Hi, I'm working on iraudamp7s based amplifier, and I don't know why but the 2092 keeps resetting when I test it (4 ohms half bridge) with a crash cymbal on a virtual drum. I removed all external protections so the only thing is left is the 2092 ocp inside. One thing to note tho is that when I turn it on (therefore when the circuit is still cold) the 2092 doesn't reset. Anyone has some tip what to check? I suspect that the discrete 4019 could be counterfeit so the gate charge could be too high, I don't know.
There is an app note with quite a bit of info on the OCP at: https://www.infineon.com/dgdl/an-1138.pdf?fileId=5546d462533600a40153559a077610d1
Do you already have that?
Also, do you have a scope?
Should we presume that 4019 refers to IRFB4019PBF ? Something else?
Curious about the test signal too. Maybe there is something peculiar about cymbal signal, say, maybe frequency spectrum, DC offset, etc.?
Do you already have that?
Also, do you have a scope?
Should we presume that 4019 refers to IRFB4019PBF ? Something else?
Curious about the test signal too. Maybe there is something peculiar about cymbal signal, say, maybe frequency spectrum, DC offset, etc.?
Last edited:
Hi, thanks for your answer. I'm new at audio, but I read the AN a couple of times. Yes, I have a 50MHz 1GSa/s rigol. I couldn't print the screen cause I couldnt make "Ultrascope for DS1000E Series" to work on my pc, but I took a picture(the output signal when I applied the crash cymbal and the csd pin referenced to 0V)
What is the source of the crash cymbal? Is it a drum machine, a wav file, other?
Can you turn down the level of the cymbal to see if that makes any difference?
What about other test signals, is it only the cymbal that causes a problem?
IOW, it may be the problem has more to do with the test signal than the amplifier. Have you ruled that out?
Where did the amp come from, and why are you concerned parts might be counterfeit?
Do you have a schematic?
The scope isn't showing the cymbal waveform very well leading up to and at the moment of the reset. It might be helpful to try to get a better look at that.
Can you turn down the level of the cymbal to see if that makes any difference?
What about other test signals, is it only the cymbal that causes a problem?
IOW, it may be the problem has more to do with the test signal than the amplifier. Have you ruled that out?
Where did the amp come from, and why are you concerned parts might be counterfeit?
Do you have a schematic?
The scope isn't showing the cymbal waveform very well leading up to and at the moment of the reset. It might be helpful to try to get a better look at that.
Last edited:
Hi, thanks for the answer.
The source of the crash cymbal is https://www.musicca.com/pt/bateria
Apparently when I turn down the level, the 2092 protection is not triggered. The schematic is the same as IRAUDAMP7s from infineon/IR, I just changed the high side ocp resistors(R2=10K,R3=6.8K)
When I'm monitoring the output signal of the cymbal and the csd pin, it appears that the protection triggers so early that no output signal is sampled in the scope, I tried again and the result is the same as the pictures I sent. But when I turn down the level, therefore the IC dont reset, I can see the cymbal signal:
The source of the crash cymbal is https://www.musicca.com/pt/bateria
Apparently when I turn down the level, the 2092 protection is not triggered. The schematic is the same as IRAUDAMP7s from infineon/IR, I just changed the high side ocp resistors(R2=10K,R3=6.8K)
When I'm monitoring the output signal of the cymbal and the csd pin, it appears that the protection triggers so early that no output signal is sampled in the scope, I tried again and the result is the same as the pictures I sent. But when I turn down the level, therefore the IC dont reset, I can see the cymbal signal:
From the schematic it looks like the input is an analog signal. How about capturing that on one scope channel, and the OCP trigger on the other channel? Set the scope to trigger on the OCP trip, with the in single sweep mode (or maybe normal mode to help get trigger settings stabilized, just not auto sweep mode). Also try to set the scope trigger display point more towards the right side of the scope display so that a longer time view of the cymbal input before the trigger can be seen.
Also, most digital scopes have a USB connector for a usb memory stick (although some scopes may not like memory sticks with long file names in them). Scope screen captures can usually be saved in picture form, maybe .bmp or .jpg
Tends to be better than taking a picture with a camera, is all.
Also, most digital scopes have a USB connector for a usb memory stick (although some scopes may not like memory sticks with long file names in them). Scope screen captures can usually be saved in picture form, maybe .bmp or .jpg
Tends to be better than taking a picture with a camera, is all.
