My constant current load - 90W , 3A 30V

[Spikee]

OLD first post in this topic:

Code: [Select]

[b]Schematic:[/b]

[url=https://dl.dropbox.com/u/100819959/Schematic%20Prints.pdf]https://dl.dropbox.com/u/100819959/Schematic%20Prints.pdf[/url]

[b]Info:[/b]
I use a Atmega328 running Arduino IDE to program the code.
I have an external rotary encoder to set the voltage to the dac ( 0-2048) and i instruct it to do a gain of 2.
So the 12bit dac MCP4728 outputs a set 0-4.096V .
The 12bit dac output goes to the first opamp (U4A).
The gain is set by a 15 turn 1k pot to achieve ~12V at 4.096V input.
R5 and C9 act as a low pass filter. I use S2 to switch between linear mode and pwm mode.

 The second opamp U4B is an regular voltage follower. The output has an pulldown resistor R4 (10K) to keep the gate of the mosfet low when needed.

J2 connects to the source of the mosfet for feedback. J1 connects to the gate.

S3 selects if A0 of the atmega is connected to measure voltage or not.
Because if if I > 5A than it wont like it. [[u]text in schematic is wrong[/u]]

The mosfet , 100W power resistor is mounted on a fan assisted heatsink.
U7 measures the temperature of the heatsink.


feedback is appreciated.
Spikee

Updates:

Update 1 : breadboard - https://www.eevblog.com/forum/projects-designs-and-technical-stuff/did-i-forget-anything-constant-current-load/msg166741/#msg166741

Update 2 : Measurements + photos - https://www.eevblog.com/forum/projects-designs-and-technical-stuff/did-i-forget-anything-constant-current-load/msg166933/#msg166933

Update 3 : Mosfet matching - https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/msg168787/#msg168787

Update 4 : status + pics - https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/msg168847/#msg168847

Update 5: made a youtube video for you guys:


Update 6: new youtube video , fixed opamp problem:


Update 7: new youtube video , testing 1ohm vs 0.1 ohm load resistor + running the load at 5V 10A.


Update 8: Big progress , 4 IRFP064N on a big CPU cooler running 63A peak + video
https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/msg195585/#msg195585

Update 9:
Schematic :
https://dl.dropbox.com/u/100819959/constant%20current%20load/Constant%20current%20EEVblog_Spikee.pdf

[Harvs]

Some thoughts:
-   The voltage regulators have no bulk capacitance following them. I would probably put a 10uf tantalum or 100uF electro on their output as well.
-   I don’t see a connection between the 12V bar on the op-amps and the output of the 12V reg.
-   I don’t see any bypass capacitance on the uC.  A 100nF ceramic cap close to the power pins will reduce noise.
-   You’ve connected both AVCC and VREF directly to the power pins.  At the very least AVCC and VREF should be decouple from the main supply noise through an LC circuit as shown in the ATmega328 datasheet.  You have a very hard time trying to keep digital noise out of the ADC as it is, so you need all the help you can get.  Further, if you are using a filtered 5V rail for the ADC just be aware that your accuracy is only going to be as good as your regulator.
-   I would swap D1 out for the fast UF4007, 1n4007 is generally too slow for switching applications. Cost difference is negligible.
-   The LM35CZ power source is best being powered by the filtered analogue power (reduce noise).
-   Instead of switching the input voltage from the analog pin, why don’t you just put a voltage divider (two resistors) between the input voltage and uC input?  A switch as a protection measure is almost certain to get you into trouble one day (e.g. a voltage spike.)
-   The gate driving op-amp needs some compensation to stop it becoming unstable.  If you have a look at my design posted a couple of days ago on here, you can see a resistor isolating the op-amp from the high mosfet gate capacitance and adding a capacitance between the op-amp output and the feedback path creating a dominant roll off in the op-amp response.
-   Can you talk more about what overall specs you’re trying to achieve, and I might be able to provide more feedback on the over parts and voltages in the design.  There’s a few areas that look like they might not do what you’re intending (but that depends on what your intentions are…)

[blackdog]

Hi Spikee,

I do not see were you sense your current...
Use a 0.1 Ohm resistor 3 a 5 Watt resistor in your Source.

I made al little schematic for you, its a part of my Electronic Load.


I hope this wil help you :-)

Kind regarts,
Blackdog

[Spikee]

I sense the current using connector j2 (who is connected like your schematic on a big heatsink off the pcb) .

I use the NXP buk652R7-30c fet 30V max  200W @25 deg. Celsius.
My Resistor is : Ohmite TEH100 W 1ohm
it is mounted on a ELV LK-40 heatsink capable of 100W when assisted whit the normal fan. 0,73 K / W @ 6.5 CFM
I bought a new fan that does 24 cfm at max speed (delta FFB0412SHN ) 6W 13000 rpm.

