With the original schematic at hand I took the liberty to make a few changes. First of all I replaced two transistor-zener regulators with LM317L/LM337L. Circuits are calculated to produce 33V positive and 3V negative voltages. Thus total supply voltage for the opamps does not exceed 36V and therefor we may use standard ones. I also made changes in the LED drive circuit and a few other minor changes.

After that I decided to simplify the schematic even further. I replaced unnecessary complicated circuit for voltage reference build with IC2 with simple resistor-zener circuit. This will give us stable reference voltage as the supply voltage is already regulated with LM317. In the original schematic voltage reference is 9.4V so I chose to use two zeners - 3.3V and 6.2V connected in series which should give us 9.5V. Also the selected zeners have opposite temperature coefficients which should eliminate each other resulting in excellent temperature stability.

This was tested on the finished board of the previous version - I removed IC2 from the socket, desoldered R5 and Z3 and connected additional zener (for the test I used one 9.1V zener) and resistor with wires. It worked very well - just as I expected.

The rectifier gets very hot when the output current is over 2A, so a small heatsink on top of it will be useful.

The transformer must be 100-120W with 27-30VAC output voltage. You must make some corrections in the schematic if the output voltage is lower or the voltage drop is higher when the current is high. R10 and R21 set the output voltage of the regulator IC3 (LM317) and they must be calculated in such way that the output voltage is 2V lower than lowest input voltage. If, for example, lowest voltage measured across C1 when power supply is fully loaded is 27VDC then output of IC3 must be 25V. With R10 = 4k3 and R21 = 220R we will have that output voltage. With 25V stabilized voltage for ICs the maximum output voltage of the power supply will be around 23VDC.
The schematics will work without these changes but the output voltage will not be so stable.
If the voltage across C1 is bellow 33VDC without load then the IC3 will be unnecessary and we may omit it.

For the current sensing resistor R7 I use two parallel resistors 0.68Ohm/10W. You may use a single 0.33Ohm/10W resistor but it will get too hot.
With R16 = 82k and R7 = 0.33Ohm the maximum current limit adjusted with P2 will be more than 3A - more like 3.3A. If we want to be closer to 3A then the R16 must be 91k.

You may add a 1k linear potentiometer in series to the P1 for fine adjustment of the voltage. Or the better way is using a multi-turn potentiometer but they are expensive.

The bizarre looking zener Z1 connected with PAD1 is used to power supply the digital voltmeter which indicates the output voltage. It requires 6-28V supply voltage and with that zener I reduce the input voltage to an acceptable level. Z1 may be omitted if not needed.

A lot of people asked me to draw a diagram how to connect digital panel meters. Here how can be connected digital voltmeter and ammeter. As you can see, in "variant 1" the voltmeter is connected serially to the ammeter, so his supply current will be added to the measured current and will present a very small error (bellow 10mA). In "variant 2" the ground cable of the voltmeter is connected to the negative output of the board instead of the negative binding post. So his supply current will not be measured, but the voltmeter will display a little higher voltage because the drop voltage at the ammeter (50-80 mV max) will be added.
Make sure the combined supply current of two meters does not exceed 15-16mA (the zener Z1 will overheat).

There was also report, that negative voltage may oscillate. This can happen if the input AC voltage is lower or at high load current it drops significantly. Then input voltage for IC4 (LM337L) gets to low to sustain stable output voltage of -3V. The cure for this is simple: increase the value of C2 to 22 or 47uF.