Power Supply Resevoir Size - Page 103 - diyAudio
 Power Supply Resevoir Size
 User Name Stay logged in? Password
 Home Forums Rules Articles diyAudio Store Gallery Wiki Blogs Register Donations FAQ Calendar Search Today's Posts Mark Forums Read Search

 Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
diyAudio Member

Join Date: Mar 2008
Steven,
this is not an audio amplifier - its a "splat tester" for measuring inductance, but I used the aforementioned construction technique.

The Splat Tester is a big bank of caps going to a coil under test, then into a FET switch to 0V (many FETs in parallel). On the other side of the PCB is +Vdc, and a slot cut across the PCb. I solder the choke I'm testing across the slot.

There is a loop of wire I clip my current probe onto, and a pulse generator drives the input of a FET driver. The diodes you see are between the FET Drains and +Vdc.

I feed the cap bank from a bench supply, up to 100V. Then whack the choke with a pulse from a pulse generator. This makes current build up in the Choke being tested:
V = L*dI/dt, so for t = Tpulse, I ramps up to Vdc*Tpulse/L_choke

the slope of the measured current dI/dt = Vdc/L_choke, so I can calculate

L_choke = Vdc*Tpulse/I

I ensure that 0.5*6.5mF*Vdc^2 is >> 0.5*L_choke*I^2 (or in other words that +Vdc doesnt droop much)

there is a BNC connector for the pulse generator input, another to connect a scope probe to the input pulse (to trigger from) and a BNC measuring Vdrain.

This way I can measure inductance vs current, and see saturation too. with ferrite cores the current slope is straight, until it saturates (which is quite sharp). iron powder cores have a permeability that varies a lot with current, so the current slope isnt straight, nor is the transition into saturation very pronounced.

attached are splat test results from Iron Powder (Moxie...) and Ferrite (NGD....) cored inductors.

Only downside to this toy is I used a plastic box. So I had to cobble together a heat shield from FR4, so I can het the device under test to 100C without melting the box, or my horribly expensive current probe

You can make a crude version of this with a decent sized cap bank, and a low inductance current sensing resistor. stick the R in series with the cap -ve terminal, and hook a scope probe to it. stick another scope probe to the + terminal. solder one leg of the DUT onto one cap terminal, and use a bit of solid wire soldered to the other terminal as a switch. charge the caps up, then close the switch - you'll soon see why its called a SPLAT test! eventually you will get sick of poor contacts (bounce = lousy waveform) and make a box like this.

I learned this trick years ago, working on big AC drives. 40mF charged to 600Vdc makes quite an impressive splat, and is a great way of getting several hundred amps flowing into a choke. ever so slightly lethal though......
Attached Images
 Moxie Choke Comparison #1.jpg (265.9 KB, 417 views) NGD dc-dc 8T EP13 splat test #14.jpg (241.6 KB, 384 views)

diyAudio Member

Join Date: Mar 2008
Steven,

here is the impedance of the cap bank. vertical scale in dB-Ohms, top line = 40dBOhms.

I also measured an SMH 25V 12,000uF = 12mF capacitor. I dont have a dedicated test fixture for this cap, so the measured lead inductance is a bit higher than ESL (I had to have enough leads to solder the darned thing to an existing test fixture). I'd estimate this cap at around 7-10nH (test fixtures are important when measuring stupidly low inductance).

But the cap bank has a total of 1.7nH - almost 10x better. and there are NO nasty resonances out to 200MHz (HP3577A limit). Nor are there any "decoupling" caps (film or otherwise)! yet the Dc bus impedance is < 100mOhms @ 2MHz, rising to 1 Ohm at 90MHz.

The cap bank is very crude - a DS-PTH pcb would shave off maybe 0.2~0.5nH. Using thinner PCB (this is 1.6mm) would have a much larger effect - I would expect a total inductance of < 0.5nH in that case.

Edit: this very low broadband impedance will do a VERY good job of attenuating any AC line noise that makes it past the rectifiers. Likewise this will effectively filter any audio signal above 1kHz or so, preventing it from reaching the transformer. assuming, of course, one does not then go and undo all this good work with lousy wiring.

