Simple? Power supply design question

[cowasaki]

My usual haunt is digital electronics, microprocessors, arduino etc but I have some experience in design.  Power supplies are something I think I know about and the enclosed design does appear to work as per spec other than the following which I could just ignore but I want to learn and want to know how it could be better.  It is 5v,12v & -12v around 300 milliamps for each rail (so 12v is effectively double that because it is where the 5v also comes from).  I'm giving it enough headroom though that 500mA is doable.

1) I'm using 2200uF electrolytic capacitors to smooth each side of the rectified voltages coming from the split output transformer.

Question - This is more habit than anything else.  How would people normally decide on the actual values.

2) I'm using 470uF electrolytic capacitors to smooth each side of the voltages coming from the rectifiers.

Question - How would people normally decide on the actual values, I have made an educated guess.

3) The output of the 7912 is -16v until a load is added at which point it drops to the right level.

Question - Would people add a load across it to force it to drop or just leave it like that?  Maybe power the "Power on" LED using the -12v line?

4) More general.

Question - Is there any real issue using too large capacitors other than general engineering principles of waste of space/cost etc.

5) Regarding electrical checks re power.

Question - If I want it to manage 500mA how would people check this.  My current plan is setting the limit on the bench power supply and using large resistors to load the output.  Is this a reasonable plan?  Is there something better?

note) The USB isolation section had been reworked to only have the ADuM IC as an option but KiCAD has changed my file so I can't load the newer version just at the moment.

[Wimberleytech]

1) I'm using 2200uF electrolytic capacitors to smooth each side of the rectified voltages coming from the split output transformer.

Question - This is more habit than anything else.  How would people normally decide on the actual values.
>>When the filter capacitor is not being charged by the AC source, the current drawn by the load cannot cause the voltage on the capacitor to dip below the required input voltage of the regulator.  I will show below a diagram that illustrates this.

2) I'm using 470uF electrolytic capacitors to smooth each side of the voltages coming from the rectifiers. <<regulators??>>

Question - How would people normally decide on the actual values, I have made an educated guess.

>>You don't need 470uF.  Typical is 0.1uf but look at the datasheet for the regulator.  The manufacturer will recommend the value.  I noticed that on the 7912 datasheet, you are warned that any value over 100uF needs a diode from input to output because on powerup, the capacitor appears as a short momentarily.

3) The output of the 7912 is -16v until a load is added at which point it drops to the right level.

Question - Would people add a load across it to force it to drop or just leave it like that?  Maybe power the "Power on" LED using the -12v line?

>>Sounds odd to me but it has a long time since I have used a negative regulator.  I looked at the schematic of the 7912 and there is already a resistive load ground from the output.  Look at it with an oscilloscope.  May be unstable.  Remove the 470uF cap and check again.

4) More general.

Question - Is there any real issue using too large capacitors other than general engineering principles of waste of space/cost etc.

>>As noted above, a large cap can create a problem.  They can cause stability problems in some circuit configurations. 

5) Regarding electrical checks re power.

Question - If I want it to manage 500mA how would people check this.  My current plan is setting the limit on the bench power supply and using large resistors to load the output.  Is this a reasonable plan?  Is there something better?

>>That is probably what I would do.

[duak]

Filter cap sizing is pretty simple, at least to the first approximation.  It neglects series resistances which tends to make things a bit worse.

Basically  dV/dt = Iavg/C where dV is the p-p ripple voltage, dt is the period of the voltage waveform from the rectifier, Iavg is the output current and C is the value of the capacitor in Farads.  Solving for C gives = Iavg*dt/dv.   I remember it by thinking that the capacitor value is proportional to the current & period and inversely proportional  to the ripple voltage.

For example, with full wave rectified 50 Hz;   1 amp * 10 ms / 1 Vp-p = 0.01 F = 10,000 uF

Cheers,

[cowasaki]

Thanks I will make some mods to a second board and see what the results are.  They sent me 11 so I've got plenty to play with

[ocset]

Please find attached an LTspice sim which shows you whats going on.
You can click for the ripple current in the caps which are downstream of the rect.
You can also adjust the situation with the cap sizing and the transformer turns ratio and loading to get your wanted compromise.

Remember to do it at min mians input.

