I followed this one, and works like a charm
pcb really is everything here, more then components themself
No more ringing? Or just no more shut down?
In any case good to know that this specific core seems to add the required damping in the 10MHz range.
There is still a little ringing but much reduced on the previous core.
It no longer shuts down.
Hm... hearing now the second time (in a very different application) that a desirable damping could be achieved with such cores, without causing undesired heat by the main magnetization.
Even if I tend to swank around, how to avoid or dampen HF ringing of class D amps by the counter measures as described above - I will keep these metal powder cores in mind. ...may be this inert damping is also fortunate for EMI.
I used a power inductor to start with and that got to 120 degrees C !
The t1-6-2 iron powder core runs cold.
There was little info on it other than it was a 22uh 16 amp power inductor
I got it from Farnell.
They will actually run on my amp but they get hot and smelly.
The new inducter I should have said is a T106-2 core with about 20-30 turns of 18swg on it.
Last edited:
When working with a class D amp try to think of it as a high speed switching power supply that drives a speaker thru a low pass filter. I design hi speed DC-DC converters for a living from 50Khz up to 3-4 Mhz. The layout of the PCB, routing traces and placement of the components are more critical than the parts you select.
There is an exception to this, the output filter inductor has to be constructed of a ferrite material that has low core loss int he frequency range that the class D is switching at. Most switch in the 200-500Khz range so if you use a core good for a switching DC supply then it should also have low core loss & low dissipation. You also need to select an inductor or design one with the correct wire AWG for the needed output power and make sure the inductor core will not saturate. The main goal in layout is keeping the loop from your power supply filter capacitors thru the switching fets, inductor & output as short as possible because this area of the circuit has very hi DI/DT transient currents. Any extra lead inductance in this loop can cause ringing.
Remember that just a large filter cap will not work for fast switching currents, the ESR of the larger caps are to hi to draw a 3-5A current pulse from them, adding some film or higher value (4.7-22uf) caramic caps in parallel with the large filter caps will keep the total ESR of the capacitor bank low & prevent some of the ringing. These smaller caps also can be placed closer to the switching mosfets which will keep the input - output current loop shorter.
There is an exception to this, the output filter inductor has to be constructed of a ferrite material that has low core loss int he frequency range that the class D is switching at. Most switch in the 200-500Khz range so if you use a core good for a switching DC supply then it should also have low core loss & low dissipation. You also need to select an inductor or design one with the correct wire AWG for the needed output power and make sure the inductor core will not saturate. The main goal in layout is keeping the loop from your power supply filter capacitors thru the switching fets, inductor & output as short as possible because this area of the circuit has very hi DI/DT transient currents. Any extra lead inductance in this loop can cause ringing.
Remember that just a large filter cap will not work for fast switching currents, the ESR of the larger caps are to hi to draw a 3-5A current pulse from them, adding some film or higher value (4.7-22uf) caramic caps in parallel with the large filter caps will keep the total ESR of the capacitor bank low & prevent some of the ringing. These smaller caps also can be placed closer to the switching mosfets which will keep the input - output current loop shorter.
dllaffer!
Unfortunately only the half is true.
The ripple current is much smaller then the audio current, this because quite bad ferrite materials or low permeability iron powders are still good for this application.
In many PSU you can find -26 or -52 cores, but these are almost unusable for ClassD audio.
But not at the output filter choke! Here di/dt is low, only 1...5 A/us, while at the MOSFET it's 200...600 A/us. The loop wich important is GND, supply rail positive, FET high side, FET low side, supply rail negative, GND.
For example I use in one of my amps 5 paralelled 1000 uF 200 V, combined ESR=5 mohm, so there is no problem with 50 or more amperes from the aspect of ESR. Layout inductance is a more serious limiting effect, but I use 4 uF SMD MKT capacitors close to the MOSFETs.
Unfortunately only the half is true.
The ripple current is much smaller then the audio current, this because quite bad ferrite materials or low permeability iron powders are still good for this application.
In many PSU you can find -26 or -52 cores, but these are almost unusable for ClassD audio.
But not at the output filter choke! Here di/dt is low, only 1...5 A/us, while at the MOSFET it's 200...600 A/us. The loop wich important is GND, supply rail positive, FET high side, FET low side, supply rail negative, GND.
