Anyone care to review this power supply circuit

[Balthazar]

Hi

I'm trying to make circuit to power a UAF42 filter IC with bipolar +/- 9v. Anyone care to look at it? Here it is:



I'm planning to use Sanyo polymer caps for C1, C2 and C4, (180uF and 82uF). The 180uF ones have an ESR of only 22mOhm, and can handle 3A ripple current. The 82uF one has 28mOhm ESR and can handle 3.3A ripple current. They are said to be more expensive than other electrolytic caps, but I thought their low ESR could help to smooth out ripple in the Vs+ and Vs- voltages. I know it will be out of the audible range, but I'm a little paranoid.....

RLoad1 and RLoad2 are there only for simulation. The transistor is there because of this section in Linear Technology's manual: "Special care must be taken in LT1054 circuits to avoid pulling [the Vout pin] positive with respect to any of the other pins" ... "Note that most op amps present just such a load" ... "To prevent start-up problems with such a load an external transistor must be added as shown in figure 1". They have a formula to calculate the value of Rx:

Rx =< (|Vout|)B
         -----------
             Iout

The B is beta. Since my limited knowledge of electronics comes from reading a book about tube amplifiers(!) I know nothing about transistors. Any comments on my choice of Rx are welcome. I just assumed Iout would be the maximum that LT1054 can produce 100mA, although I think it will be far less. The circuit will power a state-variable filter UAF42, which is an IC consisting of 4 op-amps. The 20pF C3 is there to alter the switching frequency of the LT1054 out of the audio band. So, does the schematic look alright?

[Skruffyhound]

Funny, I just posted on this subject here ; http://www.diystompboxes.com/smfforum/index.php?topic=100442.msg887609#msg887609 after reading a lot of data sheets this evening.
I'm a hack so I can't tell you if your circuit is any good, but I'd also be  interested to know if it will work and protect the IC against the evil opamps (I missed that snippet of information, well spotted) so I'll bump your thread 
and hope someone knowledgeable comes by.

[PRR]

If you have a live simulator, make Rload closer to 1K so they pull ~~9mA which is a fair approximation of what your chip will suck.

Layout is more important than dumping money in fancy caps. You can have more than 33mOhm in a badly run PCB trace. The mOhm hardly matters if switching spikes have their own path around their loops.

[Balthazar]

Thanks. I'll change Rload as you suggest. I wouldn't say I'm dumping money on the caps. I searched on google for the type, and the first quote I found was 0.87 dollars/piece from a european seller. But I also found some Chinese sellers who charged 0.1-0.2 dollars/piece, with a minimum order of one. The type is Sanyo 16SVPF180M. The tech specs for different "switched capacitor voltage converters" explain a lot about layout, so I've been reading and taking notes.

I didn't understand this sentence: "The mOhm hardly matters if switching spikes have their own path around their loops".

[Balthazar]

OK.... I have now simulated it in LTSpice with the whole shebang: The filter chip in place, a LED to indicate that it's on, and a source. I wanted to see how much current the whole circuit draws from the battery or the AC adapter. I couldn't find any numbers of how much internal resistance there is in a regular AC adapter, but I had made a previous search, and remember some estimates of 0.1-0.3 Ohms. So I tried simulating with an internal resistance of 0.1 Ohms first. I let the LED have about 20mA.

I see there is a large current spike of 31A at about 20 µs that immediately dimminishes. The curve goes like this:

Surge 31A for 20 µs
1.33A at 57 µs
1ms 500mA
3ms 300mA
6ms 100mA
10ms 50mA
20ms 36mA

With an internal resistance of 0.5 Ohm I get:

Surge 14A at 20 µs
1.9A at 120 µs
800mA at 340 µs
...continues like 0.1 Ohm

I couldn't simulate with 1 Ohm, which seems to be the value for a new battery, because it choked LTSpice. So, for an extremely short time it draws far more current than an AC adapter can handle. But within a fraction of a microsecond it's within reasonable amounts. Clearly it's drawing a lot of current to fill up the capacitors at startup. I chose large capacitors to eliminate ripple. Is this reasonably safe for most AC adapters? Maybe an internal resistance of 0.1 Ohms is unrealisticly low anyway. Also, is 20mA alright for a LED? I just took that number out of tech spec recommendations, but it means it will draw more current than both the charge pump and the filter together. If the simulation is anywhere near correct. I used a 350 Ohm resistor for the LED at 9V.

[PRR]

> Surge 14A at 20 µs

It takes many seconds to burn-up a wall-wart. 20 micro Seconds is nothing.

Note that 0.5 ohms at 14A is 7V, essentially the whole supply voltage.

There ARE issues with getting systems started. However it is quicker to just smoke-test than to collect ALL the data needed to simulate.

> Also, is 20mA alright for a LED?

Won't hurt most LEDs. May be too bright on a dim stage. As you say, with transistors, one lamp (even LED) can be more power than the whole rest of the machine. For battery work, you aim much lower. Indoors, 1mA is often plenty for LEDs aimed at the user. With wall-wart, 20mA isn't a real drag, but you may still want less for eye-comfort.

[Balthazar]

Thanks for your help!