What is going on? Strange component placement in PSU

[FriedMule]

I have stumbled upon this strange power supply for a 5GHz unit I found in an old magazine. So I can't copy more of the circuit since it may still be under copyright.
On the image have I marked the "strange" components, so you know what I am asking about.

1) the PSU contains two Voltage controllers, 7808 and 7805, why?

2) The Zener diode (D17) is in front of a LED, why and what does it do there?

3) There are two inductors (SD12K100) is they the same as SD12-102-R : https://www.mouser.dk/ProductDetail/Bussmann-Eaton/SD12-102-R?qs=%2Fha2pyFadug4QZjN0cSDOUFn6uFsp6ssRidD%2F6taA6M%3D

4) At the 9V is there two protection diodes (DD214AA), what are they doing there, are they not to protect against lightning and ESD?

Hope you'll please explain it to me. :-)

EDIT: Sorry but U7 is reversed, please ignore that.

[ataradov]

Dual stage voltage regulation is for filtering. You first drops the voltage to 8 V, but 7808 is not great at noise rejection. Then you see a whole lot of filtering and then it drops to 5 V for the final voltage. The inductors are there for the same filtering reason.

D21 / D22 logically prevent back feeding the first stage voltage regulators, but I'm not sure why it would be a concern here. They may know something about their load that would make them necessary.

No idea about the Zener diode. Are you sure there is connection to the ground? R16 makes no sense here.

[FriedMule]

Oh, thanks for your great reply. I have checked again and yes the zener is going to ground as shown, besides it is written 8V2.
Could the R16 be to make the same resistance / voltage drop as the zener and LED combined?
Do you know what data the DD214AA should have to work best?

[ataradov]

Equalizing the two parts should not make any difference, since the voltage levels will be further regulated later on. I have no idea what that resistor is doing. It is possible it is some load so that 7808 is stable, but that should not be necessary.

If Zener is 8.2 V, then this is some sort of over voltage indicator. I feel like it could be safely omitted.

Diodes should not matter too much, but low forward voltage Schottky would be best, since 7805 needs at least 2 V headroom to regulate properly.

It is also not clear why it says 9V. 7808 will produce 8V.

How old is that magazine? There are probably better ways to get clean +/- 5V supply.

[Vovk_Z]

If Zener is 8.2 V, then this is some sort of over voltage indicator. I feel like it could be safely omitted.

Yes, it looks like an over-voltage indicator, but it will work no longer then LED can withstand short-circuit current.
And in general - this circuit looks a bit like its designer not really know well what he do (so I don't see a reason to investigate it very deep). Possibly I am wrong because I haven't worked with 5GHz devices (but...).

[digsys]

Agreed with above comments, something isn't shown maybe? A tap off the center ? a second feed in there? Definitely ODD with a capital D :-)

[Zero999]

Presumably the transformer is 7.5V-0-7.5V? If so, the LM7808 and LM7908 won't do anything, because the input voltage is too low to give 8V out, with decent supply rejection.

[ataradov]

7.5 VAC will be rectified into 10 VDC, so the 8V regulators will work. Probably not very well.

The whole thing makes no sense.

[Vovk_Z]

7.5 VAC will be rectified into 10 VDC,

It depends.

[FriedMule]

The design is from the end of 2015, and made someone I understand is considered as a bit of a guru who should know a lot of what he is talking about and people are listening to.
As I understand your comments should the 7.5VAC transformer, give 10V and maybe via some voltage drop, give about 8V/9V?
If so should the zenor-LED maybe glow if voltage is ok? If Voltage is too high?

I think it may be smartest to just follow his drawing, but yes a lot of it does not make sense. I.e. the 8.2V zener, I simply did not understand what it meant for the circuit.

Also, I can not find out the impedance on the zener, the values needed on the two safety diodes or if the LED's needs to have some exact specification / color.

[tunk]

Do you know how much current that is drawn from the PSU?

[srb1954]

Oh, thanks for your great reply. I have checked again and yes the zener is going to ground as shown, besides it is written 8V2.
Could the R16 be to make the same resistance / voltage drop as the zener and LED combined?
Do you know what data the DD214AA should have to work best?

Is it DD214AA or DO214AA as shown on your schematic?

DO214AA is not a diode type number but a diode package style. It is more commonly known as an SMB package.

[srb1954]

Presumably the transformer is 7.5V-0-7.5V? If so, the LM7808 and LM7908 won't do anything, because the input voltage is too low to give 8V out, with decent supply rejection.

