EEVblog #329 - Tracking Pre-Regulator LTspice Simulation Part 2

[EEVblog]

Dave.

[nitro2k01]

R1 is pretty much unnecessary indeed. The usual instinct is that you need a resistor across the base to limit the base-emitter current, but since there's a resistor on the emitter (and the collector too) the base current doesn't need to be further limited.

[Rufus]

I simulated this on the 20th March. An improvement of the circuit BravoV posted in the EEVblog #260 thread on the 19th. And here we are 69 Vblogs later.

[BravoV]

I simulated this on the 20th March. An improvement of the circuit BravoV posted in the EEVblog #260 thread on the 19th. And here we are 69 Vblogs later.

Hey Rufus, thanks for bringing this up again and your tweak, honestly I forgot that thread already. 

For those who curious, the #260 thread -> https://www.eevblog.com/forum/blog-specific/eevblog-260-tracking-pre-regulator-simulation-in-ltspice

[EEVblog]

I simulated this on the 20th March. An improvement of the circuit BravoV posted in the EEVblog #260 thread on the 19th. And here we are 69 Vblogs later.

Wow, I don't recall that 
I got it as an email ages back from John.

Dave.

[kkp]

@R1 removal suggestions:
Before removing R1 from the design, check what happens when V2/OUT is at 0V (either the supply is set to 0V, or the current limiter circuit pulls it down, or the output is shorted for the schematic after 5:10):
The PNP has a collector-base diode as well as a emitter-base diode. With the base hard at 0V, the collector (and the 1.235V-referred feedback pin) won't go above 0.7V.
The boost converter will then try to generate a high voltage. And runaway boost converters tend to be destructive.

[LaurenceW]

Dave, I didn't hear you mention it, but at some point in the video you moved the base drive of the transistor (with or without the somewhat redundant base input resistor, we learn) from the SET pin of the LT3080 to the output pin. I have been wondering why you didn't do this in your earlier (now parked?) LT3080 linear power supply?

Now, I guess that if the 3080 regulator IC is working to spec, the output voltage should be pretty damn close to the SET  voltage anyway, so perhaps it doesn't make much, if any, difference. Possibly a marginal improvement in transient response...?

But then I got to thinking - what if the power supply output is shorted to ground? with the base connected to the SET pin of the output regulator, the preregulator/boost is going to try and continue to output the target output voltage +2V. But with the base of the transistor connected to the regulator output, if the output voltage is forced to ground, then the preregulator will try and drop its output voltage, too, so reducing the short circuit power dissipation of the circuit overall.

what have i missed?

[kaindub]

dave
I don't quite agree with your description of the way the circuit works.
Q1 is acting as an emittier follower. You sort of said that is round about terms.
The current through R4 is set by the difference between the emitter voltage and the Pre Reg voltage.
This same current is applied to R2, which develops the feedback voltage across it (the 1.25V). I don't think R2 is acting a constant current source. The voltage across R2 is developed by the current passing from Q1 (its actually the constant current source).
It may be helpful to other to know that Pre Reg voltage above the oputput can be chnaged by changing R4. If R2 is 10k ohms, then the voltage across R4  will be 126.5 micro amps times the value of R4 plus 0.6V (these are aproximate values and are depenedent on the circuit paramteres). So by changing R4, the Pre Reg voltage gap can be changed.

I also noticed that you took the control voltage  for the pre regulator from the output of the linear regulator. The original circuit had the control volatge taken from the Set pin of the linear regulator.  By using the set pin you take the control voltage to the pre regulator out of the second feedback  loop. The regaultion will be better and stability will be better.

I enjoyed this episode so keep up the great shows
Robert

[f4eru]

Hello Dave

Why not do the preregulation with a self made flyback isolating converter ? would avoid one conversion in the chain and allow for a much more efficient supply in low voltage mode, and would be really much much cheaper than a potted DC/DC

There are many cheap transformers for 500V isolation in the same housing as a SMD ferrite coil.
http://www.digikey.com/product-search/en/inductors-coils-chokes/arrays-signal-transformers/197323 >> filter for surface mount
for example : SRF1280 or DRQ73

Also, you could easily make a separate supply ( using the flyback transformer as a forward for the aux) for Vctrl of the LT3080, and reduce the dropout to 500mV ! better efficiency again !

