do apex audio amp ic's have some kinda special rules?

[coppercone2]

I fried three different apex chips. One in a dead bug PCB and two in a real pcb.the first one I broke was like a PA60 or something, obsolete. the other two were from mouser . the old one worked for a while

I got a ESD mat, I checked the PSU rails (it goes up to +-15V in like 2 seconds with no over shoot).

The last one was installed in a board with no input and no load (50 ohm termination on the input) and it made a crackling noise and I noticed it was dropping 1/2 an amp through the non inverting pins through the 50 ohm resistor.

Do op amps not like having a resistor to ground on the non inverting pin in a voltage follower without a series resistor ? I thought it was just leakage current. I assume it fried on power up.

I am not seeing why a power op amp with a measly 50 ohm load to ground and a terminated input just fucking fried on first power up. The input pins should be high impedance right? so long the voltage input is within the rails nothing should happen? am I missing something here

I played with alot of op amps from Ti, LT, AD and probobly some others, I never seen this kinda stuff happen before.

its the PA74 chip.

I am not getting how the base of the transistor can supply that much current if its OK. I thought its limited to like 500 nA.


ima solder some bullshit dual op amp in place of this thing with dead bug wire to splice it in and see if this works because IDK what the fcujk

the only thing that comes to mind is that maybe putting a 50 ohm to ground on the non inverting input pin during startup causes some kind of problem so maybe it should be like 1k resistor from the op amp to the 50 ohm to ground but I swear I never seen that be a requirement. its just a fucking random guess. I feel like I saw aalot of successful shit without any fucking resistor in series with the non inverting pin. Actually I am sure I saw op amps used like that for major projects that sold alot.

[Whales]

Neg input: tied to output
Pos input: tied to ground through 50R, somehow pulling 0.5A

What is ground?  0.5A through 50R is 25V.  The part is only rated to 40V (eg +/-20V split rail).   So you must mean ground = negative voltage rail; or your 0.5A is only approximate and I'm drawing too many conclusions from it.

The datasheet says the common mode input includes the negative supply, so that should be OK.

Perhaps the negative voltage rail is not connected to the chip or somehow less negative than the ground you are tying to the input through 50R?  Then 0.5A might flow into this chip via VCC and then out of the chip via the pos input, either through input protection diodes or some unintended substrate diode misbehaviour effect.

[magic]

The datasheet says all sorts of things which don't make sense if the schematic also included in the datasheet is true

The topology they show ought to have phase reversal, the output is likely to go to the top rail if the inputs are brought to the bottom rail or the top rail comes up before the bottom rail and the bottom rail is near ground where the inputs also are.
The schematic they show has no protection from excessive differential voltage and base-emitter breakdown of input pair transistors, although absolute ratings say there is no differential limit.

Very sloppy datasheet OR a very sloppy chip with a very sloppy datasheet. If it's the latter, the two points above could explain how it blew.

 
Send the carcass to Noopy or ask the manufacturer wtf is going on. Or both.

edit
On second thought, emitter degeneration resistors may save it from phase reversal ehm, no, they can't, apex is clearly hiding something here
But it's still unclear how it behaves when the negative rail is left floating and it's still unclear if it really has differential overvoltage protection.

If you would like to try again, add reverse biased 1N5158 or at least 1N4001 from each rail to ground and few-kΩ resistors in series with each input. Connect a scope and look for anything weird during startup.

[coppercone2]

its 500mA total (250mA per channel). Basically the input is at rail voltage at rail impedance.

I don't expect a op-amp input to go to -14V or whatever with basically no limiting. I thought a electrolytic capacitor was shorting, I could not believe it was the 50 ohm resistors on the input..................

yeah they are pissing me off I think hiding something proprietary. i think i can do this with other chips that I trust more.

I don't think I want to try again because its expensive 

for the negative voltage, maybe I can try to probe it under the socket with a grabber hook to make sure the socket is not mesed up. I know the socket pin is getting the voltage. Its new and looks good.



But the only way it could provide 250mA from a INPUT to a ground impedance setting resistor is through some kinda unlisted diode right? The base of the input transistor would have to be sourcing alot of power.  I thought itt could only source 0.5 micro ampers.

oh yeah, its from the negative rail. because the negative side of PSU was reading 0.47 amps with the input resistors connected to ground. So the negative rail is coming out of the inputs. if I remove the chip from the socket, there is no voltage on the inputs BTW. I am not soldering anything. I don't think it could be a bridge. I also don't have bolts on the chip, so its not flexing anything to cause some weird short or whatever, and its o nly a 2 layer PCB

i wonderi f its possible the socket is messed up and the positive rail is floating??

[Conrad Hoffman]

I've had good results with Apex stuff but it was many years ago. Their support was excellent so I'd see what they say. Best case is they might replace it. It should work pretty much as you'd expect, no weird requirements.

[coppercone2]

i wonder if its a sequencing issue, I need to check that the PSU follow close enough. I know its slow and not over shooting but maybe there is a delay between - and + ?

but it never broke op amps before

ok once I ruled out the common suspects, it looks like a feedback loop op amp is doing something funny (composite).

When I removed some other op amps the rail went back to normal. Maybe this chip is OK.

So to clarify:
1)first opamp PA60 on dead bug was from ebay, it died under load grilling a resistor. could be natural causes
2) second op amp : dead channel, I think DOA
3) third op amp : feedback loop op amp making the protection diodes go crazy?
3a) checked the socket by putting grabber hooks under it, no problems
3b) power supply sequencing ok
3c) no short circuits found
3d) need to bench test apex amp again with voltage follower configuration to see if it still works as unity gain buffer as it did on arrival
3e) if the feedback loop is making the protection diode blast current into the input pins (not sure how this is happening but its related to op amp next to it i pulled) its possible that the input was dead shorted on OPA#2 and then it caused something in the chip to blow up).

so maybe its some kind of solvable problem and not a broken part.

I bet what happened is that I heard the other op amp make noises on power up. it has been through 3 prototypes already.

[TimFox]

I don't know about this specific device, but I have had problems in the past with latch-up due to power sequencing on split supplies.
A typical cause of latch-up is when any pin of the device goes negative with respect to the substrate (most negative node of the circuit in normal operation) by more than a silicon-junction voltage (0.6 V or so).
With split supplies relative to ground, a possible latch-up sequence is having the positive supply present when the negative supply is still an open circuit or high impedance (such as the output resistance of a linear regulator with insufficient input voltage) to ground.
In that case, current from the positive supply through the output stage or other parts of the amplifier can pull the negative supply terminal positive with respect to ground, while at least one of the input terminals is hard-wired to ground, inducing the latch-up, which can persist after the negative supply comes up, resulting in large supply current and overheating.
(Some mixed-signal devices, such as switched-capacitor filters running on split supplies, have a "ground" pin which is actually an input, the reference voltage for the digital clock input circuit.)
The cure for that problem is to put Schottky diodes from the positive and negative supply pins of the device to ground, to avoid full diode voltage across the relevant circuit with one supply missing.

[coppercone2]

alright since Im not insane and its mostly through hole without solder screen I can add some diode to ground in a fairly easy way if the op amp passes the bench test

[magic]

Alternatively, don't use a split rail PSU. (Or use it with the center tap not connected to anything).

Generate a virtual ground with resistor divider and connect it to IN+.
Connect IN- to the output, preferably with another safety resistor in series.
See if it works (OUT = IN+).

[TimFox]

For an audio amplifier, the output pin must source a large current to the load, which either requires a grounded load and a split supply, or a grounded load and a large series capacitor with single power supply and a low-current bias supply.

[magic]

Or virtual ground with capacitors to each rail of about the same size as output coupling capacitors.
Eliminates input coupling capacitors, avoids turn-on transients, but puts ripple on the virtual ground (effectively: on both supply rails simultaneously) and exercises the amp's PSRR more...

But we are getting off topic, I think. I simply suggested a way of testing such devices with lower risk of smoke escape.

[coppercone2]

wait, I have zener diodes on the rails because the PSU I have can put out 40V and its above the regulator voltage. So I limited it to 1 volt under the regulator voltage. I have seen this PSU explode before and put out 60V, so I added the zener diode

reversed biased it should perform the same function as the shotkey diode?

[coppercone2]

hmm ok something aint right, 200mA flows from PSU center point to oscilloscope negative. it should be floating

is that some kinda ground loop hmm

what is this lol

oh ok, the alligator clip I grabbed bnc thing was cross wired (ebay addon). I wanted the grabber hook because its better for this then alligator clip but I will just use the old test lead. (this is a new problem from the last 5 min due to poor cable choice)

[coppercone2]

alright for testing with a 1KHz sine wave that is from a isolated floating source with a good bnc cable

it looks like when its powered on both channels pass a sine wave with the voltage follower configuration. when its powered off you pass a half sine with some bump on top, I imagine that is what the op amp does when its not powered but has a input. Wired with a whole lot of grabbers

the bad amplfiier passed nothing at all with the burned channel or whatever it is

so maybe the amp is fine, and the noise I heard was the little amp burning up. not sure how it passes current but at least maybe its not a 100$ burn up

ima take some time to make some better cables because i fee llike I am at the point right now where I typically blow something up due to lack of proper experiment planning (when I want to run it over)

I think a BNC to grabber, proper DMM probe with shrowd, maybe some mini-clips with the hook up dupont

and it just fried. trashed. i think i will stick to other manufacturers lol

[schmitt trigger]

Just curious; how are you heatsinking these devices? The multipin TO-3 package is not very common.

