Modify Boost Converter Circuit (MT3608)

[T_guttata]

Hi everyone

I'm using the boost converter module based on the MT3608 from aliexpress. It's very versatile, but it's limited to 28V.

Here is the module: https://de.aliexpress.com/item/32948417810.html?spm=a2g0o.productlist.0.0.74db480cbPrLP7&algo_pvid=6790c464-b2e0-4765-a241-5f8ddf6fcef8&aem_p4p_detail=202202010312151502121846646080020672586&algo_exp_id=6790c464-b2e0-4765-a241-5f8ddf6fcef8-0&pdp_ext_f=%7B%22sku_id%22%3A%2210000015705863452%22%7D&pdp_pi=-1%3B1.95%3B-1%3B-1%40salePrice%3BUSD%3Bsearch-mainSearch

The circuit is attached below.

What could I do to increase the output voltage to 40-50V without increasing footprint too much? Obviously another IC is required, as the MT3608 does not go >30V (https://pdf1.alldatasheet.com/datasheet-pdf/view/1131968/ETC1/MT3608.html)

Could anyone make a suggestion for a particular IC? Would the TI LM 5157 be suitable? https://www.ti.com/lit/ds/symlink/lm5157.pdf?ts=1643699079935&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FLM5157

Thanks!

[Zero999]

What's the input voltage?

Add another transistor, to form a cascode with the internal switching transistor. Refer to the post linked below for an example. The maximum input voltage is limited to the gate-source voltage rating of the MOSFET, but the output voltage is not. A potential divider, or zener diode could be added to resolve this issue.

https://www.eevblog.com/forum/beginners/constant-current-led-driver-309077/msg3961499/#msg3961499

The diode is only rated to 40V, so if you need to exceed that, it will need to be changed.

[T_guttata]

Thanks for your reply.

I understand what you suggest, but honestly I don't yet understand how it works.

The input voltage range for the MT3608 board is 2-24V.
Let's assume we use 12V as input voltage. What would be the maximum output voltage (if the diode is exchanged)?

[Zero999]

I understand what you suggest, but honestly I don't yet understand how it works.

M1 forms a cascode with the internal switching transistor of U1. When U1's output transistor is off, the voltage on M1's source is just below the gate voltage and only a tiny leakage current flows. When U1's output transistor turns on, M1 also turns hard on, as its source voltage is now well below, the gate voltage. Because M1 forms a source follower, when U1's output transistor is off, the voltage across the output transistor always stays below 12V.

Let's assume we use 12V as input voltage. What would be the maximum output voltage (if the diode is exchanged)?

The maximum ratings of M1 + 12V minus the gate threshold voltage, at whatever the tiny leakage current though U1's output transistor is, minus the diode's forward voltage drop, or the diode's maximum reverse voltage rating, whichever is lower.

The RVQ040N05 has a maximum drain source voltage rating of 45V. The gate threshold is specified to be between 1V and 2.5V, with a drain current of 1mA and at the tiny leakage current through U1's output transistor, it'll be even lower. It's probably safe to say the maximum output voltage will be around 54V, with this arrangement.

EDIT:
I forgot that M1 will also leak a bit of current, so its source voltage will be a bit higher than I initially thought. I doubt it's an issue, even if it does exceed the maximum voltage rating of U1's output transistor, because the current will be limited to such a low level. A high value gate-source resistor, say 10M would keep the source at 12V, if you're paranoid.

[robert.rozee]

as an aside, pins 1 and 2 of R2 should be linked to allow the adjusting of R2 to linearly change the output voltage. this also ensures there is a constant 0.6/2200 = 270uA current flowing through R1. see image below.

the omission of this link seems to be common to all the MT3608 PCBs from china, and does represent a design error.


cheers,
rob   :-)

edit: fixed incorrect reference to R1 instead of R2. added mention of R1 current flow.

[David Hess]

Replace D1 with a voltage doubler.

[T_guttata]

I did not do the modification yet, because I found another solution.

I found a board based on the XL6009, which has an output voltage of 5-50V and current is also fine, it delivers 2.5A.

