Transformer tap switching between series and parallel

[nemail2]

Hi all

would this work for transformer tap switching between serial and parallel operation of the (identical) secondary windings?
The transformer (https://www.mouser.com/datasheet/2/410/VPT18_5560-781803.pdf) is designed to be used either in parallel or serial connection, not independently.

Should I do it differently? I plan to use this in one of my lab power supply designs to avoid having to burn all the voltage on the series pass transistor.
Microcontroller and stuff would be on a seperate transformer so I don't run into any issues there.

Thanks!

[Cliff Matthews]

Looks OK 2 me..

[drvtech]

I'd be tempted to put some protection (a fuse) in line with secondaries to protect against the unlikely situation where the upper contact switches over but the lower contact remains closed. However the mains input fuse should blow so perhaps I'm being a bit 'belt and braces'

[nemail2]

I'd be tempted to put some protection (a fuse) in line with secondaries to protect against the unlikely situation where the upper contact switches over but the lower contact remains closed. However the mains input fuse should blow so perhaps I'm being a bit 'belt and braces'

I think I'd like that as well. I want to do secondary side fuses anyway so they could be very well coming before the relay.

Any other protection measures? I figured the relay would have to be a quite robust one, as I'm planning to switch between 9/18VAC @ 4A, right? I saw that the AC ratings for relays are often lower than for DC.

[bdunham7]

Definitely fuses on each secondary, but you probably want to make sure and use one DPDT relay so that the switchovers are guaranteed to happen simultaneously absent some strange failure--in which case the fuses will step in.  That really should not require a very large relay, switching 4 amps at under 30 volts is fairly easy, AC or DC.

[nemail2]

Definitely fuses on each secondary, but you probably want to make sure and use one DPDT relay so that the switchovers are guaranteed to happen simultaneously absent some strange failure--in which case the fuses will step in.  That really should not require a very large relay, switching 4 amps at under 30 volts is fairly easy, AC or DC.

Oh sorry it wasn't obvious from the schematic, but I always was planning to use one single DPDT relay, in fact I have one in the schematic as well, only the poles are separated in the symbol for schematic layout easyness.
I was planning to use this G2R-2 relay: https://www.mouser.at/ProductDetail/Omron-Electronics/G2R-2-DC12?qs=%2Fha2pyFaduhiTr278rQ5adtGSq1z2sWxdbawesLWMjhDC95nS77V7g%3D%3D
No objections regarding the rated current? There's 20k µF of bulk capacitance after the rectifier (which is an ideal diode bridge rectifier using the LT4320) so the peak current from the transformer, especially when the windings are in parallel, is going to be pretty high, under full load, until the 20k µF caps are fully recharged. Should I rate the relay for that peak current? The datasheet of the G2R-2 doesn't really give me any figures on any peak current values...

[IanB]

I'm not too familiar with relays, but I see in your diagram you have the S1 winding across the two output poles of REL2. In a typical construction those relay contacts are quite close together and I'm wondering if there is any danger of a switching arc establishing between them and continuing to burn up the contacts? Maybe someone more familiar with relay usage and specifications knows if this is a real danger or not?

[bdunham7]

No objections regarding the rated current? There's 20k µF of bulk capacitance after the rectifier (which is an ideal diode bridge rectifier using the LT4320) so the peak current from the transformer, especially when the windings are in parallel, is going to be pretty high, under full load, until the 20k µF caps are fully recharged. Should I rate the relay for that peak current? The datasheet of the G2R-2 doesn't really give me any figures on any peak current values...

That is a lot of capacitance and your peak current may be 50A or so.  Also, the relay you have chosen is just marginal for a 4A load and certainly not going to do well with the inrush surge which will occur just as the contacts are settling down.  I would suggest going up a bit on relay capacity and adding an NTC inrush limiter.

https://www.mouser.com/ProductDetail/EPCOS-TDK/B57237S0259M000?qs=AKDv8POSxR3rj6%252Bw0RTy3A%3D%3D

[tautech]

Be sure you get the secondary's for series or parallel in phase.....don't ask me how I know ! 

[Brumby]

Be sure you get the secondary's for series or parallel in phase.....don't ask me how I know ! 

Very much yes.  I've never found out the hard way because I am - and always have been - paranoid about getting it wrong.  Fortunately, the test (for two secondaries of the same voltage) is very simple:

1. Connect one wire from one secondary to one wire from the other - then measure the voltage between the other two wires.
2(a)  If the voltage is near zero, then you can join those two wires and you will have a parallel configuration.
2(b)  If the voltage is double the voltage of a single secondary, then you have a series configuration ready to roll.  (Don't join the two free wires together or you will replicate tautech's learning experience!)

[jmelson]

You must make sure the relays are guaranteed break before make with a decent gap between the open contacts.

Jon

[xavier60]

I used an OMRON MK2P-I 10A Relay  for mine, it's rather loud though.
The break before make action, although necessary caused some complication with the voltage sensing for relay control.
Changing from paralell to series for increasing output voltage sometimes caused an AC peak to be missed causing a dip in the DC voltage
which would then cause a momentary switch back to paralell.
I fixed this by reducing the threshold voltage for a short time after switching from paralell to series.

 https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873

[james_s]

You must make sure the relays are guaranteed break before make with a decent gap between the open contacts.

Jon

Have you ever seen a relay that wasn't make before break? This got me thinking about it and I don't think I ever have. The contacts are normally mounted on the same armature with the NC contacts up above the NO contacts so it's physically impossible for both to be touching at once.

[golden_labels]

If it comes to decreasing chances of that unlikely event of one of the contacts staying connected while the other makes a new connection: what about this layout?

[xavier60]

It's still possible to short circuit SEC1.

[golden_labels]

Hence “decreasing chances”, not “eliminating chances”.

But if the bottom switch is controlled separately, it can be disconnected for the time the upper one is switched. That would, however, require two independent relays. I was sticking to the originally available parts.

[Zero999]

Adding more parts reduces the reliability.

Is the extra current capability/lower impedance given by having both secondary windings in parallel really needed? If not, just connect it as a centre tapped transformer and use one switch to select between the tap and both in series. It has the advantage of the switch not having to pass the huge surge current taken by the filter capacitors. C3 only needs to keep the circuit powered for the few ms required for the switch to change state. It can be much smaller than the main filter capacitors C1 & C2.


EDIT: Another advantage is when set to the centre tap, there's only one diode drop, rather than two.

[nemail2]

Hi

thanks for all your inputs.

That is a lot of capacitance and your peak current may be 50A or so.  Also, the relay you have chosen is just marginal for a 4A load and certainly not going to do well with the inrush surge which will occur just as the contacts are settling down.  I would suggest going up a bit on relay capacity and adding an NTC inrush limiter.

https://www.mouser.com/ProductDetail/EPCOS-TDK/B57237S0259M000?qs=AKDv8POSxR3rj6%252Bw0RTy3A%3D%3D

Regarding the relay: would I rate its contact current rating to that what the NTC inrush limiter would allow to pass? Also: wouldn't there be quite some voltage loss and heat buildup in within the NTC inrush limiter during normal operation? Max. current is 4A.

Be sure you get the secondary's for series or parallel in phase.....don't ask me how I know ! 

Haha well the only way to mess up is to not pay attention to the dots in the transformer's datasheet, right?

You must make sure the relays are guaranteed break before make with a decent gap between the open contacts.

Well i hope the relay will do its job properly. I guess sizing it appropriately will help. For the worst case, I think the fuses should do their job?

I used an OMRON MK2P-I 10A Relay  for mine, it's rather loud though.
The break before make action, although necessary caused some complication with the voltage sensing for relay control.
Changing from paralell to series for increasing output voltage sometimes caused an AC peak to be missed causing a dip in the DC voltage
which would then cause a momentary switch back to paralell.
I fixed this by reducing the threshold voltage for a short time after switching from paralell to series.

 https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873

now that is a big ass relay, and a nice PSU, by the way. But I'm not seeing any relay action in the schematics, or am I blind? :-) I would control the tap switching via the MCU and measuring with the ADC so I'd simply insert a delay after switching before any new decisions regarding switching the taps are being made.

Adding more parts reduces the reliability.

Is the extra current capability/lower impedance given by having both secondary windings in parallel really needed? If not, just connect it as a centre tapped transformer and use one switch to select between the tap and both in series. It has the advantage of the switch not having to pass the huge surge current taken by the filter capacitors. C3 only needs to keep the circuit powered for the few ms required for the switch to change state. It can be much smaller than the main filter capacitors C1 & C2.


EDIT: Another advantage is when set to the centre tap, there's only one diode drop, rather than two.

The datasheet of the transformer says that the windings are designed to be used in parallel or in series, not independently. I guess that's a show stopper. Also, I'm using a LT4320 ideal bridge rectifier controller with MOSFETs, so there isn't any diode drop in my design anyway...

[Zero999]

The datasheet of the transformer says that the windings are designed to be used in parallel or in series, not independently. I guess that's a show stopper.