You can get ltspice from linear.com
its free
just open .asc file and hit running man icon.
Write back here if you want a quik start guide
its very simpel

[Wimberleytech]

Nice work.  That should help cowasaki.

[ocset]

Thanks, hope so, as  you know,  i  used a too low  primary inductance to be realistic , but cowasaki can play with that.

Actually ive corrected it here, and added a ground connection for the primary, which is needed for the sim.
And just for giggles, ive added another with a really high Lmag(pri)  so the mag current gets lower

[cowasaki]

Thanks everyone. I’ll play with it but making a decent power supply is key to so many things that I thought it would be good to get some more expert input. Most of my stuff is powered by the host device so I don’t normally have to worry about the power supply other than to make sure it is adequate and maybe a capacitor that smooth it out. I’ve just got home with the wood for my new bench so no electronics for me today but I should have a really nice space by Monday.

[Wimberleytech]

LTSPICE is pretty amazing.  I have only recently begun to use it.  For years I have used a Spice I paid for with arguably a better interface...by my standards at least.  But once I have gotten more familiar with LTSPICE, I have grown more fond of it, and since so many people use it in posts on this forum, I find it a necessity.

[cowasaki]

I received the isolator IC and so decided to build a second isolation board this time with the USB isolator and with the mods which were suggested and worked out.

The second board started with the same issue ie -16.8V from the -12V rail.  It would appear that the -12V regulator cannot handle a lack of current drain so I made a slight modification to the circuit and powered the LED from the -12V rail through the 220r resister rather than from the 5v rail.  One cut track and a short jumper wire and its working fine as is the original board (I'm not totally happy with the workmanship of the mod on this board but my circuit specialist's magnifying lamp fell to pieces due to becoming brittle over time and it was hard to do it using a loupe! - Just waiting for either the new USB microscope or new magnifying lamp to arrive)

Unfortunately I can't check the isolator as I have no 820r SMD resistors but hopefully I should have these in a couple of days.

[cowasaki]

Thanks, hope so, as  you know,  i  used a too low  primary inductance to be realistic , but cowasaki can play with that.

Actually ive corrected it here, and added a ground connection for the primary, which is needed for the sim.
And just for giggles, ive added another with a really high Lmag(pri)  so the mag current gets lower

I will have to check out LTspice so that I can see what you've obviously taken time over doing.

[cowasaki]

Well I've installed it and had a bit of a look.  I'm going to have a look tomorrow as it looks very useful.  The power supply is working as per spec but I'd still rather learn how to make the best stuff I can.  The vast majority of my electronics experience has been in digital electronics with the only real analogue stuff being repairs or in attaching sensors and the like to a digital devices. 

Because of this there are certain components that I don't have a real understanding of but I do want to learn.  For example coils, I understand that they charge up through magnetism holding current and effectively try to drag the current back up in a similar way to a capacitor does with voltage.  I could see how this would assist a power supply but I don't really understand where, how and what value I would use or more importantly in what circumstances.  I also understand that my description might well be fundamentally inaccurate.  It's stuff like that that I am trying to learn.

Thanks everyone..

[Yansi]

Filter cap sizing is pretty simple, at least to the first approximation.  It neglects series resistances which tends to make things a bit worse.

Basically  dV/dt = Iavg/C where dV is the p-p ripple voltage, dt is the period of the voltage waveform from the rectifier, Iavg is the output current and C is the value of the capacitor in Farads.  Solving for C gives = Iavg*dt/dv.   I remember it by thinking that the capacitor value is proportional to the current & period and inversely proportional  to the ripple voltage.

For example, with full wave rectified 50 Hz;   1 amp * 10 ms / 1 Vp-p = 0.01 F = 10,000 uF

Cheers,

Not quite. It is usually better to assume how much % of a period the diodes are conducting (say k=20%) and then for a bridge rectifier calculate like this:

C_filt = I_dav * (1 - k) / (2 * f * dU)

where k is the above mentioned ratio, f the mains frequency  and dU the required voltage ripple p-p. I_dav is the average rectified current.

Set the voltage ripple to ~10% of the nominal no load voltage.

[ocset]

if you want my SMPS course with well annotated ltspice sims, then ill send it to you by google drive if you email andymassey22@gmail.com