"Peer Review" on Dual Tracking Linear PSU with Pre Regulator

[Mechatrommer]

this is what i'm working on for more than a month now, i thought i can complete this in one week in one PCBoard. but now it evolved to 3 boards due to enclosure limitation, 7x LM324 duh. countless number of times of amendment, trashing, redo etc. i hope i can get this thing right before laying components on the PCB for the Nth time after Nth mistakes. so if anyone with keen eye may help to comment and critics. if not much serious mistake, i'll proceed with PCB layout, i can cut traces on copper plate latter on , in the mean time i think i need to rest for other things these few days as this thing has halted my normal daily routine for weeks. i wont elaborate as maybe no one will be interested anyway, and this thread may turn into a build log. but... any doubt just ask, cheers. (complete schematics available in the attached zipped pdf)

[Mechatrommer]

these are the hierarchical elements in the schematics above...

[Mechatrommer]

What are the desgn goals.10W or 100W? What is the source for your supply are you running off a residential AC line.

somewhere 100W output power, but to be exact will be empirically tested. for example power will be highly related to:

1) Vce of the power mosfet (Vin - Vout of LM317) = Vpa
2) sagging or ripple of supply voltage. pre regulated voltage (Vpi node) and the main 60V capacitor (V60+ node) under various load characteristic.
3) the magnitude of V60+ minus Vpi. the bigger the difference, i believe i can suck more amperage from V60+, more sagging is ok since we have plenty of room between V60+ and Vpi and hence Vro.
4) the limit of current is also bottlenecked by power bjt spec, ie 20A for each rail. 2x TIP142 is all that i can provide and the pcb can fit right now. i wish more

from empiric testing, mcu program will be made to set or limit Vpa (difference between Vin and Vout of linear regulator, Vp = Vr + Vpa, default value 5V) and Ircb (maximum current) for a particular Vref (PSU's voltage output set by user using trimpot).

the source of power is from mains 50Hz transformer that i salvaged from somewhere. a hi-fi set iirc. its size is roughly 7x7x10 cm go figure. it has 2 separate secondary coils, 60Vpp and 6Vpp. the 60Vpp coil has center tap, thats where i'm getting V30 supply from, mainly to power ±15V controlling opamps and comparators. 6Vpp coils for supplying logic ICs, attiny261 and 74hc08n. each end of both coils' will be connected for common ground. from rough testing, i can suck 2A from 60V 3300uF capacitor with acceptable sagging. so i guess 100W is a reasonable ballpark, i will lower this spec if the system cannot do the job from empiric test. this will be a fun experimentation for me.

power supply can be clearly seen in this schematics (top left)...

[Mechatrommer]

attached are the rough look of the pcb from earlier revision (linear regulator board and pre (switch) regulator board), i've been doing this like 10 times of amendment i felt like much effort has been wasted. so its enough i said, i need to get this thing right in the schematics level first. but all is not lost, earlier pcb layout gave me rough idea on how much components can fit in. i'm suspecting the latest "switch regulator" schematic will not fit in a single board, its highly likely that i will have to move some parts to auxiliary control board, later...

[Mechatrommer]

sorry i forgot to mention the spec: its a 0 to ±50V, 0 to 20A Dual Tracking PSU. If i have more room, i will add 2 Quadrant Bipolar Source in there as a feature, but its not in the schematics yet, it depend on available room left... expected PSU output power is as follow, but this is just "expected" or theoritical, depending on how actually strong the "unspecified salvaged transformer" is..

[MosherIV]

its a 0 to ±50V, 0 to 20A Dual Tracking PSU

Thats 400W to dissipate worst case !

Looking at your 3d cad layouts, I would say that you do not have enough room for the massive heat sink you are going to need.