I genuinely do not understand why the DC bus layout of audio amps is so horrendously bad. it need not be, and the cost is almost zero.

cheers
Terry
Attached Images
 parallel cap impedance.jpg (205.6 KB, 396 views)

Last edited by Terry Given; 7th September 2012 at 04:35 AM.

 7th September 2012, 05:57 AM #1023 diyAudio Member     Join Date: Aug 2012 Location: Los Angeles, California Terry, Okay you fooled me. I see what you are saying about using this to test inductors. But could this basic principal be used to tie all of the capacitors together for an audio amplifier power supply?
 7th September 2012, 07:50 AM #1024 diyAudio Member   Join Date: Mar 2008 Absolutely! as the impedance plot shows, its a 6.5mF cap with 17mOhm ESR and 1.7nH of inductance, including the "wiring" - an order of magnitude better than a single large cap (without any wiring!)
 7th September 2012, 11:52 AM #1025 diyAudio Member     Join Date: Nov 2007 I've noticed something on the Honey Badger. It has 470u//220u aboard. This will give you a wonderfully balanced tone. For personal preference, I would add 1 of 2u 250v cap (polyester, electro, or RC) v+ to v- for noise block (takes some experimentation to find the right cap for desired clean cool laid back effect of noise blocking), but that is not in question. Here's some question: The 220u is greatly responsible for the clarity and tone of the amp (as always, the smallest electro on an amp board governs the tone), and I think an excellent choice. The 470u is not greatly responsible for the clarity and tone of the amp--they are defense. Could I not replace the 470u with 3300u? Perhaps if I combined a high efficiency 220u with a standard 3300u, the ESR would come closer to matching and one cap wouldn't fight the other for charge. For clarity and tone, it is necessary that the 220u Always Wins the fight for charge. Ideas? Other way to say it: The Honey Badger board, and the LM1875 chip alike, lump together the predrive bypass caps (220u) and the output bypass (470u) into one paralleled section. This is inconvenient and a bit confusing, both to myself and to the capacitors when they compete for charge. The simplistically paralleled grouping is the only option for the LM1875 chip due to limited number of power input pins. However, the Honey Badger is a discrete amp with no such limitation. I'd like to know how to use the 220u for predrive and locate the 470u (or larger) more effectively to accomplish their individual tasks. I'd like to assure that the 220u predrive caps aren't needlessly drained when a bass beat is delivered by the outputs. The layout contains no such guarantee, and the loop area is very, very big. __________________ ♦ DIY Models & Apps ♦ ClipNipper ♦ LM1875 ♦ TDA7293 ♦ Powerful Parallel TDA7293 kit ♦ My post has opinion. Last edited by danielwritesbac; 7th September 2012 at 12:06 PM.
 7th September 2012, 01:29 PM #1026 diyAudio Member   Join Date: May 2007 The 470uF and 220uF are not in parallel, as there is a 22ohm resistor between them according to the circuit diagram. When capacitors compete for the available charge they never get confused but always follow the laws of electromagnetism and Ohm's Law etc. 470uF appears to be an on-board local decoupler for the output, and the 220uF does a similar job (plus a small amount of smoothing) for the driver. Presumably the 470uF is in parallel with the main PSU caps. Increasing its value would have the main effect of bringing more charging pulse current onto the amp PCB - perhaps a bad idea?
 8th September 2012, 07:30 AM #1027 diyAudio Member   Join Date: Jul 2004 Location: Scottish Borders The size of the "pulse" current will be limited by the emf driving that pulse and the impedance trying to resist that pulse. The cables are very effective pulse limiters. Look at Peter Daniel's implementations where he completely foregoes all PSU smoothing and uses local decoupling of 1mF to 1m5F on the amp PCB. He says great for mid and highs (yes the MF and HF decoupling is doing it's job). __________________ regards Andrew T. Sent from my desktop computer using a keyboard
 8th September 2012, 09:13 AM #1028 diyAudio Member     Join Date: Feb 2001 Location: USA A commercial amplifier design I did a few years back had a pair of 3m3F + 33µF all on the same board with the amplifier (50W x2), everyone was stunned at how much better it sounded than the run-of-the-mill commercial amplifiers (Adcom, etc.). A 160VA transformer was used. __________________ Candidates for the Darwin Award should not read this author.
 8th September 2012, 02:10 PM #1029 diyAudio Member   Join Date: Nov 2006 Location: Indiana Blog Entries: 1 Hi again, Progress update: i have been running simulations with the new per-unitized transformer model, which is able to be scaled to different Volt-Amp ratings and different Output Voltage ratings simply by changing the values of two corresponding spice parameters. (Thank you, Terry Givens!!) I have found at least two interesting and suprising patterns while finding the minimum reservoir capacitance that doesn't allow the transformer current pulses to cause the rail voltage to make overt incursions into the output signal voltage. These were found after I refined the methods used so that the output amplitude and offset were initially set very precisely and then the calculated output expression was precisely matched to that, so that the calculated difference between them gave much more accurate error and distortion measurements. Also, because of the easy-to-use scalability of the VA and Vrms ratings of the new xfrmr model, i was able to try, for example, large VA xfrmrs with high-enough Vrms ratings for cases with relatively-low output power specs, and was able to simulate broader ranges of output power values for a given transformer size. I have only begun to scratch the surface, because my available "play time" has been severely limited, lately. So i have more to do before I can hope to have any comprehensive or definitive results to share. I mainly wanted to let Nico know that the work is still progressing. But, so far, it looks like whenever the VA and the Vrms ratings are "large enough" then the absolute-minimum reservoir capacitance is much lower than I would have guessed. This is not really surprising, maybe. But I was surprised by the consistency of the numbers, between cases with widely-varying transformer ratings and output power specs. For example, with a 44 Vrms xfrmr output rating and various appropriate VA and signal output power specs, the minimum C values were all in a narrow range, under 500 uF. The other thing I noticed is probably also not as surprising as I thought: i can predict (or find) the minimum C value by looking at the minimum difference between the maximum signal output peak voltage and the minimum power rail voltage. For the 44 vct case, with VA all the way from 1000 to 240 and output power of 100 Watts, 75W, 50W, and even 25w (all different peak output voltages), that difference doesn't change, staying at about 3.35 volts, but going down to just above 3.0 volts for some of the 25W output cases. When I simulate a series of C-value steps for a new configuration of VA, Vrms, and Pout, I usually first look at the tabulated measured vrailmin minus vsignalmax that I now have spice find for me. I can _always_ select the two C value steps where the last few hundred uV of the overt charging pulse distortion finally disappears, that way. In the meantime, i am also collecting all of the data to eable plotting the distortione versus C value for each case. When I get some more time i will probably first see how the results i have so far correspond to what would be predicted using a constant DC load, since I am finally at a point where the measurements are staying consistent. More later, Tom Last edited by gootee; 8th September 2012 at 02:22 PM.
diyAudio Member

