Attempt at a fixed psu

[Vgkid]

In my attempt to create a rather stable 12v power supplyI have decided to use the ultra basic design of a opamp driving a darlington bjt which in term is driving a power transistor. Since I do not have much experience with this here is my circuit. Note the AC power supply is a 16V5A unit. In simulation the circuit will work down to about 14.5Vac.
Please post your critiques, note the V-ref/opamp/components are what I plan on using, suggest better TH substitutes.

[Deathwish]

Never seen an AC side that goes to earth/ground, isnt that going to go bang fast ?.

[MatthewEveritt]

Never seen an AC side that goes to earth/ground, isnt that going to go bang fast ?.

Not necessarily, it basically just shorts out one of the diodes in the bridge, making it act as a half wave rectifier, which could be fine. I'm guessing that isn't deliberate through.

[Deathwish]

Thank you Mathew, i have learnt something I never knew, I always thought if you fixed one side of an AC supply to earth it went bang or at least did not work.

[fubar.gr]

Some simulation programs require that all parts of a circuit must be ground referenced, else the simulation cannot be carried on.

LTspice is one such example. In that case you can connect the floating parts of the circuit to the ground via a very big resistor, eg 1GOhm.

This keeps the simulation algorithms happy and won't have any measurable effect to the circuit.


Also in the schematic you are using a huge 22000 uF capacitor. If you plan to implement this power supply in real life, you should strive to keep the smoothing cap value to a minimum. Capacitors are like a short circuit and the bigger they are, they cause huge current spikes.

First decide what the max output current will be. Since you have a 20 Volt peak input, and assuming a 3 volt dropout voltage, you can have up to 7 Vpp ripple.

Use the formula DV=I/(C*f) to calculate your minimum capacitor value. DV is the ripple, I is the max current and f is the frequency, that's double the line frequency or 120Hz

[Vgkid]

It will have a max load of 2.5A, though that will drop to about 1.0A max(closer to 850mA under steady state conditions). It will be powered by a 16v5A transformer. It will be powering a gpsdo and a few ocxo's.
Multisim likes to have everything grounded.

[BennVenn]

you're right Deathwish, its a short on the AC input for half its cycle.

[Vgkid]

Good catch on the grounding. My previous simu version had a transformer, and i recall much less ripple(and capacitance).

[c4757p]

Some simulation programs require that all parts of a circuit must be ground referenced, else the simulation cannot be carried on.

LTspice is one such example. In that case you can connect the floating parts of the circuit to the ground via a very big resistor, eg 1GOhm.

The other side of a rectifier isn't floating, though.

[onlooker]

The schematics has too many unnecessary zig-zag lines. Some are overlapping in a bad way. Maybe, this is just due to careless. But, some looked more like intentional (to make it looks more complicated than what it is?).   

[lowimpedance]

The schematics has too many unnecessary zig-zag lines. Some are overlapping in a bad way. Maybe, this is just due to careless. But, some looked more like intentional (to make it looks more complicated than what it is?).

Actually that is an annoying usage issue with multisim when placing connections, and too tiresome to rectify when just wanting the sim result.

[codeboy2k]

1) you need to compensate the feedback path to be sure you have a stable loop.  To do so properly, you should use a Frequency Response Analyser, for example from Venable Instruments or Ridley Engineering.. however most people don't have one, they are expensive, and for hobby use you can get by with just a rule of thumb making sure your loop is a little on the slow side, and over-compensated. So pick a frequency around 25-50Khz and make your loop filter fit that.  For example, you currently have the output voltage divider going straight to the inverting input of the OPA177 with no loop compensation. Instead, you should put a 1k resister in that path (from the divider) then follow that with a 2.2nf to 10nf capacitor from the resistor to the output of the opamp (so the cap is now connected from the output to the inverting input, and the 1k resistor is from the inverting input to the 20k/100k divider center node).  This gives a loop filter between 70kHz (2.2nF) down to 15kHz (10nF). This should help to stabilize the loop from most oscillations that will occur at much higher frequencies.  (1/2*pi*R*C)



2) instead of the ZTX601B/2N3055, I'd go with a single TIP140, unless you happen to have those parts just lying around already.  The TIP140 is a complete NPN power pair in a TO220, which is easier to mount then a 2N3055 (because the collector of the 2N3055 is the case, so you need to isolate it from your metal case and your heatsink which might be grounded).  Also the 2N3055 is just ancient, and not really used anymore.