[Mechatrommer]

its a 0 to ±50V, 0 to 20A Dual Tracking PSU

Thats 400W to dissipate worst case !
Looking at your 3d cad layouts, I would say that you do not have enough room for the massive heat sink you are going to need.

how that would be? i guesstimate switching loss from V60+ to Vpi+ will be very little. also loss in current sense and bjt balancing power resistors are kept minimal due to i've put very low resistance, so most power will go to linear elements and to the output (usefull power). since this is pretrack regulator, linear element will only mostly constant heat loss of 5V * 20A = 100W (50W per TIP142), the remaining power will go to the output load. if both rail is used then its twice the number, i dont think the transformer can provide 400W. as i said, this is just earlier expectation, i will know it precisely in practice. and the decision i've made of using mcu to calculate maximum allowable current (Ircb, from Irc_pwm nodes) that the system can give. so this will be easily adjustable. may be i'll set 50W heat loss internally, and 50W output per rail... as for the heatsink, if there is not enough room inside the enclosure, i will put them at the back (heatsink, fan and linear power bjts). the above graph plot will be internal loss + output power, yeah maybe i missed this point i plotted it wrongly as if 100W is output power ignoring internal loss. which can be misleading.

[Kleinstein]

I think there are quite some parts still really wrong:

The switched mode pre-regulator does not look good. It's rather difficult to make it work over a large voltage range of something like 10-60 V. Also synchronous operation is not that simple. With a +-30 V raw voltage, a buck converter will not give you 60 V, but more like 28 V maximum. 

It is not a good idea to use LM317 or similar to do the regulation: the dropout is quite high. In the shown circuit this is about 3-3.5 V. More important stability against oscillation is tricky if you don't have a known well behaved load. Also combination with current limiting is tricky - transients can be really hard. Finally there is no separate sense line, so accuracy can not be good.

Due to the large dropout, there will be too much power at low voltages / high currents. So I doubt you can go above 10 A at low voltages. For the transformer I would expect something like a 200-300 VA rating. So after rectifier without PFC this is something like 140-200 W at most. So it is more like a 50-70 W power limit per channel, maybe slightly more at high voltage and less at low voltage.

I would really start with a smaller version - enough to learn, but with less expensive parts to burn and less danger to mechanical blow parts apart. So something like 1 A and 25 V purely linear might be enough. Having a switched mode (pre)regulator is than a second project. The first part is getting the circuit plan right - the layout should wait until after a discussion / simulation.

[Mechatrommer]

thanks for your advice. it seems you are talking (concluding) out of your experience. i can see your conclusion (as you posted) but i dont have your experience, that maybe i will get from the hard way when testing the circuit on the pcb..

I think there are quite some parts still really wrong:

i will be glad if people can point that out, this is this thread is for.

The switched mode pre-regulator does not look good. It's rather difficult to make it work over a large voltage range of something like 10-60 V.

i dont have your experience so i'm not sure the reason why. either you care to elaborate which i will appreciate, or i will figure out the reason when i build this stuff, thanks.

Also synchronous operation is not that simple. With a +-30 V raw voltage, a buck converter will not give you 60 V, but more like 28 V maximum. 

i'm not sure what you mean by "synchronous" but the buck convertor in the circuit is getting supply from ±60V (V60+ and V60- in p3.png), not from ±30V. V30+ and V30- is used to feed ±15V linear regulator for controlling opamps/ICs and some transistors that doesnt need full ±60V swing.

It is not a good idea to use LM317 or similar to do the regulation: the dropout is quite high. In the shown circuit this is about 3-3.5 V.

i agree about the large dropout, but since this is dual tracking (negative rail), i cant find complement for the AMS1117 LDO that i have in stock. and i'm not aware of other LDO chip with its complementary negative regulator, which are significantly better or popular (examples in the net) than LM317/LM337 pair. anyway, i believe the high dropout LM317 with tracking preregulator is better than a low dropout chip with no preregulator that has to take 60V in and 1V out in full linear mode.... even at 1A at best, let alone 20 or 10A.