[hans]

Maybe an idea to investigate/test in Dave's particular setup:

I have been trying the same concept on a buck converter (as far as I can see, only the feedback resistor voltage is adjusted, and that's very similar to all type converters).
I first had a low frequency converter, I believe a LT1076 (as a test). It's only 100kHz, so the inductor was 100 - 180 uH and a lot of output capacitance (470uF was recommended). I see Dave has 220uF as well, so this doesn't make the converter very 'quick' on transients.

I tried stepping the voltage up and down (V2). Rising edge is  fine; the feedback voltage drops lower (usually 0V if the transient is very steep), meaning the converter has to increase it's output. The SMPS determines how steep the slope is of the PSU output, because the LT3080 can't do magic and has to wait.
Falling edge may cause issues with the large capacitance because the output voltage can't drop very quick, especially with no load attached. I saw the feedback pin voltage in my case going as high as 15V, but I am designing a PSU with 30-36V input voltage and an output that hopefully can reach 25V. I had a resistor ratio of 1k on 1k, so it's basically 30V/2 worst case (with the PNP fully open).

Trouble is most converters datasheets tell you that the feedback pin voltage may not exceed x V. In my case it was only 7V. So I added a zener diode across it to limit it at 4.7V. I'm pretty sure that in my case I need it, or the SMPS chip will not live a long life..
In Dave's case of the power supply, a 10V output means 12V output for the pre-tracking regulator. A resistor ratio of 10k:10k means 6V on the feedback pin worst case, which isn't that far off  it's maximum (or maybe above the chip Dave is using, I haven't looked it up).
So maybe it's a good idea to throw in a zener diode for more robust setup.. A power supply suddenly shooting in CC is basically will simulate the same situation of suddenly dropping the control voltage from maximum to zero.

[Bored@Work]

Falling edge may cause issues with the large capacitance because the output voltage can't drop very quick, especially with no load attached.

That is why some power supplies have a feature called active down programming. It is a beefy transistor (plus collector resistor to limit the current through the transistor) in parallel with the output.

The control logic drives the transistor to get the output voltage down faster when needed.

[Rerouter]

now i dont quite follow why you are using such a high input capacitance, having a 0.1 ohm buffer resistor following the diode allows you to get all the way down to 40uF with bugger all noise, its easier to filter the Vcontrol for noise and let that mop up the ripple, also allows the use of ceramic capacitors instead,

also if one where to aim to isolate the output, what would you guys reccomend sticking with a convetional switcher connected to a HF transformer, or another topology?

as for the down programming if a seperate constant current source was made it wouldnt be all that hard to set it up to shoot up to a much higher current to pull it down :/ hmm, back to the drawing board i go, also wouldnt make for that bad of a self diagnosis function, though i am starting to once again be bested by feature creep,

[hans]

I was using the recommended capacitors as per the datasheet, and I saw Dave was using very similar values. Look at Rev C schematic; it has 22uF ceramic with 220uF aluminium 25V. So it probably has very similar issues.

Yes I was aware of a 'dump' resistor or transistor, I encountered a similar issue in the past before and 'solved' it with a 68R resistor. It wastes a ton of energy all the time, but not a problem for that case (transformer powered).

I just wanted to make awareness this can be an issue in this setup. Personally I don't like adding active control transistors. The LT3080 will dissipate extra power in this situation anyway. So I guess you still need a decent sized heatsink, just in case.

[konfu]

I don't really get why Dave is telling about 3 volts pre-regulation. The LT3080 needs somewhat around 500mV between input and output voltage. So why aiming for 3 volts here and smashing all the efficiency again?

Shouldn't we aim to get the VIN of the LT3080 to about 0.5V insteal of 3V then?

Sorry to bring that thread back to light now but I am wondering about that fact for days now and just don't get it.

[Jay_Diddy_B]

Hi Konfu,

The answer lies in the LT3080 datasheet.

There are two dropout voltages:

Vin to Vout is 0.5V

Vctrl to Vout at 1.1A is 1.6V max. This is because Vctrl supplies the bias to the NPN output transistor.


Jay_Diddy_B

[konfu]

Ah. Got it. So Dave wants to feed both "VIN and VCONTROL" at the same time. Roger that ;-)