3) as an alternative to #2, I generally don't like darlington output stages anyways, they have high drop-out, high-gain and low speed which makes them susceptible to feedback oscillations.  This can be compensated for (see #1 above) but if you can avoid a darlington output stage then you will be better off.  The darlington output stage also means you have to have an opamp that can drive the first base almost as high as your desired output voltage level (minus the 2 Vbe drops). You can get around this requirement, and also avoid the darlington feedback nasties by using a PNP/NPN output stage instead.  I did this with an example in another PSU thread. See point #4 in this thread, there is a schematic of the power output stage I suggested there.

[sync]

You should use remote sensing or the cable's voltage drop will ruin the precision. Also the 7815 wants a minimum load of 5mA and decoupling caps.

[codeboy2k]

... Also the 7815 wants a minimum load of 5mA and decoupling caps.

Considering this, then you have two choices. (1) put a load resistor at the output of the 7815 to get it to draw a constant 5mA, or (2) don't waste a 7815 and simply use a 15V zener diode drawing 5-10 mA of current instead.  Here's why a zener will work:  The line regulation of the reference is 0.002%/V, which put another way means that an input voltage change of 1V causes .002% change in output of the reference.  The reference is 0.05% initial tolerance. If you don't even calibrate the PSU output, you are still going to be well within 0.05% even if the zener cannot hold its output stable +/- 1V , but it can do much better than that.  The load for the zener regulator is only the reference, and it is going to be a relatively constant load at 1 mA or so (and the load for the reference output is only the input of an opamp which draws very little current).  So the zener regulator will not see much load change, so it's voltage output will remain relatively flat. Most zeners can hold within a few hundred mV over a 10 mA load change or so, and this application will see very little dynamic load changes at the zener regulator output. Even if it DID change by 10 mA, the zener voltage will change by about 100 mA or so, and that will result in just 0.002% of that getting through to the reference output, or just 2 uV of output change in the reference. It's insignificant and within your error budget.

As for the input to the zener regulator, a quick simulation showed that even a 4V input change, dropping the input from 22V to 18V, which can happen under heavy PSU output loads, caused a change in the zener regulator output by just 200 mV. This will be seen through the reference at 0.002%/V or just 4 uV of change in the reference output voltage.

My point of going through all that is to show you that a few observations and calculations can determine that a 7815 is not really needed here, and a zener is good enough for pre-regulating the reference input voltage.  Remember, the reference itself is already a voltage regulator.

I just used a 1K resistor in series with a 15V zener, to get the most basic 15V zener regulator.  You can do better, but it's all overkill.

[Vgkid]

Thanks for the suggestions. I will not use the 7815, but since i do need a bipolar portion, I will just use the transformerfor that(+\- to center tap) and the full tap for the 12v rail.
@ Sync how important is the sensed rail, considering the power draw is under 850ma. Note 14/16 awg wire will be used with total length under .25m.

[edavid]

This seems like a great application for an LM350T or LM338K.

[codeboy2k]

Thanks for the suggestions. I will not use the 7815, but since i do need a bipolar portion, I will just use the transformerfor that(+\- to center tap) and the full tap for the 12v rail.
@ Sync how important is the sensed rail, considering the power draw is under 850ma. Note 14/16 awg wire will be used with total length under .25m.

16 AWG is 13.17 mOhm per meter, so 0.25 meters is only about 3.29 milliOhms .  the I*R drop is 0.850 * 0.00329 = 2.8mV or 0.02% of 12V
It's not necessary to remote sense this application, unless you need accuracy at the load better than that.

[codeboy2k]

This seems like a great application for an LM350T or LM338K.

Agreed, a monolithic regulator seems like the appropriate choice here. 

The OP might be building from the parts bin, as I find myself often doing the same   I don't have an LM350T or LM338K handy myself, but I have opamps and power transistors here.

[sync]

Thanks for the suggestions. I will not use the 7815, but since i do need a bipolar portion, I will just use the transformerfor that(+\- to center tap) and the full tap for the 12v rail.
@ Sync how important is the sensed rail, considering the power draw is under 850ma. Note 14/16 awg wire will be used with total length under .25m.

16 AWG is 13.17 mOhm per meter, so 0.25 meters is only about 3.29 milliOhms .  the I*R drop is 0.850 * 0.00329 = 2.8mV or 0.02% of 12V
It's not necessary to remote sense this application, unless you need accuracy at the load better than that.

I somehow thought this is an external power supply and that the load varies (0.85 - 2.5A). Then the cabling and connectors can easily cause a voltage drop of 100mV or more. For an internal PSU with short cables remote sensing is likely not needed as Codeboy2k shown.

PS: I would use a monolithic regulator too.

[Vgkid]

Looks like I will run one of the monolithic v-regs controlled by the op177 circuit. I thought I still had those transistors, but couldn't find them.