Mic. phantom power decoupling

[aneevuser]

I have a couple of questions about the correct decoupling of mic. phantom power from an attached preamp. I was planning to use the arrangement shown on page 8 of this document:

http://www.ti.com/lit/ds/symlink/ina163.pdf

but with my homebrew NE5532/LF411 in-amp arrangement (see https://www.eevblog.com/forum/beginners/measuring-amplifier-noise-with-scope/ for more details).

This looks straightforward but...

1) why does note 1) suggest the use of nonpolar capacitors if phantom power is to be switched off - presumably there is the possibility of reverse polarity on the caps, but I can't see how from that circuit. What am I missing/misunderstanding?

2) we have the limiting diodes on the amp inputs presumably to prevent them being pulled to -48V when power is removed. This looks good but...I've discovered this document:

http://www.thatcorp.com/datashts/AES5335_48V_Phantom_Menace.pdf

which seems to cast doubt on the efficacy of that diode arrangement. I don't have the technical nous to be able to tell if their arguments are valid or not. They seem to be suggesting (if my understanding is correct - possibly doubtful) that you should at least aim for "low energy" decoupling + fast Schottky diode protection. Would anyone like to comment on that paper? I'm somewhat confused, as I have seen elsewhere the kind of diode limiting setup in the INA163 datasheet, so if op amps were routinely failing with it, I'd expect people to notice. What gives?

[fcb]

THAT paper makes alot of sense, and I've referenced it many years ago when designing outside broadcast gear.

INA163 datasheet Fig.5 runs the risk of damaging the the 1N4148's (see THAT paper) if there is a sudden short to ground - they aren't great diodes for pulse applications. I think you'll find experienced audio designers are likely to use something beefier with an INA163 (assuming they can afford to design one in - pro audio BOM's are TIGHT), or fitting series (Rs) resistors.

Answers:
1) When not using phantom power, there will be no overall bias on the electrolytics and they will be subject to reverse voltages.  If just a microphone levels then not really a real-world problem - but if the front end is designed to also take line levels, then there could be (+12dBV is 11Vpp) substantial reversing taking place.  If you are bothered, put two 100uF series back-back.
2) 48V Phantom is +48V w.r.t system 0V, not -48V.

[aneevuser]

INA163 datasheet Fig.5 runs the risk of damaging the the 1N4148's (see THAT paper) if there is a sudden short to ground - they aren't great diodes for pulse applications. I think you'll find experienced audio designers are likely to use something beefier with an INA163 (assuming they can afford to design one in - pro audio BOM's are TIGHT), or fitting series (Rs) resistors.

Will simply replacing the 1N4148s with Schottky diodes be enough?

Answers:
1) When not using phantom power, there will be no overall bias on the electrolytics and they will be subject to reverse voltages.  If just a microphone levels then not really a real-world problem - but if the front end is designed to also take line levels, then there could be (+12dBV is 11Vpp) substantial reversing taking place.  If you are bothered, put two 100uF series back-back.

By "front end", you mean the voltages at the amp inputs? Else I can't see how reverse polarity can occur. Regardless, I don't think that it'll be a problem for me.

2) 48V Phantom is +48V w.r.t system 0V, not -48V.

When phantom power is on, the mic. side of the caps are +48V w.r.t the amp inputs. If mic. side of the caps are switched to 0V, then instantaneously the amp inputs are pulled to -48V - that is the scenario that was on my mind, and which I'm assuming the limiting diodes are put in to protect against. Is this not correct?

[floobydust]

The input protection depends on your project's budget. It is common-place to have cost reductions there so products fail due to phantom faults.

The clamp diodes dump the -48V 47/100uF impulse into the -15V rail, which can kick down and destroy the IC or others on the rail if it goes to say -30V.

The THAT paper Schottky diodes SB160 have higher capacitance ~50pF/15V, so adding 100pF on each leg. 1N4148's are 1/5 of that, although they cannot take a 3A transient. Neither can a 1W zener diode.

[fcb]

If you can live with a few dB extra noise in your application I’d increase ‘Rs’ to reduce the peak current under fault conditions.

You can model the effect of the extra pF that beefier diodes would add.  If it gets desperate then put a couple of diodes in series, the capacitance divides.

As for over-voltage on the supply rails - you have to run the maths. It’s pretty common for audio applications to have quite a bit of bulk capacitance on the rails, so not that risky (think about 47uF dumping into 470uF - you won’t get more than a few volt rise), so likely no need for crowbars on the rails. Bulk capacitance might be cheaper than crowbars.

[aneevuser]

You can model the effect of the extra pF that beefier diodes would add.  If it gets desperate then put a couple of diodes in series, the capacitance divides.

Am I right in thinking that the only effect of the diodes at signal frequencies is to put \$\frac{C_d}{2}\$ capacitance per pair between the amp inputs? (with \$C_d\$ being the capacitance per diode)

As for over-voltage on the supply rails - you have to run the maths. It’s pretty common for audio applications to have quite a bit of bulk capacitance on the rails, so not that risky (think about 47uF dumping into 470uF - you won’t get more than a few volt rise), so likely no need for crowbars on the rails. Bulk capacitance might be cheaper than crowbars.

At the moment, it's a battery powered implementation, so I don't think I have to worry about power supply crowbars and the like.

[fcb]

If you stay well below the diode forward voltage, diode reverse breakdown voltage, have fairly low resistances in the circuit (so you don't have to worry about leakage currents), and aren't using glass diodes in a bright/fluctating enviroment - then just imagine the diode as a small capacitor.