Join Date: May 2007
Quote:
 Originally Posted by gootee But, so far, it looks like whenever the VA and the Vrms ratings are "large enough" then the absolute-minimum reservoir capacitance is much lower than I would have guessed. This is not really surprising, maybe.
True. It is exactly what I would expect. With a big enough off-load rail voltage you can have lots of droop/ripple without causing problems. Hence my belief that there is no absolute minimum capacitance.

Quote:
 Originally Posted by gootee staying at about 3.35 volts,
So my wild guess of 2V was not too far out. It explains why the cap values I calculated were a bit different from the ones you found from simulation.

 Posting Rules You may not post new threads You may not post replies You may not post attachments You may not edit your posts BB code is On Smilies are On [IMG] code is On HTML code is OffTrackbacks are Off Pingbacks are Off Refbacks are Off Forum Rules

 Similar Threads Thread Thread Starter Forum Replies Last Post SanderW Power Supplies 25 4th January 2013 04:12 PM Burnedfingers Tubes / Valves 25 5th January 2012 12:23 AM BigE Power Supplies 11 5th July 2011 02:59 PM diymixer Power Supplies 1 10th October 2010 05:47 AM spooney Car Audio 3 6th December 2007 11:50 PM

 New To Site? Need Help?

All times are GMT. The time now is 08:05 PM.