There's no reason why you can't connect the transformer, as I suggested. The windings are not being used independantly, but are connected in series. When the switch is connected to the centre tap, the current is drawn alternately from each winding. Lots of power supplies use this configuration to produce positive and negative voltages. The subtle difference here is the 0V reference is taken from the negative side, rather than the mid point and a switch selects between the two output voltages.

The only advantage of connecting the windings in parallel is it doubles the current rating. There's no point, if you need the same maximum current, at both voltages.

[Wolfram]

Here's another option, as diodes are often cheaper than relays. You need twice as many as the original solution, but they only need to be rated for half the current. There are also no issues with contact sequencing.

[nemail2]

There's no reason why you can't connect the transformer, as I suggested. The windings are not being used independantly, but are connected in series. When the switch is connected to the centre tap, the current is drawn alternately from each winding. Lots of power supplies use this configuration to produce positive and negative voltages. The subtle difference here is the 0V reference is taken from the negative side, rather than the mid point and a switch selects between the two output voltages.

The only advantage of connecting the windings in parallel is it doubles the current rating. There's no point, if you need the same maximum current, at both voltages.

is there any term I can search for which explains that center tap stuff which I'm still not able to wrap my head around entirely? Would be appreciated :-) I don't quite understand how that works, what you have suggested...

Here's another option, as diodes are often cheaper than relays. You need twice as many as the original solution, but they only need to be rated for half the current. There are also no issues with contact sequencing.

Well, this is how my rectification looks, I'm not sure your solution would blend in that good or would it? At least I guess I'd need a second LT4320 and they are freakin expensive, unfortunately. But there's nothing better than a 4A delivering rectifier which has Tambient :-D

[jmelson]

You must make sure the relays are guaranteed break before make with a decent gap between the open contacts.

Jon

Have you ever seen a relay that wasn't make before break? This got me thinking about it and I don't think I ever have. The contacts are normally mounted on the same armature with the NC contacts up above the NO contacts so it's physically impossible for both to be touching at once.

Yes, certainly.  But, any relay with totally fixed contacts should be break before make.  Some telephone-style relays have long springy arms, and the "fixed" contacts are more springy than the "moving" contact.  They can be made to be make before break for audio purposes, or just get out of adjustment over time.

But, a lot of signal-level and low power relays have really small contact gaps, and while truly break before make, thay may be prone to arcing, especially where inductance is present.

Jon

[james_s]

I would assume that one would not be trying to use a small signal relay to switch several amps. Any suitable power relay ought to get the job done. If it does fail, the fuse on the primary will blow.

[xavier60]

I don't think that there are significantly different ways to wire the relay to the secondary windings.
With my PSU, D2 supplies a 12V regulator for powering the fan and relay coil.

[Zero999]

There's no reason why you can't connect the transformer, as I suggested. The windings are not being used independantly, but are connected in series. When the switch is connected to the centre tap, the current is drawn alternately from each winding. Lots of power supplies use this configuration to produce positive and negative voltages. The subtle difference here is the 0V reference is taken from the negative side, rather than the mid point and a switch selects between the two output voltages.

The only advantage of connecting the windings in parallel is it doubles the current rating. There's no point, if you need the same maximum current, at both voltages.

is there any term I can search for which explains that center tap stuff which I'm still not able to wrap my head around entirely? Would be appreciated :-) I don't quite understand how that works, what you have suggested...

It's normally use as in the links below:
https://www.circuitlib.com/index.php/schematics/product/75-bipolar-power-supply/category_pathway-26
https://sep.yimg.com/ay/glass-ware/tr-ps-3-5.gif

By the way, please keep it to one thread. I thought some of my posts went missing, until I realised I was looking at the other one!

Here's another option, as diodes are often cheaper than relays. You need twice as many as the original solution, but they only need to be rated for half the current. There are also no issues with contact sequencing.

That will work, but is overly complex. Why not just add another diode? Now the switch doesn't have to carry the huge current surge and only passes current in one direction, so it can be a transistor. The additional diode can be, Schottky, or a MOSFET and ideal diode controller IC, if low voltage loss is a requirement.

[Wolfram]

Here's another option, as diodes are often cheaper than relays. You need twice as many as the original solution, but they only need to be rated for half the current. There are also no issues with contact sequencing.

That will work, but is overly complex. Why not just add another diode? Now the switch doesn't have to carry the huge current surge and only passes current in one direction, so it can be a transistor. The additional diode can be, Schottky, or a MOSFET and ideal diode controller IC, if low voltage loss is a requirement.

I'm used to working at voltage and power levels where I have to parallel devices to get the required ratings, and voltage levels where you don't want your relay to switch DC currents, so my solution might be a bit biased by that. Note that my proposal uses twice the number of rectifiers, but they only need half the current rating, which might in some cases be cheaper but probably not here.

Your latest solution is the best one proposed so far in the thread in my opinion, it nicely solves the "relay between capacitors at different voltage" issue that the previous one had, and it also works directly with the MOSFET based bridge rectifier as mentioned in the other thread.

[nemail2]

It's normally use as in the links below:

Thanks for the links! Unfortunately I still do not understand how this works, when there is which voltage and why - also: are the secondary windings permanently connected in series in the schematic with the center tap buggering off to between C1/C2?
Where would I have the 20000uF? I chose to have that much capacitance so i don't have pretty much any ripple even at full load so I'd like to keep those 20k uF under any operating condition...

Also: Wouldn't lead mixing the ideal bridge rectifier with the fifth diode to issues? And how would I insert another controller (besides the fact that it is very expensive) with only one diode instead of another four diodes?

I'd like to keep the ideal bridge rectifier with MOSFETs so I don't have to worry about cooling of the rectifier diodes. I once was using diodes in this design and they did heat up to more than 100 °C.

By the way, please keep it to one thread. I thought some of my posts went missing, until I realised I was looking at the other one!

sorry, usually I don't do that, I felt that the second topic was related but still different from this topic so I thought creating a second one would be better than going haywire in this one.

[Zero999]

It's normally use as in the links below:

Thanks for the links! Unfortunately I still do not understand how this works, when there is which voltage and why - also: are the secondary windings permanently connected in series in the schematic with the center tap buggering off to between C1/C2?

Yes, the idea is to connect the secondary windings together, forming a centre tap, which goes between the two capacitors.

Where would I have the 20000uF? I chose to have that much capacitance so i don't have pretty much any ripple even at full load so I'd like to keep those 20k uF under any operating condition...

That's a large capacitance. How much current does it use? A centre tapped configuration would require two capacitors in series, which can be half the voltage rating.

Also: Wouldn't lead mixing the ideal bridge rectifier with the fifth diode to issues? And how would I insert another controller (besides the fact that it is very expensive) with only one diode instead of another four diodes?

No, it wouldn't create any issues. The ideal diode controller circuit, just goes where the fifth diode is.

I'd like to keep the ideal bridge rectifier with MOSFETs so I don't have to worry about cooling of the rectifier diodes. I once was using diodes in this design and they did heat up to more than 100 °C.

100 °C might be fine. The internal junction temperature of most diodes is something like 150 °C or more. Did you consider using Schottky diodes, which would have half the voltage drop?

[nemail2]

That's a large capacitance. How much current does it use? A centre tapped configuration would require two capacitors in series, which can be half the voltage rating.

4A max. current and I wanted to eliminate ripple to an absolute minimum.

No, it wouldn't create any issues. The ideal diode controller circuit, just goes where the fifth diode is.

but I'd then have to have two of them, wouldn't I? also, the diode controller circuit has 4 MOSFETs, replacing 4 diodes, what would I do with the other thee and which of the four would I use?
Sorry for the stupid questions...

100 °C might be fine. The internal junction temperature of most diodes is something like 150 °C or more. Did you consider using Schottky diodes, which would have half the voltage drop?

Originally I was printing the enclosure of this PSU with my 3D printer (PETG), that stuff gets all wobbly when heated up to about 80°C. Currently I'm using a PS (Polystyrol) case which I'm not too confident heating up too much either.
But wait, there's more reasons: The PSU is quite of a precision device (at least in self-made amateur measures). It allows to set and hold voltages within 1mV and it usually is bang-on over 99% of the whole voltage range. You don't want heat and temperature drift in your case if you're doing stuff like this (afaik).
Also, I already tried with schottkys, their Vf rises with current and all of them heated up to around 100°C. In retrospective, I would have had to put in some massive block all in one full bridge rectifier and put some big-ass heatsink on it (+ airflow) to keep the temperature down. But I went with the ideal diode controller and never looked back (Tcase = Tambient makes me happy :-))

[Zero999]

That's a large capacitance. How much current does it use? A centre tapped configuration would require two capacitors in series, which can be half the voltage rating.

4A max. current and I wanted to eliminate ripple to an absolute minimum.

I thought it was powering a linear regulator? The minimum voltage due to the ripple just needs to be high enough to avoid the drop-out region. Huge capacitors result in a massive inrush current, which can cause other problems.

