CCCV MOSFET Buck

[Leeroy]

Hello.

I am trying to make this project work for months now, but I am an amateur and it is not cooperating. Although I've consumed a lot of theory info about DC-DC converters, there are still so many variables that it is way above my expertise to even attempt to debug such circuits.

Idea: make ~150W passively cooled LED light source for home use and for educational purposes.
I am going to use 4 strings of 3W 750mA 3.24Vf "Cree" LEDs from Aliexpress. Each string consists of 14 LEDs in series. PSU: MEAN WELL LRS-200-48. Each string needs DC-DC Buck converter with constant current output controlled with MCU+DAC providing reference voltage for gm error amplifier (that last part is out of the scope of this thread).

So far I succeeded in designing a working circuit around TL494 using datasheet approach with BJTs. Tested with real LEDs and works flawlessly and what's most important dead silent. But of course after measuring input and output power I see that at low brightness levels efficiency is around 69%, which is terrible for LED lighting IMO.

I decided to try switching to FET approach. I modified working schematic to utilize high side n-channel switch with an integrated driver and something went wrong. I'm using mosfet driver for the first time, so there's that fulfillment on the educational purposes part and I probably do not understand what am I doing..
MOSFET circuit appears to be working in that it regulates output current, but either transistor or inductor is making awful audible noises, which follows me in any FET based buck circuit that I try (yes this is not the first one):



And moreover inductor is getting hot very fast, I haven't powered it for more than 30 sec and still started to smell the compound between the coil and shielding. Tried changing 47uH part to 220uH, hot as before.
Tried to scope output of driver IC and saw some waveforms with a frequencies always less then 1kHz, which is basically a K.O. for me:



Souldn't it be the same as input frequency that is an output of TL494 (and I verified that it produces ~110kHz square waves as it should) ??

Also you can see 4 channels on this scope screenshots, although it had only 1st one hooked up. DS1054Z turns off channels that are not hooked when pressing "AUTO" trigger mode button, but this time it turned them on and showed same frequency waveforms. That tells me that I made one hell of an EMI source if an open BNC connector is enough to see this. At this point I feel like my experiments are doing more harm than good..
What am I doung wrong?

You can find PDF schematics in my nextcloud folder along side with setup photos:

https://cloud.plox.uk/s/B4WKkgK3bawjwtd

I understand photos are not suitable for debugging, should just give an idea of conditions. Basically all the components are soldered to each other; switch, diode, inductor and caps connections are kept short and their ground is a 1mm diameter inch long piece of copper wire. My load for unstable circuit consists of automotive lamps (4x 24V5W in series + 4x 12V10W in series) that draws a bit more current than my LEDs (~800mA); required voltage is a slightly less than sum forward voltage of 14 LEDs (more like 11-13, I believe) but still comparable.
Partnumbers not specified in schematics:
D1 MBRS360T3G
L1 SCDS125T-470M-N (47uH), B82477G4224M (220uH)
C1 TKR221M2AJ20
C2, C3 2.2uF X7R 10% 1812 100V (Taiwan)
C4 TKR102M2AK35
C9 GRM319R71H104K

[Leeroy]

Anyone? 

[hussamaldean]

get yourself this one
https://www.aliexpress.com/item/32639621677.html
this buck converter has CCCV modes which will help you to get started how to design
but if you are interested see this video from greatscott which explain how to add current limiting feature for buck-boost converter

[wraper]

I wonder why do you even bother with TL494 instead of modern specialized IC? It's not meant for buck converter as it's main purpose and is outdated. Not to say maximum rated voltage is 40V but you run it at 48V. And BJT circuit is crap because you get voltage drop of two transistors because of pass transistor arrangement. Not that there is any reason to use BJT to begin with as it's guaranteed to be suboptimal. You also can use buck IC with integrated switch like LM2592HV. Or better specialized buck LED driver like TPS92512HV.

[wraper]

BTW there is no good reason to use high side current sense.

[Leeroy]

Thank you for replies.

get yourself this one
https://www.aliexpress.com/item/32639621677.html
this buck converter has CCCV modes which will help you to get started how to design
but if you are interested see this video from greatscott which explain how to add current limiting feature for buck-boost converter

This video was the first thing that I started my research with. I can not use this converter because it's only 40V input, 35V output as any other converter on aliexpress that I've found. That would require me to use 24V AC-DC PSU and have 7 LEDs per string, so around 8 strings.. Also it is not guaranteed that 24V PSU won't have a fan as it should provide twice the current. I already got 48V PSU and it is fanless. These converters are not very efficient as well, I own a couple, they heat up quite a bit under load.

BTW there is no good reason to use high side current sense.

