Just wondering how these amps work in real life. At the moment, I am simulating them to get an understanding. Anyway, during simulation with 0V input the noise floor is real flat around -120 dBV (up to around the switching freqency). But, change the input to something real, say 2khz tone, the noise floor jumps up drastically to around - 50dBV.
Is it my simulation, and maybe I am missing something, or do Class D amps behave this way in real life?
Oh, at the moment, the simulation is using all ideal components and no feedback.
Is it my simulation, and maybe I am missing something, or do Class D amps behave this way in real life?
Oh, at the moment, the simulation is using all ideal components and no feedback.
You mentioned the test frequency, but neglected everything else like gain, power level, type of amp (self oscillating etc)
At higher power the noise floor might increase slightly, but I don't think it's any kind the sort of offender a class B amp would be.
Possibly due to increased EMI, especially with a bad layout. More likely due to parasitics being worked a little harder at higher power levels, which is what a simulator might actually be showing. In reality your filter inductor /core material would play a major roll in how it changes with power levels.
Either way on an extremely dynamic recording with a good D amp, you can crank the volume right up and the noise floor you end up hearing is still that of the recording itself. So that of the amp must be a good measure lower than the majority of recordings out there.
At higher power the noise floor might increase slightly, but I don't think it's any kind the sort of offender a class B amp would be.
Possibly due to increased EMI, especially with a bad layout. More likely due to parasitics being worked a little harder at higher power levels, which is what a simulator might actually be showing. In reality your filter inductor /core material would play a major roll in how it changes with power levels.
Either way on an extremely dynamic recording with a good D amp, you can crank the volume right up and the noise floor you end up hearing is still that of the recording itself. So that of the amp must be a good measure lower than the majority of recordings out there.
I left out the details because I was really just looking for a real world comparison vs simulation. Since you ask:
1. not self oscillating, 400Khz triangle wave feeds comparator.
2. Open loop. No feedback until I understand this portion of the design.
3. All components are "ideal".
4. Both 1/2 bridge and full BTL topologies.
Can you tell me how much of the switching noise leaks out to the speaker? This test circuit isn't optimized by any means, but with no audio input, the amplitude of the switching noise (at the speaker) is around 0.5 Vp-p. Seems high to me, but I have no point of reference.
1. not self oscillating, 400Khz triangle wave feeds comparator.
2. Open loop. No feedback until I understand this portion of the design.
3. All components are "ideal".
4. Both 1/2 bridge and full BTL topologies.
Can you tell me how much of the switching noise leaks out to the speaker? This test circuit isn't optimized by any means, but with no audio input, the amplitude of the switching noise (at the speaker) is around 0.5 Vp-p. Seems high to me, but I have no point of reference.
.5V pk-pk is great.
That would be the switching noise leaking out to the speaker there. It should also be at fairly low power.
The speaker's impedance at that frequency is high, time constant of it is long, there's no physical way the speaker can react to it at all, anymore than you could hear it if it did.
The biggest problem of that high frequency ripple is EMI from the speaker wires. If you have long speaker wires or just want to attenuate that ripple even further to help prevent EMI you can simply employ a common mode choke on the speaker outputs.
Some of the older threads in this forum really cover alot of material, even if they're kind of long, worth reading, should help in your quest alot.
That would be the switching noise leaking out to the speaker there. It should also be at fairly low power.
The speaker's impedance at that frequency is high, time constant of it is long, there's no physical way the speaker can react to it at all, anymore than you could hear it if it did.
The biggest problem of that high frequency ripple is EMI from the speaker wires. If you have long speaker wires or just want to attenuate that ripple even further to help prevent EMI you can simply employ a common mode choke on the speaker outputs.
Some of the older threads in this forum really cover alot of material, even if they're kind of long, worth reading, should help in your quest alot.
.5V pp is good, eh? Well that's good to hear (so to speak)! Again, it's in simulation with ideal components.
how would a common mode choke work in a single-ended app (half bridge)? If I use an L-C filter with more poles, that should help for single ended.
I will look through those posts - thanks.
gene
how would a common mode choke work in a single-ended app (half bridge)? If I use an L-C filter with more poles, that should help for single ended.
I will look through those posts - thanks.
gene
getting back to the results . . .
With simulation step size set to [.0001 / 16384} (good compromise between accuracy and simulation time), and the FFT set to 16384 points, there results are reasonable. The floor is at around -90 dBv (litte trick here, don't start the FFT at time=0 because it includes the startup stuff). Vin is a 2KHz sine wave, Vswitching is 400KHz triangle.
The SFDR is really lousy at around 65 dB for the BTL version and around 48 dB for half-bridge. The harmonics are everywhere. Is this also a feature of class-D or perhaps I am doing something wrong.
Take a look at the plots here is the half-bridge output. . .
With simulation step size set to [.0001 / 16384} (good compromise between accuracy and simulation time), and the FFT set to 16384 points, there results are reasonable. The floor is at around -90 dBv (litte trick here, don't start the FFT at time=0 because it includes the startup stuff). Vin is a 2KHz sine wave, Vswitching is 400KHz triangle.
The SFDR is really lousy at around 65 dB for the BTL version and around 48 dB for half-bridge. The harmonics are everywhere. Is this also a feature of class-D or perhaps I am doing something wrong.
Take a look at the plots here is the half-bridge output. . .
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.