I was quite surprised to find that chinese board also on digikey https://www.digikey.com/en/products/detail/dfrobot/DFR0123/6588566?s=N4IgTCBcDaICIDEBKAGAjGAzCAugXyA

Now, I'm looking also for an alternative IC because of two reasons:
-XL6009 is chinese and I don't see I could use it in a commercial product
-XL6009 is quite big

Do you know any alternative ICs which are low cost (<$2)?
The LMR64010 is small and low cost, but output current is significantly less than for the XL6009. (<0.4A for V=40V according to the datasheet). Off course I had a look at digikey, but it's not easy to find something suitable. There are many ICs which are overkill, in terms of features and cost (>$5).

Maybe this IC could be interesting? https://www.ti.com/lit/ds/symlink/mc33063a.pdf?HQS=dis-dk-null-digikeymode-dsf-pf-null-wwe&ts=1646379427371&ref_url=https%253A%252F%252Fwww.ti.com%252Fgeneral%252Fdocs%252Fsuppproductinfo.tsp%253FdistId%253D10%2526gotoUrl%253Dhttps%253A%252F%252Fwww.ti.com%252Flit%252Fgpn%252Fmc33063a

[T3sl4co1l]

MC33063...

How much do you want again?  Like 50V 2.5A?  At the same time?  From what input?  Down to 2V?  12V?  That's a lot of power for just one converter.  I'd be looking at a D2PAK like that thing, or a TI SimpleSwitcher or etc., or something synchronous (maybe not at this voltage though), or a controller (external FET + diode).

That XL6009 by the way looks alright, probably a cloned SimpleSwitcher.  Key features to look for on the block diagram are: switch current monitored; error amplifier, with compensation (usually internal; external is more flexible, but also more faffing around to adjust); if oscillator and current sense into flip-flop, then it's peak current mode control, and needs slope compensation (ramp + summing block) to allow stable operation with lower ripple fraction (CCM, continuous current mode).  Which that all has, so it should be okay.  I mean they could always be lying or something, who knows.

BTW if you find yourself looking at the likes of MC33063 (5-30V, 1A ballpark), at least look at NCV3063 and friends -- which have real current limiting, not the, whatever stupid hack the MC- have that bends duty cycle but doesn't actually stop the output.  It's also faster switching so you're not wasting everything on enormous L and C.  It's still hysteretic, so expect possible funky output behavior resulting in higher ripple (compared to a proportional control type), but good regulation otherwise, and not really any compensation to worry about.

Tim

[T_guttata]

The LMR64010mentioned before is a TI SimpleSwitcher?

Off course I don't expect 50V @ 2.5A continuously (would be 125W). The XL6009 is very versatile but there must be alternatives? I will search again and try to focus on the features you mentioned.

[T3sl4co1l]

Quite.   There are more powerful ones than the LMR64010 as I recall. 

Looks like XL6009 is available from the usual suspects, emphasis on "suspect".  Surprised it's not at LCSC.  Oh well.

Tim

[T_guttata]

Ok, on digikey you can find several hundred ICs wich are called "simple switcher".

I filtered according to the output voltage, and the LMR64010 is the cheapest one <$2.
The next one is the LM2577, but there is a huge price gap ($7).

The XL6009 seems to be quite unrivaled at a price of 40 cents. The LM2577 is almost 20x the price!

[T3sl4co1l]

Like I said, not much in that range for integrated switch regulators.  You'll easily save that with a controller, UC3843 or better for example.

Tim

[T_guttata]

The UC3843 is a PWM controller?
But there are also Boost Controllers such as the TPS43060. What are the main differences?

[Someone]

needs slope compensation (ramp + summing block) to allow stable operation with lower ripple fraction (CCM, continuous current mode).  Which that all has, so it should be okay.  I mean they could always be lying or something, who knows.

"they" preferred slop:

[Someone]

Like I said, not much in that range for integrated switch regulators.  You'll easily save that with a controller, UC3843 or better for example.

plus assembly, plus stock/part/supply handling, plus board real estate, suddenly the integrated solutions start looking cheap. Biggest problem is the OP trying to find commodity small/cheap/simple parts in a non-commodity application (what volume application needs boost to >24V and several A of current?)....  waiting for the next drip feed of requirements to be easily hand solder-able, and available through local retail stockists in qty 1.

