Elvee said:
Gottlob Frege spent all his life on "Die Grundlagen der Arithmetik". The important was to find solid theoretical axiomatic grounds for arithmetic: only after, mathematics would have had solid applications.
At the end of his life a young 25 years old English guy, Bertrand Russel, wrote to him showing a couple of paradoxes logically deductible from Frege axioms. Frege had to admit that his beautiful construction had still empirical basis.
In 1935 Kurt Goedel published the logical demonstration about the impossibility of an axiomatic for arithmetic.
After Goedel, the dream of "technical perfection" before experience definitively ended.
Elvee said:
Your description is what I mean by sounding better, and I totally agree with what you said. Additionally, even with the same instruments, there are different test methods that should be explored. Normally after you are satisfied with your measurements, it's the ears that are not satisfied. That's when it's necessary to explore deeper. I think class4sure is not talking at that going deeper stage yet, just at the theory and PC grade measurement stage. He's just yanking chains to dig for mor information.
At this point, only Lars has presented some data. Of course he also has a desire to improve his product.
Purpose of the output inductor of class D amps is more like designing filters for a DAC. You are trying to recreate a continouse signal from discrete signals by smoothing out those discrete portions so that they are continuous. The difference is that the time between each disctrete segment right out of a DAC is close to the same whereas in class D this will vary. So it's not just winding and core material, but how the design matches the switching portion of the class D amp.
soongsc said:
The output of a DAC will look more like a staircase waveform, discrete in time and voltage level. That's a bit different from the PWM output signal which has only two voltage levels and an infinitely variable (in theory) duty cycle. Although the purpose of the filter is still the same - to reconstruct the input signal - the PWM filter will most likely be operating at a much higher power level.
Since we are dealing with much more power in this area, we cannot use any kind of notching as we do with voltage sources, because this means consuming power at those frequencies by applying low impedance.
If we just use a normal choke, there is no way the high frequency energy can be dissipated, so it will just reflect back into the circuit, which is not something we want. What we want is for the energy of the audio band to pass though, but the switching related frequency content energy be lost. Normally this would be in the form of heat.
If we just use a normal choke, there is no way the high frequency energy can be dissipated, so it will just reflect back into the circuit, which is not something we want. What we want is for the energy of the audio band to pass though, but the switching related frequency content energy be lost. Normally this would be in the form of heat.
BWRX said:
If I remember correctly, the two voltage PWM signal is up to the gates. Into the load it's already a reconstructed wave form that still has some PWM signal in it depending on how fast the switches are. The faster the switches, the worse the case. So the switching need to match wiith the output filters, and varying of the switching frequency is also an important part. So you cannot just look at the output filter alone and hope for the best.
The very purpose of a class D amplifier is to improve efficiency by reflecting back any superfluous power into the supply rails. It is the role of the diodes across the switching elements.soongsc said:
Optionally, a dissipative notch filter can be added after the main choke to improve the cleanliness of the output, but the power dissipated at this stage will be very small.
LV
Elvee said:
We don't want any energy reflected back from the chokes. As far as reflecting back into the power rails is something I need to look into a bit more, but it does not seem good if this reflection causes fluctuation in the power rails. Probably this is why the PSRR is so low in class D designs?
PSRR is low in many class-D designs because they don't use post-filter NFB. In an open-loop design any change is supply voltage will cause an equivalent change in output voltage.
The diodes connected across the switching FETs are there to provide a path for the current flowing in the filter inductor at the point the FET switches off. This current path is an essential part of a Class-D design. Compare the role of the diode in a buck dc-dc regulator PSU design. This is all basic knowledge for SMPSU designers and should be for anyone designing Class-D amps.
The diodes connected across the switching FETs are there to provide a path for the current flowing in the filter inductor at the point the FET switches off. This current path is an essential part of a Class-D design. Compare the role of the diode in a buck dc-dc regulator PSU design. This is all basic knowledge for SMPSU designers and should be for anyone designing Class-D amps.
Ouroboros said:
This is a common misunderstanding. The pre / post filter scheme has absolutely no bearing on PSRR. However open / closed loop has.
One way to get rid of the RF transmission problem in the coil, is to place a small open aircoil to 'eat' the high slopes before they hit the real filter coil. This way you can wind a normal simple filter coil without considering the RF properties.
Oh my, what part of 'keeping objective comments about sound quality out of this discussion' did you not understand?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Lars,
I disagree, you're just inserting another passive pole in the loop, in the feedback loop in case of post filter feedback. Either way this is a solution that seems to stem from a misunderstanding of the fact that a properly calculated control loop will behave properly with RF too, provided you have a proper layout and a properly wound coil. I hate to bring it up again as it makes me sound like a fanboy, but Bruno seems to be able to manage this quite well.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I disagree, you're just inserting another passive pole in the loop, in the feedback loop in case of post filter feedback. Either way this is a solution that seems to stem from a misunderstanding of the fact that a properly calculated control loop will behave properly with RF too, provided you have a proper layout and a properly wound coil. I hate to bring it up again as it makes me sound like a fanboy, but Bruno seems to be able to manage this quite well.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Lars,
And obviously one other way to keep RF down or from propagating through the coil is to make sure RF spiking is minimized by using snubbers and a proper dead-time setting for the MOSFETs, but I'm guessing the importance of snubbers is clear to you, no?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
And obviously one other way to keep RF down or from propagating through the coil is to make sure RF spiking is minimized by using snubbers and a proper dead-time setting for the MOSFETs, but I'm guessing the importance of snubbers is clear to you, no?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.