Well it is the recommended way in MAX4080 datasheet. It is also marketed there as an advantage:

High-side  current  monitoring  does  not  interfere  with  the  ground  path of the load being measured, making the MAX4080/MAX4081 particularly useful in a wide range of high-volt-age systems.

Not sure if it would make a difference for my application though. Should I think otherwise?

I wonder why do you even bother with TL494 instead of modern specialized IC? It's not meant for buck converter as it's main purpose and is outdated. Not to say maximum rated voltage is 40V but you run it at 48V. And BJT circuit is crap because you get voltage drop of two transistors because of pass transistor arrangement. Not that there is any reason to use BJT to begin with as it's guaranteed to be suboptimal. You also can use buck IC with integrated switch like LM2592HV. Or better specialized buck LED driver like TPS92512HV.

Two reasons:
1. I suck at finding the most suitable IC for the job
2. Modern and specialized stuff is not very common in local IC stores, thus 3-4 weeks of delivery, possible postal charges and much higher prices for low quantities that is if they are willing to sell to individuals.
This was initially the plan. I though, oh since I design circuit myself I'm gonna use  something modern and fancy, it will be efficient and so on.. Chose TPS54360, buried myself in its datasheet & theory and failed to even make it look alive before I burned it due to reverse polarity or something stupid like that because I got impatient during troubleshooting. That was the schematic I was building back then:
https://cloud.plox.uk/s/p4peJ3zDFyLQig9

Compensation circuitry totally threw me off, theory behind it was too complex for my understanding.
Than I tried LM2576HV which is the same thing as LM2592HV but operating at 52kHz, I believe. That worked, but distinct audible switching noise in the range of 2-3kHz was there when IC was operating in CC mode and I burned it again while trying to solve the problem. I was basically making wild guesses about the cause of the noise so there was zero confidence that I could overcome it.

But I've got to admit that TPS92512HV looks very nice since it has this IADJ pin and dimmable by PWM. It is not readily available at my place, but the IC is definitely noted. Then again that COMP requirement.. The only close things I found myself were LM3404 which uses some odd current setting resistor instead of current shunt, this is a showstopper for me. Moreover, judging by block diagram it is based on BJT anyway as LM2576HV is, and LM2592HV is..

Even though it is decades old, I quite liked TL494 because it is an ultimate set of building blocks for such a converter and every part is exposed for you to wire it as needed unlike specialized ICs where not only you have no control over it's theory of operation but you also have zero understanding how it really works under the hood since it is IP of the manufacturer and you only get that much info from functional block diagram. 40V max VCC is totally overlooked by me, have no idea how it survived ~50V since I was also experimenting with adjust pot of AC-DC PSU. That old stuff is indeed rugged. Now I'm wondering if it there is a chance for it to not function properly after that abuse..

How come do you think TL494 is not suitable for buck topology? Buck is literally its typical implementation in datasheet.
I just feel like I am wrong somewhere in my calculations & assumptions and not using the right passives or something like that.
May anyone tell me what are the possible causes for inductor to get hot in a buck topology? Googling only resulted in some cases with boost converters and that does not apply to my case, right?
Has it something to do with switching frequency?

[wraper]

It seems all of your problems were due to overthinking current feedback. Audible noise most likely was because of feedback loop oscillating at such frequency. Unless you really need negative output to be tied directly to GND, using high side current sense is simply stupid and waste of parts. Also you made feedback loop wrong, it's CV with CC slapped on top of it. But all it needs to be is CC.

[wraper]

BTW what the hell is this? It basically has infinite gain at DC / comparator. No wonder you had problems.

[wraper]

How come do you think TL494 is not suitable for buck topology? Buck is literally its typical implementation in datasheet.

It's suitable. But don't expect it to be best performer and it cannot drive MOSFET directly. In all of your circuits I see intention of overcomplicate circuit and resulting in something broken by design.

But I've got to admit that TPS92512HV looks very nice since it has this IADJ pin and dimmable by PWM. It is not readily available at my place, but the IC is definitely noted. Then again that COMP requirement.. The only close things I found myself were LM3404 which uses some odd current setting resistor instead of current shunt, this is a showstopper for me. Moreover, judging by block diagram it is based on BJT anyway as LM2576HV is, and LM2592HV is..

If you actually looked into datasheet, it's obvious it has MOSFET.

Then again that COMP requirement.

Dunno what you mean.

LM3404 which uses some odd current setting resistor instead of current shunt, this is a showstopper for me.

Again. As simple as it gets. Usual shunt resistor which is selected according to current needed. Usual feedback input, you can modify circuit if you want current to be adjustable.

[Leeroy]

Audible noise most likely was because of feedback loop oscillating at such frequency.