No, it wouldn't create any issues. The ideal diode controller circuit, just goes where the fifth diode is.

but I'd then have to have two of them, wouldn't I? also, the diode controller circuit has 4 MOSFETs, replacing 4 diodes, what would I do with the other thee and which of the four would I use?
Sorry for the stupid questions...

Then use an ideal diode controller designed for one MOSFET, for the fifth diode.

100 °C might be fine. The internal junction temperature of most diodes is something like 150 °C or more. Did you consider using Schottky diodes, which would have half the voltage drop?

Originally I was printing the enclosure of this PSU with my 3D printer (PETG), that stuff gets all wobbly when heated up to about 80°C. Currently I'm using a PS (Polystyrol) case which I'm not too confident heating up too much either.
But wait, there's more reasons: The PSU is quite of a precision device (at least in self-made amateur measures). It allows to set and hold voltages within 1mV and it usually is bang-on over 99% of the whole voltage range. You don't want heat and temperature drift in your case if you're doing stuff like this (afaik).
Also, I already tried with schottkys, their Vf rises with current and all of them heated up to around 100°C. In retrospective, I would have had to put in some massive block all in one full bridge rectifier and put some big-ass heatsink on it (+ airflow) to keep the temperature down. But I went with the ideal diode controller and never looked back (Tcase = Tambient makes me happy :-))

As long as the case isn't in direct contact with the diodes, I don't see how that's an issue.

It sounds like the Shottky didoes didn't have a high enough current rating. The reason why I'm questioning the need for a low loss rectifier is, it's driving a linear regulator and all you're doing is moving the heat there. The main advantage is it will work with a lower transformer voltage, but if it's already more than high enough, it won't make any difference.

[nemail2]

I thought it was powering a linear regulator? The minimum voltage due to the ripple just needs to be high enough to avoid the drop-out region. Huge capacitors result in a massive inrush current, which can cause other problems.

I'm having a bit of a confusion/epiphany right now - does that mean, that the linear regulator would be able to "regulate out" that 50 Hz ripple as long as the input voltage doesn't go below the dropout voltage? Thinking about it, it seems very logical to me for the regulator (two BD139/TIP3055 pairs in this case) to be at least as fast as 50 Hz. That would mean, I was doing it wrong all the time, lol.
I have an equation here:
Ripple Voltage = Max Current / Capacitance [F]

If my transformer voltage is 18VAC * 1,414 (in fact it is 27V even under load, measured, don't know why), even 8V ripple (is that peak to peak?) should work ok, which would be only 5000uF of capacitance needed instead of 20000uF. Is that possible?

Then use an ideal diode controller designed for one MOSFET, for the fifth diode.

Didn't know such a thing exists, will have to search. Thanks for the hint (you don't know one, by accident?).

As long as the case isn't in direct contact with the diodes, I don't see how that's an issue.

Well like I said, general heatup within the case, temperature drift of all the precision devices, and so on made my concerns... Also, temperature was still rising, the 100°C i had reached only after a couple of minutes. I just didn't want to deal with that kind of heat on the diodes anymore.

It sounds like the Shottky didoes didn't have a high enough current rating. The reason why I'm questioning the need for a low loss rectifier is, it's driving a linear regulator and all you're doing is moving the heat there. The main advantage is it will work with a lower transformer voltage, but if it's already more than high enough, it won't make any difference.

I was using these and I just realized they weren't Schottkys, sorry for that misinformation: https://www.mouser.at/datasheet/2/427/vs-16edu06-m3-1769326.pdf
However I remember searching for diodes with lower Vf but if I recall correctly, they were all around at least 0.6V or so, when it came to currents around 4A.

But you are of course very right about "only moving the heat to the linear regulator", I wasn't looking at this from that point of view. On the upside is that I have a large heatsink plus a fan on the dual darlington pair so I have no issues of getting rid of the heat there. Also, I was thinking about lowering the transformer voltage (18VAC transformer output vs. 16VDC PSU output, which is not good) to reduce dissipation in the first place or at least raise the output voltage of the PSU, as ~16VAC transformers don't seem to be common, at least not with that kind of amperage rating. That lowers my tolerance on voltage drop on the rectifier in any case.

[Zero999]

I thought it was powering a linear regulator? The minimum voltage due to the ripple just needs to be high enough to avoid the drop-out region. Huge capacitors result in a massive inrush current, which can cause other problems.

I'm having a bit of a confusion/epiphany right now - does that mean, that the linear regulator would be able to "regulate out" that 50 Hz ripple as long as the input voltage doesn't go below the dropout voltage? Thinking about it, it seems very logical to me for the regulator (two BD139/TIP3055 pairs in this case) to be at least as fast as 50 Hz. That would mean, I was doing it wrong all the time, lol.
I have an equation here:
Ripple Voltage = Max Current / Capacitance [F]

If my transformer voltage is 18VAC * 1,414 (in fact it is 27V even under load, measured, don't know why), even 8V ripple (is that peak to peak?) should work ok, which would be only 5000uF of capacitance needed instead of 20000uF. Is that possible?

Yes, that's the idea. A linear regulator will output the same voltage, as long as the input is high enough.

Then use an ideal diode controller designed for one MOSFET, for the fifth diode.

Didn't know such a thing exists, will have to search. Thanks for the hint (you don't know one, by accident?).

It's the most common form of ideal diode circuit. I don't have any part numbers in my head, but it's fairly easy to find.

As long as the case isn't in direct contact with the diodes, I don't see how that's an issue.

Well like I said, general heatup within the case, temperature drift of all the precision devices, and so on made my concerns... Also, temperature was still rising, the 100°C i had reached only after a couple of minutes. I just didn't want to deal with that kind of heat on the diodes anymore.

It sounds like the Shottky didoes didn't have a high enough current rating. The reason why I'm questioning the need for a low loss rectifier is, it's driving a linear regulator and all you're doing is moving the heat there. The main advantage is it will work with a lower transformer voltage, but if it's already more than high enough, it won't make any difference.

I was using these and I just realized they weren't Schottkys, sorry for that misinformation: https://www.mouser.at/datasheet/2/427/vs-16edu06-m3-1769326.pdf
However I remember searching for diodes with lower Vf but if I recall correctly, they were all around at least 0.6V or so, when it came to currents around 4A.

Schottky diodes tend to have lower voltage ratings, than silicon diodes. Generally those with a lower voltage drop, also have a lower voltage rating.

But you are of course very right about "only moving the heat to the linear regulator", I wasn't looking at this from that point of view. On the upside is that I have a large heatsink plus a fan on the dual darlington pair so I have no issues of getting rid of the heat there. Also, I was thinking about lowering the transformer voltage (18VAC transformer output vs. 16VDC PSU output, which is not good) to reduce dissipation in the first place or at least raise the output voltage of the PSU, as ~16VAC transformers don't seem to be common, at least not with that kind of amperage rating. That lowers my tolerance on voltage drop on the rectifier in any case.

Yes, lowering the transformer voltage will reduce the power dissipation, just as long as there's still sufficient voltage headroom.

[nemail2]

Hi

sorry for not coming back earlier, I had non-hobby stuff to do unfortunately, besides making a first prototype of this project.
As expected, inrush current of even "only" 10.000uF instead of 20.000uF capacitance like planned earlier doesn't make the tap switching relay sound very good when switching under full load. Also I'm getting quite some ripple so I'd like to add even more capacitance. The ripple however may as well come from the probably too low voltage transformer but I guess it is a mix of both causes.

Anyway, I came back to the idea of Zero999 and I wonder whether this (see 1st screenshot) will work as intended. I have chosen a MOSFET like suggested, for switching the taps. I don't need the current capability of two windings, but only the lower/higher voltage to lower the dissipation on the pass transistors.

Specific questions for this design are:
- first of all I don't really understand how this works and where the voltage is at and where the current is flowing in which case. understanding would be a nice bonus
- will I have 10.000uF or 20.000uF of capacitance?
- are the fuses placed correctly?
- will the LTC4357 (ideal diode controller) work as intended if connected exactly like this?
- should I use a RC hold up circuit like mentioned in the LTC4357 datasheet (see 2nd screenshot) and if so, which values would be to choose?
- can I really use a MOSFET for switching the taps (T1 in the screenshot)?

Thanks, your help is very appreciated!

[nemail2]

anyone? pleeeeaazzee
I'll just build it up and try it out if nobody answers my questions but then I'll still have no idea whether the fuses are in the correct places...

[Circlotron]

Another method is simply parallel the two transformer secondary windings and feed it to this circuit which will function as a normal bridge rectifier for low voltage output and as a voltage doubler for high voltage output. Only needs a single N/O relay contact, so no problem with syncing multiple contacts. This circuit has in the past been widely used for switch mode power supplies that had to run on 120 or 240V.

[Zero999]

Come on, be pateint. Most people here have other things to do that post. I don't have time to answer all the questions, at the moment. The capacitance is 10mF in both modes and T1 must be a P-channel device, with a potential divider on the gate, to ensure the gate-source voltage rating isn't exceeded.

A voltage doubler is another option, but you're back to using a relay again to switch between voltages and much larger capacitors are required, to achieve the same ripple.