The cascode discussed above is probably worth exploring, as it reuses the current limit inside an integrated regulator. A nice intermediate step between fully integrated and a controller with everything external.

[T_guttata]

The first question is if there is any alternative, which is equally versatile. Seems not to be the case.

I can sacrifice some of the voltage input range, output voltage and current. If I do so, the LMR64010 seems to be an option. I will probably try it, but it's really not that easy. I'm not an electronics engineer.

...(what volume application needs boost to >24V and several A of current?)..

How about a spot welder?

[T3sl4co1l]

There are solutions for that, even directly (example: that battery welding kit that uses a whack of transistors in parallel, and a RC lipo battery, for spot welding!).  It will take work to achieve such currents from a switching supply, of course. Or do you mean a pulse welder, just charging a cap?

Tim

[T_guttata]

I mean the latter which should be quite sinple, just charge the cap and discharge it over a low resistance mosfet. Thats one application I was thinking of to use the XL6009. I even searched for capcitor charging ICs, but I only found those for photo flash circuits (>300V).

How would you design a circuit to charge a cap to 50V in a reasonnable time? Average power should be much lower than peak. From what I understand you can get much better efficiency if you charge with CC instead of CV.
I found a mod for the MT3608:
https://hackaday.io/project/28432-coin-cell-jump-starter/log/72237-how-to-charge-a-capacitor-efficiently

Certainly there are better solutions around for that task?

[T3sl4co1l]

Not using an SCR?  The pulse ratings are much higher, and they're easier to use.  Hm, I wonder if that might not actually be cheaper anymore...

Anything that works for photoflash, works scaled down to lower voltages.  Note that a boost converter is undesirable as the inrush (up to VIN) is unlimited; if you still want a ~1:1 ratio, use SEPIC instead (preferably with a coupled inductor, which are readily available in these ratings).

Nothing special about the application; a current limited boost/flyback type regulator/controller will do fine.

Tim

[T_guttata]

What do you think about the TPS4021?
https://www.ti.com/lit/ds/symlink/tps40210.pdf?HQS=dis-dk-null-digikeymode-dsf-pf-null-wwe&ts=1646992698872&ref_url=https%253A%252F%252Fwww.ti.com%252Fgeneral%252Fdocs%252Fsuppproductinfo.tsp%253FdistId%253D10%2526gotoUrl%253Dhttps%253A%252F%252Fwww.ti.com%252Flit%252Fgpn%252Ftps40210

[T3sl4co1l]

Sucks, don't use it.

The fault logic is broken.  Namely, fault is activated when V(COMP) exceeds some limiting threshold.  Which is reachable by the error amp within its normal operating range.  Including during startup.  Then it faults out, does a soft start, which immediately fails out again and so on, sitting there farting in its own misery, never producing enough average current to start up properly.

Maybe I didn't try correct values; maybe it starts up properly with a much larger start cap or something, but I found best results with none at all, giving an auto-restart cycle of some 100s Hz I think, and a startup current of about half nominal output, which was enough to get that project into production.

UC3843 and family are still very much viable options these days.  Maybe not if you need lower startup voltage, higher frequency, or lower V(Isense) (or lower component count thanks to internal slope comp and etc.).  Afraid I don't have any recommendations offhand for boost controllers meeting those requirements.

Tim

[T_guttata]

I also tried TI WEBENCH. One alternative is: LM5156.

But to be honest: I don't really trust this tool. Why? See image!

[T3sl4co1l]

Well, it's just a pile of heuristics, you can't expect it to be right all the time.

Yeah, that controller looks nice.  Screaming fast too!  Careful how you use that thing.   If you can find any, it doesn't look like it's well stocked right now... (which, probably means they're all being used up, a good sign..?).

Tim

[T_guttata]

Well, everything which is not pure crap is currently out of stock^^

I will take that IC on my list as a possible candidate.