Hello
i've made a PCB to build a amp (UCD topology ) with IR2110 as driver.
it starts to oscillate with a 6.8 ohms power load connected at output.
F=280 KHz,low side gate measurement is quite good,no ringing.
after 10 mn,i have try to check pre-filter signal with a digital scope (without ground )
i saw a little spark when the scope probe touch the Vs pin of IR2110.(ground clip was attached to star ground,far away from 0V onboard )
10 seconds after,output MOS (IRF540N ),fuses and driver blown.
maybe,this a problem with parasitic capacitor probe,i don't know..
if anyone got an idea,advice...to avoid IR2110 destruction.
i've made a PCB to build a amp (UCD topology ) with IR2110 as driver.
it starts to oscillate with a 6.8 ohms power load connected at output.
F=280 KHz,low side gate measurement is quite good,no ringing.
after 10 mn,i have try to check pre-filter signal with a digital scope (without ground )
i saw a little spark when the scope probe touch the Vs pin of IR2110.(ground clip was attached to star ground,far away from 0V onboard )
10 seconds after,output MOS (IRF540N ),fuses and driver blown.
maybe,this a problem with parasitic capacitor probe,i don't know..
if anyone got an idea,advice...to avoid IR2110 destruction.
Attachments
hello,
thanks for reply,
there is no capacitor (100nF ) as close as possible to the Vs-VB and Vcc-com pins.
only 1 uF plastic film,1.5 cm (0.7 inch ) from the pin to capacitor
ground clip was connected to 0V,on the capacitor bank.
the scope earth is disconnected.i can make measurement on floating voltage.
i will try to put 100nF between Vs-VB,Vcc-com,
maybe i have to discharge the probe (shorting ground clip and probe ) between each mesurement.
is it possible that the ground clip was to far from the measurement point,it makes a large loop ?
i saw IR application note,they talk about diode between Vs and COM (for undershooting voltage ),but in this case
this diode is connected in // with the low side MOS,it acts as a freewhelling diode !
thanks for reply,
there is no capacitor (100nF ) as close as possible to the Vs-VB and Vcc-com pins.
only 1 uF plastic film,1.5 cm (0.7 inch ) from the pin to capacitor
ground clip was connected to 0V,on the capacitor bank.
the scope earth is disconnected.i can make measurement on floating voltage.
i will try to put 100nF between Vs-VB,Vcc-com,
maybe i have to discharge the probe (shorting ground clip and probe ) between each mesurement.
is it possible that the ground clip was to far from the measurement point,it makes a large loop ?
i saw IR application note,they talk about diode between Vs and COM (for undershooting voltage ),but in this case
this diode is connected in // with the low side MOS,it acts as a freewhelling diode !
Attachments
Last edited:
The oscilloscope is only floating at low frequencies, like 50/60Hz for line isolation, but nothing is floating at RF or even just at 10khz. Same applies to the power supply of your amplifier.
That's why you will be always creating an high frequency short circuit and destroying your amplifier every time you try to connect the ground of a conventional oscilloscope probe to a node being switched at high frequencies (or just with high dV/dt voltage slope).
You need a differential probe to do that, a special kind of active probe with one ground and two tips that senses the voltage difference between the tips.
And now, do you understand how dumb was the idea of connecting the ground of the probe to VB, VS or HO?
That's why you will be always creating an high frequency short circuit and destroying your amplifier every time you try to connect the ground of a conventional oscilloscope probe to a node being switched at high frequencies (or just with high dV/dt voltage slope).
You need a differential probe to do that, a special kind of active probe with one ground and two tips that senses the voltage difference between the tips.
And now, do you understand how dumb was the idea of connecting the ground of the probe to VB, VS or HO?
Why the oscilloscope probes must be floating to measure something between GND and testpoint? I understand, that it is required to measure the high side driver output voltage (connect between Vs and Ho), but the measure using normal GND does NOT require floating probes. And the sensing probe was connected to Vs - switching node with huge current capability - this would more probably destroy the oscilloscope input, rather than mosfets.
Please, notice the absence of the groundplane on input part in the photo. I think, the reason was in parasitics, which were produced by the osc probe (GND wire was connected far away from switching node with high inductance inbetween). These parasitics have turned the amp to be not an UcD, but to be a simple self osc amp with prefilter feedback... And such uncontrollable prefilter feedback can cause very high switching frequency, causing overheat of the mosftes (in few seconds). Also, experimenting with class D amp with output mosfets without any heatsink is pretty dangerous. Any uncotrollable behaviour will cause uncontrollable heating of them. I have also killed pair in such way
Please, notice the absence of the groundplane on input part in the photo. I think, the reason was in parasitics, which were produced by the osc probe (GND wire was connected far away from switching node with high inductance inbetween). These parasitics have turned the amp to be not an UcD, but to be a simple self osc amp with prefilter feedback... And such uncontrollable prefilter feedback can cause very high switching frequency, causing overheat of the mosftes (in few seconds). Also, experimenting with class D amp with output mosfets without any heatsink is pretty dangerous. Any uncotrollable behaviour will cause uncontrollable heating of them. I have also killed pair in such way
Hi
Some here above have hinted at the possible failure modes.
For the future when sensing high dV/dt low impedance source/s. Use an on board test point with a resistive voltage divider say at least 10-20 dB with a 50 ohm termination at the scope to keep the bandwidth. ( an old RF trick we call a sniffer probe)
Some here above have hinted at the possible failure modes.
