The 48V Phantom Menace Returns

[mediatechnology]

Here's a paper Ros Bortoni and I recently wrote about the phantom menace. Ros has just presented it at AES 127.

http://www.thatcorp.com/datashts/AES127-000183.pdf

We managed to discover a differential fault mode that's particularly destructive.

Previously, the diode bridges to the supply rails have been assumed by many to clamp phantom fault transients. They don't - at least without some "help."

[JR.]

Hi Wayne... congrats on getting published in AES.

The observation about transient interfering with PS rail may explain one difference between my (low) failure experience in mixers and your bench.. I suspect the consoles may have more rail capacitance to soak up the current. (Also I typically used 22 uF).

In you final recommended circuit (fog 24), did you consider the possibility of either a resistor between the ground and common point of the two added PS 47uF? Or perhaps a 3 cap version where one cap spans between +/- rails with two smaller (or not) caps from rails to ground.

Interesting discussion.

JR

[mediatechnology]

Thanks John.

The observation about transient interfering with PS rail may explain one difference between my (low) failure experience in mixers and your bench.. I suspect the consoles may have more rail capacitance to soak up the current. (Also I typically used 22 uF).

I agree. And there's a lot of other load absorbing that energy too in a large format console like all the NE5532s, TL07Xs and LMXXXs. Things that survive when installed in the console might not survive similar faults in FT.

If there's no place for the energy to go, like a USB-powered box, digital camera, or small-format gizmo, then the rationale is to give it some place to go.

There's quite a wallop but if it's spread out over a lot of devices and Cs (large console) or steered to another capacitor (some place for energy to go) far less destructive.

I haven't thought about the alternate capacitor connection or third C. The third C does make sense. We want to do a third episode.

[JR.]

Wayne's paper seems to be focused on the energy stored up in the phantom blocking capacitors.

For line level inputs in my experience the common faults come from static hits or mains power (at least in live sound reinforcement). The diode clamps can do a good job absorbing static hits, but 120-240V mains voltage will usually release the smoke from the protection circuits and the parts...

In live SR with the popular use of 1/4" cables for both line level signals and speaker feeds, there is also the opportunity to hit line inputs with speaker level audio signal... the diode clamps are usually adequate to mange this, but unprotected inputs could fail.

JR

[mediatechnology]

We figured that input and output phantom protection of line-level stages was a lot to throw in. There may be a sequel.

To answer Roger's question, the published 1646 protection using bridge to rails, works fine. One thing I have found with the 1646 is that OutSmarts makes the part tolerant to stored charge. I haven't sorted that out yet.

[JR.]

?? I was talking input and output... In general outputs require diode clamps to rails just like inputs. In fact one pretty common failure mode is output mute FETs if they don't have protection diode clamps. Again probably from static hits...

Of course I don't know what THAT have built in. I recall years ago talking to the sundry dog an pony shows that would come through Peavey once every few years trying to sell their latest and greatest output balanced drivers.

Since, they were trying to replace <$0.50 worth of opamps and glue parts, I needed them to build in more of the glue... protection, turn on/off muting/and even a few dB of voltage gain. In my ideal output driver definition they would accept hot unregulated rails. detect rail collapse for automatic turn off muting. Include a few dB of gain so they would be able to use full voltage swing of unreg rail, etc.

I never saw such a part and never gave them their high volume design in.. Of course then was then.. I don't know what PV is doing now.

JR

[JR.]

In fact when wireless microphones were new, the engineers who designed them didn't account for phantom blocking caps. Since these wireless receivers were actually plugged into mic inputs on purpose, and most value consoles/mixers use global phantom power switches so you can't turn off the phantom on individual inputs. Many of these first generation wireless mic receivers would hand grenade before reaching old age.

JR

[JR.]

The discrete output mute fets could be damaged by simple voltage spikes. If using a N channel mute JFET with gate sitting at minus unregulated (unmuted) it won't take the full +48 added on top of say -24 to exceed its Vgs max and release it's magic smoke, (it's only a technicality that 48v is attached to FET drain, since fets are symmetrical.)

Most (modern) IC failures are probably parasitic (or intentional) substrate diodes that conduct back into the PS rails...As Wayne's paper measured, the fully charged Phantom caps can dump a few amps of current transiently... Outputs usually have some build out resistance but enough voltage can make enough current to kill unprotected circuitry.

I don't have specific knowledge about THAT chips so will defer to Wayne or others.

JR

[mediatechnology]

The 48V Phantom Menace returns has just been published in the March, 2010 Journal of the AES.

EDIT: This link is now broken: http://www.aes.org/tmpFiles/JAES/201004 ... _PG197.pdf

[JR.]

nice job...

JR