My specs are just to try some supply's of mine. My main Agilent psu does 30V @ 3A and it would be nice to fully test it.
My design got some inspiration from jjoster and dave.

The test pcb is beeing manufactured as you speak [for free ] i will find / fix the problems after it gets back (a week or so) .



U7 is actually on the heatsink, not on the board.
U1,2 are on a heatsink , i just didn't make it 3d.

Reply's :

I added the switch (S3) because i rarely would be going over 5 amps.
And i will replace the output cap of the voltage regulator whit a bigger one.
The LM36 just gives a ballpark figure of the temperature. I don't need it to be super exact.
The opamp (one lm324N ) is connected to the top rail of 12V and bottom rail of GND.
I will replace the 1n4007 whit  a faster equivalent . I will test if the gate capacitance gives me a hard time. If so i will solve it accordingly (bodge  it) . The ADC could give me a hard time because of noise but i will have to experience it first hand and solve my mistake than.

[blackdog]

Hi Spikee,

If you use a 1 Ohm resistor and the design is for 5A, you wil lose 25 Wats in this resistor!
And, the minimum voltage you can use 5A wil be around 6V.
Also wil the opamp U4B have not enough drive voor 5A and 1 Ohm resistor.
The drivelevel wil be: Voltage drop Sensresistor, Wires, Gatedrive voltage.
This wil be around 10V.
The output commonmode voltage of the LM324 on the + site is around 3V.

Kind regarts,
Blackdog

[Spikee]

i can replace the LM7812 whit its 15v equivalent if it makes it any better. The max voltage i can get out the transformer is 23V .
the current won't exceed 3A 99% of the time. i just had the heatsink laying around.

[Harvs]

I use the NXP buk652R7-30c fet 30V max  200W @25 deg. Celsius.

So you want to test a 30V power supply, yet the absolute max Vds of that MOSFET is 30V...  bad idea.  Find a MOSFET with a max Vds of >1.5x that.

My Resistor is : Ohmite TEH100 W 1ohm
it is mounted on a ELV LK-40 heatsink capable of 100W when assisted whit the normal fan. 0,73 K / W @ 6.5 CFM
I bought a new fan that does 24 cfm at max speed (delta FFB0412SHN ) 6W 13000 rpm.

So the heatsink is "capable" of 100W. I can't find it on the web but I'm guessing by that they mean at an ambient temp of 25C, you're heatsink will be near on 100C (100*0.73+25), and if you go any hotter the paint will have problems.  If you want to understand why this is a bad idea, do some reading on thermal resistance for devices (junction to case, case to heatsink etc.)

My specs are just to try some supply's of mine. My main Agilent psu does 30V @ 3A and it would be nice to fully test it.

As above, if that heatsink has a 0.73 degrees/W thermal resistance, you don't have enough dissipation for 90W.  The idea that you plan on doing it with a single MOSFET is really pushing shit up hill.

The test pcb is beeing manufactured as you speak [for free ] i will find / fix the problems after it gets back (a week or so) .



U7 is actually on the heatsink, not on the board.
U1,2 are on a heatsink , i just didn't make it 3d.

Reply's :

I added the switch (S3) because i rarely would be going over 5 amps.
And i will replace the output cap of the voltage regulator whit a bigger one.
The LM36 just gives a ballpark figure of the temperature. I don't need it to be super exact.
The opamp (one lm324N ) is connected to the top rail of 12V and bottom rail of GND.
I will replace the 1n4007 whit  a faster equivalent . I will test if the gate capacitance gives me a hard time. If so i will solve it accordingly (bodge  it) . The ADC could give me a hard time because of noise but i will have to experience it first hand and solve my mistake than.

To be honest I don't understand the mentality of putting a PCB out for production with what I consider basic errors (decoupling capacitance and ADC noise reduction are poking you in face if you read the datasheet.)  If you wanted to experience these errors for yourself then by all means do so, but why wouldn't you just do it on a breadboard before you had a PCB messed up?  At a guess, I get the feeling you were under a time pressure to get the board out so you could add it on a panel going for production free of charge.  But then if you didn’t have time to fix up errors instead of resorting to bodging it up, why bother asking us for input?

Edit: Appologies for sounding negative...

[Spikee]

The pcb is a sample of a new pcb manufacturer i'm trying. So i needed something out fast to test the service .
I am using a fan who pushes around 4 times the air of the normal fan (100w rating of heatsink) .
Because i don't have the exact numbers i will need to test it out myself.
I know that the mosfet is rated for 30V max  and that the heatsink might not provide enough cooling capacity for 90W continious.
But peek value will suffice for pwm mode.

I agree that i should have made the complete circuit on a breadboard before getting a pcb built. But i just needed to test the service ,  the final pcb will be for another day. And i ordered a few extra breadboards because i ran out of room.

