Inductor values and placement. Power stage filter.

[Dan R. Hansen]

Greetings from Denmark.

I am a beginner regarding electronics and components values. I am am having trouble figuring out what value/values I need for filtering the AUX power output of an ICEpower 1200AS2 module.

I will write a short resume of the values at the end of my post.

According the the manual, they recommend the following for both the regulated 5V, and the +/- output:
"to add extra supply line series feed impedances and decoupling capacitors on the connected PCB to attenuate the power supply voltage ripple".

The manual says the power stage switching frequency when idle, is 500KHz, and that the extra capacitance added to the modules output can't exceed 1000uF for the 5V, and 470uF for the +/- output.
The maximum allowed power draw from 5V is 1A, and 500mA for the +/-.


What I have problems with, is how make the correct and best "LC" filter for the +/- AUX power?
The Values also confuses me, because when I look at impedance graphs of various inductors on Mouser, values like 470uH and 680uH looks to be ok, but I also see schematics for 500KHz powers supplies, where they use values like 220uH to filter the output?

I understand the answer could lay in the fact that the noise I need to filter is ties to the rise-time of the 500KHz signal, but I am in way to deep at this point.

I just can't figure it out...


The criteria the manual gives me is:

All power stages is 500Khz frequency.
5V is reregulated, but needs additional noise filtering.
5V can't exceed 1A draw (40 mA in standby), and 1000uF max.
+/- stage is not regulated
+/- can't exceed 500mA draw, and 470uF max.

What values should I use for the inductors? On what lines do I place them? +, -. AGND and +5V? Or?...

I will be putting 7812 and 7912 regulators on the AUX +/- lines AFTEr the inductor and capacitor, so I can get a steady +/-12V for the line driver op-amps and buffer.

I am sorry if my explanation is confusing, I am not doing will with this stuff.   
Thank you for your time, and thank you in advance.

Cheers.

[Dan R. Hansen]

I had a look at a typical 78xx regulator, an it seems like it can handle ripple noise somewhat ok up to a few kilohertz.
Would that mean making a low-pas filter that let noise under 1KHz (for example) through, would be OK?

[PGPG]

If you select L to be used in DCDC converter than 470uH or 220uH can be important difference.
If you select L for low-pass filter 470uH to 220uH (with the same capacitor) will give you only 1.4 times lower pole (frequency at which signal (noise, ripple) damping starts.
The higher L at first look can seem to always be better, but the higher L have also higher parasitic paralel to it capacitance and so lower self-resonant frequency so parasitic elements influence problems can start at lower frequencies and you want mainly to damp 500kHz and its many harmonics.

20 years ago I was using SPICE mainly to simulate supply filters but using in simulation L and C models containing their parasitic components (C has serial R and L). For C (I was interested in SMD ceramic capacitors) i got those time from Murata the program that was showing their ceramic capacitors (X5R, X7R) parameters. For L you should find in datasheet their self-resonant frequency and their DC resistance. In model you can have this resistance in serie with L and add paralel to them C you calculate from self-resonant frequency.

My advise is to do some SPICE simulations and you will get some feeling about what is going on in filters. Filter that can be looking being better at low frequencies will be worse at high frequencies and you don't want to filter out AC 50Hz but DCDC 500 kHz (and its harmonics).
When you have filter in Spice you can see its AC characteristic, but you can also see what from 500kHz square (or triangle) signal at input you get at output.
There is commonly used free LTSPICE.
If you plan to learn PCB design than consider learning KiCad. It has also SPICE simulation integrated (probably not so good as LTSpice (yet) but certainly enough for supply filters simulation).
I'm not sure of comparison as I have never used LTSpice, and Spice in KiCad I have just check an year ago that it works and nothing more.

At 12V inputs to my devices I use pi filter (470uF - 100uH - 470uF).
As you consider 470uH and I am using 100uH when you will select 220uH you will feel you made a good compromise.

You don't need L at your GND connection.

