The problem is i got the circuit below manufactured already. So i am stuck with the LT1028 and the bjt's.
The first approach has nothing to do with LT1028 nor with BJT.
If you want 50mAp on a 5Vp input, then there is no ther way than R4 = 5V / 50mA = 100 Ohms. Don´t ever change R4 to any other value.
I told you to keep all other values as in the original circuit. (Hopefully the schematic of post#1 shows the originall vales).
You did not follow this rule with your new schematic of post#33. Thus I won´t comment on it unless you clearly write why and show the according math.
******
You say "I got 140 mW RMS". There is nothing like "mW RMS". Sadly the audio industry confused us. They used the nonsense phrase "output power RMS" as a synonyom for "output power for minimal distorted sine". Voltage may be RMS, current may be RMS, but not power.
RMS means root-mean-square. But power is not calculated this way. Power gets simply integrated / averaged.
-> so what you mean is "140mW average power"
*****
Power dissipation.
Mind that power dissipation in a linear circuit can not be changed. This means:
(as example)
* 3V and 20mA at a resistor causes 60mW of heat
* 3V and 20mA at a diode causes the same 60mW
* 3V across CE of a BJT with 20mA collector current causes 60mW
* and so on.
This is true for every device that can not store or convert energy (capacitors, inductor, battery, motor ...)
Since your circuit is rather fixed on the 50mAp current, and 5Vp input voltage, you are fixed with R4 = 100 Ohms.
Also the supply voltage of is fixed - at least we don´t have any other information
Also your coil inductance is fixed - at least ....
Thus your overall power dissipation is rather fixed, too. The only thing you can electrically change to reduce the overall power dissipation is to reduce the bias current of the BJT stage. But this will result in higher distortion.
Also one can reduce the dissipation of power on one part .. but this automatically will increase the power dissipation of other parts.
And to reduce temperature (hot spots) you can improve heat spreading: Fan, heatsink, bigger parts, bigger PCB copper area ...
****
The biggest improvement in power dissipation will be to move to "switching circuits" instead of your linear approach.
I guess there are many audio class D amplifiers that could be used here.
****
If I´m not mistaken: the voltage across the coil is just a couple of mV. so Why use +/-15V supply? ... for sure this huge difference causes a lot fo heat.
We don´t know your options, neither your true requirements.
Indeed it took about 30 posts to find out that your main goal is to get 50mAp coil current. This says a lot.
Klaus