If you're getting cold feet on the air-cored Ls, I have a filter design using Coilcraft ferrite cored inductors which will be considerably quicker to build. Can't speak for its THD performance though. My filter's designed to attenuate as well rather than pass the full amp output signal amplitude, but does a whole lot better than -52dB in its stopband.
Hi All: Don’t try to use any core other than AIR. The best linearity from a ferrite material was about 38dB, most samples were below 27dB.
I designed the AUX0025 filter when I was @ AP. I did many hours of research and testing of many Class D amplifiers and components that went into the filter. Total linearity of the total filter needed to be better than the AP. The COILS are air core 1.00 mHy and placed inside a NONFERIOUS box. Now I know some of you may not think so, however inside the steel covers is the “magic stuff”. The alum shields keep the magnetic fields from finding the ferrous materials and destroying linearity. The R's & C's were also verified. The filter can be placed on top or bottom of the AP and the cables supplied are also part of the filter package.
Duke
I designed the AUX0025 filter when I was @ AP. I did many hours of research and testing of many Class D amplifiers and components that went into the filter. Total linearity of the total filter needed to be better than the AP. The COILS are air core 1.00 mHy and placed inside a NONFERIOUS box. Now I know some of you may not think so, however inside the steel covers is the “magic stuff”. The alum shields keep the magnetic fields from finding the ferrous materials and destroying linearity. The R's & C's were also verified. The filter can be placed on top or bottom of the AP and the cables supplied are also part of the filter package.
Duke
Wow. What a gift - thanks for your post Duke.
I've experimented a bit with the design of these filters and found that the 5th order C/L/C/L/C made the most sense with everything considered. It was fun to learn from a link in this thread that the AUX-0025 generally uses that as well.
What still interests me is how it was decided to aim at 50dB+ attenuation around 250kHz and then at 400kHz with the AUX-0040. Maybe a scope was used to measure (and ultimately limit) the maximum switching noise slew rate allowed to pass to the analyzer?
Thanks!!
Hi Ben
The Coil data:
DCR 3.4 OHMS
SIZE: CUP CORE BOBBIN ID=0.487, OD=0.8475, TALL=0.608
COIL ID=0.50, OD=0.770, TALL=0.528
Hi Holland
Bruce made the changes and did the AUX0040. I think that with the improvements in Class D chips the switching frequency went up and the RFI/EMC products have greatly improved.
I started the AUX0025 in 2002 and the 60 plus Class D amplifier samples I tested had the carrier freq were 30kHz to 250kHz and the carrier output voltages ranged from 2mv to >16 volts.
Duke
The Coil data:
DCR 3.4 OHMS
SIZE: CUP CORE BOBBIN ID=0.487, OD=0.8475, TALL=0.608
COIL ID=0.50, OD=0.770, TALL=0.528
Hi Holland
Bruce made the changes and did the AUX0040. I think that with the improvements in Class D chips the switching frequency went up and the RFI/EMC products have greatly improved.
I started the AUX0025 in 2002 and the 60 plus Class D amplifier samples I tested had the carrier freq were 30kHz to 250kHz and the carrier output voltages ranged from 2mv to >16 volts.
Duke
Hi Duke,
Can you comment on the pots on each phase and a rough way to cal them?
Thanks,
Ben
Those linearity figures on the surface look overly pessimistic to me. I've been using gapped ferrite coils for years and whilst I've not got an AP to make measurements with them, if I was getting -27dB for distortion I reckon that would be quite audible.
Marchand use ferrite cores in one of their crossover units and they're claiming distortion better than 0.01% (so -80dB) : XM46 Passive Line Level Crossover Network, PLLXO, 24 dB/oct
Common mode noise rejection is an important feature for the inputs of an audio analyzer. The ideal behavior for the analyzer's balanced inputs is to ignore complex (magnitude and phase) voltages that are identical on the signal carrying pins 2 and 3. Only the difference voltages are of interest and the ratio of the accepted difference voltage to the rejected common voltage is called the common mode rejection ratio or CMRR.
If the impedances of the circuit running from the DUT to the analyzer are not the same, CMRR will be compromised by the extent of that difference and part of the common mode (noise) voltage will be converted to differential and thus be included in the measurement. AP wanted to minimize this imbalance in the AUX-0025 due to component tolerances by tuning those variable resistors as needed.
