-Using-IC-sockets-with-amplifier-drivers
Sorry to hear about your woes, krilli, and sorry that you'll be under the spotlight until this problem is resolved...
I agree with V-bro that the TA3020 amps don't pop when you plug the source in. There would have to be something else wrong or incompatible.
Are you ABSOLUTELY CERTAIN that plugging in the macbook coincided with the pop? For example, was there a second delay? Why did you say explosions (plural)?
Popping as a result of over driving the input is speculation hence a "rule" to clamp inputs is premature at this stage.
Forget about static damage for now. I pursued it for a long time with comatose Truepath and it was completely futile. The components are tough.
This is how I would approach the problem. It contains a lot of questions.
As I see it, something caused either the high or the low mosfets to pop. These then took out the other two because high and low side mosfets share the same die on AMP15. In popping, the mosfets went into a short circuit amorphous blob state (this is what mosfets normally do). This created a short between VPP and VNN which took out the current sense resistors (and may well have damaged your bulk caps). The amorphous blob also shorted the TA3020 gate drivers (or perhaps a boot rail) which ravaged the TA3020.
Now, what caused the first mosfets to pop? Our experience to date is that this invariably caused by doing something wacky with the power supplies such as shorting or reversing them, but this doesn't seem so likely on AMP15PS since it has all on-board PSs. Nonetheless, here's a thought. Do (did) you have a short between one of your power rails and chassis? Check heatsinks for shorts. Is your chassis earthed? What sort of plugs are you using for your input connectors? Are they isolated from chassis?
Here's some other preliminary quick checks.
Take a close look at the PCB and you may see a little black speck as a result of the initial short. What were you using as a load? Did you move the amp in the period between its final pop and the previous time when it worked properly. Did you change anything else? Any components? Did you notice any strange hums when it was working (this could suggest a chassis short per the Mr T incident)?
How good is your soldering? Was the PCB clean of debris (eg slithers of solder) that could have caused a short? Can you see any slithers? (Soldering would have to be pretty rough before there would be such slithers and if they are there they would be visible. It is very unlikely that this alone would have caused the fault on it's own but it is worth considering if you made any of the aforementioned changes).
If none of the above leads to the cause then I have other more sinister ideas.
I just named those input protection diodes because -some- protection on a good amp is a lot better than -nothing- at all.
Is 1 euro to much to protect the inputs of -any- amp?
I will put it on my amp-1 as well and the 32 and 4 too if nothing on there.
Better safe than sorry.
And maybe a high-R resistor to drain inputs from possible high static as well...4M in front of input cap maybe. Need to think about that. I do-not want my amps blowing up because of weird stuff happening on the inputs. Could be from a broken pre-amp too.
And was K's amp15 already properly cased?
I wonder what happens if a fly/bug lands somewhere on a working amp-15/2/8/7/TP PCB....
because high and low side mosfets share the same die on AMP15.
Just a small correction here, this is about the XP version and this one has separate high/low side mosfets in normal to220 package....
EDIT: Oh, no it is the standard AMP15 after all, sorry...
Thanks for all the super useful replies! I won't be able to act upon them for a few days, but I want to express my gratitude at all the super helpful info and suggestions.
Static is still a posibility, since this was no comatose amp.
I wonder why are there no pull down resistors for this amp, when the Tripath datasheets specify them for the other TA3020 based amps? Was there a conscious decision to omit them in the AMP15 design?
Re. flux residue - The board was IMO not clean enough, in retrospect.
I think K is saying it could have possibly contributed to the fireworks.
Yes, I wonder about that dirt/residue on the PCB.
Doesn't that give an effect in the first min. already when you've powered on the amp?
Or can that be weeks later? Depends on type of dirt/flux maybe. And the voltage/frequency at that point.
It does raise the question: if the board works perfectly and is clean, would it be wise to varnish it to keep it proper?
Some PCB's I come across are varnished. Keep a spray-can on hand?
(yeah, bit late reaction....)
I think I just found the reason for this. One of the gate resistors, R5-R8, actually had a broken lead. I must have nicked it when I was bending it for fitment, originally.
Finding this was a surprising process. I had replaced the output transistors, the big flat SMD resistors (current sense?), and the chip, and was testing. One channel basically didn't come out of mute - it had 2.6V on it, while the other one was fine.
I measured around the output transistors, and one half of one transistor didn't show any interconnections, while the other transistor pins showed Mohm resistances or ~10K to other pins.
Measured some more and found that one gate pin wasn't connected to anything ... I traced it, fortunately it's a simple path: Each gate goes through one 5.6 Ohm resistor to its own pin on the Tripath. I was getting strange measurements, especially across the resistor, seemed like a dry joint. When I fluxed and resoldered the resistor, it just swung loose! Just ... no lead attached to the resistor on one side. It looked totally right before I heated it up!
I can't quite put it all together. I'm thinking I've nicked the resistor in the original assembly, and that it has been a dodgy connection but good enough for normal operation. Then a spike from plugging the source must have melted the half-severed resistor lead? Because is FOR SURE related to plugging in the source - the explosion was quite sudden and dramatic!
It's strange because the all those resistors are spotless, and don't seem to have had any potent voltages across them. The resistance across the resistor itself is correct.
Oh well! I'll try to put a hack job in there and see if it measures OK and is mechanically stable ...
YES! It lives ... At least it measures OK! Both output offset voltages down to 0.0V :)
I had almost given up on this one ... the actual physical damage was pretty imposing at first. Phew!
edit -- Yep - speakers connected - it works. YES!!!!!!
Oh my god it sounds amazing. (As they do of course!)
I'm especially happy because this is the one with the surface-mount caps and non-inductive resistors a la V-Bro, and the unusual PS4 power supply config - and that's the one I was concerned about having an impact on the sound ... but if it does I can't hear it. God it's good. It way outclasses a pair of Amp11's that were connected to the same speakers a few moments ago. The higher voltage plays a part in that for sure though. Nothing wrong with the bass handling of this amp though!
Congratulations Krilli!! I sure know how that feels!!
Against a pair of AMP11 I can understand, but the AMP11-HV is about in the same league, got some monoblocks here and an AMP15 to compare with and I can't say it's a major difference....
Powered by vBulletin® Version 4.2.1 Copyright © 2013 vBulletin Solutions, Inc. All rights reserved.