What does this capacitor do in this PSU?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hey fellas (and felletes)

I've got two questions about In CircuitBasic's LM3886 PSU (schematic attached).

1. What is the purpose of C10 (outlined)?
C10 is a single film cap connected between the positive and negative rails., rather than smoothing and bypass caps connected between a rail and ground.

2. I assume that the RC networks near the outputs are snubber networks - the RC pairs R3/C5 & R4/C6.
These values here are a little confusing - CarlosFMs snubber tests came up with values of 0.47R and 1.5nF, much lower than the 1R / 100nF here. In addition, CircuitBasics uses a 5W resistor here, which seems like overkill, rather than Carlos's 2W.
How are these snubber values calculated, or is there a general "rule of thumb"?

Thanks!
 

Attachments

  • Schem1.png
    Schem1.png
    27.6 KB · Views: 657
Administrator
Joined 2007
Paid Member
^ Pretty much. It just maintains a low impedance at HF between the points linked. All the values are pretty much at the whim of the designer. A 5 watt rating for the 1 ohms seems crazy. A 0.5 or 1 watt carbon or metal film would be infinitely preferable imo.
 
That "Circuit Basics" is quite a comprehensive manual indeed.

There's another guy that just completed some rather 'ground breaking' work on the LM3886 amp and it's Tom Christianson and his Neurochrome website - it's well worth a visit and of particular note is how the speaker output terminals/connections are arranged
 
Read the the circuit basics guide well, it is very informative regarding the lm3886tf circuit (I have not found any guide that explained the circuit in a simpler way). After that read neurochrome guide about the lm3886. I say this because the transformer for basically the same application differs by a factor 2.5, which is quite a alot.

circuitbasics 250VA (rule of thumb / based on continious maximum powerdraw)
neurochrome 100VA (refined method / based on the crestfactor of music)

I am not saying circuitbasics method is wrong, but powerwise the CB-designmethod does not feel as refined as in the neurochrome-method. Maybe this massive difference also translates into the unregulated powersupply because it is upstream from a transformer perspective.
 
Last edited:
A PSU with bypasses and snubbers looks so much more 'professional' than one with just a reservoir cap, so surely it has to sound better?

The important point is the voltage across the amplifier not the power supply.
The amp should be well decoupled.
I often add extra RC decoupling for the front end of amps so on high powers power rail distortion isn't getting through.
 
A PSU with bypasses and snubbers looks so much more 'professional' than one with just a reservoir cap, so surely it has to sound better?

My apologies in advance if I am hijacking this topic.

Do you have a better psu-design in mind for a similar 20W lm3886tf amplifier? I have also used the circuitbasics-guide to select components for my lm3886tf amplifier. I want to try to design the pcb's myself, so I still have alot of room to manouvre (also pcb wise if components should be on the lm3886 pcb instead of the psu pcb).
 
Last edited:
I would put the rectifier and reservoir caps in the PSU, and decoupling caps next to the amp chip. Getting the grounding right is far more important than using very large caps.

Parenthetically it always seems odd to me the degree with which solid-state designers go to “reservoir out” their power supplies. Look at any chip-amp (from the dawn-of-time to present!) and you find PSRR's¹ exceeding 80 dB. So, both in practice and in theory, any residual ripple on the good ol' power supply is handily “taken care of” by the PSRR of the amp.

AKA “remarkably high negative feedback”.

Oh… if we're casting about the castor bean beads and shaking our dried unicorn wands vigorously, it seems that convention guilts us into believing that Big Reservoirs are Markedly Imperfect at doing their reservoir job, so need endlessly hopeful HF bypassing, per the OP's original schematic. Dual 8,200 µF caps bypassed by a 100 µF cap, bypassed by a 1.5 µF, bypassed by a 0.1 µF, and then finally “sweetened” by yet another 0.1 µF in series with a snubber resistor.

Mmmm… ridiculous.” Comes to mind.

All that, and not even a passing bit of attention to using smallish chokes to mercilessly squash HF hash coming from the rectifier banks. Its like assembling a massive army and forgetting to outfit the soldiery with bullets AND guns. I suppose one could always throw more snow cones at the arriving offensive force.

Anyway, snippy sniping aside, just KISS is the real answer, whether directly stated so or satirically hinted at by other postings. KEEP IT SIMPLE, SURELY! (wouldn't want the censors to shortchange my commenting status due to a st*p*d reference to a banned word…).

GoatGuy

¹ - Power Supply Ripple Rejection, measured in dB, since tiny fractions are so hard for the mind to handle. –70 dB = 10 to the (–⁷⁰/₂₀ = –3.5) power. Or 0.000316 … small number.

This means that a 1 VP-P power supply ripple will only result in VOUT RMS × 0.000316 ripple-on-signal intermodulation distortion². Pretty darn good. Pretty darn good indeedy.

² ≡ IMD
 
Last edited:
@GoatGuy, @DF96 and others.

The circuitbasics writer did address grounding in that article (link to that section), but the exact details of the psu-circuit were not in the scope of the article like the lm3998tf circuit was. Look at the picture below, it comes from the same article and does not have those extra capacitors in the circuit. Maybe you guys/girls and the author are on the same level regarding the importance of grounding and how to do accomplish good grounding?

9dm1jGY.png


In my first attempt of mastering easyeda website-utilities, I made an attempt to copy the above circuit in easyEDA (shared as readonly hopefully). This without those extra capacitors you guys/girls mentioned and having the decoupling capacitors on the pcb of the lm3886tf according to your comments and the circuitbasics/neurochrome design (is similar:1000uF/22uF/4.7uF X7R).

HpFSFMj.png


Hopefully these TDK capacitors are of a high enough quality to serve as reservoircapacitors. TDk has more expensive ones also of which "resistance specifications" is sometimes mentioned on mouser, but these were a bit the cheaper.

So my question basically is: Is the easyEDA-circuit above what you guys/girls ment designwise (sort of) or does the circuit need more improvements? (I am not the sharpest tool in the shed so please be civil ;-)




Edit: I did not use any smallish chokes to mercilessly squash HF hash coming from the rectifier banks, because I basically do not understand what you mean and how to addres the issue in a practical way. If you could help me out with some tips I would be grateful.
 
Last edited:
I am unclear what is the supposed function of the snubber. In the original circuit it was correcting for the mistake of adding a bypass cap, but now there is no bypass so nothing to correct.

Yep.

Three things (being constructive):

№ 1: lower PS bridge is wrong This is a big error, if the labeling is correct. Note that the pins on the far right are "V+" and "V-"? Its a very important labelling issue.

The original drawing is of two independent power supplies.
Your EasyEDA mockup is of two supplies whose outputs are hooked in series.
But the problem is that “the bottom one” is backwards.

№ 2: full wave bridge backwards on bottom supply. Pin 4 and 2 are the issue of № 1 above. Remember the old mnemonic: “points to positive” for diodes and rectifiers. The symbol conceptually points towards the positive and away from the negative function in power supplies.

№ 3: No choke between the Big Reservoirs per section. In all the power supply designing I've done, putting a choke between the large pair of reservoir caps on the "hot rail" significantly squashes power supply buzz. It doesn't need to be a big choke either. 1 millihenry is enough. Buzzkill! (How often do we get to borrow a commonplace word and use it quite accurately?)

Anyway, if you were to use a choke, consider https://www.mouser.com/datasheet/2/177/5c0032-57977.pdf at the 1 mH / 10 amp type. You'd need a pair of them. Maybe outside your budget at $17 ea.

Good luck!
GoatGuy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.