Inductors with ferrous cores have a linear current handling range determined by the magnetic permeability of that core. If the current exceeds that range, the inductance will vary in a non-linear fashion called saturation as conceptually shown here:
Air core inductors don't saturate, they can get hot and increase their resistance a bit before they melt, but that's about it. When dealing with high level currents that power amplifiers are capable of, only air core inductors can maintain their linearity. At lower levels a properly spec'd ferrous core inductor can remain linear as well.
A note on inductor distortion (non-linearity); you need to look beyond simple THD and into IMD as well.
Last edited:
If one were competently designing a gapped ferrite inductor for power amplifier current levels then it would remain as linear as another properly spec'd one designed for lower currents. It would of course be much bulkier though. The core sees only magnetization (H) it doesn't care what combination of current and turns number puts it there.
Hi All
The DISTORTION levels, 0.0001% or better is what I was looking for. It was “Not to degrade the AP SYSTEM measurements”. If the linearity was not >-120dB then I had a problem. My tests were THD%, IMD% and then I settled on an FFT test. I apply 26 volts @ 18 kHz & 20 kHz using my Cascade for the tests. I used the FFT and look for any spikes out of the noise. This was what I used to test each part of the filter. Capacitor(s) linearity was also a problem.
The CALIRATION was a special power amplifier and special adapters to drive single ended & differential modes. See attachment.
Duke
The DISTORTION levels, 0.0001% or better is what I was looking for. It was “Not to degrade the AP SYSTEM measurements”. If the linearity was not >-120dB then I had a problem. My tests were THD%, IMD% and then I settled on an FFT test. I apply 26 volts @ 18 kHz & 20 kHz using my Cascade for the tests. I used the FFT and look for any spikes out of the noise. This was what I used to test each part of the filter. Capacitor(s) linearity was also a problem.
The CALIRATION was a special power amplifier and special adapters to drive single ended & differential modes. See attachment.
Duke
This is huge Duke, I can't thank you enough for sharing your hard earned knowledge - you filled in a lot of gaps in my understanding on this stuff!
Nothing is as simple as it looks in this sport.
I have one of Chris Strahm's designs called the LF280 that uses cored inductors and when I get some time I'm going to use your measurement methods and report what I find. It's not really relevant anymore because you can't buy these things, but it'll be interesting. AP should send you a thank you note because I may end up buying one of those things you designed.
Nothing is as simple as it looks in this sport.
I have one of Chris Strahm's designs called the LF280 that uses cored inductors and when I get some time I'm going to use your measurement methods and report what I find. It's not really relevant anymore because you can't buy these things, but it'll be interesting. AP should send you a thank you note because I may end up buying one of those things you designed.
Thanks all, built it today and works great! Unfortunately i dont have any Class-D's sitting around lol. (not a huge fan, but repair them from time to time for people..)
Anyway, I had some longer than ideal nets back to ground for the caps that kind of made me sigh.. but all works as expected.
I hit it with fs signal from the dscope (19v?) and swept all the way up. Thdn is as expected compared to normal loop back and filter performs as expected too.
Inductor winding was somewhat a learning experience.. Seems to me ~287 windings is about right. Hint.. start with more, check with LCR, remove a loop or two, resolder, check again, etc...
Anyway, I had some longer than ideal nets back to ground for the caps that kind of made me sigh.. but all works as expected.
I hit it with fs signal from the dscope (19v?) and swept all the way up. Thdn is as expected compared to normal loop back and filter performs as expected too.
Inductor winding was somewhat a learning experience.. Seems to me ~287 windings is about right. Hint.. start with more, check with LCR, remove a loop or two, resolder, check again, etc...
Its possible to get really low distortion from ferrites. however not easy. The CLT-1 uses ferrite cores in its low pass and high pass filters as well as the matching transformer. The unit is capable of 3W out (use care on the 1 KV output setting) with at least -150 dB HD3 and up. However the transformer is the size of a bowling ball for 10 KHz and the other inductors are quite large.
If someone makes up a PCB I would be interested. I think point to point could work but will be very sensitive to layout. Passive filters are not easy. However active low pass filters will be challenged with HF when the run out of gain. Maybe some of the new ultrawide band opamps could prove useful however you then get noise and dynamic range issues.
If someone makes up a PCB I would be interested. I think point to point could work but will be very sensitive to layout. Passive filters are not easy. However active low pass filters will be challenged with HF when the run out of gain. Maybe some of the new ultrawide band opamps could prove useful however you then get noise and dynamic range issues.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.