[Dan R. Hansen]

Thank you for your reply PGPG. It put me a little at ease.

I feel like the datasheet for the 1200AS2 might not give me all the data that I need to figure it out. It says:

"The AVDD and AVSS supply outputs are loosely regulated. They may require additional voltage regulators on a frontend board depending on the type of circuit supplied by these pins.
It is therefore recommended to add extra supply line series feed impedances and decoupling capacitors on the connected PCB to attenuate the power supply voltage ripple.
The series impedance can be either a low value resistor or a small inductor. The extra decoupling capacitance must not exceed the value specified in Table 13. (470uF)".

It mentions that the power stage switching frequency is 500KHz, and that is it. So I assume it goes AVDD and AVSS as well. Since it's not regulated, the power is just rectified 500KHz with some capacitors, so I got the impression, that the values aren't super critical since I just need to remove the voltage spikes? Though they only mentions ripple. I will be regulating power after the filter.

May I ask you about how to go about putting filers in a dual supper with 0v (GND) in the middle? DO I need to put an inductor on +, -1 and GND?

[PGPG]

May I ask you about how to go about putting filers in a dual supper with 0v (GND) in the middle? DO I need to put an inductor on +, -1 and GND?

Have you read last sentence of my previous post?
I have posted my previous post without this last sentence, but I have added it not later than 1 minute after posting.

[Dan R. Hansen]

Have you read last sentence of my previous post?
I have posted my previous post without this last sentence, but I have added it not later than 1 minute after posting.

I don't know how I missed that. 
Sorry about that.

I have tried TI's "Power stage designer". When I put in the values I can see, it calculates that there will be a nasty gain in attenuation around 1Khz. It also suggest that i should put a non-polar capacitor of 10uF or more, with very low ESR, in parallel with the inductor, to remove that spike.
Such foil capacitor will increase the price significantly too. And I am not even sure if the 1200as2 unit will require special values in order to function correctly. I know of the extra capacitance limit, but would the capacitor in parallel with the inductor somehow affect the capacitance limit as a whole? I mean, in some reactive way?

[PGPG]

It also suggest that i should put a non-polar capacitor of 10uF or more, with very low ESR, in parallel with the inductor, to remove that spike.

I have never seen such solution.
It looks they want to have paralel resonance somewhere close to problematic frequency. It can be interesting, but this capacitor will dramatically degrade filter characteristic at other frequencies and we are interested in these other frequencies and not 1kHz.

It is typical that low pass filter at the pole frequency has gain higher than 1. Electrolytic capacitors have relative high ESR so this gain should not be very high. But why you care. You don't have 1kHz in your ripple. You have 500kHz so you should look at your filter characteristic at 500kHz and odd harmonics of this frequency.

[Dan R. Hansen]

I have never seen such solution.
It looks they want to have paralel resonance somewhere close to problematic frequency. It can be interesting, but this capacitor will dramatically degrade filter characteristic at other frequencies and we are interested in these other frequencies and not 1kHz.

It is typical that low pass filter at the pole frequency has gain higher than 1. Electrolytic capacitors have relative high ESR so this gain should not be very high. But why you care. You don't have 1kHz in your ripple. You have 500kHz so you should look at your filter characteristic at 500kHz and odd harmonics of this frequency.

No wonder you haven't seen such configuration, since I misunderstood what the program was displaying.
It showed 3 types of values withing a dotted box, which symbolized the 3 factors of an inductor which included capacitance. It wasn't part of a filter layout.

I have added 2 pictures of the charts for the 5V rail unregulated), and the +/- AGND rails (That will be regulated after the filter). I do not have the inductance values of the capacitors and the value doesn't matter at all for the low side of the filter, but they matter significantly on the high side.
'I still don't understand how the placement of the LC filter can be ok with on on the plus rail and one on the minus rail. What I mean is; that if I were to disconnect the plus rail, and only use the minus rail, wouldn't the center tap (AGND) now technically be the positive part of the circuit where the filter goes on the positive circuit?