Transformer Tap Changer - Reducing TPD In Linear Bench Power Supplies

[James98]

Hi my name is James I'm 17 and have been studying electronics as a hobbyist for a few years now.

I know alot of comercial bench power supplies use transfromer tap switching to reduce the input voltage to the voltage regulator at lower output voltages and therefore reduce the power dissipated in the regulator enabling them to achieve higher output currents.

I thought I'd have a go at designing my own circuit to switch 4 transformer taps automatically in an adjustable bench power supply which i have included as an attatchment. Please consider it a concept as oppossed to a final design. I'd like to get some feedback and hopefully outline any mistakes or improvments that could be made to this idea. I suspect some unwanted switching could occur when low impedance load is connected to output of powersupply and voltage at output may drop for a short period I'm not sure. Maybe some capacitance at non-inverting inputs could help prevent this but may cause the circuit to have poor response.

Thanks from James

Circuit Description:

I have used two centre tap transformers in with output wired in series (T1,T2) to represent multitap transformer however i guess they could also be used as shown to create a multitap transformers as they are not easy to find (as long as the transformers are identical). I have included fuses on every tap so that they will blow should there be a short circuit between any taps. RL6,5,7,8 are being used to represent relay contacts by engaging these relay contacts different transformer taps can be selected to then power a bridge rectifier, Filter and linear regulator etc. (Not shown). The two terminals labled 'AC TO D1' are used as to provide the refrence voltage to the control circuitry. This is rectified and filtered by D1 and C1 Then to a 10:1 voltage divider R5 should be a 20k Trimpot to allow the switching threshold to be adjusted (Neccessery to allow switching to higher voltage tap at a lower output voltage than that tap due to losses it would never reach threshold voltage and never switch.) Then onto a voltage divider to set the switching intervals these voltages are fed into the inverting input of the op-amps configured as comparators these compare the output voltage from the power supply and 'decide' which tap to use. The NPN's provide current to swich relays and the capacitors on the base should help against intermintant switching near threshold voltage due to power supply output ripple/noise. The relays are configured to disconect the previous tap before engaging the next to prevent any shorts between taps.

[wraper]

Few things that catched my eye. C2-C5 are not necessary, 10k resistance for R9-R12 might be too much, basically they aren't needed, except if you want to protect outputs of LM324. Voltage drop on transistors will be quiet high too, about 2V (if enough current flows into the bases through R9-R12, otherwise even worse) which will cause them to heat, plus reduced voltage on the relays.  Also you need positive feedback to make some hysteresis, otherwise you relays will be clicking like hell in certain conditions.

[James98]

Okay thanks for advice. I'll need to do some research on positive feedback and hysteresis, I've only recently started to work with op-amp circuits. 

[dom0]

Consider using thyristors and opto-thyristors to 'extend' a bridge rectifier. I consider relays mostly obsolete for this application.

[unitedatoms]

For 4 taps the 4 relays is better to be replaced with 2 relays.
Tap 1 -\-- Mid 1 Relay 1
Tap 2 -

Tap 3 -\-- Mid 2 Relay 1
Tap 4 -

Mid 1 -\-- Out Relay 2
Mid 2 -

This will guarantee no shorts whichever state the relays are in

[Circlotron]

D2-D5 are not necessary either because the relay coils are being driven by emitter followers. At switch-off the relay coils will try to spike negative but this pulls on the emitter, turning the transistor partially on for a fleeting moment and preventing the emitter going more than 0.6v below the base. No diodes needed.

[James98]

Thanks for all replies. I have attatched a redesigned schematic here based on your feedback. I've removed a relay from the circuit aswell which was'nt necessary, I was using it to disconnect 8V tap when all the switching was done via comparators however RL3 does that switching now.

[Mark Hennessy]

In addition to the suggestions so far, I think I'd clean up the input to the reference inputs to the comparators. Right now, they come from an unregulated diode bridge+smoother, so will have a small 100Hz ripple component. Just replacing R6 with a 3.3V zener would probably be enough, or you could use a 78L05 or similar to replace R5+R6.

Another thought is to consider what might happen when the power supply is being used dynamically - that is, when you have a load that is regularly triggering the current limit - it's more common than you might expect. Do you want the relays chattering when this happens? You can introduce a time constant to the circuit - somewhere in the R7/8 region - that will hold up the relays for a second or two after the output voltage has dropped. Will save a lot of wear and tear on the relay contacts (and the main bridge and smoothing capacitors take quite a "hit" in these applications as they voltage is ramped up).

A safer way to do this is to monitor the voltage across the regulator. It's a bit more complex, so perhaps for a later design, but it allows you to really minimise the drop, irrespective of all other variables.

BTW, I was doing exactly this when I was 17. Building a variety of linear power supplies is a really good way to learn about electronics, and you can never have too many bench power supplies in a lab. Good job so far 

[Mark Hennessy]

Hi again, James,

I just re-read your first post and see that you did indeed think about dynamic conditions - apologies for missing that! The key is to ensure that the circuit is fast to rise, but slow to fall. Adding diodes to charging circuits is usually a good first move 

Mark

[Mark Hennessy]

Thanks for all replies. I have attatched a redesigned schematic here based on your feedback. I've removed a relay from the circuit aswell which was'nt necessary, I was using it to disconnect 8V tap when all the switching was done via comparators however RL3 does that switching now.

