diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   PSU RC multistage filtering (http://www.diyaudio.com/forums/power-supplies/184068-psu-rc-multistage-filtering.html)

gpapag 27th February 2011 09:23 PM

PSU RC multistage filtering
 
3 Attachment(s)
Hi all

I did some ripple and noise measurements on 3 different versions of a simple 18Vdc PSU and I thought to share it with you.

The same materials were used on all three versions.
5W wirewound 4R7 resistors, 2W carbon 470R, and electrolytic caps.

AC Voltage measurements (through an 1uf MKT cap and a 47K carbon potentiometer to an M-Audio USB Audiophile sound card) were recorded as wav. files and then imported to RMAA 5.5 for spectrum analysis.

gpapag 27th February 2011 09:26 PM

3 Attachment(s)
What comes IN the 7818 voltage regulator and what comes OUT of it.

Regards
George

wintermute 28th February 2011 01:56 AM

Hi George,

in the sims I've done having the bigger cap closest to the rectifier seemed to work best, could you reverse the order of the caps in A) and post the difference (if any)?

Tony.

Joachim Gerhard 28th February 2011 03:04 AM

The new Sutherland RIAA stage uses massive RC filtering.

Joachim Gerhard 28th February 2011 03:20 AM

http://www.elusivedisc.com/SUT2020-HFW-review.pdf

agdr 28th February 2011 04:15 AM

Quote:

Originally Posted by Joachim Gerhard (Post 2485738)

Wow! From that article:

"The power then goes through ten cascaded RC stages of passive RC filtering..."

Massive indeed!

Wintermute: I would enjoy seeing the results with the cap order reversed too. I also used to think having the big one (usually one or more 10,000uF in my case) on the end of the chain was best to supply the load current between the 8.3mS chargings without having to pull it through the series resistor(s). But lately it does seem the opposite may be true. The waveshape seems a lot better (lower overall charging current peaks and more sinusoidal) having the big cap first.

Maybe best to think about the RC chain in terms of ripple reduction rather than supplying load current. The leftmost cap is hit with the highest level of ripple, of course (and should be rated for that). The ripple then decreases at each node to the right down the chain, presumably requiring smaller caps with lower ripple ratings (?).

Gopher 28th February 2011 07:41 AM

Something's not right here.

Circuits B and C have the same amount of capacitance (1760uF) after the rectifier but B has additional RC filtering action due to the 14.1 Ohms of resistance before the caps. The cutoff frequency of the RC filter in B is a low 6Hz.

You would expect, therefore, the spectrum at P14 to have much lower HF content than that at P24, but they're almost identical. This cannot be correct.

DF96 28th February 2011 10:37 AM

P14 is a bit lower at HF than P24. Don't forget to include diode forward resistance and transformer resistance in the analysis, then the difference between B and C is small.

Comparing A with B reminds me of something I learnt in school chemistry about 40 years ago (when we were studying dilution). To get something clean, using a fixed amount of water (or other solvent), wash it lots of times using a small amount of water each time rather than one big wash using all the water. The same principle applies to RC smoothing.

Gopher 28th February 2011 11:16 AM

DF96 - you must have sharper eyesight than I do. I can't tell any difference worth mentioning.

Two diode drops should only be an ohm or so max, similarly the transformer resistance. With an additional 14 ohms of resistance in circuit B there should be a noticable reduction in the HF spikes compared to circuit C, especially beyond 1kHz, but I see none. If the Fc is 6Hz, they should drop off at 6dB per octave or 20dB per decade.

wintermute 28th February 2011 11:24 AM

5 Attachment(s)
Hi agdr, I hadn't done any fft's of the sim but did so tonight. What I discovered is that the arrangement of caps with lowest to highest results in a lower voltage at the last cap than the other way around.

Originally when doing my sims (of LM317 reg) I didn't realise that I needed quite a bit higher than the dropout voltage in-out to get the best performance. The differences I saw were mainly due to the lower voltage more so than because of the difference in ripple pre-reg.

The results of the fft's is interesting. The noise floor (which is unrealistic anyway) is lower with the large cap first and I think the harmonics are slightly lower but the primary is exactly the same.

I'll post the various plots and the circuit diagram.

I've only shown the fft's at the point before the reg for each config and the comparison after the reg. Note that if I up the voltage pre rectifier the blue trace also goes down to -280db flat line, I know that this is completely unrealistic but it shows how even a small voltage difference before the reg can make a difference after!

The thing that surprised me the most was the voltage difference, it was something I didn't pick up before because I was looking at the post reg performance and basically ignoring what was happening before provided I had at least 3V higher than the output voltage. What I've since found was that at least 5V difference was necessary for best performance (at least with the model I have for the LM317).

Tony.


All times are GMT. The time now is 12:51 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2