Some background info:
I'm a Student in the first year of college. I have had no electronics classes or anything like that. This is just hobby.

Thanks for the feedback , i will adjust my circuit and test it next week.
gr. Spikee

[Harvs]

Fair enough mate. Just keep a list somewhere of stuff you need to change. It's always difficult to remember when you come back to it.

[kaindub]

What is the point of the Atmel?
Since there is no feedback to the uP, you are just using it to chnage from digital setpoint to anaolog.
get rid of the uP and just use a potdirect to the opamps.

[Spikee]

What is the point of the Atmel?
Since there is no feedback to the uP, you are just using it to chnage from digital setpoint to anaolog.
get rid of the uP and just use a potdirect to the opamps.

The micro measures the heatsink temperature (can be programmed to shut off the output at overload) .
And adjust the pwm to the heatsink fan. Because 13000 rpm continuously will not be fun.
It also has an lcd to display the set current , measures current , heatsink temperature , pwm value etc ...

Also a 10 turn pot will be more expensive than a 2 euro micro .

[Spikee]

how much current does the adc use ? i looked at the Electrical characteristics but didn't find much.
I was thinking of putting a low pass rc / lc filter after the LM7812 and putting a precision 5v reference to power the adc and use it as the analog reference. Like the LT1027 precision 5v reference . It can source about 15mA so it should be enough. I will need to add some filter caps but i have to look what value.


Thanks,

Spikee

[T4P]

Scary ... a sunon high dB turbofan by any chance?

[Spikee]

Scary ... a sunon high dB turbofan by any chance?

Old fan was a sunon slow speed . New one is delta FFB0412SHN around 55dB at 13000 rpm.

[Spikee]

i have now made my circuit on a breadboard.
some test results:

25V 1A
fan at 5V
heatsink temp ~35-38 deg C
room temp 21-22 deg C
heatsink Rth (j-a)   ~ 0.56 - 0.68 deg. C/W
Without fan ~53 deg. C

25V 1A [short test cuz of noise ]
fan at 10V
heatsink temp ~30 deg C
room temp 21-22 deg C
heatsink Rth (j-a)   ~ 0.36 deg. C/W

25V 2A [short test cuz of noise ]
fan at 10V
heatsink temp ~40 deg C
room temp 21-22 deg C
heatsink Rth (j-a)   ~ 0.36 deg. C/W

little heatsink can hadle alot .
The mosfet is dead now . Probably because it could not transfer enough heat fast enough.
I have two Buz 11 mosfets here (50V 33A 90W max @ 25 deg. C ) and i will give them both a opamp channel and parallel the outputs.
I will post the result later. The pcb should arrive in a few days.

gr. Spikee

Problems:

1. Changing arduino pwm timers  messes up the rest of the program . - have to do some extra research. (default you hear the harmonics)
2. Fan gave noise to the temp sensor so i tried an rc circuit. It didn't solve it so i hardwired the fan to 5V rail and it was gone. I will have to find an better solution for the final pcb.
3. Blew up mosfet because it could not transfer the power fast enough. I have two Buz 11 mosfets and i will parallel the output of them to share the power.
4.  The current reading was a bit fiddly. I will have to change a bit in my Arduino code.
5. The current setting whit the rotary encoder would act weird after some time. I will change the 100nF caps whit 10nF tomorrow. Maybe it will fix it.

Thats kinda it for today , i will try some new things tomorrow.

gr. spikee

[Spikee]

I made some upgrades , the one nxp fet has been replaced by two BUZ 11 whit gate , drain , source connected together.

some pictures:

setup:
https://dl.dropbox.com/u/100819959/2012-11-28%2023.00.45.jpg

Display:
https://dl.dropbox.com/u/100819959/2012-11-28%2023.01.06.jpg

Heatsink:
https://dl.dropbox.com/u/100819959/2012-11-28%2023.01.17.jpg

Power supply (Agilent U8001A) and mosfet temp. measurement (Rigol DM3058) :
https://dl.dropbox.com/u/100819959/2012-11-28%2023.03.05.jpg


Some more tests:

15V 3A - 45W  fan at 5V:

duration: 10 minutes
heatsink temp : 50 deg. C
mosfet (2) temp : 55 deg. C
room temp: 21 deg. C
heatsink Rth (j-a) ~ 0.73 C/W


20V 3A - 60W  fan at 5V:

duration: 15 minutes
heatsink temp : 59.5 deg. C
mosfet (2) temp : 65 deg. C (still self balancing)
room temp: 21 deg. C
heatsink Rth (j-a) ~ 0.73 C/W


Mosfet (2) failed at 60W 70 deg c. on package 60 deg c. heatsink. Because the the second fet failed the first one also did.
The balancing was not good , mosfet (2) was 70 deg. c while mosfet (1) was 60 deg. c  .
I could ad balance resistors but i deoubt that it will fix much.