That's what I though, but was not smart enough to figure out the cause.

using high side current sense is simply stupid and waste of parts

I'm not sure how exactly I would implement current sense with less parts, I already have just a shunt and all-in-one amplifier. The only reason I went with MAX4080 and not discrete opamps is that it is still uncharted territory for me and having an IC here made sense since it is inexpensive for a non-mass produced circuit and I have one less thing to worry about. The reason I went straight with high side current sense is I have no idea if MAX4080 supports otherwise and never tested it myself. Are you saying that high side sensing can cause oscillations? I'm confused here.

BTW what the hell is this? It basically has infinite gain at DC / comparator. No wonder you had problems.

Replace DAC and C5 with trim pot and you get what was reverse engineered from this module:
https://www.aliexpress.com/item/1084552308.html

Module works and it is silent in CC mode.. Some Chinese voodoo.

Also you made feedback loop wrong, it's CV with CC slapped on top of it. But all it needs to be is CC.

I was thinking about it while taking that China approach with LM2576HV and as I recall it wouldn't work without CV feedback for some reason. Then I though that in case of open circuit instead of load CC would push output to the maximum since shunt is reading 0 and CV feedback acts as something like OVP. I do not know how this should really be done. TL494 even has separate error amplifiers for that.

How come do you think TL494 is not suitable for buck topology? Buck is literally its typical implementation in datasheet.

It's suitable. But don't expect it to be best performer and it cannot drive MOSFET directly.

As I stated earlier dynamic driving FETs is new to me. And I thought using gate driver IC would do the trick.

In all of your circuits I see intention of overcomplicate circuit and resulting in something broken by design.

Please keep in mind that I am only a couple of years in electronic circuits and have no electrical engineering education. I do not do this on purpose, it's a result of making things done with what I know. So saying things like "no wonder you had problems" does not give me suggestions for improvement, I still remain confused. Hope you understand.

If you actually looked into datasheet, it's obvious it has MOSFET.

I did. Was talking about LM2592HV

[wraper]

Replace DAC and C5 with trim pot and you get what was reverse engineered from this module:
https://www.aliexpress.com/item/1084552308.html

And what makes you think you can replace trim pot with capacitor and DAC? 

[wraper]

I'm not sure how exactly I would implement current sense with less parts, I already have just a shunt and all-in-one amplifier.

Just place shunt on low side. Just like in TPS92512HV circuit in datasheet.

[Leeroy]

And what makes you think you can replace trim pot with capacitor and DAC? 

The fact that it was just producing variable reference voltage from 5V reg via voltage divider.

Just place shunt on low side. Just like in TPS92512HV circuit in datasheet.

Will do.

[wraper]

And what makes you think you can replace trim pot with capacitor and DAC? 

The fact that it was just producing variable reference voltage from 5V reg via voltage divider.

Don't you still get that you put reference voltage into comparator which you made by omitting any DC feedback from opamp? It literally outputs logic ones and zeros with slew rate limited by 0,33uF capacitor and output impedance of DAC.

[Leeroy]

Again. As simple as it gets. Usual shunt resistor which is selected according to current needed. Usual feedback input, you can modify circuit if you want current to be adjustable.

I mean look at this formula:


This is ridiculous for a simple feature of current setting.
Sub 1Ohm resistors are only easily accessible here in standart some shunt values like R100, R050, R005 and so on..
Ok let's say I went through all the trouble to calc and obtain the correct resistor for my 750mA. Then some time later I need to assemble myself a similar circuit for a 350mA LED and I should go through all that again every time? Fuck that..

Don't you still get that you put reference voltage into comparator which you made by omitting any DC feedback from opamp? It literally outputs logic ones and zeros with slew rate limited by 0,33uF capacitor and output impedance of DAC.

Yes. And my understanding is that it is enough since when current sense voltage is higher than reference IC either lowers the Ton or skips cycles depending on its capabilities. I don't know how hard it affect hysteresis and if it actually introduces the oscillation and noise. But judging by the fact that aliexpress module is stable it is acceptable. Maybe I should verify that I re-engineered it 100% correctly, since it was long time ago and in a hurry.
Take a look at this thread:
https://www.eevblog.com/forum/beginners/adding-cc-to-a-cv-buck-to-make-it-cccv/

[wraper]

This is ridiculous for a simple feature of current setting.
Sub 1Ohm resistors are only easily accessible here in standart some shunt values like R100, R050, R005 and so on..

Nope, you are ridiculous by calling that ridiculous. It's completely normal and even preferable to select shunt resistance according to intended current. Because in that way you can minimize losses in shunt while keeping circuit stable and simple. And if there is no standard value, connect two resistors in parallel or series to get right value, oh my.

Ok let's say I went through all the trouble to calc and obtain the correct resistor for my 750mA. Then some time later I need to assemble myself a similar circuit for a 350mA LED and I should go through all that again every time? Fuck that..