Even if the LM7808 and LM7908 somehow barely manage to get enough voltage to regulate their outputs the ripple rejection performance will be terrible. The LM7805/LM7905 won't fare much better in the ripple rejection numbers as they will have barely enough input-output differential to work with.

In my experience the ripple rejection of this type of regulator performs well below spec unless there is at least 3V input-output differential to work with. With such a low AC input voltage you would probably get better overall ripple rejection from this circuit by dispensing with the LM7808 and LM7908 pre-regulators altogether and just running the smoothed DC supplies straight into the the LM7805 and LM7905.

[Gyro]

R16, the Zener and the LED make sense if there is no intermediate connection to 0V, maybe you mistook a crossover for a join (especially if it's a print from an old magazine).

Maybe the circuit also makes use of the intermediate  +/- 8V rails as well as the +/- 5V. Impossible to say without seeing the whole thing.

[FriedMule]

You are all great! Thanks!!
The schematic you can see is one I have drawn from a copy of a copy of the original article.
I have checked and rechecked and I have not forgotten any part or lines, but by looking again, can it maybe look like I have mistaken a smudge for a connection. Therefor have I deleted the ground connection between R16 and D17.
I have looked again and luckily is "DO214AA" very clear, it's not "DD" but on the original schematics are they shown as diodes.
As I can read it, is the PSU a 3VA6 7.5-0-0-7.5 with the two zero's connected.

[tunk]

I think Zero999 and srb1954 are right.
7.5Vac has a peak of 10.6V and about 9.5V after rectification.
The dropout voltage of the L7808 is ~1.6V, so it won't regulate
at all (it's even worse if you add the capacitor ripple). Are you
sure it says 7.5V?

Edit:
It could possibly work if it draws 2x0.1A and that the transformer
voltage at that current is 8.5V: ~10.8V after rectification and
subtracting ~0.5V ripple gives 10.3V, which is sufficent for regulation.

[FriedMule]

I'll try to find a better copy of the article, about 90% are missing, maybe there are some clues in there.
To me does it all look strange and a bit like something an amateur would have drawn, but this person is apparently looked at as someone who knows his stuff. Maybe a bit like if Dave had drawn the schematic, nothing would be on the schematic without a perfect reason:-)

[Mechatrommer]

Presumably the transformer is 7.5V-0-7.5V? If so, the LM7808 and LM7908 won't do anything, because the input voltage is too low to give 8V out, with decent supply rejection.

Even if the LM7808 and LM7908 somehow barely manage to get enough voltage to regulate their outputs the ripple rejection performance will be terrible. The LM7805/LM7905 won't fare much better in the ripple rejection numbers as they will have barely enough input-output differential to work with. In my experience the ripple rejection of this type of regulator performs well below spec unless there is at least 3V input-output differential to work with...

thats why inductors and diodes (D21/D22) are there. as Alex said, its input 20Vpp rectified.

With such a low AC input voltage you would probably get better overall ripple rejection from this circuit by dispensing with the LM7808 and LM7908 pre-regulators altogether and just running the smoothed DC supplies straight into the the LM7805 and LM7905.

LM7x05 will get hot! you dont know how much load they are supplying. D21/D22 can also help reducing heat. i believe that circuit had been tested, including R16/D17 (equalizing load seen by +/-, clamping -ve rail who knows? only the one who put there can tell), i believe some thought had been put into that the designer awared of the load behaviour. if its me i wont remove R16/D17 from pcb, i can "do not populate" later no problem if proved useless or unecessary. ymmv.

[srb1954]

Presumably the transformer is 7.5V-0-7.5V? If so, the LM7808 and LM7908 won't do anything, because the input voltage is too low to give 8V out, with decent supply rejection.

Even if the LM7808 and LM7908 somehow barely manage to get enough voltage to regulate their outputs the ripple rejection performance will be terrible. The LM7805/LM7905 won't fare much better in the ripple rejection numbers as they will have barely enough input-output differential to work with. In my experience the ripple rejection of this type of regulator performs well below spec unless there is at least 3V input-output differential to work with...

thats why inductors and diodes (D21/D22) are there. as Alex said, its input 20Vpp rectified.

With such a low AC input voltage you would probably get better overall ripple rejection from this circuit by dispensing with the LM7808 and LM7908 pre-regulators altogether and just running the smoothed DC supplies straight into the the LM7805 and LM7905.