The best solution would be one big TO264 package or so.
I will need some help to find the good one.

gr.spikee

[peter.mitchell]

did you use some thermal paste/grease or thermal pad? that would really help/hinder, also, make sure you tighten them on sufficiently, bigger packages (to-247 and to-264) will help a bit, but 60W should be fine with a TO-220, provided the heatsink is cool enough and thermal conductivity is high enough.

[mrflibble]

I made some upgrades , the one nxp fet has been replaced by two BUZ 11 whit gate , drain , source connected together.

You can parallel those mosfets, but connecting it all together like that is not the best plan. ;-)

At the very least use seperate gate resistors. See for example: http://www.irf.com/technical-info/appnotes/an-941.pdf

Anyways, maybe parallelling is more trouble than it's worth. If you have enough cooling + a decent thermal contact then 1 device should be able to survive that 60 W.

[blackdog]

Hi Spikee

You are making mistake's...
Place the 2 powerfets in the middle of your heatsink, thats the place were your vancontroler sensor is.

Mount the powrfets on the other side of the heatsink, that is turn is 90C.
NOT on the edge were you coupled the heatsinkparts togetter...
One on each side.
If you do it this way your heatsink wil be ~0.73c/Watt
Be shure the Fet is mounted flat en use a good compound!

Use a small Source resistor say 0.1 Ohm, use a current of 1A en measurethe voltagedrop over these resistors.
If the voltage is within 15% its OK, other pair some fets or use a little higher source resistor.

Mount the gateresistor direct on the Fet.
Keep those wires short!

Kind regarts,
Blackdog

PS (Dutch Mode)
Als je ook op het forum van "Circuits Online" zit kan ik je wat beter helpen.
Engels is nog steeds knap lastig voor mij :-)

[Spikee]

The mosfets were mounted this way: bolt - plastic washer - mosfet - thermal grease - mica plate - thermal grease - heatsink .
I read in some study that it was the best way. After a second look i saw that i could move mosfet (2) , thats probably the reason that one got so much hotter than the other one.

The mosfets i have laying around:

5x MTP3055 - 50V 12A ~48W
7X IRF510 - 100V 5.6A ~43W
30x IRLB8721 - 30V 44A  ~ 65W @ 25 deg. C    33W @ 100 deg. C (logic level)

@blackdog ,
Ik zit ook op Circuits online ( Spikee ) maar heb daar momenteel geen topic lopen.

I will try to match a pair of those MTP3055 . What amount of gate resistor should i be looking for ?
And i will mount the mosfets according to your post.

Thanks for all the info and feeback.
Spikee

[Harvs]

Without an insulator and using Arctic Silver thermal paste, I had 60W per TO220 mosfet being transferred into a CPU cooler (120W total), not getting above about 60 degrees C.  But that's with a heatpipe cooler, I couldn't get close to that result with a standard alu heatsink.

When paralleling mosfets I used two control loops.  Tied the drains together to the input, then had separate load sense resistors and op-amps. If you just hook them together like you have, then you're going to struggle to get equal current sharing.  Sense resistors and op amps are cheap...

[Spikee]

I prefer one big TO-264 package n channel mosfet. I saw some high power ones from IXYS at farnell at a reasonable price.
I do not mind paying around 5-9 euro for one of those mosfets. The bigger packaging also is nicer to solder to.

Now i need to find one that is suited for my application.

Gate drive voltage : 0 -> 12V
Voltage rating: from 75V and up
Current rating: 10A and up
Watt rating: 100W and up

[blackdog]

Hi Spikee

Buy 5 a 10 of these Fets:
http://www.dickbest.nl/index.php?_a=viewProd&productId=2164

Select the fet's, see picture


Kind regarts,
Blackdog

[Harvs]

I prefer one big TO-264 package n channel mosfet. I saw some high power ones from IXYS at farnell at a reasonable price.
I do not mind paying around 5-9 euro for one of those mosfets. The bigger packaging also is nicer to solder to.

Now i need to find one that is suited for my application.

Gate drive voltage : 0 -> 12V
Voltage rating: from 75V and up
Current rating: 10A and up
Watt rating: 100W and up

None of that is particularly difficult.  The cheapest TO-264 mosfet on digikey, IXFK230N20T, could easily do it.

With the junction to case and typical case to heatsink figures you're only looking at a 24 degree difference at 100W from junction to heatsink with a 175 degree max temp.  Your difficulty will just be to keep the localized heating in your heatsink under control.  That's where heatpipes are king.

[Spikee]

I can't get the IXFK230N20T at farnell so that would be a bit of a problem.
To prevent localized heating multiple fets would be the best solution.
I have ordered the fets blackdog suggested and i will use his method to match a pair of them.

They will arrive tomorrow or saturday . I will test them on sunday / monday .