Possible improvement of output voltage tracking circuit

[prasimix]

I'm wondering if simple tracking circuit as marked red in the schematic below could be somehow improved to provide more constant "rms" difference between control signal (Vout that comes from the post-regulator) and the pre-regulator output (POUT or Vpre):



I tested it with various load (and output voltage and current) and got the following results (P1=240K):



When for example 8R2 load is connected and Vout (yellow) is set to 5V POUT (cyan) looks like this:



... and with Vout set to 40V the following difference can be measured between Vout and POUT:



As you can see in the first measurement the difference between POUTmin and Vout is ~5V and in second one ~1V what is a lower limit that linear post-regulator have some function. If I try to decrease difference for smaller voltage (and current) it become unusable for higher voltage (and current).

The question is how to ensure let say 1V difference regardless of voltage and current?

[Kalvin]

Have you considered using a zener on the base of the Q1 instead of the adjustable R1 / P1. Also, using a darlington as Q1 might improve the circuit's regulation. Instead of the RMS you could survive with the average voltage difference. An averager of the VOUT is quite simple to add to the circuit.

[T3sl4co1l]

Is this supposed to be a phase control circuit, or one of those "inductorless" switching circuits, or just a linear regulator?

It doesn't seem like it would be especially good at any of the three... 

Tim

[prasimix]

Have you considered using a zener on the base of the Q1 instead of the adjustable R1 / P1. Also, using a darlington as Q1 might improve the circuit's regulation. Instead of the RMS you could survive with the average voltage difference. An averager of the VOUT is quite simple to add to the circuit.

Situation is similar with zener when P1 is replace with 4V3.



For Vout=5V (load=8R2):



At Vout=40V POUT is too small what naturally results with increased ripple:





That could be resolved with increasing zener voltage (i.e. 5V1 or higher) but again voltage difference between PREGmin to Vout is not uniform.

[Kalvin]

Your zener is wrong way Now the zener acts like an ordinary diode and the voltage drop across the zener is approx. 0.7V. But when you turn it around, it will operate as a zener, and the voltage drop across will be the 4.3V as specified.

[prasimix]

Is this supposed to be a phase control circuit, or one of those "inductorless" switching circuits, or just a linear regulator?

It doesn't seem like it would be especially good at any of the three... 

Tim

It has to be first: "phase control circuit" or let say approximation of such device . I'm aware that is inferior in comparison with circuit that can be found in HP/Agilent PSUs built around TCA785 or SG3525 but it's cheap and simple. Ok, maybe too simple and due to that this topic is opened

[prasimix]

Your zener is wrong way Now the zener acts like an ordinary diode and the voltage drop across the zener is approx. 0.7V. But when you turn it around, it will operate as a zener, and the voltage drop across will be the 4.3V as specified.

Yes, that true but if I reverse it I got even worser "overlap" on upper end (Vout=40V for the same load 8R2). Here is with 5V1 zener:



... and this is with 6V2 zener:

[Kalvin]

I propably missed something here. I thought that the Q3 should conduct and recharge the C2 only when the voltage level of the X1-9 is approx. 5V below the  voltage of the C2. And when the voltage level of the X1-9 is less than 5V thershold, the Q3 should be turned off and the C2 will provide the energy to the circuit. Now the ciruit seems to work a bit differently or my assumption is way wrong.

The current through the 8ohm load resistor at 40V is around 5 Amps, so the C2 will be heavily discharged during the mains periods. Either you need to reduce the load current, increase the C2 or increase the mains frequency

[prasimix]

I propably missed something here. I thought that the Q3 should conduct and recharge the C2 only when the voltage level of the X1-9 is approx. 5V below the  voltage of the C2. And when the voltage level of the X1-9 is less than 5V thershold, the Q3 should be turned off and the C2 will provide the energy to the circuit. Now the ciruit seems to work a bit differently or my assumption is way wrong.

The current through the 8ohm load resistor at 40V is around 5 Amps, so the C2 will be heavily discharged during the mains periods. Either you need to reduce the load current, increase the C2 or increase the mains frequency

Yep, I tested extreme and C2 at least twice as bigger could be a good improvement

[T3sl4co1l]

Well the first thing that's grossly missing in a phase controller circuit: no filter inductor!  It should look something like this:
http://seventransistorlabs.com/tmoranwms/Circuits_2010/Phase_Angle_Supply.png
(Heh, only two SCRs are necessary, the top or bottom pair can be regular diodes.)

The L and C have to be dimensioned for line frequency, so this gets very heavy, very quickly!

You can't do a phase controller without a filter inductor because the capacitor voltage will quickly charge to peak voltage, and the peak current will be a surge limited only by supply impedance, very likely to destroy most semiconductors.

Tim

[prasimix]

Well the first thing that's grossly missing in a phase controller circuit: no filter inductor!  It should look something like this:
http://seventransistorlabs.com/tmoranwms/Circuits_2010/Phase_Angle_Supply.png
(Heh, only two SCRs are necessary, the top or bottom pair can be regular diodes.)

The L and C have to be dimensioned for line frequency, so this gets very heavy, very quickly!

You can't do a phase controller without a filter inductor because the capacitor voltage will quickly charge to peak voltage, and the peak current will be a surge limited only by supply impedance, very likely to destroy most semiconductors.

