Hello aabc!
After an extensive period of just lurking and reading the book, I finally got around to making a post, yay for you(or me, depends on the point of view actually )

I am currently designing a 9 volt power supply for a few of my homemade guitar effect boxes. Since it's for audio I would like to have as little ripple as possible and be as stable as feasible.

The schematic started out as a non-regulated "normal" supply with 4x4700u + 4x100n capacitor filtering bank, then I read alot about regulating, tried different things in LTSpice and one design kinda "stood out" among the crowd.

Now, this design is by no means "finished" and I am hoping for inputs on how to improve it(within reasonable terms of course).

I have chosen for the output transistor the FJN965 from Fairchild, some due to it being rated at 5A(in a TO92 casing, I'll take those 5A with a grain of salt) and mostly due to the fact I have those in my inventory already. (The schematic actually says SS8050 witch I also have, but that is a 1.5A unit and yes I know the pinout is different so the board is wrong )

I have 2N3904/2N3906/SS8050/SS8550/FJN965 to work with, any pointers on the design, am I doing something horribly wrong? Can I do something better within reasonable limits? Am I overcomplicating an other simple problem?

Appreciate any input, comments, rants or outbursts

-
Kolbjørn

 

Hi,

the image files are coming up so small they are illegible.

 

Hi,

the image files are coming up so small they are illegible.

Uhm, try clicking on them?

 

What we get is just the thumbnail.

Repost the images editing your post.

If you can't, provide a link. I use Photobucket. Simple.

 

Overcomplicated, a simple 9V regulator would do this job, either a LM317 or a LM7809.

 

Fused for 1.5 amps in and 500 ma out? All those filter caps will not help much with ripple. That's the job of the regulator.

 

Overcomplicated, a simple 9V regulator would do this job, either a LM317 or a LM7809.

Indeed it is overcomplicated, my solutions tend to be over-engineered.
But as it stands now I currently don't have any variable or 9v fixed regulators in my inventory, so this started kinda as a learning experience and grew from there. (I've been through both zeners and opamps controlling transistors, this was kinda in the middle.)

Fused for 1.5 amps in and 500 ma out?

Aye, the diodes have a nasty tendency to draw current at the top of the curves, so the fuse is a bit bigger and will be the slow type aswell.
(During my simulation in LTSpice the first spike hits over 45A, filling the caps is ofc mostly, if not all to blame).

All those filter caps will not help much with ripple. That's the job of the regulator.

Sure, the regulator does, as it stands now, take alot of the remaining ripple away, but if I start removing the bulk filter I tend to get bigger ripple. I will admit all those caps are a bit over the top for the application, but thats kinda how I like it. (And I do have a few caps I can't think of anything else to do with *shrug*).

C11 is helping the regulator to take away a lot of the remaining ripple, but if I start to up it any bigger than I already have I'm running into the problem where I have to wait several seconds before I get a stable voltage.
As it stands now, with 470uF, I pass 8v in ~220ms. If I up C11 to 1000uF we are looking at 460ms. So I am a little hesitant to up it to much.

Again, most of this info is based on my simulation with LTSpice and may not apply in the real world, feel free to point at where my understanding of things are lacking so I can better my understanding of this.

 

All those filter caps will not help much with ripple. That's the job of the regulator.

Dang, seems I have to call defeat on this one, although the caps have an effect, we are talking mV at near full load, which mean that in practice it won't matter.

Cutting away 4700uF gets us up to 1.5mV ripple, if we leave just one 4700uF cap we are looking at 4mV ripple, so you were quite right to point that out to me.

-
Kolbjørn

 

I've simplified the circuit a bit, removed the current limiting transistor and the darlington setup. Now it's just two transistors where one is regulating and the other is getting feedback.

The circuit is doing ok in LTSpice and I have a working prototype hooked up. Of course I know that if I short the output now, having removed the current limiting transistor(or what was a try on a current limiter), at least Q1 will be sacrificed to the transistor-god, maybe Q2 as well.

I'm a little bit unsure on what input voltage I should base the output on; I live in a country that can have up to 10% variation on mains voltage, getting us anything from 8.1 to 9.9v ac.
With light loads we have even higher voltage, so getting a uniform voltage out of the regulator, at different loads, seems an interesting challenge.

If you have any ideas on how to improve this circuit, don't be shy, I won't bite

Pictures!

