Electronics > Projects, Designs, and Technical Stuff
Preregulation of a linear bench PSU
<< < (4/12) > >>
not1xor1:

--- Quote from: Kleinstein on December 25, 2018, 06:33:04 pm ---To avoid fast switching up and down, one should have quite some hysteresis and delay between turn on and turn off. Other wise there is a chance the variable loads in CC mode and thus variable voltage can lead to frequent switching of the relay. Some relays don't like this as is can lead to welded contacts. The critical test would the so called file test - so an fast intermittent short like the contact on a file. Just directly a load after the linear regulator to make the pre-regulator oscillate should not be possible. It is more like a load that is already oscillating so that the oscillation get through to the pre-regulator.

So a solution switching with MOSFETs could be a good idea, as this would have less problems with frequent switching.

The circuit from not1xor1  might work, though I would prefer a diode instead of the 2.nd MOSFET. The little extra loss at the lower voltage tap should not be such a big deal.

I had just shown an example of the linear use of 2 taps: https://www.eevblog.com/forum/projects/200v-200300ma-power-supply/msg2060992/#msg2060992

--- End quote ---

To avoid faults due to intermittent short-circuit (file test) one might use an additional circuit to slow down CC/CV switching without affecting the CV/CC one.

One might use a sort of buffered peak detector connected to the output of the current control opamp, with the detector output connected via a 3rd diode (i.e. besides V control and C control) to the base/gate of the power device.

When the current control opamp switches off, the peak detector keeps the output voltage close to the previous value for a bit longer making it increase slowly to the set output voltage as the hold capacitor discharges.

In simulations the spikes on short recovery were greatly reduced or completely cancelled depending on the capacitor value/discharge time.

Unfortunately the discharge time needed to avoid any spike even when the CV is set to fraction of volts, makes the recovery time last several seconds when conversely the output voltage is set in the order of several tenths of volts.

In any case, I might be wrong, but while I see the advantage of having a fast CV/CC switch, I do not see any advantage from a fast CC/CV switch.

The disadvantage is that you have to use 2 additional opamps.
Although a complementary 2 BJTs buffer + electrolitic capacitor also make a coarse but yet fairly effective circuit (at least in simulations).
not1xor1:

--- Quote from: Atom on December 25, 2018, 06:33:58 pm ---
--- Quote from: not1xor1 on December 25, 2018, 08:31:57 am ---I'm in a hurry at the moment so have not yet checked carefully your circuit (I find some symbols quite odd).
But here is a proof of concept I simulated with LTspice a while ago.
That would work even with just a TL431 as a comparator to switch the rails, but it is for negative rail regulation. You have to reverse it and use P-MOS (with usually higher Rds-on) for positive rail...

--- End quote ---

So i tried it and it worked at first wit no load then i don't know what really happend but i blew up the mosfets ..you mentioned that this circuit was for a negative and that i have to swap the mosfet to a p channel.

what is the problem with the circuit and why isn't suitable as it is ? is it for the body diodes because i really have no idea of what else could it be...please explain |O

--- End quote ---

Did you noticed that the center tap mosfet is reversed so that the body diode prevents conduction when it is off?
not1xor1:

--- Quote from: Atom on December 25, 2018, 07:15:59 pm ---
so the problem with the circuit is that there isn't enough dead time and all the 2 mosfet could be conducting ad the same time shorting out COM with AC2 (and maybe themselves since there is a lot of current passing in a short)?

 is my understanding correct?

--- End quote ---

I made a more complete simulation with a 1Hz 0-30V triangle wave to simulate the variation of a PSU output voltage.
I found that even with a 1ms deadtime there may be cross conduction if the switch doesn't occur on zero-crossing.

So to work properly the circuit would need a zero-cross detector, a histeresis comparator and a flip-flop to be set/reset when both AC is 0V and voltage is over/below the given threshold.
duak:
The first ranging power supply I encountered was an hp 6002A.  It used a state machine to select the best combination of transformer secondary taps for the output voltage and current.  The design was written up in the June 1977 hp Journal: http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1977-06.pdf

Figure 3 on page 4 shows the basic transformer and rectifier circuit.  The takeaway from this is that for the low voltage range, a standard bridge rectifier provides all the current and when a higher range is needed, a higher voltage tap is switched in and the LV rectifier doesn't conduct.  In this supply, thyristors were used for tap switching but I don't see any reason that a MOSFET can't be used.  The hp supply is complicated because it has four ranges and a 200 W maximum power limit.  I have one that I repaired but I don't use it much so I can't say if its complexity is worth the capability.  I can't even say how clean its output is compared to non-ranging or SCR supplies.

Cheers & Best Wishes,
Atom:

--- Quote from: not1xor1 on December 26, 2018, 04:18:05 pm ---I made a more complete simulation with a 1Hz 0-30V triangle wave to simulate the variation of a PSU output voltage.
I found that even with a 1ms deadtime there may be cross conduction if the switch doesn't occur on zero-crossing.

So to work properly the circuit would need a zero-cross detector, a histeresis comparator and a flip-flop to be set/reset when both AC is 0V and voltage is over/below the given threshold.

--- End quote ---

So you are telling me that switching should be done at the zero crossing point...and i would need some external circuitry to get everythig to work ... that adds substantial complexity to the circuit..i'll se what i can do.

stupid idea but if i can induce a proper delay and drive the mosfet gates from a stabilized dc supply(without any ripple) those problems would go away right? or i'm still missing something

since the trasformer is a toroidal one i can add an aux winding pretty easily for driving mosfet gates. Could you share the LTspice file i'm not familiar with it but since it's what everyone uses it's time to learn something new.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod