Hi All!
I am currently designing a Class D Amp based on TIs TAS5504 and TAS5152. I use the TAS5152 in 2x BTL configuration with 2x 10uH and 1x 0.47uF for each BTL. PVdd is 36V
So far it's working, but most of the coils I tested for the output filter heat up for 20-30°C above Tamb in idle (!) mode.
And those Coils are listed as "Coils for Class-D Amps". So I guess the resonant frequency of the core shouldnt be a question.
But to check it: (1) what should be the resonance frequency of the core when switching is done at 384kHz?
Interesting is that the coils do not heat up that strong when I place them in the TI eval-Board. So (2) my layout could be a reason? (I have a very compact board)
Or could there be some jitter on the PWM caused by a decoupling failure? (3).
I measure betweeen 0.5-1.0V Vss ripple at 384kHz. This seems to be a normal value to me (which was also confirmed by a Spice simulation).
Any help, hints or notes would be highly appreciated!
Thx and best regards,
Klaus!
P.S.: I foud the "heating in inductor thread", but this ended in nowhere
I am currently designing a Class D Amp based on TIs TAS5504 and TAS5152. I use the TAS5152 in 2x BTL configuration with 2x 10uH and 1x 0.47uF for each BTL. PVdd is 36V
So far it's working, but most of the coils I tested for the output filter heat up for 20-30°C above Tamb in idle (!) mode.
And those Coils are listed as "Coils for Class-D Amps". So I guess the resonant frequency of the core shouldnt be a question.
But to check it: (1) what should be the resonance frequency of the core when switching is done at 384kHz?
Interesting is that the coils do not heat up that strong when I place them in the TI eval-Board. So (2) my layout could be a reason? (I have a very compact board)
Or could there be some jitter on the PWM caused by a decoupling failure? (3).
I measure betweeen 0.5-1.0V Vss ripple at 384kHz. This seems to be a normal value to me (which was also confirmed by a Spice simulation).
Any help, hints or notes would be highly appreciated!
Thx and best regards,
Klaus!
P.S.: I foud the "heating in inductor thread", but this ended in nowhere
The core losses take place also at idle because of the 50% PWM signal. I guess the temperature rise you are experiencing is normal (just for empirical reasons). I had a problem with a coil that heated up and after a while a really annoying noise prevented listening to the amplifier. I repaired it by installing a small computer fan
But anyway one should check the core losses with given frequency & voltage. If I remember right, those are the parameters that affect the core losses. But check them somewhere!
I have also had that kind of ripple in my amplifiers.
But anyway one should check the core losses with given frequency & voltage. If I remember right, those are the parameters that affect the core losses. But check them somewhere!
I have also had that kind of ripple in my amplifiers.
When I first stated designing class-D amps, I tried various types of output filter choke core. Most of the dust-iron cores sold for winding output inductors in buck regulators have far too high hysteresis losses when used at the 300kHz or so in a class-d amp. For the last couple of designs I have done for the company I work for, I have used gapped RM cores in Ferroxcube 3F3 material. (Epcos N87 ferrite also works well). The white ferrite toroids with the precision-cut gap from Ferroxcube also work well, but are difficult to buy in small quantities.
The white "Polo Mint" gapped toroids from Feroxcube are in 3C20 material. the characteristics of this ferrite grade are at:
http://www.ferroxcube.com/prod/assets/3c20.pdf
http://www.ferroxcube.com/prod/assets/3c20.pdf
I don't know what kind of resonance do you refer to, but anyhow, this is irrelevant.
Well, I just know from speaker-cabinets that you should use them only/ideally at frequencies above double then the resonant frequency. So i expected that you should use coils in a similar way. With voltage/current frequencies below the resonant f.
But if it doesnt matter, it's fine!
The core losses take place also at idle because of the 50% PWM signal. I guess the temperature rise you are experiencing is normal (just for empirical reasons). I had a problem with a coil that heated up and after a while a really annoying noise prevented listening to the amplifier. I repaired it by installing a small computer fan
But anyway one should check the core losses with given frequency & voltage. If I remember right, those are the parameters that affect the core losses. But check them somewhere!
I have also had that kind of ripple in my amplifiers.
I see - I have a basic misunderstanding here: I though: no current through the coil, so no heat! But the heat is generated by the toggling "re-magnetisation" through the switching voltage within the core?!
(or am i again wrong...?)
(but the energy creating the heat in the core must come from somewhere. So there must be some current flow. isnt it? hmmm.....
)I see - I have a basic misunderstanding here: I though: no current through the coil, so no heat! But the heat is generated by the toggling "re-magnetisation" through the switching voltage within the core?!
(or am i again wrong...?)
(but the energy creating the heat in the core must come from somewhere. So there must be some current flow. isnt it? hmmm.....
Yes, absolutely, there is a significant current! Ipeak=PVdd/(8*L*fsw)=1.2A. But this current would be reactive (non-energic) in ideal case.
Yes, there is current flowing backwards and forwards through the inductor at the switching frequency. This is what gives the ripple voltage across the filter capacitor. If you use a core with the wrong characteristics (such as the dust-iron toroidal cores used for low-frequency buck regulators) this high-frequency current can cause quite a lot of core loss, causing the core to get hot. T-106 toroids and MPP toroids are better, but I still far prefer gapped ferrite cores in proper high frequency ferrite materials such as 3F3 and N87.
I believe the coil appears as a high impedance to the PWM signal - is that incorrect? There should be very little current flow through the coil as a result of the 50% PWM signal.
If the amp is idling, then the output will be at nominally 0V.
Assume the amp is running on +/- 45V rails, and has an output inductor of 20uH.
di/dt=V/L (45/20E-6 gives 2.25 amps per microsecond when a MOSFET turns on)
If the oscillation frequency is 300kHz, then the high and low MOSFETS are each on for approx 1.67usec. The alternating current in the inductor will have peak values of approximately +/- 3.7A.
Assume the amp is running on +/- 45V rails, and has an output inductor of 20uH.
di/dt=V/L (45/20E-6 gives 2.25 amps per microsecond when a MOSFET turns on)
If the oscillation frequency is 300kHz, then the high and low MOSFETS are each on for approx 1.67usec. The alternating current in the inductor will have peak values of approximately +/- 3.7A.
Yes it's very little - compared to the maximal signal current. About 10% of it. And then it is shunted by the capacitor, this because it doesn't go out, but inside the amplifier it flows.
Apologies, you're quite right Pafi, 3.7A pk-pk. It's still enough of a flux change in the core to heat it up quite noticeably if the core is either too small, or if the material isn't adequate for operation at these frequencies.
This is something I expected! But my PSU and another digital ampere-meter just show 200-300mA current in idle mode. This is about the current consumned by ADC+Clock+uP+TAS+LEDs.
No more current left for heating any coils...
I think I have to use another measuring method and find out the correct value of the current.
If the frequency of the current would be 380kHz. Maybe this is not suitable for the PSU and my ampere-meter...?
(Hope to find the old analog amperemeter somewhere in the cellar...
)P.S.: Ipeak=PVdd/(8*L*fsw): is this I = U / R with R = 2*PI*f*L (2*Pi rounded up to 8)