[nemail2]

Another method is simply parallel the two transformer secondary windings and feed it to this circuit which will function as a normal bridge rectifier for low voltage output and as a voltage doubler for high voltage output. Only needs a single N/O relay contact, so no problem with syncing multiple contacts. This circuit has in the past been widely used for switch mode power supplies that had to run on 120 or 240V.

ty, didn't know this! however i must admit that no relay and less capacitance needed seems more attractive to me at the moment

Come on, be pateint. Most people here have other things to do that post. I don't have time to answer all the questions, at the moment. The capacitance is 10mF in both modes and T1 must be a P-channel device, with a potential divider on the gate, to ensure the gate-source voltage rating isn't exceeded.

A voltage doubler is another option, but you're back to using a relay again to switch between voltages and much larger capacitors are required, to achieve the same ripple.

sooorray, i knew sh*t could hit the fan if I start pushing but i juust couldn't resist
i'm on vacation right now so i got time to do stuff

i really appreciate anyone putting time into my very newbie questions, i do not take this for granted.

i'm looking forward for you having some more time to answer the other questions, if you'd like and ty in the meantime for the already answered ones.

unfortunately another question came up with your last answer xD
depletion mode or enhancment mode p-fet? also how (potential divider??) will i be able to control the p-fet from my microcontroller which has 3.3V GPIOs?

thanks again, yall!

[fcb]

Another method is simply parallel the two transformer secondary windings and feed it to this circuit which will function as a normal bridge rectifier for low voltage output and as a voltage doubler for high voltage output. Only needs a single N/O relay contact, so no problem with syncing multiple contacts. This circuit has in the past been widely used for switch mode power supplies that had to run on 120 or 240V.

This is a great circuit - but take time to understand how it works.

With the switch open, the circuit behaves 'normally' - you'll get AC*1.414 less diode drops.

With the switch closed, you'll get approx DOUBLE the voltage out.  You couldn't use this circuit as the rectifier on a 'common' mains SMPSU as you'd end up with (in UK) either the 'normal' bus voltage (340VDC) or 680VDC(!) with the switch closed.

[nemail2]

This is a great circuit - but take time to understand how it works.

With the switch open, the circuit behaves 'normally' - you'll get AC*1.414 less diode drops.

With the switch closed, you'll get approx DOUBLE the voltage out.  You couldn't use this circuit as the rectifier on a 'common' mains SMPSU as you'd end up with (in UK) either the 'normal' bus voltage (340VDC) or 680VDC(!) with the switch closed.

from what I have learned, I'd rather use the version with a p-fet instead of a relay and less capacitance for the same amount of ripple. nonetheless, I have learned something new and I am very grateful for that.

I have now changed some things in regards to my previous proposal:
- P-channel FET RSH070P05GZETB instead of N-channel FET for tap switching
- BC847 NPN transistor for controlling the P-channel FET from the microcontroller
- added a voltage divider for the P-channel FETs gate

V(gs) is +-20V max for this P-FET so with my voltage from the 12/24VAC rated transformer being around 33V rectified, I guess I should be good after halfing that...?
The RDS(on) of this FET is max. 15mOhms @ -10V(gs) so at 5A load it should be dissipating around 375mW which shouldn't really bother a TO-263.

Please let me know whether this would work as intended or if I have forgotten something...

Thanks!

[Zero999]

R3 was in the wrong place.

[nemail2]

R3 was in the wrong place.

Thanks, seems obvious, now that you pointed it out, stupid me 
My version would blow the P-FET by exceeding the V(gs) voltage.

Otherwhise this should work or are there any objections? Have you got time to look over it whether I missed something?

- placement of the fuses
- the way I connected the LTC4357? datasheet: https://www.mouser.com/datasheet/2/609/4357fd-1269235.pdf

[tautech]

- placement of the fuses

Makes more sense to be on pins 5 & 8 on TR1.

[nemail2]

- placement of the fuses

Makes more sense to be on pins 5 & 8 on TR1.

hmm seems legit to me, thank you!

edit: this (screenshot attached) is how it currently looks like, in case anyone is interested. everything will be open source, like my previous project (https://github.com/mamama1/LabPSU_CA1).
It works a treat, besides the ripple due to the too low voltage transformer and the tap switching relay not sounding good (that's why I'm trying to implement the currently discussed solution).

[nemail2]

hmm i guess I'll just grab the money, buy the parts (LTC4357 and P-channel FET) and try it out (with R3 in the right place) on a prototyping board or something...
will report back...

[nemail2]

Hi

ok so I have built this up using ordinary diodes for starters (see the piece of art in the attachments) and this seems to do it (you wouldn't have expected otherwhise, i guess), so thank you very much for that advice, I will most probably implement this into my circuit for tap switching :-)
Now the only thing I have left to test is the same stuff but using the ideal mosfet rectifiers instead of diodes. I have ordered one LTC4357 and I have a LT4320 in stock to play with, as well as plenty of N-FETs so I will build this up next and report back.

In the meanwhile I'd highly appreciate any hints or advices what I may have overseen or wired wrong (or not ideally) in the last corrected schematics which Zero999 posted.

Thank you all again for your support and help!

[nemail2]

Hi

i did build this now like in the attached schematics using the ideal diode stuff (but using FQP50N06L N-FETs and one SQD50P06-15L P-FET) and I'm getting weird voltages at the output, see scope shots SDS00001 and SDS00002. Using normal diodes in the very same circuit produces the correct results -> 13V and 26V depending on whether one or two windings are being used.

With the ideal diode controller stuff, I get around 9V mean for one tap and around 24V mean for both taps. Currently there are only 2x100 uF caps for filtering but also no load at all. In the version of the circuit where I used normal diodes, I had a 1k resistor across the output and 100uF caps as well, just like here.

Can this be due to the breadboard and/or the partially long jumper wires or is there probably some other wiring issue?

Please help, thanks!

edit:
I have double-checked the datasheets of the ideal diode controllers and the wiring as well and I couldn't find any issue. So I reconnected the piece of art from the previous post again which uses ordinary rectifier diodes and measured the output with the scope, see SDS00005 and SDS00007. I measured in DC mode so we can see the absolute output voltage as well. low pk-pk output ripple in comparison to the version with the ideal diode controllers where the voltage seems to go almost to zero as if there was absolutely no capacitance??

I'm wondering whether this is a design failure on my side, maybe this solution for tap switching is not compatible with ideal diode controllers, especially with the LT4320, because of the way it works?

I'm clueless, ideas anyone?

I would really prefer to keep using the ideal diode controllers if possible, so I don't have to worry about heat dissipation and space for a bigger rectifier + heatsink. I've got a big ass heatsink + fan on the power transistors anyway so no worries dissipating the heat there instead.

[fcb]

How does it behave if you increase C53 to something like 1000uF?

[nemail2]

will try that later today. the datasheet explicitely mentions 1 to 10uF ceramic or tantal to be mandatory as close to the output pins as possible iirc. i'll try 1000uF in parallel as well as 1000uF instead of C53.

also i will try the ideal bridge rectifier alone without all the other stuff to see whether that works on its own on the breadboard. on PCB it certainly works, i'm using it in my PSU design successfully for years now...

thanks in the meanwhile, i'll report back.

[nemail2]

How does it behave if you increase C53 to something like 1000uF?

ok i tried now a few things:
- 1000uF additionally to or instead of C53 flattens out the ripple for the output voltage when the two windings are in series, but not when only one of the windings is used, obviously, as those 1000uF are disconnected, as soon as T1 opens.
- 1000uF across the output of the whole circuit (so after the T1 switch, basically on pin 7/8 of IC1), makes the ripple go away and voltages stable around 15V and 30V depending on whether T1 is open or not.

anyone got an idea why this works with the standard diodes but not with the ideal diode controllers? With the ideal diode controller it just seems like C2 isn't being used at all when only one winding is active and if both windings are used, C2 also seems to not hold much voltage which makes the overall output voltage drop appearently.

if i disconnect C1, C2, IC1, T1 and everything and only leave the ideal bridge rectifier with 1uF ceramic + 100uF electrolyte cap across the output, I get the expected 30V so the rectifier itself seems to work fine even on the breadboard.

edit:
just verified again: using a standard diode bridge rectifier instead of the ideal bridge rectifier (IC10) with mosfets makes this work, even with the one single ideal diode controller (IC1/LTC4357). as soon as I put in the ideal bridge rectifier IC10 instead of a normal diode bridge rectifier, sh*t hits the fan. no matter if IC1 is there or a standard 1N4007 diode.

even if i connect a cap of 1000uF additionally to C53, one winding alone doesn't deliver more than around 9.5V. it seems like C2 isn't charging at all. i guess the issue could be that the ideal bridge recitifier controller isn't connected at all to the center tap of the transformer and therefore somehow some current path is missing half the time or something??

[fcb]

My guess is (not sure I'll explain it properly)  that there is a 'shift' in the secondary voltages w.r.t. to your system ground which is interfering with the way your ideal-bridge IC is functioning.