For the future when sensing high dV/dt low impedance source/s. Use an on board test point with a resistive voltage divider say at least 10-20 dB with a 50 ohm termination at the scope to keep the bandwidth. ( an old RF trick we call a sniffer probe)
Then sorry, I didn't understand you properly, and from time to time I have to deal with people that are trying to connect the ground of a conventional scope to a switching node and not understanding why it's such a bad idea.
The spark is a sign of a circuit being closed and current flowing through a low impedance path, so there could be something wrong with the probe or the input of the oscilloscope, for example, exceeding voltage range.
For accurate measurements the probe ground and tip should be connected to close places. The long ground lead of the probe can act as an antenna and will radiate and pick up stuff and disturb the waveforms that you see in high speed switching circuits. A trick is to bypass it using a short piece of solid wire to connect the ground ring near the tip directly to a suitable place near the measurement point.
I usually create test points where the probes fit by soldering two pieces of resistor (or other component) lead to the PCB and bending them to hold the probe while touching the tip and the ring. This should be done with care so that a short is not easy to create if the probe slips a bit.
The spark is a sign of a circuit being closed and current flowing through a low impedance path, so there could be something wrong with the probe or the input of the oscilloscope, for example, exceeding voltage range.
For accurate measurements the probe ground and tip should be connected to close places. The long ground lead of the probe can act as an antenna and will radiate and pick up stuff and disturb the waveforms that you see in high speed switching circuits. A trick is to bypass it using a short piece of solid wire to connect the ground ring near the tip directly to a suitable place near the measurement point.
I usually create test points where the probes fit by soldering two pieces of resistor (or other component) lead to the PCB and bending them to hold the probe while touching the tip and the ring. This should be done with care so that a short is not easy to create if the probe slips a bit.
Last edited:
thanks for helping
probe is 10:1 in this case
i will do some measurements points onboard,(with piece of metal to grip ground clip and probe correctly ) the way as you describe EVA.
btw:screenshoots we can see for high side gate measurement are always made with differential probe ?
probe is 10:1 in this case
i will do some measurements points onboard,(with piece of metal to grip ground clip and probe correctly ) the way as you describe EVA.
btw:screenshoots we can see for high side gate measurement are always made with differential probe ?
Last edited:
The ringing is a LC resonance between power MOSFET, PCB inductance and local power supply decoupling. Resonances in output inductor and capacitor may be contributing too. It has to be damped with R(L)C networks if you want decent EMI performance.
Removing resonances from supply rails and from the switching and output nodes is a kind of science on itself.
First step is to optimize layout for low parasitics so that the resonances are shifted as high in frequency as possible. For example, in my current project they are around 100Mhz before damping is added. It's so high because the lead inductance of the TO-220 MOSFET dominates, I made layout inductances almost negligible. Damping resonances in power paths below 50Mhz becomes more and more difficult as snubber losses become progressively higher. On the other hand, my 100Mhz resonance is damped with just a 0.25W thru-hole resistor (which does the RL part) and a small chip ceramic capacitor (easy considering that the amplifier does 5KW at 2 ohm).
Removing resonances from supply rails and from the switching and output nodes is a kind of science on itself.
First step is to optimize layout for low parasitics so that the resonances are shifted as high in frequency as possible. For example, in my current project they are around 100Mhz before damping is added. It's so high because the lead inductance of the TO-220 MOSFET dominates, I made layout inductances almost negligible. Damping resonances in power paths below 50Mhz becomes more and more difficult as snubber losses become progressively higher. On the other hand, my 100Mhz resonance is damped with just a 0.25W thru-hole resistor (which does the RL part) and a small chip ceramic capacitor (easy considering that the amplifier does 5KW at 2 ohm).
Last edited:
Hi,
your simple amp is important to increase your knowledge.
I have small suggestions for you. (see pic).
I also have two questions (my curiosity only):
What you used just before IR2110?
You can see pic with 1kHz input signal and 80Hz/40Hz, just before clipping?
before you decide to make changes?
your simple amp is important to increase your knowledge.
I have small suggestions for you. (see pic).
I also have two questions (my curiosity only):
What you used just before IR2110?
You can see pic with 1kHz input signal and 80Hz/40Hz, just before clipping?
before you decide to make changes?
Attachments
Hi AP2,
yes,the change you re' suggesting have been done exept the MOSFET snubber and the serial resistor with Vb capacitor.
power supply capacitors are close as possible to the MOSFET,on the other side of PCB
carrier snapshot above was taken with these modifiacations.
i've tried with a 4.7 ohms on bootstarp capacitor,no change on output ringing.
voltage ringing is not very important,maybe a snubber gonna cancel it,i will try after testing the overall PCB and behaviour with audio.
i will do somme THD and IMD measurement today.
saturation with audio signal,on 6.8 ohms 300 w dummy load
it's quite clean.
yes,the change you re' suggesting have been done exept the MOSFET snubber and the serial resistor with Vb capacitor.
power supply capacitors are close as possible to the MOSFET,on the other side of PCB
carrier snapshot above was taken with these modifiacations.
i've tried with a 4.7 ohms on bootstarp capacitor,no change on output ringing.
voltage ringing is not very important,maybe a snubber gonna cancel it,i will try after testing the overall PCB and behaviour with audio.
i will do somme THD and IMD measurement today.
saturation with audio signal,on 6.8 ohms 300 w dummy load
it's quite clean.
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.