Tim

Thanks Tim for your input. Interestingly this circuit works for me now for months with various loads and without any damage! If something could be heavily "under pressure" is mains transformer and its "buzzing" can be audible especially when control switch from 100Hz to 50Hz what is mentioned here.

[Kalvin]

prasimix, Did you get any improvement you were looking for?

[prasimix]

Not for now. I can live with this for current version of the PSU and just started this topic wondering how to decrease dissipation over the whole range by having more uniform voltage difference between pre-regulator (PREG) output and post-regulator (Vout).
I'm aware that due to the nature of this pre-regulator huge low frequency ripple can be expected on higher current and huge L could manage that as Tim suggested, but have no plan at the moment to increase PCB for hosting such inductor. Still thinking about "magic circuit" that can provide almost constant voltage difference between PREGmin value and Vout regardless of set voltage and current.

[T3sl4co1l]

PCB area?  You have that backwards: the inductor mounts the PCB, not the other way around.

Most in this class would be chassis mounted, with wire leads back to the PCB (secured in whatever reasonable fashion: soldered in, screws and lugs, quick connect..).

If you're having a lot of dissipation, then that would explain it: it's not a phase controller at all, and all the pulsing is a lot of wasted effort, or worse, an illusion that prevents one from seeing it as it is.

If you don't want to wind the inductor yourself (and you're probably better off that way, since filter chokes of that value are quite unusual, indeed -- perhaps Hammond has a few close enough?), you'll be absolutely all-around better off doing a switchmode circuit.  Boring, yes, but it's just a power supply, that's how it should be.

Tim

[Kalvin]

The Linear Technology Application Note 32 (Page 3) has a nice SCR/Thyristor-based Pre-regulator for 15V / 1A linear regulator. The inductor size is only 500uH. That circuit would be a good starting point to tweak the suggested pre-regulator for improved performance. Note the peak current, which can be quite high.

http://cds.linear.com/docs/en/application-note/an32f.pdf

[prasimix]

PCB area?  You have that backwards: the inductor mounts the PCB, not the other way around.

Most in this class would be chassis mounted, with wire leads back to the PCB (secured in whatever reasonable fashion: soldered in, screws and lugs, quick connect..).

If you're having a lot of dissipation, then that would explain it: it's not a phase controller at all, and all the pulsing is a lot of wasted effort, or worse, an illusion that prevents one from seeing it as it is.

If you don't want to wind the inductor yourself (and you're probably better off that way, since filter chokes of that value are quite unusual, indeed -- perhaps Hammond has a few close enough?), you'll be absolutely all-around better off doing a switchmode circuit.  Boring, yes, but it's just a power supply, that's how it should be.

Tim

Tim, I'm talking here about saving possibly a few watts when output is supplying 100 to 200W. Each volt makes difference on the post-regulator in the worse case of 5W. Yes, regular way is to build PCB around components and not vice versa. Since PCB is already done and I have workable solution in this moment, I don't want to spent more time and money on major hardware redesign. A lots of works is waiting now on the software side.

The Linear Technology Application Note 32 (Page 3) has a nice SCR/Thyristor-based Pre-regulator for 15V / 1A linear regulator. The inductor size is only 500uH. That circuit would be a good starting point to tweak the suggested pre-regulator for improved performance. Note the peak current, which can be quite high.

http://cds.linear.com/docs/en/application-note/an32f.pdf

Thanks Kalvin for that. Sure, that could be a good basis for a major redesign of PSU's pre-regulator part sometimes in the future.

[Kalvin]

The Linear Technology Application Note 32 (Page 3) has a nice SCR/Thyristor-based Pre-regulator for 15V / 1A linear regulator. The inductor size is only 500uH. That circuit would be a good starting point to tweak the suggested pre-regulator for improved performance. Note the peak current, which can be quite high.

http://cds.linear.com/docs/en/application-note/an32f.pdf

Thanks Kalvin for that. Sure, that could be a good basis for a major redesign of PSU's pre-regulator part sometimes in the future.

Sure. I like your modular approach in which the pre-regulator is in a separate PCB to the actual linear part of the power supply, thus making possible testing different circuit ideas quite easily. 

[T3sl4co1l]

Tim, I'm talking here about saving possibly a few watts when output is supplying 100 to 200W. Each volt makes difference on the post-regulator in the worse case of 5W. Yes, regular way is to build PCB around components and not vice versa. Since PCB is already done and I have workable solution in this moment, I don't want to spent more time and money on major hardware redesign. A lots of works is waiting now on the software side.

Ok, so you want to save a couple of watts, but you don't have the correct approach to save it; what more can you do, then? 

I can tell you ways that can (or, I guess since we're talking >> 90% efficiency, might) save those few remaining watts, but they will be very different from the circuit that you have presented here.  And if it is your absolute desire to save those watts, without changing the circuit.. I guess you'll have problems that "a lots of works" [sic] in software cannot possibly address.  You know what I'm saying?

Tim

[prasimix]

You are welcome to suggest something completely different as Kalvin already did referencing to the LTC app note. I possibly will not be in position for the pre-regulator remake soon. I'm aware of the fact that software cannot fix hardware, but still software even for "questionable" hardware require real effort especially if we don't want to stay on the "Hello world, set and see my voltage and current" level.

[nichga]

Hello Prasimix

you like an improvement

pull out the zenner

and wait for solder smoke

C°  F°  / cool down