More important stability against oscillation is tricky if you don't have a known well behaved load. Also combination with current limiting is tricky - transients can be really hard....
I would really start with a smaller version - enough to learn, but with less expensive parts to burn and less danger to mechanical blow parts apart. So something like 1 A and 25 V purely linear might be enough. Having a switched mode (pre)regulator is than a second project. The first part is getting the circuit plan right - the layout should wait until after a discussion / simulation.

this is my 2nd dual rail PSU project. the 1st is 15V, no current limit, but using opamp+bjts voltage follower setup, no dedicated linear regulator chip. i still can find the log... https://www.eevblog.com/forum/beginners/please-check-my-simple-psu-circuit/msg78890/#msg78890 check psu1f.png i can say regulation is not great (report also in that thread) but did the job for years for all my crude projects. but as i'm moving into more detailed, demanding and many projects, i think i may need a better/multiple PSU to work side by side with my crude dual tracker PSU ver 1. and this salvaged transformer that i usually used to supply me with 60Vac long time ago suddenly popped out in front of my eyes, so i think this is the time, i wipe my incomplete differential probe project that i'm working on in the middle to give room for this "critically needed" project. hence i choose LM317/LM337 to help with better regulation of various load demand, its the best to my knowledge. compared to my earlier opamp+darlington setup.

Finally there is no separate sense line, so accuracy can not be good.

i'm not sure what you mean. but i have sense lines in everywhere, shunt/current resistor sense (to auxiliary control board, 3 opamp differential amplifier sensing Vpi and Vri, gain 150X ), voltage divider at Vp (preregulated voltage or V input to the LM317) and at Vr (linear regulator output) on the linear regulator board (p5.png near the title block at the bottom). these voltage divider reading feedback will be buffered and then passed back to preregulator board for tracking and protection (drain) purpose.

ps: the need for voltage divider (divide by 5 for voltage and divide by 2 for current) is due to all controlling components are low voltage ±15V components. i basically have no high voltage parts/ICs, even transistors i need to make new orders (2SA1013, 2SN2383, TIP142/TIP147, IRF9540/IRF540) to cope with high voltage level. i'm not intending to buy high voltage opamps or specialized chips. because i know i will seldomly use those parts. transistor is ok, they are more multipurpose and cheap, so i have high probability i can use them in other project or as replacement damaged parts in stuffs. and i've stick to my own requirement to use jellybeans parts thats already accumulating in my stocks, and as a challenge for a cheap hobby project. fwiw.

best regards.
Shafri

[Mechatrommer]

AcHmed99: nice project for SMPS, i bought the china small modules for my need http://www.ebay.com.my/itm/10PCS-Super-mini-3A-DC-DC-Converter-Step-Down-Power-Supply-Module-3V-5V-16V-/201414945974?hash=item2ee5442cb6:g:kJQAAOxyXVBSL~xa 3A 50 cents each. in anyway, i can connect it to a battery or to my diy PSU later...

for the enclosure, i have also salvaged red 22x16x10 cm enclosure from damaged SLA charger (just diodes and a burnt transformer inside), minus the transformer that i'm going to use, i can only fit 10x14cm pcb inside, one board is not enough from the resulting schematics, luckily i have vertical space, so 3 boards will be stacked together from bottom up to form my new PSU.

[Kleinstein]

For a higher power lab supply it is a poor idea to use a 3 terminal power regulator chip. They are just not made for this. The proper way is a circuit based on reference and OPs or maybe the old µA/LM723 regulator chip. This one is at least made for an external power transistor. The 3 Pin regulators sense the voltage between two of there pins, so with quite some current at at least one of the pins.

For a low output voltage it is OK to have the linear stage as an OP and emitter follower. However for more than about 30 V, this would need an extra amplification stage or special high voltage OPs. The alternative way would be a kind of floating regulator, using the power transistor in emitter circuit. This way the OPs can run from something like +-5 V or +-12 V for any output voltage.

Building an SMPS is not that easy. So if you want to learn, I would test it as a separate small project first.

A 60 V-peak to peak AC voltage will result in something like a 58 V DC voltage after a bridge rectifier, not an +-60 V. With a center tap you have something like +-29 V.

For the learning stage it also a good idea to stay below about 40-50 V, just for safety reasons.