Can you look at the current being drawn from an ordinary bridge-rectifier and the ideal-rectifier?  Also, 'scope the actual secondaries (scope ground=output 0V), you might see a difference between the ordinaryBR and idealBR.

[nemail2]

My guess is (not sure I'll explain it properly)  that there is a 'shift' in the secondary voltages w.r.t. to your system ground which is interfering with the way your ideal-bridge IC is functioning.

Can you look at the current being drawn from an ordinary bridge-rectifier and the ideal-rectifier?  Also, 'scope the actual secondaries (scope ground=output 0V), you might see a difference between the ordinaryBR and idealBR.

yes, something like a "shift" i was also putting together in my mind from what i have seen and measured. only way i currently could explain why C2 seems to be completely uneffective.

current being drawn from the positive output terminal of the rectifiers to the rest of the circuit you mean?

scope the secondaries: scope ground to the negative output of the rectifiers and then measure directly a the top terminal of the transformer and then the center where the middle terminals of the transformer are connected together?

also: do you think there could be a way around this? i really like the possibility of the space saving switching of the windings with a P-FET and not having to worry about the contacts of a switching relay at all but having to stick with a standard diode based rectifier for this seems too much of a drawback for me. i'd then have to cope with heatsinking on the rectifier again + deal with the voltage drop. the zero voltage drop came in quite handy...

[nemail2]

just found this thread on the LT forums - maybe related to this issue...? https://ez.analog.com/power/f/q-a/115706/lt4320-with-center-tapped-transformer

i also posted this question there, it is currently in the moderation queue though...

edit: did some measurements.

Ch1: IC10 pin 2
Ch2: IC10 pin 7
Ch3: IC10 pin 6

SDS00008: everything connected as in schematics, no button pressed
SDS00009: IC10 pin 6 (positive output) disconnected from C1 and thus from the rest of the circuit. Voltage flattens out at around 29V as ist should be
SDS00010: everything conencted as in schematics, button for top winding disconnection pressed. the output of the whole circuit falls to about 9.5V (not seen in scope shot), the output of IC10 drops a liiiiittle bit and the mosfet gate drive changes a tiny bit as well

pins 3 and 4 of IC10 (ground mosfet gate driving pins) look perfectly square, non of that weird business which is going on at the positive mosfet gates. not sure what's that about but as that doesn't really change even with pin 6 of IC10 disconnected from the rest of the circuit, where the voltage stabilizes at the expected leve, i reckon that's just as it should be?

geez, I'm out of ideas.. anyone got any clue? please

[nemail2]

dios mio, can this be the solution?
https://www.diyaudio.com/forums/group-buys/333844-ideal-bridge-rectifier-gb-post5911437.html

i am scared to try, though  especially on the toroidal transformer which exactly specifies how to put the windings in series or in parallel...

[fcb]

I've spent a few minutes simulating the LT4320 in roughly your configuartion in LTSpice (see attached). Including reversing one of the secondaries.

TLDR; LT4320 won't work with a split supply.

[nemail2]

I've spent a few minutes simulating the LT4320 in roughly your configuartion in LTSpice (see attached). Including reversing one of the secondaries.

TLDR; LT4320 won't work with a split supply.

thanks for your time and effort!
regarding the result: well, crap...

any ideas? I really liked the possibility to switch with a FET instead of a relay and also not having to heatsink the bridge rectifier...
I'm out of ideas, obviously. If I had to decide between cooling the BR or going with a relay instead of a FET for switching taps, I guess I'd go with the relay. And I think that's probably not a smart choice. But I just really don't want to deal with any additional major heating within the case and having to worry about airflow around the BR and stuff.

edit:
I revisited that diode topic in my mind and was wondering (again, but harder), why the rectifier diodes in my scenario got so hot, before I chose to use the LT4320 controller.
I was using these https://www.mouser.at/datasheet/2/427/vs-16edu06-m3-1769326.pdf and they had around 80-100°C under full load which was "only" 4A (currently I am aiming towards 5A). Looking at the datasheet, they should have around 0.9V of Vf drop each and that multiplied with 4A gives me 3.8W which then again leads to 6,46°C temp rise according to the datasheet (RthJ to solder pad). I even had heatsinks on them back then but still they were cooking.
Could this be due to the 20000uF of capacitance right after them? The transformer in use is afaik not underrated (4.17A) so I don't think there should be too much ripple which would lead to constant capacitor charging/discharging at high currents...?

Thanks!

[fcb]

I think there are a fewof options if you care about heat dissipation over BOM cost, and don't want to visit switching techniques.  But probably the easiest is to give each secondary it's own LT4320 and effectively build two isolated rectifier circuits, stack the outputs and then feed then switch between them with MOSFET's.

If you wanted to use a few more MOSFET's then I'm sure you could arrange some sort of circuit that allowed you to parallel the two seperate rectified outputs (for more current) or stack them (for more voltage).  Or perhaps just use a DPDT relay.  Of-course if you had a DPDT relay, you could just switch the transformer secondaries and have a vastly reduced circuit with a single LT4320 and no requirement for more MOSFET.

[Ian.M]

Another alternative for a center-tapped bridge is to simply use separate dual ideal diode controllers for the positive and negative side MOSFET rectifiers.

I've attached a sim of the negative side of such an arrangement using a LTC4354 controller.  Unfortunately LTspice doesn't seem to have a model for a positive side dual ideal diode controller that's as simple and easy to use as the LTC4354, although A.D. do sell some.

[nemail2]

I think there are a fewof options if you care about heat dissipation over BOM cost, and don't want to visit switching techniques.  But probably the easiest is to give each secondary it's own LT4320 and effectively build two isolated rectifier circuits, stack the outputs and then feed then switch between them with MOSFET's.

If you wanted to use a few more MOSFET's then I'm sure you could arrange some sort of circuit that allowed you to parallel the two seperate rectified outputs (for more current) or stack them (for more voltage).

That both sounds good but also rather expensive and quite board space intensive. I think that would be a classic example of "overdoing" stuff if I chose to do it that way for my intended purpose (although I'm kind of tempted tbh). I know I said I'd like to keep the ideal diode controller stuff, just an example of how undecided I am. I mean if the diodes would settle at around 50 or 60°C at full load, I guess I'd go for that + the P-FET for switching the second tap. probably way cheaper (P-FET + diodes way cheaper than ideal diode controller + SPDT relay). only thing I'm a bit worried, is whether it is allowed to use only one winding of a toroidal transformer.. the datasheet claims otherwhise... https://www.mouser.at/datasheet/2/410/VPT36_4440-781806.pdf

Or perhaps just use a DPDT relay.  Of-course if you had a DPDT relay, you could just switch the transformer secondaries and have a vastly reduced circuit with a single LT4320 and no requirement for more MOSFET.

That's how it is now, and it is working (see screenshot). But I'm worried about the current and arcing the relay has to withstand when switching taps under full load.

Regarding the hot BR diodes, before I was using the LT4320:
Playing with LTSpice showed me that the bridge rectifier diodes have to deliver way more current than what the load is sinking - can this be the reason why my diodes were heating up that much at only 4A load?

[nemail2]

Another alternative for a center-tapped bridge is to simply use separate dual ideal diode controllers for the positive and negative side MOSFET rectifiers.

I've attached a sim of the negative side of such an arrangement using a LTC4354 controller.  Unfortunately LTspice doesn't seem to have a model for a positive side dual ideal diode controller that's as simple and easy to use as the LTC4354, although A.D. do sell some.

Thanks but I'm not sure if I understand this correctly - I don't need any negative voltage, I only need to switch between series and parallel connection (or at least between series and single) of the transformer taps. Would your proposed solution still work?

Thanks!

[Ian.M]

The problem with the LT4320 is that the load current must be purely between its OutP and OutN terminals, and cannot use a center tap of a secondary feeding In1 and In2 for a split supply.

My proposed LTC4354 circuit is the negative side of a classic bridge rectifier + secondary center tap split supply, which is best regarded as separate two diode full wave rectified positive and negative supplies.

Any split supply circuit can be used for half/full positive voltage supply - just move the ground from the center tap to the former negative rail and connect all the former ground pins together and to the center tap.

[bdunham7]

Regarding your heating concerns, are the transformer and voltage regulators going to be in the same enclosure?  If so, I can't imagine the heat from the bridge rectifier being significant compared to the heat from those.  Perhaps just reassessing what a proper, robust bridge would be might dissuade you from pursuing an ever more complex solution here.

https://www.allelectronics.com/item/fwb-352/35-a-200-piv-bridge-rectifier/1.html

[nemail2]

The problem with the LT4320 is that the load current must be purely between its OutP and OutN terminals, and cannot use a center tap of a secondary feeding In1 and In2 for a split supply.

My proposed LTC4354 circuit is the negative side of a classic bridge rectifier + secondary center tap split supply, which is best regarded as separate two diode full wave rectified positive and negative supplies.