And I know, I know. you mentioned that one blew up from the powerup, so it may not be a thermal failure.

[coppercone2]

whatever i am not doing further research on apex parts. trash can

i might solder in a dual op amp in there just to exonerate my PCB but whatever if the regular things I do for all my parts don't work for apex and i don't have some kinda crazy specific requirement for apex then im not gonna use apex. and yes the heat sink is inconveniant and expensive enough that it makes me further not want to use apex (and it was over powered for my application anyway). the savings from the integration minus the pain in the ass from the heat sink in the design minus whatever special thing I am missing makes the net worth negative as fuck. maybe the satilite aerospace people can figure this out but its not what i need

im not gonna be terrorized by this huge bastard anymore

[TimFox]

wait, I have zener diodes on the rails because the PSU I have can put out 40V and its above the regulator voltage. So I limited it to 1 volt under the regulator voltage. I have seen this PSU explode before and put out 60V, so I added the zener diode

reversed biased it should perform the same function as the shotkey diode?

To prevent latch-up, the diode drop (when it is forward biased, i.e. power supply reversed) must be less than 0.6 V, so a Schottky is required.
Zener diodes when forward biased are equivalent to a normal P-N junction.
Before retirement, we used TO-3 power op amps from Apex with split supplies (required for DC operation) successfully to drive precision 40 W DC motors.  I think we had local protection diodes on the board.
The only need for such amplifiers is if you need relatively large output current (as in our DC motor drivers), or for simple audio power amplifiers.

[coppercone2]

they should put a foot note because thats a pretty god damn huge requirement

[TimFox]

I don't know that that device is more prone to latch-up than others, but the diodes are a good precaution.
Devices that did latch up on me, before a useful engineer at LT told me about the diode trick, ended up almost shorting the +6 to -6V lines (for that switched-capacitor filter), and if the power supply was capable of large currents the devices (DIP package) became painful to the touch and failed.
The failure mode is equivalent to grounding the op amp input terminals, open-circuiting the negative pin (which will then go positive wrt ground), and supplying +15V from the positive power pin to ground.

[coppercone2]

alright but where do you put these diodes. if I make another board with different parts (probobly LT) I might as well add these diodes.

Can they be by the power input connector where I have the RF caps, zener diode, TVS? or must I put it near the amp.

or should they be after the LDOs?

[TimFox]

Connect each of two diodes from a power pin to ground, so that each is reverse-biased in normal operation.
I would put them close to the device, but it may not make a difference.

[coppercone2]

but before or after LDOs??

this board has pos/neg LDO and powered from DC source (LAMBDA). (LT3015/LT3083)
I don't like clutter near the device!

I have an accidentally ordered PA74 too. I guess I could give it one more try by doing some fancy footwork with the pins on that so it can solder to the bottom of the socket. 

maybe some nye fireworks

aFTER that one breaks too maybe I can wire the TIG welder into the board and do a stress test

anyone notice that by the time you are doing cruel experiments on chips bending the leads and mounting them upside down in grotesque PCB structures the problem ends up being solved? just reroute the elephant spleen into the badger ear and its all good!

[TimFox]

After the LDOs, directly to the pins of the op amp.
The circuit I was working on, lo these many years ago, where I learned about latchup had a 7806 and a 7906 on the board driven from the external +/- 15 V power rails.
I found that the 7906 negative regulator had a very high output impedance when the input voltage to it was below dropout.
The device was a CMOS switched-capacitor analog filter IC, where the signal voltage was symmetric about zero and the filter needed split supplies to follow it.

[coppercone2]

what diode do you recommend.

I thought a 1n4148 would work but I see that still a silicon diode. the voltage drop is too high. for some reason I thought that was a sk

I get like 0.6 on that one

[TimFox]

Maybe a 1N5819 DO-41 TH part (nominal 40 V, 1 A).
Max Vf = 0.6 V at 1 A.
Typ Vf = 0.25 V at 0.1 A.
1N5818 rated at 30 V, slightly lower spec Vf.