Wide input voltage boost converter - How to do slope compensation?

[Miyuki]

Hi folks

I want to ask how to best implement slope compensation or what suitable controller to use with internal one

Vin = 30-150V
Vout = 170V

With this high input range can duty cycle vary from 0.1 to 0.8
Ipk is constant as output is kin of Constant current and is controlled by microcontroller

So I can have dedicated current mode pwm controller like classic uc3842 or something more modern with integrated clope compensation
Or can drive it completly from mcu with comparator

What do you think will be best way to solve it ?
Can I just calculate maximal suitable duty for curent voltage and run in cycle by cycle current mode with max duty limited by actual voltage ratio plus some margin. I think this can keep duty in some reasonable boundaries, some variation is acceptable

[jbb]

Err, it's been a while since I've done the maths for slope compensation, but I think getting newer chips with slope compensation built in might be best.  You've got a wide input range, which is going to compromise your efficiency a bit.

Some scoping questions:

• What power level do you need?
• Do you need full power at 30V input? (High inductor current!)
• EDIT: Do you need continuous operation over the 30V - 170V input, or is there a 'dead zone' where operation is not required? (e.g. are you supporting a "48V" power system and a "120V" power system?)  This can allow for some topology tricks...
• Are you under pressure for cost, size, efficiency or some combination?
• How many will you make?

[T3sl4co1l]

You can run the oscillator from VIN instead of VREF, i.e., use a (much larger) resistor from VIN to RTCT, and then do slope comp as usual.  Note this varies the switching frequency, which may be undesirable.  However, it is precisely necessary, if ~constant current ripple is desired.  Probably, you should design for nominal Fsw at max VIN, then check how things work out at min VIN where the frequency will be much lower.  Adjust filters accordingly, to control voltage ripple (on both input and output).

Note that Fsw also varies with supply ripple, if this is an off-line application.  Heh, well, it's spread-spectrum for free, so that's one thing, but you will see ugly waveforms when probing the circuit, as both period and pulse width will be varying during operation.

You'll probably want an independent aux supply, as UC3842 draws too much current to really be practical running from a linear regulator at voltages that high (e.g., 10mA requires ~1.4W dissipated, not at all impossible in a SOT89 or DPAK transistor, but definitely a cost), and the boost inductor won't provide any useful voltages from a tapped winding (i.e., the switch-on and switch-off voltages are all over the place).

There are better suited regulators out there, which have a timing or slope compensation signal derived from the supply.  TPS40211 and LTC3810 come to mind as examples I've used (not necessarily applicable here).  Actually, that may not be "better suited" so much as just another incarnation of the same strategy.  I forget.

Also consider a zeta or SEPIC topology, which may be more convenient for certain reasons (like aux power derived from a winding), if probably not best for overall efficiency (the inductor handles more reactive power, and is split into two windings).  YMMV.

Tim

[Miyuki]

Err, it's been a while since I've done the maths for slope compensation, but I think getting newer chips with slope compensation built in might be best.  You've got a wide input range, which is going to compromise your efficiency a bit.

Some scoping questions:

• What power level do you need?
• Do you need full power at 30V input? (High inductor current!)
• EDIT: Do you need continuous operation over the 30V - 170V input, or is there a 'dead zone' where operation is not required? (e.g. are you supporting a "48V" power system and a "120V" power system?)  This can allow for some topology tricks...
• Are you under pressure for cost, size, efficiency or some combination?
• How many will you make?

Power level raise with input voltage (it is for solar system so more cells give more voltage and this should handle different cell count) so input current remain similar and output will rise significantly (to hundreds of watts)
I was thinkink about different topologies but everything I try to calculate give me terrible efficiency this have like 95% even for medium input voltages
I can change output voltage but it is chosen to be high enought to have low current even at high power levels and also can use cheap shottky diodes and fast mosfets with low rds and work at high switching speed to keep inductors and capacitors small
Size doesnt matter, efficiency is not a main issue but try to keep decent syse passive cooling, and prise cheaper i better
Now I will make like 5 pieces to show how whole system will work together (their output will be parallel connected to one voltage bus), then will see

I have design for power part of it an by calculation it can easy handle up to 10A of input current with just small heatsink for transistor and diode (like 5 to 10W loss each depend on voltage and like 4 to 6 watt magnetics loss)
Problem is how to controll it right as they must act as current source to be easy to parallel connection and should keep this high input range to be universal

[Miyuki]

As I was reading some interesting papers about this topic

What do you think about hysteretic current mode with variable frequency
Have you any experience with this solution ? To have average current with constant ripple it look promising to solve all issues or is there again some devil in detail ?

[jbb]

For high speed slope compensated microprocessor control, you can look at the TI Piccolo line (some have special hardware for it).