PS: Yes, there is a lack of small bypass caps, they have been omitted for simplicity sake and will be in the final version(the one getting a pcb, in 2099 probably )

-
Kolbjørn

 

Hopefully getting my package from mouser today, so did some calculations on the circuits regulation(read: LTSpice did some calculations on my numbers):

Varying input voltage with 5k load and 5k6/1k voltage feedback divider(R2/R3):

Voltage In:     Voltage Out: 
11.0                8.33       
12.0                8.88       
12.7                9.18     
13.0                9.28
14.0                9.61
15.0                9.89

Mains can vary up to 10% here and still be within spec, so that gives us 8.1 to 9.9v AC from the transformator, multiplied with 1.414 gives us 11.45 to 13.99 DC voltage, with 12.7 beeing the spot on voltage.
I did calculate it up to 15volts cause transformers generally put out higher voltage than spec when loaded lightly.

These numbers seem fine, I'm pretty confident any 9v apparatus will cope with that, but what happens if we start loading the circuit?

Varying load with 5k6/1k divider, 12.7v input:

Load:                  Current:        Voltage out:
5000                    1.8mA               9.18
2500                    3.7mA               9.16
(Skipping ahead to the interesting part)
 100                    89.8mA              8.98
  50                   178.0mA              8.90

I must say myself happy with these figures, not much drift at all.
Now, as I am not using the exact same type of transistors in LTSpice as in real life, this will only be an approximation.
In real life the voltage out of the transformer will vary according to the load, with no load we are expecting 13-14v.

I have put back the transistor in the darlington stage again, if I started to load the circuit I was lacking gain to keep the voltage sufficiently stable(According to LTSpice).

-
Kolbjørn

New print layout:

 

Try placing a low-value high-wattage resistor between the bridge and 1st filter cap.
For example, a 3.3 Ohm resistor between the bridge and the 1st filter cap will limit your inrush current to around 2A.

To reduce the ripple further, use an inductor between the 1st filter cap and the 2nd filter cap. Try a value in the range of of 220uH to 1000uH.

Look up "pi filter"

[eta]
Threw some numbers in Elsie and came up with a 50Hz 5-element Chebychev filter.

Started off with R1 being 3.3 Ohms, but if you replace that with an automotive bulb rated for around 10w, you will get a nice non-linear response.

Note that Rload is just simulating a 180mA load; it wouldn't be in your actual circuit.

Image and .asc attached

 

Very interesting, I would never have thought of using a light bulb for this application, but it does make sense, looking at the resistance curve of a bulb.

I will probably skip the inductors though, seems adding more filtering before the regulator does not really do anything major in this circuit, I removed 3 of the 4700uF caps and was still under 5mV ripple. The major part of ripple rejection is done in the regulator.

I'm kinda new to inductors, but in what grade will we see a voltage jump like the ones we get in relays and motors(and joule thieves)? I'm guessing it's not a issue as the inductance is very small compared to a relay-coil(is it? I'm guessing wildly here).

You kinda poked my curiosity with mentioning the Chebychev filter, I have to investigate inductors some more as they intrigue me; And the fact that I had never heard about a Chebychev filter

-
Kolbjørn

 

The benefits of the inductors will become more apparent as the regulator nears maximum load. The Chebyshev design I used has a nearly brick-wall response, allowing frequencies below 50Hz to pass through with 3dB ripple in the passband. At 50Hz, there is at least 9dB rejection. It's over 30dB down at 60Hz.

The inductors used in the filter are 6.2mH (milliHenries), which are fairly large. You could salvage some large ferrite toroids from junked equipment (like old amplifiers, DVD players, satellite TV decoders) and wrap 50 turns or so of 18-20 gauge magnet wire around them to get in the ballpark.

You wrote:

'm kinda new to inductors, but in what grade will we see a voltage jump like the ones we get in relays and motors(and joule thieves)? I'm guessing it's not a issue as the inductance is very small compared to a relay-coil(is it? I'm guessing wildly here).

6.2mH is pretty large. Motors have more inductance than that, unless they're pretty small.

Inductors keep the current flowing through them stable. If the current path is interrupted, the magnetic field around them collapses to keep the current flow going. If there is no path for current, the voltage across the inductor can become extremely high. That is why diodes are used across relay coils; to provide a current path for the reverse-EMF that occurs when the switch controlling the current is turned off.

High frequency DC-DC converters (and other types of boost, buck-boost, and buck regulators) can use quite small values of inductance. The lower the frequency, the larger the value of inductance is required.