For example I use in one of my amps 5 paralelled 1000 uF 200 V, combined ESR=5 mohm, so there is no problem with 50 or more amperes from the aspect of ESR. Layout inductance is a more serious limiting effect, but I use 4 uF SMD MKT capacitors close to the MOSFETs.
But you can see the colors, and you can count the number of turns, hopefully! Also you may be able to measure the sizes.
It is a pitty that chokes are very often completely underspecified.
The core losses do depend mostly on two factors of the application.
1) Voltage x time integral, for rectangular voltages this results in a triangular current ripple (as already stated by Pafi).
2) Frequency
Both parts have an overproportional effect on the losses.
There are many DC/DC converters in the lower voltage range with moderate operating frequencies, where powder cores like material mix -26 or -52 are advantageous. Especially in low voltage range the continuous mode operation of DC/DC converters does not cause much headache with reverse recovery peaks (because low voltage diodes are available with small Qrr). Continuous mode allows operation at high currents and still with small voltage time integral. Most of the current is DC and does not cause core losses, but high DC flux density. And high DC flux density is a condition where -26 or -52 materials are much better than ferrite.
This also helps to settle a design with low DC resistance, which is also important in for low voltage and high current DC/DC conversion.
For magnetics it is always a question about the application details which solution is fortunate.
In class D amps we do have frequencies which are in the medium range.
(Well considering the power levels, it is even more on the high side. Handling 1kWs at 400kHz is a word...)
Voltages are in the medium range.
Consequently metal powder with low losses and low permeability do work.
Also standard ferrites do work. I got very good results with N67 - of course N87 or N97 might be superior. But I think even N27 would work pretty OK.
One nice thing of N27 is the combination of low price high saturation flux density.
Winding capacity should be kept low, typically single layer constructions are fortunate for the output choke of a class D amp.
Hm, from the pic on Nigel's web page the 'power inductor' could be made with material 52 or similar. 52 has the color code green/blue.
Smelly 120°C ? Not a really nice temperature for iron powder.
The organic components tend to evaporate at such temperatures, over time the magnetic properties will change. Also I think the evaporated organics are not healthy.
Good that you changed it.
The core losses do depend mostly on two factors of the application.
1) Voltage x time integral, for rectangular voltages this results in a triangular current ripple (as already stated by Pafi).
2) Frequency
Both parts have an overproportional effect on the losses.
There are many DC/DC converters in the lower voltage range with moderate operating frequencies, where powder cores like material mix -26 or -52 are advantageous. Especially in low voltage range the continuous mode operation of DC/DC converters does not cause much headache with reverse recovery peaks (because low voltage diodes are available with small Qrr). Continuous mode allows operation at high currents and still with small voltage time integral. Most of the current is DC and does not cause core losses, but high DC flux density. And high DC flux density is a condition where -26 or -52 materials are much better than ferrite.
This also helps to settle a design with low DC resistance, which is also important in for low voltage and high current DC/DC conversion.
For magnetics it is always a question about the application details which solution is fortunate.
In class D amps we do have frequencies which are in the medium range.
(Well considering the power levels, it is even more on the high side. Handling 1kWs at 400kHz is a word...)
Voltages are in the medium range.
Consequently metal powder with low losses and low permeability do work.
Also standard ferrites do work. I got very good results with N67 - of course N87 or N97 might be superior. But I think even N27 would work pretty OK.
One nice thing of N27 is the combination of low price high saturation flux density.
Winding capacity should be kept low, typically single layer constructions are fortunate for the output choke of a class D amp.
Hm, from the pic on Nigel's web page the 'power inductor' could be made with material 52 or similar. 52 has the color code green/blue.
Smelly 120°C ? Not a really nice temperature for iron powder.
The organic components tend to evaporate at such temperatures, over time the magnetic properties will change. Also I think the evaporated organics are not healthy.
Good that you changed it.
With those low permeability cores it may be possible that capacitance (a lot of turns) and quite some skin effect in MHz range actually snub a bit if one i lucky, but filtering properties are gone anyway because it's way above choke's resonance anyway
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.