On my, it's so difficult. For LM3404 calculations for standard circuit are more complicated because current through LED is not constant as there is no capacitor at all on the output (can be simplified as further if you can deal with a bit of imprecision by assuming current is constant). But if you add capacitor or use TPS92512HV calculations become extremely simple.  Also nobody prohibits complicating feedback circuit a little bit and make it adjustable. Though it that case it's a bit easier to use regular buck driver IC because of higher reference voltage. But it's not needed really as both LM3404 and TPS92512 have PWM input.

Take a look at this thread:
https://www.eevblog.com/forum/beginners/adding-cc-to-a-cv-buck-to-make-it-cccv/

That's a crappy workaround for adding crude current limit to CV circuit as already mentioned in that thread. And already mentioned it will be oscillating (you complained about noise).

[mariush]

I don't understand sometimes, are people allergic to "led driver ICs" ?

Just get some LED driver IC and control the brightness using a microcontroller (to send a pwm signal to chip)...

Get a couple LT3956 : https://www.digikey.com/product-detail/en/linear-technology-analog-devices/LT3956EUHE-TRPBF/LT3956EUHE-TRPBFCT-ND/8126002

Parallel two leds at a time and make a chain of 12-13 pairs ... 13 x 3.25v = 42.25v , 26 leds x 3w = 78w , 2 drivers x78w =156w
~12$ x 2 for chip and all components in the schematic below, all parts are available on Digikey

You can basically copy the application circuit from the datasheet, just adapt some resistor values if needed to account for the lower voltage:

[Leeroy]

Get a couple LT3956

This one is a total overkill. I won't have neither 4 layer PCB to properly trace this IC, nor solder mask to reliably solder QFN package. Not to mention that LT parts are the most rare in my country and price is not appealing at all.

On the other hand, I found the biggest problem with my circuit. It was a weak internal pullup resistor on the pin 2 (~HIN) of IR2102.
PWM signal was just above 1V peak to peak with -500mV offset... After adding 1k external pullup output frequency became as expected, inductor stopped overheating and Buck is now alive.



Moving current shunt to low side did not work. Seems like IC does not support this mode.

Switching noise is mostly gone. But in some narrow range of load current it comes back, see screenshot (ignore low frequency noise, it is a scope fan). This range lies somewhere near full load, just before CC mode is about transition in CV. Oddly enough oscillations go away before still in CC mode and not with the transition itself.



Silent gaps that you see on the screenshot are the result of random discovery:
While (and only while) my scope ground lead (which is not connected to the earth btw) is connected to source of the mosfet, if I touched or placed my finger near the exposed wires connected to some points (like pin 4 of TL494; "Vdiv" on my schematic; "5Vref"; probably GND too) oscillation was going away and noise too.
Any ideas how to stabilize?

[mariush]

Get a couple LT3956

This one is a total overkill. I won't have neither 4 layer PCB to properly trace this IC, nor solder mask to reliably solder QFN package. Not to mention that LT parts are the most rare in my country and price is not appealing at all.

Eh, you'd do it just fine with a 2 layer board, but fine... here's some simpler stuff...

AL8862 : <1$ each
Datasheet: https://www.diodes.com/assets/Datasheets/AL8862.pdf
Available in lots of places you could order from : https://www.findchips.com/search/AL8862
 
Up to 60v, up to 1A current , set maximum current through a resistor and then use a PWM signal or DC voltage to adjust output current up to set limit.  So you'd need to use one for each chain...

MP24833A : <1.2$ each
Datasheet: https://www.monolithicpower.com/pub/media/document/MP24833-A_r1.01.pdf
Available everywhere : https://www.findchips.com/search/MP24833  - also on eBay , 8-10$ for a 5 pack

up to 55v, up to 3A  (so you could have pairs of leds or maybe even groups of 3 leds)
same deal , enable, pwm or dc input to adjust current after setting maximum current through a resistor.

LED6000PHR  .. up to 61v 3A ...

and lots of others...

If you'd be willing to drop your maximum voltage to 24v, you'll find lots more chips that are cheap and simple to use. A lot of chips can handle maximum 36v or 40..42v.
24v power supplies are very common, and you could have 12-18-24 leds per driver easily (6 x (1..4) in series)


ex: PAM2863ECR (max 40v 2a) : https://www.diodes.com/assets/Datasheets/PAM2863.pdf  .. also on eBay something like 4..7$ for a 5 pack
 
ex : MP2840 ( max 36v 3A) : https://www.monolithicpower.com/pub/media/document/MP2480_r1.01.pdf   .. eBay has them as 10 for 8$, 50 for 20$ .. though who knows if they're genuine or counterfeit.

LED5000PHR (up to 48v 3A) (datasheet