Any split supply circuit can be used for half/full positive voltage supply - just move the ground from the center tap to the former negative rail and connect all the former ground pins together and to the center tap.

thanks! I will try to wrap my head around this. any DaveCad sketch would be appreciated, though

Regarding your heating concerns, are the transformer and voltage regulators going to be in the same enclosure?  If so, I can't imagine the heat from the bridge rectifier being significant compared to the heat from those.  Perhaps just reassessing what a proper, robust bridge would be might dissuade you from pursuing an ever more complex solution here.

https://www.allelectronics.com/item/fwb-352/35-a-200-piv-bridge-rectifier/1.html

yes, they are going to be in the same enclosure. the thing is that:
1) there shouldn't be much heating of the diodes at all in the first place (mathematically, that would be 1,4V * 5A = 7W * 3K/W(1) = 21°C + 25°C (Tambient) = 46 °C)
2) when I had the design built up with a diode bridge rectifier, the diodes got around 100 °C hot, even with a heatsink

Note 1: 3K/W is for GBU8 with a 4x4x0.15inch copper plate mounted

So I'm wondering why they even got so hot and back then I just decided to go with the ultra-cool ideal diode controller and just don't bother with that issue anymore.
Back then I was using TO263AC-3 diodes (VS-16EDU06-M3/I) like mentioned before and even with those there shouldn't have been significant heating at the specified load of 4A.

The cooling solution of the pass transistors is mature or even overdone a bit. I have two BD139/TIP3055 Darlingtons mounted on a huge heat sink with silpads and even a 80mm fan. silent operation kinda guaranteed :-) heatsink: https://www.mouser.at/ProductDetail/ohmite/cr201-75ve/?qs=EU6FO9ffTweI3RXL9kLvyw==&countrycode=DE&currencycode=EUR

Regarding cooling of the diodes: other than slapping some heatsink on them there is no solution currently in my design nor much of free space in the case. I guess if I'd be using some GBU10 or something like that (or even a bigger rectifier like you suggested), I'd be able to put some heatsink on + somehow put that heatsink into the airflow of the main cooler. But first I'd really like to understand why the diodes got so hot in the older version of this design.

i have attached a screenshot of how the diodes were placed back then. on top of them i had also some small heatsinks (once I even tried with one larger BGA heatsink which covered all four of them). I was assuming that all the vias and copper should be able to dissipate quite some heat but still the diodes warmed up to 80-100°C...

[bdunham7]

yes, they are going to be in the same enclosure. the thing is that:
1) there shouldn't be much heating of the diodes at all in the first place (mathematically, that would be 1,4V * 5A = 7W * 3K/W(1) = 21°C + 25°C (Tambient) = 46 °C)
2) when I had the design built up with a diode bridge rectifier, the diodes got around 100 °C hot, even with a heatsink

Note 1: 3K/W is for GBU8 with a 4x4x0.15inch copper plate mounted

So I'm wondering why they even got so hot and back then I just decided to go with the ultra-cool ideal diode controller and just don't bother with that issue anymore.
Back then I was using TO263AC-3 diodes (VS-16EDU06-M3/I) like mentioned before and even with those there shouldn't have been significant heating at the specified load of 4A.

Something went wrong in your thermal design or calculations, but it's not uncommon for rectifiers to have a T_j quite a bit over 100C if you don't derate them.  Is there any room on your heatsink?  Just bolt on the part I linked and I'm very sure your problem will be solved forever.  You could also use TO-220 sized Schottky rectifiers if you want to reduce voltage drop a bit.  If you stick with your split-CT design with a one diode voltage drop, these would work well for all 5 diodes and might fit on your heat sink more easily.

https://www.mouser.com/datasheet/2/395/SRAF1620_SERIES_I1512-1918579.pdf

[nemail2]

yes, they are going to be in the same enclosure. the thing is that:
1) there shouldn't be much heating of the diodes at all in the first place (mathematically, that would be 1,4V * 5A = 7W * 3K/W(1) = 21°C + 25°C (Tambient) = 46 °C)
2) when I had the design built up with a diode bridge rectifier, the diodes got around 100 °C hot, even with a heatsink

Note 1: 3K/W is for GBU8 with a 4x4x0.15inch copper plate mounted

So I'm wondering why they even got so hot and back then I just decided to go with the ultra-cool ideal diode controller and just don't bother with that issue anymore.
Back then I was using TO263AC-3 diodes (VS-16EDU06-M3/I) like mentioned before and even with those there shouldn't have been significant heating at the specified load of 4A.

Something went wrong in your thermal design or calculations, but it's not uncommon for rectifiers to have a T_j quite a bit over 100C if you don't derate them.  Is there any room on your heatsink?  Just bolt on the part I linked and I'm very sure your problem will be solved forever.  You could also use TO-220 sized Schottky rectifiers if you want to reduce voltage drop a bit.  If you stick with your split-CT design with a one diode voltage drop, these would work well for all 5 diodes and might fit on your heat sink more easily.

https://www.mouser.com/datasheet/2/395/SRAF1620_SERIES_I1512-1918579.pdf

dunno what went wrong. isn't large capacitance after the rectifier diodes affecting them negatively in terms of heat generation?
I'd have to put another heatsink in, the current one is occupied with the two TIP3055 and i don't think it could carry the additional heat from 5 diodes...
Attached a photo of the situation

[bdunham7]

dunno what went wrong. isn't large capacitance after the rectifier diodes affecting them negatively in terms of heat generation?
I'd have to put another heatsink in, the current one is occupied with the two TIP3055 and i don't think it could carry the additional heat from 5 diodes...
Attached a photo of the situation

If you can line them all up along the top, I doubt that they would have much impact on a fan cooled heat sink like that.  Or bolt a small aluminum plate to the top that overhangs and then bolt them to the underside.

The filter cap size does affect the diodes to a certain extent because it narrows the conduction window by reducing the per-cycle voltage droop.  That means that your average current has to flow in less time--so if the conduction time is 33% at full load, your peak current will be 3X full load.  This means your voltage losses are higher and the diode ends up dissipating more than 3X the average during the peak--and therefore more than if the current was constant.  I think your original diodes were not underrated, just a combination of small and minimally heat sinked--and you want them to run cool.  If you look at the datasheet for the Schottky I linked, you'll see that at 12A V_f is 0.6V, so since  you have two diodes conducting at any one time, you have 4*0.6*2 = 4.8 watts dissipated at full load.  5 watts can cook small parts even with some PCB heat sinking, but 4 TO-220s on a fan cooled aluminum heat sink, not likely.

[nemail2]

If you can line them all up along the top, I doubt that they would have much impact on a fan cooled heat sink like that.  Or bolt a small aluminum plate to the top that overhangs and then bolt them to the underside.

that top area you mean? quite nice idea I must admit, although I'd be concerned of the long wires + maybe some interference with the pass elements...? This PSU is quite precise, only around 1-8mV absolute error from 0-24V, settable in 1mV steps and I wouldn't want to introduce a huge error/interference source into that, if that'd even be a concern... (i honestly don't know).

BUT probably the biggest issue is, that there is very little clearance to the top of the case there, so hot rectifiers would very likely be melting the case there (assuming they'll still get considerably hot while being on such a heatsink, which they probably wouldn't)...

Anyway, I guess I'll figure out some way for this to happen.

The filter cap size does affect the diodes to a certain extent because it narrows the conduction window by reducing the per-cycle voltage droop.  That means that your average current has to flow in less time--so if the conduction time is 33% at full load, your peak current will be 3X full load.  This means your voltage losses are higher and the diode ends up dissipating more than 3X the average during the peak--and therefore more than if the current was constant.  I think your original diodes were not underrated, just a combination of small and minimally heat sinked--and you want them to run cool.  If you look at the datasheet for the Schottky I linked, you'll see that at 12A V_f is 0.6V, so since  you have two diodes conducting at any one time, you have 4*0.6*2 = 4.8 watts dissipated at full load.  5 watts can cook small parts even with some PCB heat sinking, but 4 TO-220s on a fan cooled aluminum heat sink, not likely.

Thanks! Is there some rule of thumb formula how I can calculate the conduction time and the peak current? I wasn't able to find anything on the internet, maybe I googled wrong...
What about a load which rapidly switches a high current (to be precise, I have a very cheap electronic load from aliexpress which does this)? That load seems to be switching to achieve the set current draw, I could imagine that this load could be maybe drawing loads of current more from the caps than it should which in turn causes the caps to be recharged much "harder" by the diodes?

I guess I'll give it another shot with the diodes, veeery likely with Schottkys like you suggested. Any advice on the reverse voltage rating with a 12/24VAC toroidal transformer? Would something like 40V be sufficent?
Unfortunately every try costs quite some significant bucks and time, unless I go ahead and desolder all the small, expensive parts and reuse them on the next prototype (not knowing whether an what damage I might have caused to them buy desoldering/resoldering them).
Sometimes I think I should have separated the power board from the analog board so I could respin them independently but then again I always think "doesn't matter, it will be the last revision" 

Thank you again for your help!