Power level raise with input voltage (it is for solar system so more cells give more voltage and this should handle different cell count) so input current remain similar and output will rise significantly (to hundreds of watts)

I have design for power part of it an by calculation it can easy handle up to 10A of input current with just small heatsink for transistor and diode (like 5 to 10W loss each depend on voltage and like 4 to 6 watt magnetics loss)

Ahhh, solar input.  So I'm guessing the spec could be summarised as '30 - 150V input @ 10A.'

When you consider Maximum Power Point Tracking (MPPT), it makes sense to use a microcontroller.  This is because good MPPT requires adaptive behaviour - the best panel voltage changes with sunlight, temperature, and of course the panel in use.  Also, it's easier to fiddle with control software than control hardware.

If you do go for a microprocessor, I thoroughly recommend using an isolated JTAG adaptor.

If we have a look at a panel IV curve (e.g. https://goo.gl/images/QNYhL6), we see that a panel is much like a constant current source.  We need to control the (boost) converter so that the panel voltage is Vmp to get peak power (see PV curve).  The good news is that the top of the PV curve is somewhat flat, so a small voltage error isn't the end of the world. (A very smart controls engineer pointed this one out to me.)

It is possible to draw a PI (power vs current curve) and control the panel current, but the slope of the power peak is much steeper.  This means a small current error is going to cause a lot of power loss.

What controls your 170V power rail?  Perhaps it's the DC link for a single phase grid inverter? Or battery charger?  Or maybe a battery bank?  In those cases, we end up with the following arrangement:

Panel (I source) ==== Converter ==== DC link (V source)

... and it's good to connect an I source to a V source.  So maybe you can operate the boost converter with voltage mode control, knowing the input current is limited by the panel.

Panel (I source; Ipanel) ==== (V sink; Vdc * duty) Converter (I source; Ipanel * duty) ==== DC link (V source; Vdc)

Problem is how to controll it right as they must act as current source to be easy to parallel connection and should keep this high input range to be universal

Paralleling can be tricky.  If you want to connect a 20A panel to two 10A MPPT units, you need current sharing.  A current mode control scheme is very effective for this.

I was thinkink about different topologies but everything I try to calculate give me terrible efficiency this have like 95% even for medium input voltages
I can change output voltage but it is chosen to be high enought to have low current even at high power levels and also can use cheap shottky diodes and fast mosfets with low rds and work at high switching speed to keep inductors and capacitors small
Size doesnt matter, efficiency is not a main issue but try to keep decent syse passive cooling, and prise cheaper i better
Now I will make like 5 pieces to show how whole system will work together (their output will be parallel connected to one voltage bus), then will see

I was involved with some PV work in the past and we found that the boost converter (and refinements) is one of the best candidates for the job.

EDIT: forgot last post:

As I was reading some interesting papers about this topic

What do you think about hysteretic current mode with variable frequency
Have you any experience with this solution ? To have average current with constant ripple it look promising to solve all issues or is there again some devil in detail ?

Hysteretic current control is a bit out of fashion.  This is because the switching frequency varies hugely with input voltage and load current (once you go to discontinuous mode), which makes EMC filtering difficult.

Also, as I said above, you're probably more interested in controlling the input voltage rather than current.

[Miyuki]

What controls your 170V power rail?  Perhaps it's the DC link for a single phase grid inverter? Or battery charger?  Or maybe a battery bank? 

Also, as I said above, you're probably more interested in controlling the input voltage rather than current.

DC link voltage control is simple.
Units will have limited output voltage to 180V at no load and at 160V is backup power (battery), so voltage can varry a little
This gives me easy paralleling and at side of load I can detect if I have enought power or running from backup

Hysteretic current control is a bit out of fashion.  This is because the switching frequency varies hugely with input voltage and load current (once you go to discontinuous mode), which makes EMC filtering difficult.

With microcontroller I can keep swithing frequency in reasonable boundaries, but I have no real experience with it

[ocset]

The latter pages of this give a good precis on slope compensation
Also, by using constant off time control, you dont need slope compensation.
COT is good for wide vin ranges.
You can 'hack' a uc3842 to behave in constant off time. this is very easy to do...just a wee fet and resistor from gate.

ok, it wont  let me attach it because its 1.074MB.

Its called "current mode control modelling" by ray ridley

[Miyuki]

The latter pages of this give a good precis on slope compensation
Also, by using constant off time control, you dont need slope compensation.
COT is good for wide vin ranges.
You can 'hack' a uc3842 to behave in constant off time. this is very easy to do...just a wee fet and resistor from gate.

ok, it wont  let me attach it because its 1.074MB.

Its called "current mode control modelling" by ray ridley

I agree it will be better approach
constant off-time peak current control solve my issue and relative simple to implement

[Miyuki]

I do some calculations and constant off time gives me nice almost constant Pout for constant Ipk with variable load
So mcu can simple set Ipk by DAC for controller and it will work exactly as I need it

[ocset]

so do you want me to send you the circuit for converting uc384x into a constant off time controller?

[Miyuki]

so do you want me to send you the circuit for converting uc384x into a constant off time controller?

No need, I found it and it can be done with single transistor