The use of the 18V-to-regulator circuit was an afterthought for the H2V2 use so even though the PO89 filter is very close to H2V2 (same small case) the regulator circuit is 6 inches away from the PO89 filter (shortest cable I could buy) and I didn’t want to rebuild the entire thing to fit in one case …..lazy me ;>)
Well, I don't think I want to do that.
Way back when I did have two threads going -- "A precision LED/LDR-based Attenuator" plus "BTFSystems precisionLDR modules for digitally-controlled LDR Passive Preamp" and it was a lot of work which, as a really senior citizen, I don't care to repeat.
I have no formal education in electronics, and I was just trying to get understanding from people who understand this filter better than I do. In fact, I don't understand it much at all. I was not trying to hijack the thread, but I take your point and will do my own experimenting and developing, thanks for the gentle advice.
Wapo, I've been using a similar set up (just single supply).
I've been going from SMPS - P0z - CRC - LM317 - AMP, but have tested a variety of configurations and SMPS.
I've been filtering the voltage first before the regulator because each filter stage drops voltage. This way not only can I guarantee that the voltage set at the regulator is what the amp receives, but I also lower the heat generated by the LM317. I go from 19v to ~17 then drop 5v at the regulator which keeps it nicely within SOAR and in the Vin>Vout stable region. The caveat is to make sure the voltage dropped by the CRC at full load leaves ~4v Vin-Vout to ensure stability.
I found that this filter is good for reducing switching noise, but did not attenuate 50/100Hz at all, so you may need to think about a CRC stage for your setup.
You can see the 50/100Hz peaks on this with/without comparison:
Does that apply here?
I reread Mark's original post where he states that total resistance of the filter is 0.15 ohms which seems right given the inductors and resistors DC total resistance.
Wonder if the output of an SMPS requires filtering at 50/60 Hz and the immediate harmonics of those frequencies? Wouldn't the intrinsic design of the SMPS eliminate powerline frequencies?
The visuals at post 40 and 86 don't seem to show any powerline frequency noise? (unless it's there and I don't recognize it . . . )
I tested 3 SMPS of differing quality levels, and how much 50Hz that came through varied with each. But it did come through to some extent with all of them.
The power supply whose measurement I posted had 50Hz noise at -92dB (after CRC and regulation), another supply was at -72dB unfiltered and unregulated. The worst one peaked at -56dB and was audible. So the point is, tailor the filtering to either the power supply combination that you are going to use, or the worst that you expect to use, and bring the noise floor down to a level that you are happy to accept.
The P0z filter only drops a negligible amount of voltage. I was referring to the voltage drop over the CRC as a driving factor for the sequence of filtering.
You can probably go SMPS - CRC - Linear Reg - Poz filter - Amp, but for me I feel that the voltage regulator should be the last step in the chain before the load.
If you have some way of measuring noise or supply ripple, I'd suggest testing different combinations to find out what works best for you before making something permanent.
SMPS work with rectified and smoothed line voltage, which has 50&60Hz hum in it. The all cannot remove that ham by 100%, hence every smps will have some of it at the output.
If you want to place a linear regulator after a SMPS, clearly the filter should be before it, so that the regulator can do a better job.
If you want to place a linear regulator after a SMPS, clearly the filter should be before it, so that the regulator can do a better job.
Since active regulators have much lower output impedance than passive filters or capacitance multipliers / "ripple eaters" ... I usually prefer to install the active regulator LAST in any cascade (series connection) of supply junk-removers. Now the signal handling circuits will be driven by the lowest possible impedance power supply.
If you can afford its complexity and power dissipation, a shunt regulator with a common collector / common drain output, plus a high gain error amplifier, is an excellent choice for the final stage in the cascade. Very low output impedance AND offering the possibility of very low noise if you handle the reference circuitry with care.
_
If you can afford its complexity and power dissipation, a shunt regulator with a common collector / common drain output, plus a high gain error amplifier, is an excellent choice for the final stage in the cascade. Very low output impedance AND offering the possibility of very low noise if you handle the reference circuitry with care.
_
Last edited:
Still struggling with this.
So, SMPS -> PO8 = minimal hash, but line frequency ripple remains.
Therefore, the wisdom is to use a CRC filter after the PO8 filter in order to reduce the LF ripple before the regulator which is the final stage of the power supply?
But if the entire p-p ripple range is well above the input voltage required to deliver the regulated output voltage. that ripple would then have no effect on the regulated output, would it?
If that's the case, why use a CRC filter after the PO8 instead of simply using the final downstream capacitor of the PO8 series (perhaps using a bigger value for that?) as the input capacitor for the linear regulator?
Yes.
A voltage regulator such as the LM317 will have a PSRR that varies with frequency (Datasheet Figure 8.)
In the case of LM317, it is about -60dB down to 100Hz, reducing in effectiveness above 15kHz.
Some of the ripple at the input of the VR will pass through to the output proportional to the PSRR ratio. The more ripple you knock down before the input, the less will be present at the output. How much you can tolerate to pass through is up to you as part of your design.
Mark's filter is designed to reject the higher frequencies from the SMPS. Whether you need to reject more of the lower frequencies than the VR can handle is up to you. CRC is a cheap and easy way to drop some ripple before the VR, and can be easily bypassed if it is determined that it is not required.
You can combine filters. You'd essentially be making a CLCLCRC.
There is an excellent set of filter tools here: Filter Design and Analysis. You can simulate various values to see the cutoff frequencies and play with the values to see what affect they'll have on the filter.
I would like to shrink the filter as much as I can for use in projects where the current requirement is well under 1A, mostly under 200mA.
For that, I'd like to use a smaller inductor with a lower current rating and am wondering if there is a problem with that. The inductors I have in mind are 2.2uH 250mOhm as opposed to the 0.1R of the specified inductor. Would this require a change in the parallel resistor as well? Any information appreciated.
I know that Mark was very specific that each component value was critical, but I did not read the 'why' or maybe I read it but didn't understand the implication of what I read. If it can't be done with my level of expertise then that's that.
I think this is a great filter, every smps used in audio should have it in the chain. A lot of small signal boxes like preamps, DACs, etc would really benefit from taking a single voltage wall wart and splitting it in the chassis to a +/- supply, but this level of application just doesn't need a 6 amp inductor and space is often at a premium. Smaller 1A~2A inductors could be close to the size of a resistor and be placed just like a resistor, and that was what I was hoping to achieve.
I also believe (and know from experience as I built 12 of them) that this is a great little device. In fact I shall even try a few of them in a non conventional way in a few hours...
But I find it small enough TBH... plus the recommended coils are already very small, the biggest parts are by far... the caps!
Anyway, good luck with your build!
Claude
But I find it small enough TBH... plus the recommended coils are already very small, the biggest parts are by far... the caps!
Anyway, good luck with your build!
Claude