It would be possible to use single-pole changeover relays for this. What's nice is that as you work up, the "earlier" relays become "don't care", so you can de-energise them to save a few 10s of milliamps if you wish. It's simple, but this computer lacks a drawing package, so I'll post something later if no-one else does...

Again, perhaps one for the future, but it's also worth remembering that this can be done in the "DC domain". Yes, there are increased losses in multiple rectifiers, but depending on the expected currents, these might not be a problem. For the basic idea, look up class G audio amplifiers, which use multiple DC rails and diodes and pass transistors to steer different voltages. It's more complex and has a higher component count than the AC switching, but there is complete freedom from switching spikes and relay chatter. It can work well for critical applications.

[Circlotron]

It would be possible to use single-pole changeover relays for this. What's nice is that as you work up, the "earlier" relays become "don't care", so you can de-energise them to save a few 10s of milliamps if you wish. It's simple, but this computer lacks a drawing package, so I'll post something later if no-one else does...

This setup should work. The relay contacts come on as upper and lower pairs. As the higher voltage ones come on the lower ones get overrun and the diodes of the lower voltage legs allow this. There is no need to have break before make and the worry about arcing the contacts if you don't.

Conceptually the idea has a parallel to a General Motors TH400 auto transmission and it's derivatives, with it's one-way sprags and clutch packs allowing the next higher gear to come in without disengaging the previous one first.

[codeboy2k]

High side emitter followers are generally not recommended for relay drivers for various reasons:

1) The base drive voltage threshold is generally quite high, almost near to the rails.  This doesn't leave much headroom and some relays may not ever pull in with the voltage drop Vce in play. 

2) high side emitter follower's turns the relay into a voltage threshold detector.  i.e. the relay will turn on when the base voltage reaches some threshold V_t(on) and this can vary between relays... which not so good for production... and given #1 above you may not even be able to reach the necessary V_t(on)

3) Because relays have a much lower holding current than their pull-in current, it introduces some hysteresis into the relay's turn on and turn off voltage (+1V_be again at the driver's base).  This can make it difficult to cause the relay to turn off when you want it to, since you have to bring the base down below the voltage that causes the necessary holding current

Low side NPN drivers sinking current to ground are always going to be better. They have near 0 hysteresis, it does not change the voltage threshold as seen by the relay (which always sees the rail voltage, so it always has enough voltage to turn on) and they are always happy turn off when told to.

And my opinion is that you always need the diode across the relay, whether you use a high-side or low-side driver.  I'd rather use a beefy diode as close to the relay terminals as possible, rather than introduce trace/path inductance to the transistor, and then relying on the somewhat fragile transistor to withstand these transients.    Vbe can easily exceed 100V for that fraction of a second until it turns on and limits the voltage to 0.6V , and many small signal transistors can't take that abuse for very long.  Finally, if the transistor is turned on during the flyback time, then it can have current flowing from collector to emitter again just near the point of the 2nd turnoff event, and then you can get relay chatter.  It's so much better to just use a proper diode at the relay coil and give the current the ability to dissipate within the coil resistance itself as the magnetic field collapses.

Finally: relays are old school for voltage taps now. Dom0 said it already here:

Consider using thyristors and opto-thyristors to 'extend' a bridge rectifier. I consider relays mostly obsolete for this application.

Put your full-bridge rectifier at the output.  An alternative design uses the SCR's seen in the drawing as one leg of the bridge rectifier.

[James98]

Thanks Mark for the replies, I will definatly include a time constant in there to help resolve some issues with dynamic conditions and  will also clean up that reference voltage. The Class G amplifiers are a really interesting topic and will have to do some reading on those. 

[James98]

Thank you circlotron and codeboy2k for your replies both design look like a much better solutions than using relays alone to do the switching thanks for the suggestions. 

Thanks for suggesting the Low side NPN driving aswell it will be a very usefull thing to bear in mind in the future.

[codeboy2k]

Thank you circlotron and codeboy2k for your replies both design look like a much better solutions than using relays alone to do the switching thanks for the suggestions. 

Thanks for suggesting the Low side NPN driving aswell it will be a very usefull thing to bear in mind in the future.

You're welcome.  Yes, low side current sinking drivers are generally better as I've said. Since their bases only need to be driven 1 Vbe above ground to turn on, then they can be easily driven by any LVTTL, LVCMOS, CMOS or TTL logic level I/O.  The high-side of the relay can be 5V, 12V , 24V, etc... it doesn't matter what voltage rating the relay has (up to a reasonable upper limit).

Since it seems you are doing this for home use and as a learning experience, you can still use relays, it's a simpler design.  Later you can improve it to gain more knowledge and replace the relays with an SCR solution.

[Mark Hennessy]

I agree that it's best to use the "layered" approach to a project like this, where learning is one of the goals. The relay solution might be sub-optimal to some, but it is the simplest approach and is almost certainly guaranteed to work, leaving you free to concentrate on the control logic. With that done, SCRs or something else can be tried later. Always best to break any problem down into simple, manageable chunks and deal with each in isolation where possible.

One note about the NPN low-side drivers. I agree they are the best solution, but be aware that the op-amp isn't a rail-to rail type, so when its output is "low", it might not be low enough to turn off the transistor. And even if it is, I wouldn't rely on it for production. A resistor between base and emitter is the solution. Simple, but worth saying as it could be a trap for the unwary