LM3875 abuse vulnerability

[Puffin]

The LM3875 has output protection from a short to ground, and the SPiKe protection is said to protect a short to the supplies.

I managed to blow one up when checking the rail voltages by attaching the DMM to V- before securing the other probe to grd.

Should this have happened?

[danielwritesbac]

Two possibilities that occur to me:
There was no cap at CI, or
Electrostatic discharge.

Its good that the chip simply quit--because the usual malfunction is full rail voltage into the speaker.

[Puffin]

Ci is an optional feeback cap to ensure unity gain at DC?

[danielwritesbac]

Originally posted by Puffin
Ci is an optional feeback cap to ensure unity gain at DC?

The cap at CI may unlevel the frequency response or cause other aberrations common to signal going through capacitors. This is why it is often omitted. That's less durable, but its easier.

To your question: Yes. The amplifier is more durable and dynamic with that cap in place. Although its difficult to select a capacitor for seemly results at CI, it IS possible and you do get the benefit of larger dynamics, a variety of input impedances available (see page 6), as well as more powerful low bass (See overture design spreadsheet) along with less clipping; and, you get a more durable amplifier.

[danielwritesbac]

Well man, really, I think the reason it broke is that a power cap suddenly discharged, probably with the help of the voltmeter probe and some static electricity as a bonus (ESD).

What I was trying to say, is that without CI in place, the LM3875 would have tried to amplify the zap by about the gain factor, and blew itself out. That's my guess on it.

[Puffin]

anyone else got a take on this? The PS rail cap (1,000uf) survived.

[pacificblue]

Originally posted by Puffin
The LM3875 has output protection from a short to ground, and the SPiKe protection is said to protect a short to the supplies.

I managed to blow one up when checking the rail voltages by attaching the DMM to V- before securing the other probe to grd.

Should this have happened?

Here is a quote from National's Application Note 898 (John deCelles, 1993)

It should be noted that
SPiKe protection was enabled after 200 us of current limiting
in Figure 23 and Figure 24, but is in general dependent upon
the case temperature, the transistor operating current and
voltage, and its power dissipation versus time.

The protection circuit has a reaction time. The short could have blown the transistor(s) before SPiKe could react.

Here is another quote from the same document

One note to make about this protection scheme is that the
current limitation is not sustained indefinitely. In essence, the
output shorts to either supply rail should not be sustained for
any period of time greater than a few seconds. Frequent
temporary shorts from the output to either supply rail will be
protected, however, continued testing of the circuitry in this
manner is not guaranteed and is likely to cause degradation
to the functionality and long-term reliability of the device.

This means that the short must not be too long either (no pun intended), and that SPiKe is neither immortal nor infallible.

[danielwritesbac]

We were just discussing fuses on another thread, so howabout some fast-blo at the speaker output + side? The no-load operating tolerances are so much higher.

OMG! This could help me too! I managed to somehow purchase a boxfull of counterfeit LM1875 and while usable as TDA2050's, they are prone to mysterious failures. Now counting in a fuse holder, the wonderful sale price was not so economical after all. Oh well. Its probably more-seemly to blow a fuse instead. It might give me some clues as to the cause of failure, because of such immediate notice that a fuse can give.

Thanks for the ideas!!

[AndrewT]

a fuse is too slow to protect the semiconductors.

Where a fuse can be good is for long term overload when a current limiting circuit prevents immediate damage to the semiconductor, but continued power dissipation increases the device temperature.

[Puffin]

This was a split second flash as the probe connected!