[nemail2]

https://www.mouser.com/datasheet/2/395/SRAF1620_SERIES_I1512-1918579.pdf

How about 5 of these? https://www.mouser.at/ProductDetail/Diodes-Incorporated/SDT30100VCT?qs=5666Nh5RPmax9ZPmI%2FKCtg%3D%3D
On the Mouser product page it says single but in the datasheet there seem to be two diodes - but I guess I could use them in parallel as one diode, probably they are meant to be used that way as the datasheet nowhere mentions it being two diodes in one case...

The diodes you suggested are kind of unavailable, so I searched the cheapest >=65V >=15A TO220 schottky diodes and this model is what i have found...

[bdunham7]

that top area you mean? quite nice idea I must admit, although I'd be concerned of the long wires + maybe some interference with the pass elements...? This PSU is quite precise, only around 1-8mV absolute error from 0-24V, settable in 1mV steps and I wouldn't want to introduce a huge error/interference source into that, if that'd even be a concern... (i honestly don't know).

BUT probably the biggest issue is, that there is very little clearance to the top of the case there, so hot rectifiers would very likely be melting the case there (assuming they'll still get considerably hot while being on such a heatsink, which they probably wouldn't)...

I doubt the wires would cause an issue at mains frequency, although you never know until you try!  If the top is tight, my suggestion of an additional plate might work.  Just cut a 2 or 3mm thick aluminum plate and attach part of it to the heatsink and have part of it hang over and screw the diodes on the bottom side, then insulate the topside of everything to keep the case cool  You could even bend the plate down towards the board to keep the wires short.

[/quote]
Thanks! Is there some rule of thumb formula how I can calculate the conduction time and the peak current? I wasn't able to find anything on the internet, maybe I googled wrong...
What about a load which rapidly switches a high current (to be precise, I have a very cheap electronic load from aliexpress which does this)? That load seems to be switching to achieve the set current draw, I could imagine that this load could be maybe drawing loads of current more from the caps than it should which in turn causes the caps to be recharged much "harder" by the diodes?

I guess I'll give it another shot with the diodes, veeery likely with Schottkys like you suggested. Any advice on the reverse voltage rating with a 12/24VAC toroidal transformer?
[/quote]

Rules of thumb are not all that accurate in a case like yours because there are too many unknowns--it's easier just to measure them. For example, how much does the peak voltage sag under full load?  You would have to check with a scope to really know.  But you should know that the AC RMS current will be higher than the DC current (1.5-3X)  which has implications for your transformer selection and that the filter cap size does have an effect on this ratio, just not a big one in any design with reasonably sized capacitors. 1-2mF per ampere of output seems reasonable.

As far as your fast switching load, it shouldn't really put any significant stress on your capacitors or rectifiers, but the linear regulator may have an inadequate frequency response and may not be able to keep up, so you would want a decoupling capacitor right at the input of the switching supply.  And some linear regulators don't like capacitive loads because it slows down their feedback response and can cause oscillation, so....  Again, check with a scope.

A reverse voltage of 40V would probably be just barely enough, but I'd go with 60V just for some safety margin.  This increases your forward voltage and power dissipation, but probably not enough to matter.

[bdunham7]

How about 5 of these? https://www.mouser.at/ProductDetail/Diodes-Incorporated/SDT30100VCT?qs=5666Nh5RPmax9ZPmI%2FKCtg%3D%3D
On the Mouser product page it says single but in the datasheet there seem to be two diodes - but I guess I could use them in parallel as one diode, probably they are meant to be used that way as the datasheet nowhere mentions it being two diodes in one case...

The diodes you suggested are kind of unavailable, so I searched the cheapest >=65V >=15A TO220 schottky diodes and this model is what i have found...

Those will work.  There are two diodes in each case, but there's no equivalent common anode design so you may as well just parallel the two cathodes.  One drawback of this part is that the metal tab is not insulated but is connected to pin 2, which makes attaching it to a heat sink a little trickier. 

[nemail2]

that top area you mean? quite nice idea I must admit, although I'd be concerned of the long wires + maybe some interference with the pass elements...? This PSU is quite precise, only around 1-8mV absolute error from 0-24V, settable in 1mV steps and I wouldn't want to introduce a huge error/interference source into that, if that'd even be a concern... (i honestly don't know).

BUT probably the biggest issue is, that there is very little clearance to the top of the case there, so hot rectifiers would very likely be melting the case there (assuming they'll still get considerably hot while being on such a heatsink, which they probably wouldn't)...

I doubt the wires would cause an issue at mains frequency, although you never know until you try!  If the top is tight, my suggestion of an additional plate might work.  Just cut a 2 or 3mm thick aluminum plate and attach part of it to the heatsink and have part of it hang over and screw the diodes on the bottom side, then insulate the topside of everything to keep the case cool  You could even bend the plate down towards the board to keep the wires short.

Thanks! Is there some rule of thumb formula how I can calculate the conduction time and the peak current? I wasn't able to find anything on the internet, maybe I googled wrong...
What about a load which rapidly switches a high current (to be precise, I have a very cheap electronic load from aliexpress which does this)? That load seems to be switching to achieve the set current draw, I could imagine that this load could be maybe drawing loads of current more from the caps than it should which in turn causes the caps to be recharged much "harder" by the diodes?

I guess I'll give it another shot with the diodes, veeery likely with Schottkys like you suggested. Any advice on the reverse voltage rating with a 12/24VAC toroidal transformer?
[/quote]

Rules of thumb are not all that accurate in a case like yours because there are too many unknowns--it's easier just to measure them. For example, how much does the peak voltage sag under full load?  You would have to check with a scope to really know.  But you should know that the AC RMS current will be higher than the DC current (1.5-3X)  which has implications for your transformer selection and that the filter cap size does have an effect on this ratio, just not a big one in any design with reasonably sized capacitors. 1-2mF per ampere of output seems reasonable.

As far as your fast switching load, it shouldn't really put any significant stress on your capacitors or rectifiers, but the linear regulator may have an inadequate frequency response and may not be able to keep up, so you would want a decoupling capacitor right at the input of the switching supply.  And some linear regulators don't like capacitive loads because it slows down their feedback response and can cause oscillation, so....  Again, check with a scope.

A reverse voltage of 40V would probably be just barely enough, but I'd go with 60V just for some safety margin.  This increases your forward voltage and power dissipation, but probably not enough to matter.
[/quote]

I doubt the wires would cause an issue at mains frequency, although you never know until you try!  If the top is tight, my suggestion of an additional plate might work.  Just cut a 2 or 3mm thick aluminum plate and attach part of it to the heatsink and have part of it hang over and screw the diodes on the bottom side, then insulate the topside of everything to keep the case cool  You could even bend the plate down towards the board to keep the wires short.

Rules of thumb are not all that accurate in a case like yours because there are too many unknowns--it's easier just to measure them. For example, how much does the peak voltage sag under full load?  You would have to check with a scope to really know.  But you should know that the AC RMS current will be higher than the DC current (1.5-3X)  which has implications for your transformer selection and that the filter cap size does have an effect on this ratio, just not a big one in any design with reasonably sized capacitors. 1-2mF per ampere of output seems reasonable.

As far as your fast switching load, it shouldn't really put any significant stress on your capacitors or rectifiers, but the linear regulator may have an inadequate frequency response and may not be able to keep up, so you would want a decoupling capacitor right at the input of the switching supply.  And some linear regulators don't like capacitive loads because it slows down their feedback response and can cause oscillation, so....  Again, check with a scope.

A reverse voltage of 40V would probably be just barely enough, but I'd go with 60V just for some safety margin.  This increases your forward voltage and power dissipation, but probably not enough to matter.

Ok, thanks! I'll try and develop a solution with five schottky diodes then

Those will work.  There are two diodes in each case, but there's no equivalent common anode design so you may as well just parallel the two cathodes.  One drawback of this part is that the metal tab is not insulated but is connected to pin 2, which makes attaching it to a heat sink a little trickier. 

initially i was worried about thermal drift of the two diodes but then i figured there are many of those TO220 devices which have two diodes inside but don't mention that fact in the datasheet. I guess they are supposed to be used as a single diode and there are two in them to meet the Vf goal..? So I guess they'll be ok...

I was also looking for a complementary common anode device but wasn't able to find one either (would have saved some space). but then, more heat in one case so maybe it is better to utilize 5 devices (cases) anyway.

Would you recommend to utilize devices with a plastic package? I was deliberately looking for diodes with a metal tab to ensure better heat transfer. if that wouldn't be of any concern, plastic cases would make mounting easier (and cheaper), of course (i was planning to use quite expensive silpads, just like on the pass elements)...

edit: filtering for plastic packages, these were the cheapest diodes if i wanted single diodes in a case: https://www.mouser.at/ProductDetail/Vishay-General-Semiconductor/VFT2080S-E3-4W?qs=HSPD0Bff7iaqHXrROfvV8A%3D%3D
I'm really not sure whether paralleling two diodes within a case would be a good idea (if they are specified as two single diodes) in terms of thermal runaway and frying themselves eventually...