LM7x05 will get hot! you dont know how much load they are supplying. D21/D22 can also help reducing heat. i believe that circuit had been tested, including R16/D17 (equalizing load seen by +/-, clamping -ve rail who knows? only the one who put there can tell), i believe some thought had been put into that the designer awared of the load behaviour. if its me i wont remove R16/D17 from pcb, i can "do not populate" later no problem if proved useless or unecessary. ymmv.

The inductors are only 100uH so they will provide negligible filtering of any 100Hz/120Hz ripple on the supply rails. With this low value they only provide useful filtering above 10KHz. To get significant filtering of low-frequency ripple with an inductive filter you need an inductor in the range of hundreds of milliHenries.

The diodes themselves provide no useful filtering of noise at low frequencies. In fact they have a negative effect on the ripple rejection because the voltage drop across them leaves less input-output voltage differential across the LM78Lxx/LM79Lxx to perform their function as regulators.

Since the output regulators are 78Lxx/79Lxx the load can't be more than 100mA. At this load current, and with the low input-output voltage differential, power dissipation in the regulators is not going to be a major problem.

[tkamiya]

I think a lot of stuff in this circuit is neither necessary nor beneficial.

I'd simplify it a lot and use more modern LDO type regulators.  They tend to have better noise/ripple suppression.  Also, with 3 terminal regulators, capacitors across input and output to ground needs to be carefully selected.  Otherwise, they may oscillate.  These things likes to oscillate in VHF range.  My favorite way is to use 0.1uF and 10uF in parallel. 

[Mechatrommer]

The inductors are only 100uH so they will provide negligible filtering of any 100Hz/120Hz ripple on the supply rails. With this low value they only provide useful filtering above 10KHz..

ditto! lets do some exercise.. PSRR @ 100Hz is maxed at -80dB, that make like what? 50uV out from 0.5V in ripple? does that really matter? who want to design a 0.5V input ripple (retarded small caps) he must be bad, no? PSRR starting from 10KHz onward is degrading bad, so i guess... thats why the inductors are there... btw i guess the design is old when LDOs are not common thing but i'll still use these jellybeans "HDOs" any days for some non-demanding tasks. ymmv cheers.

[Zero999]

The inductors are only 100uH so they will provide negligible filtering of any 100Hz/120Hz ripple on the supply rails. With this low value they only provide useful filtering above 10KHz..

ditto! lets do some exercise.. PSRR @ 100Hz is maxed at -80dB, that make like what? 50uV out from 0.5V in ripple? does that really matter? who want to design a 0.5V input ripple (retarded small caps) he must be bad, no? PSRR starting from 10KHz onward is degrading bad, so i guess... thats why the inductors are there... btw i guess the design is old when LDOs are not common thing but i'll still use these jellybeans "HDOs" any days for some non-demanding tasks. ymmv cheers.

But that graph is probably when the difference between the input and output voltages is between 3V and 15V, not when it's operating close to drop-out, which will have poor ripple rejection.

[Mechatrommer]

the dropout in the OP circuit is 2V i wouldnt say that is close, that is the max dropout at 1A at room temperature. edit: well before anyone spot it, there's 2 diodes dropout in rectifier so we can continue argue, putting a little bit higher output transformer will be the easy fix, or as anyone suggested except that original design. ymmv cheers.

[FriedMule]

So I have finally found out a bit but if it's plausible, makes sense or is understood correctly by me, is another matter. :-)

1) The two inductors should be to stop hf signal from the unit to go back out in the outlet, to keep it silent in the lab.
2) The 9V should apparently be an indicator for MAX Voltage, normally will the Voltage be 8V
3) the two LED's D15 and D16 is to make an "on board" indicate that there is power on both rails.
4) The LED on the zener diode is to show that there are sufficient power, a sort of "power on" led meant to be mounted visible.

But I could still not get information about the value of the SD12K100, it's a SD12 type but K100 is that a 1000uH like the SD12-102-R : https://www.mouser.dk/ProductDetail/Bussmann-Eaton/SD12-102-R?qs=%2Fha2pyFadug4QZjN0cSDOUFn6uFsp6ssRidD%2F6taA6M%3D

[ataradov]

1. It is dubious whether they will actually do anything to reject the HF noise.  But I guess they don't hurt, so it is fine.
2. Ok, sure.
3. That makes sense. Along with D25/D26. I would say it is an overkill, but whatever.
4. With 8.2 V Zener it will never be on if things are fine. And it will burn out almost immediately when voltage gets higher. I bet there is no connection between the lower part of R16 and the ground. In that case that LED will indeed show that voltage between +8 and -8 V rails is bigger than 8.2 V. And I guess the idea is that if either of the rails fails, then LED will be off.