 

Elsie can be downloaded here:
http://www.tonnesoftware.com/elsie.html
It is very good, particularly for a freeware-enabled program. You can design filters up to seven elements, and simulate the results.

LC filter design can be a pretty specialized field. There are engineers who spend their entire careers doing nothing but designing LC filters.

 

Been trying out the filter in LTSpice, and I must say I'm impressed, it's performing very well.

At 50Ohm load(~180mA):
                    1x4700u:    3x4700u:    3x4700u+2x6.2m
Voltage Out:         9.0377V     9.0416V       9.0372V
Ripple:               7.31mV     2.435mV       458.3uV

At 25Ohm load(~350mA):
                    1x4700u:    3x4700u:    3x4700u+2x6.2m
Voltage Out:         8.7567V     8.7637V       8.7551V
Ripple:              13.76mV      4.83mV       1.76mV

At 7Ohm load(~1A):
                    1x4700u:    3x4700u:    3x4700u+2x6.2m
Voltage Out:         7.8021V     7.8356V       7.3035V
Ripple:              35.6mV      22.45mV        35.7mV
(NOTE: The ripple on the last measurement with the inductances was not
really ripple at all, just slow variations in the voltage, kinda varying 35mV
during .1 of a second or so. 
Very interesting fenomenon, somewhat sure of it beeing not audible 
at all, or atleast the ripple without the inductors are worse.)

So, the golden question, is this something I plan on putting into action? Not on this build as I'm kinda missing the important bits from my parts-bin, but I have plans for a gainclone amp thingie with a regulated supply(approx +-42v regulated down to +-35) that could benefit from this, sometime in the future.

I'll check out elsie as well, I love fiddling with these kind of programs, eventually some knowledge leaks into the brain. Atleast I'm not bored

-
Kolbjørn

 

For some good scoop on inductors and various experiments, have a look at Ronald Dekker's page:
http://www.dos4ever.com/flyback/flyback.html

I'm not suggesting that you build a HV power supply; if you do, be careful.

Ronald's created a very handy resource for people who are interested in dabbling with inductors, boost converters and flyback converters. His examples are certainly not the most efficient, but they serve as a very inexpensive way to "get your feet wet" with a few common parts and items from your junk bin.

For example, I'll bet you didn't know that you can open up a burned-out CCFL bulb and salvage the inductor from it... (be careful to not get cut on the glass...)

 

For some good scoop on inductors and various experiments, have a look at Ronald Dekker's page:
http://www.dos4ever.com/flyback/flyback.html

I'm not suggesting that you build a HV power supply; if you do, be careful.

Hahahaha, I'll be sure to not touch the dangerous bits if I do make one
Thanks for the link, I love learning new stuff.

As a little fun test I did some LTSpice testing on ripple vs high loads on the circuit, I started at 20 and worked down to 1 ohm, recording voltage, ripple and current. I made two schematics due to one of the readings was completely off the scale, destroying the range of the chart.
(The last reading on the 3x4700uF / 2x6.2mH is due to me taking the largest voltage jump, it's not really ripple, it's a slow voltage variation of some kind, not sure how to describe it )

-
Kolbjørn

 

Did a voltage vs. load aswell, to see how the different circuits hold up under unrealistic loads.

 

Keep in mind that you are likely using "ideal" capacitors and "ideal" inductors, which don't exist in the real world.

There are a number of capacitor models available in LTSpice. Try right-clicking on a cap, and then click on "Select Capacitor" for a list.

Add some parasitics to the inductors, too. 56 turns of AWG20 on a 31mm*18mm*6.3mm iron toroid would take about 100 inches (2.5 meters) of wire, for roughly 85m Ohms.

If you want to decrease the cutoff frequency, try increasing the inductance to 9.2mH.

 

To improve load and line regulation to a fixed output 9 Volt regulator, try the following...

1. Add 1 ohm in front of the 4700u cap to limit inrush current.
2. Get rid of C4, it is preventing the regulator control loop from working properly.
3. Get rid of D4 and replace with a zener, or precision reference. Adjust the value of R3 to suit.
4. Replace C2 with a 220u cap.
5. Use 10u to 100u as the output filter to ensure stability and improve transient load response.
6. Put 0.1u across R2 for stability if required.

In lue of a zener a LED with a 2.5V fwd drop and a diode drop would do nicely. Basicaly, the higher the reference voltage, the better, but not more than 5V. Any ripple on the reference voltage will show up on the output.

Have Fun,
Ifixit