[bdunham7]

edit: filtering for plastic packages, these were the cheapest diodes if i wanted single diodes in a case: https://www.mouser.at/ProductDetail/Vishay-General-Semiconductor/VFT2080S-E3-4W?qs=HSPD0Bff7iaqHXrROfvV8A%3D%3D
I'm really not sure whether paralleling two diodes within a case would be a good idea (if they are specified as two single diodes) in terms of thermal runaway and frying themselves eventually...

Those look good.  Plastic cases are fine--the devices can tolerate a significant thermal gradient in this application.  Paralleling diodes can be an issue but not in this case for two reasons--they are thermally bonded and will always be at almost the same temperature and in any case, either one can take the full current.

[nemail2]

edit: filtering for plastic packages, these were the cheapest diodes if i wanted single diodes in a case: https://www.mouser.at/ProductDetail/Vishay-General-Semiconductor/VFT2080S-E3-4W?qs=HSPD0Bff7iaqHXrROfvV8A%3D%3D
I'm really not sure whether paralleling two diodes within a case would be a good idea (if they are specified as two single diodes) in terms of thermal runaway and frying themselves eventually...

Those look good.  Plastic cases are fine--the devices can tolerate a significant thermal gradient in this application.  Paralleling diodes can be an issue but not in this case for two reasons--they are thermally bonded and will always be at almost the same temperature and in any case, either one can take the full current.

ok thanks!

one last question about something I'm not totally sure about - would i place the fuses like in the screenshot or one on the top and one in the middle after the series connection of the two windings?

[bdunham7]

one last question about something I'm not totally sure about - would i place the fuses like in the screenshot or one on the top and one in the middle after the series connection of the two windings?

It doesn't matter as long as the fuse isolates the winding.  In an unbranched circuit, the current has to be the same everywhere, so the fuse will blow in the same manner regardless of where it is.  And if it disconnects one side of the winding, that winding is effectively isolated.  You probably want a primary side fuse as well and I'd spec them so that the primary blows first in case of overloads, while the secondaries are a sort of last resort in case of some internal malfunction like a shorted rectifier that grossly overloads just one winding.

[tautech]

You probably want a primary side fuse as well and I'd spec them so that the primary blows first in case of overloads, while the secondaries are a sort of last resort in case of some internal malfunction like a shorted rectifier that grossly overloads just one winding.

Yes that maybe an ideal scenario but in practice matching primary fuses for a max secondary output is subject to fuse value availability and a suitable blow specification (fast or slow) and then there are ON inrush currents that also need be considered.
Multiple secondary outputs will make fusing the primary totally impractical.

IMO any output current limiting should be done in silicon in conjunction with secondary fusing close to the transformer.
 

[nemail2]

It doesn't matter as long as the fuse isolates the winding.  In an unbranched circuit, the current has to be the same everywhere, so the fuse will blow in the same manner regardless of where it is.  And if it disconnects one side of the winding, that winding is effectively isolated.  You probably want a primary side fuse as well and I'd spec them so that the primary blows first in case of overloads, while the secondaries are a sort of last resort in case of some internal malfunction like a shorted rectifier that grossly overloads just one winding.

You probably want a primary side fuse as well and I'd spec them so that the primary blows first in case of overloads, while the secondaries are a sort of last resort in case of some internal malfunction like a shorted rectifier that grossly overloads just one winding.

Yes that maybe an ideal scenario but in practice matching primary fuses for a max secondary output is subject to fuse value availability and a suitable blow specification (fast or slow) and then there are ON inrush currents that also need be considered.
Multiple secondary outputs will make fusing the primary totally impractical.

IMO any output current limiting should be done in silicon in conjunction with secondary fusing close to the transformer.
 

ok thanks! I do have a primary side fuse as well, which is rated for the main toroidal transformer + the second meanwell switching brick (IRM-05-12). In this case, the primary fuse is more like a last resort, I guess.

[nemail2]

Hi

yasq - yet another (probably very) stupid question:
In the attached screenshot you'll see a pulldown for the "SW-WINDINGS2" net. are there any objections? I don't want that pin to float during any failure mode cause that might leave T8 in linear operation which might fry it.

I kinda don't like R27 and R26 being the same value, don't know why (really not sure about this). It certainly is not a voltage divider but is there any other issue I might run into with this?

thanks!

[Zero999]

Hi

yasq - yet another (probably very) stupid question:
In the attached screenshot you'll see a pulldown for the "SW-WINDINGS2" net. are there any objections? I don't want that pin to float during any failure mode cause that might leave T8 in linear operation which might fry it.

I kinda don't like R27 and R26 being the same value, don't know why (really not sure about this). It certainly is not a voltage divider but is there any other issue I might run into with this?

thanks!

Adding a pull-down is a good idea. I advise moving R27 to the base, which will reduce the current taken from the circuit driving it, as well as provide a better pull-down. There is no issue with the resistors being the same values.

I've learned a bit from this thread. I didn't know that the idea diode bridge controller doesn't work with a tapped transformer.

[nemail2]

Adding a pull-down is a good idea. I advise moving R27 to the base, which will reduce the current taken from the circuit driving it, as well as provide a better pull-down. There is no issue with the resistors being the same values.

ok thanks! unfortunately it was too late for this spin of the PCB but it fortunately my version also "works".

I've learned a bit from this thread. I didn't know that the idea diode bridge controller doesn't work with a tapped transformer.

happy to see that my thread managed to teach something even to a pro obviously I didn't know either that this ideal diode bridge rectifier doesn't work with a tapped transformer. I went with schottkys now and have just finished building the PSU up, all files and photos attached.
the rectifier diodes now don't get too hot (around 60 to maybe 70°C abs. max), even at 5A current.

Features of this linear Lab PSU (no switching prereg, completely linear):
- 0 - 24V in 1mV steps
- 0 - 5A in 1mA steps
- accuracy: pretty accurate :-D no idea how to characterize and measure everything properly
- 100W max
- 2x 12V 100VA toroidal transformer
- Teensy 4.0 - virtually unlimited resources for going wild on software features
- crystal clear, bright and crisp 128x64 2.7" OLED display
- output relay (disconnects V+ and GND)
- tap switching from this thread using a P-FET, no relay clickedy-action
- dual BD139/TIP3055 darlington pair
- active cooling for output stage with temperature monitored (PID) and PWM controlled 80mm fan
- 2x TI DAC70501 14 bit with 5ppm internal reference for Voltage and Current setting
- 2x ADS1115 16 bit ADC for voltage and current measurement
- OPA4197 and OPA2197 precision opamps for control loop/gain stage
- precision/low drift resistors for high initial and temperature-stable accurracy
- 10 mOhm 15ppm 0.1% current shunt with LT6102 Precision Zero Drift Current Sense Amplifier for low heat generation and low voltage drop
- Schottky diodes for voltage rectification to minimize voltage drop and heat
- display as well as front panel diodes brightness is software-controllable
- piezo buzzer
- there is no ripple up until about 10.8V at full load (5A) with one winding and at up to 23V at 5A with two windings (IIRC, maybe a bit less).

now imagine the possibilities if one would have time to go wild on the firmware... I have only implemented the very basic features for now.

This is the n-th incarnation of my lab psu design which I started by copying and changing Dave's old µSupply project. Years and many forum threads here and on mikrocontroller.net later and thanks to many people helping me there, I have come to the point where I'm quite happy with the result. Me on my own would have never been able to accomplish this and I have learned a metric crapton on the way.
Sure, this thing has its flaws and it is nowhere near a professional Lab PSU but it is the very best I'm able to create and I'm happy with it as it is

[tautech]

- there is no ripple up until about 10.8V at full load (5A) with one winding and at up to 23V at 5A with two windings (IIRC, maybe a bit less).

Looking good but pray do tell us what qualifies as no ripple ? 

[nemail2]

- there is no ripple up until about 10.8V at full load (5A) with one winding and at up to 23V at 5A with two windings (IIRC, maybe a bit less).

Looking good but pray do tell us what qualifies as no ripple ? 

Thanks! Maybe I did not express myself correctly. When I wrote no ripple, I meant that the scope shows a completely flat line in AC coupled mode, set to 20mV/. Except for noise and crap it is picking up, of course. When tuning the voltage up a bit over the values I mentioned, you start to get dips in 100Hz frequency so that's where the input caps are too small or the voltage headroom of the transformer is to little...
At least that's how I understood the concept

[tautech]

Yes zero ripple at full load is not easy to achieve but in reality it's not like that will be absolutely required when instead very low ripple at say 2A would be expected and is a more realistic target.

A DIY linear PSU I did years ago used a LM339 for 5A and it would do that however the transformers weren't a stiff enough supply so voltage did sag some near full load. It used 2x 12V transformers manually switched for parallel and series connection to provide 15 and 30V modes although no current limiting other than what a LM339 does.

Maybe by probing yours close to the PSU is influenced by EMI where a scopes 20 MHz BW limit can help get accurate readings. Use 1x probes and some averaging too.

[Barneyview]

I think this (attached diagram) relay / switch connection removes any danger of a short during switching