IRS2092 Failure Modes

Hello everyone,

I have come across some strange behaviour of the IRS2092 class D driver IC that I have not seen reported before. In the documentation the only damage warning that I can find relates to Vss falling lower than the COM (B-) supply:

VSS Negative Bias Clamping
An excessive negative Vss voltage with respect to
COM could damage the IRS2092(S). VSS can go
below COM when a negative supply is missing in a
dual supply configuration. To protect the IC from this
possibility, a diode is recommended for clamping
potential negative biases to VSS. A standard
recovery diode with a current rating of 1A such as
the 1N4002 is sufficient for this purpose.

I have an ES1D diode in my design as per the above. My current project is based around the IRAUDAMP9 reference design, with the following changes:

Gate resistors 10r instead of 4r7, reverse turn-off diodes across gate resistors, and separate power supplies for +/- 75V, Vaa, Vss, and Vcc (from individual transformer windings, rectifiers, and filter capacitors). Vcc is created with a 20V supply feeding a 7812 regulator, referenced to B-. I have reverse polarity protection diodes between B- and the VCC supply (at the input side of the 7812), B- to GND, and GND to B+. Each of the five power supplies has a protection fuse. The amplifier works perfectly with low noise and distortion, switching characteristics are all good, and I have no problems using it for the past few weeks. Yesterday I encountered something that I cannot explain: I was performing some measurements and accidentally powered up the amplifier without the fuse installed for the Vcc power supply. Obviously the amplifier did not start and there was no fanfare. I switched off the main power switch, waited for all the supplies to discharge, and then reinstalled the Vcc supply fuse. Bringing up the amplifier using a variac, the B+ and B- fuses quickly blew...they have never blown before. Upon inspecting the amplifier I found a ~200 ohm short between B- and system GND. I removed the IRS2092 IC and found the short was now gone. I double checked all of the surrounding components, including the BJT totem poles and the MOSFETS. Nothing else was found to be damaged, so I installed a new IRS2092, slowly brought up the amplifier, and it is again working perfectly. I decided to do an experiment to test the UVLO for Vcc and see if I could duplicate the failure...this is supposed to stop switching if Vcc falls below approximately 8V. I replaced the Vcc fuse with a high current switch, powered up the amplifier until it reached normal operation, switching present, etc., and then flipped the switch, disconnecting the 20V Vcc feed supply. Switching stopped immediately once Vcc reached just over 8V relative to B-. So far so good. I then switched off the main power switch, with the supply to the Vcc regulator still disconnected (simulating a blown Vcc supply fuse) and again the IRS2092 was damaged in exactly the same way. Replacing it got the amp working properly again. One thing to note is that on turn off the Vaa and Vss supplies (and Vcc, when it is connected) discharge much faster than B+/B-. Has anyone encountered this or any other failure conditions? Can anyone recommend what I might look at or implement to prevent this failure mode? I suppose I could just eliminate the fuse from the Vcc supply, but I like to have fusing when testing and measuring a new design...for obvious reasons.

Thanks!
 
A follow-up to my post yesterday. I decided to dig into some more detail last night to try and find the source of the damage to the IRS2092 under power off conditions with no Vcc present. I decided to try powering the IRS2092 floating input power supplies (Vaa and Vss) derived from the main B+/B- rails rather than from external power supplies. I know that by doing this the discharge time where Vaa and Vss fall to below the UVLO condition would be much longer than with my external supplies because of much higher filter capacitance on the main rails. I used resistors from B+/B- to Vaa/Vss to provide 10ma as specified in the 2092 application note. I then repeated the same test as above: powered up the amplifier, waited until everything stabilized, turned off the Vcc supply, and then powered off the amplifier. As I suspected, Vaa and Vss took much longer to fall than with the external supplies and, thankfully, the 2092 was NOT damaged! So, I have some more investigating to do, but I am now thinking that there are some supply discharge conditions/timing that must be met that are not outlined in any of the IR documentation, otherwise damage will occur...wouldn't be the first time! Next I am going to look closely how Vaa and Vss discharge relative to one another, and then look at temporarily increasing the supply capacitance for the external Vaa and Vss supplies to see if slowing down the discharge rate slightly will prevent this fault condition. Again, if anyone has any input or suggestions, they would be most welcome.

Thanks!
 
Always use a single power source! + - up to 100V is suitable. The VB supply is derived from the +B supply and negative clamping via a diode stops the incorrect polarity on power down. VCC must be derived from the 0V line with respect to COM and -B, this avoids any incorrect polarities by way of the same diode. CSH (cut off/ mode point) is detected with a diode to +B. Many major manufacturers use this chip with the most basic of supplies and don't have issues with catastrophic failures. A complimentary pair as buffers on the drive outputs to the Hi and Lo gates also helps with speed issues on some FET gate capacitances.
Here is an explanation of its functionality; http://www.infineon.com/dgdl/an-1138.pdf?fileId=5546d462533600a40153559a077610d1
 
Last edited:
Hi Jon,

Thank you for your response. I am aware of most of what you have outlined and the amplifier works well. I was just surprised that I was able to damage the 2092 in the manner explained above as in no case can I see voltages or polarities violating what is outlined in the datasheet or the AN-1138 app note. In fact my idea to use external supplies for Vaa and Vss comes directly from page 14 of AN-1138:

For best audio performance, it is preferred to
produce VAA and VSS with external regulators, such
as the three terminal regulators. To keep internal
clamping zener diodes from conducting, the supply
voltage should be VAA < VCLAMPM+ and VSS > VCLAMPM-
. Standard 7805 and 7905 regulators are suitable.

So, if there were warnings required when those external supplies might behave differently than those derived from B+/B-, you'd think that information would be important to include!
 
Try to feed VAA/VSS with 1mA derived from +/-B supply, so that VAA/VSS supply will still be properly referenced if an external +/-5V supply does not exist or was powered down before +/-B supply.

From IRAUDAMP9 R50/R52
 

Attachments

  • Capture.PNG
    Capture.PNG
    42.3 KB · Views: 577
I have seen 2092's fail when a supply is missing, although I cant remember now which one was missing.

I had a problem with the 2092 on power down with great than 10,000uf per rail.
I would get a siren type noise through the speaker and just as the supply died a huge thump through the speaker. I had a chat with IR about it and they suggested a circuit that monitors VCC and holds the 2092 in reset if VCC goes below 12 volts.
I used a PIC micro with a2d and a opto isolator to do the job and it worked fine.

I found if for some reason an output mosfet goes then it often takes the 2092 with it.

As previously advised just use one power supply or you will have problems.
 
IRS2092 Damage

Hi,

I am using IRS2092 & follow design IRAUDAMP5.

Circuit works fine with output RC filter. But when I am connecting LC filter, sometimes at the time of power off / on, IC is not working. Pin 15 (Vb) & Pin 13 (Vs) found short. Pin16 (CSH) also showing low resistance some times.

Modification in ckt:
1. Resistance between Vb & +36rail is 26K.
2. I am making +- 36v using SMPS. (twoSMPS ech side - 24v & 12V). I am taking Vcc from lower side 12V SMPS..

Anybody can help?
 
Hi,

Regarding to Vaa/Vss supply discharge timing, from my experiment, at any moment there should always be Vaa available.
In other words, during power off, it's ok for Vss to discharge before Vaa.
But IRS2092 will be damaged if Vaa discharges before Vss.

I have another question regarding -B supply experiment.
Does anybody have a problem with -B supply left floating that damages IRS2092?