Last edited:
In my Dropbox for a limited time: https://www.dropbox.com/sh/esx0e3yvroycbv1/AABI-oFN4yEZyo_mE7HA0sAka?dl=0
I captured the input signal of the crash cymbal at the RCA connector and I don't think the amplitude was the problem, cause after the 2092 reseted a couple of times and I was still hitting the cymbal, I also noted that the amplitude was higher than it was when the the protection triggered. I used https://www.szynalski.com/tone-generator/ to scroll through all frequencies and see if it was triggering the OCP, which didn't happened. It only happens(as far as I know) with this crash cymbal
Please try that same test again with with a higher sweep speed. The cymbal waveform is too coarse. Can't see enough detail. We don't need to see the whole OCP trip event, only when it triggers. IOW, looks like the scope was set for 100ms per division horizontal sweep rate. Maybe try 1ms, or less. Even at 1ms/div, a 1kHz signal would be compressed into one horizontal division. A cymbal might have much higher frequency components than 10kHz, so at this very slow sweep speed too many details of what might be tripping the OCP are being lost.
Once we know what the input signal looks like in enough detail, then we need to figure out if the amp should be able to reproduce the input signal, or if tripping is correct under the circumstances. For that we will need to know the input signal amplitude in peak volts, the amplifier gain, and what the amplifier output current would be into a 4-ohm load.
Once we know what the input signal looks like in enough detail, then we need to figure out if the amp should be able to reproduce the input signal, or if tripping is correct under the circumstances. For that we will need to know the input signal amplitude in peak volts, the amplifier gain, and what the amplifier output current would be into a 4-ohm load.
Last edited:
I realise that this should probably take place on a different sub-forum but 0is the onset of digital clipping and having only 3.5 dB headroom is not ideal in my opinion however, as I have already said that is a matter for another thread of forum so I will just view this thread in future.
0dBFS is not the start of digital clipping in most modern dacs. Please see the article at: https://benchmarkmedia.com/blogs/ap...means that there are,peaks that exceed 0 dBFS.
Dont run 4 ohm bridged.
Its what Class D does in general. Shutdown at low or high frequency at high current.
If this is just a resistive 4 ohm test load.
On real world passive crossovers, specially live PA
Horrible crossover impedance curves for high frequency shut them down.
If you are using a speaker load, the impedance curve for the tweet likely dips
down to 3 ohms or lower. Bridged....plop it wont do that
That is also known real world problem.
Specially MI or Guitar software with high frequency distortion.
Or live with guitar effects like common distortion.
Class D just wants to shut down at high frequency.
Whatever rail voltage is ,half it, then blast it with full speed noisy fan.
all fixed
Its what Class D does in general. Shutdown at low or high frequency at high current.
If this is just a resistive 4 ohm test load.
On real world passive crossovers, specially live PA
Horrible crossover impedance curves for high frequency shut them down.
If you are using a speaker load, the impedance curve for the tweet likely dips
down to 3 ohms or lower. Bridged....plop it wont do that
That is also known real world problem.
Specially MI or Guitar software with high frequency distortion.
Or live with guitar effects like common distortion.
Class D just wants to shut down at high frequency.
Whatever rail voltage is ,half it, then blast it with full speed noisy fan.
all fixed
I tried with your cymbal track and after some hours, the protection triggered the same way. I captured on the scope the first micro seconds of the trigger along with the input signal. I was thinking that it could be a ressonance in the LC output filter, I don't know.
Seen from the amp an LC output filter of a class-d amp looks like a series resonant circuit. Without load it would present a short-circuit at its resonance frequency. Fortunately the load resistor gives relief in this respect by damping it. You were talking about a load resistance of 4 Ohms. Is it a purely resistive load that guarantees damping at higher frequencies also ? Have you considered using a lowpass filter before the amp ?
What happens at lower frequencies ? I.e. does the amp behave similarily in cold and warm state at lower frequencies ? How hot do your inductors get after some time of operation ?
Regards
Charles
It would be better to set the scope to display peak to peak volts on the blue channel. RMS is fine too. How it trips may depend on impending peak current, based on the input signal.
Also helpful to know what is the gain of the amplifier? IOW, how many peak amps into a 4R load is being demanded by the input signal?