Electronic Load Design

[Bootalito]

I've been spending the past couple of months on and off designing an electronic load.  I'd like to have some critique to finalize the schematics before I Start PCB layout.

Design features:
30V
5A
93Wmax calculated (will software limit to 70W)
12bitDAC
16bitADC
Single IRL3813 mosfet cooled by an LGA1156 CPU cooler
Nextion 3.5" display
ESP32 uC
     Fully fuctioned Control using javascript with WebSocket connections
     Full Size USB A connection with full control via command lines
SD card for data logging
RTC for accurate timestamps for SD card data logging.
Front panel to feature two sets of banana plugs for load and sense, as well as pluggable terminal block connections for load and sense. (See mocked up Front Panel Design PDF)
Design will include features similar to those displayed by the Rigol DL3021 in Daves teardown EEVBLOG #1023: CV, CC, CP, CR, battery tests, pulsed modes, transient, short, OCP test, OPP tests.

My intention is to get my feet wet with a first full design and build, to get the software features built up, then make a much better version with an 18bit DAC, 24bit ADC, with multiple fets with their own op amps and proper load balance resistors, adjustable slew rates, and rise and fall time calculations.  But who knows how far I'll get.

Front panel design to quickly get a sense of the features:
https://www.dropbox.com/s/mydfejg0uw35jbh/ElectronicLoadFrontPanelDesign.pdf?dl=1

A quick picture I took of the enclosure, CPU cooler, banana plugs, terminal blocks, and Nestion display to quickly get a good idea of the final build
https://www.dropbox.com/s/zhud5vt743f3bxk/ElectronicLoadEnclosure.jpg?dl=1

The schematics didn't fit on one page...
https://www.dropbox.com/s/khigjyvebx007x3/ElectronicLoadPowerSchem.pdf?dl=1
https://www.dropbox.com/s/ql23vm1aptpiimx/ElectronicLoadDigitalSchem.pdf?dl=1
https://www.dropbox.com/s/jcuk90zd9km4x5j/ElectronicLoadAnalogSchem.pdf?dl=1

Feedback is most welcome to help me avoid glaring design flaws and important features that are missing!

[llkiwi2006]

You might want to add a fuse in the main current path just in case your software limiting or some component fails. If the mosfet is turned fully on, you could end up with some burnt tracks.

[Bootalito]

You might want to add a fuse in the main current path just in case your software limiting or some component fails. If the mosfet is turned fully on, you could end up with some burnt tracks.

Thanks. I had that thought of using a Ptc. However, I'm going to leave off the solder mask and load on the solder by hand to vastly increase the current handling. The diodes are the only thing I worry about. I'm using some diodes I had on hand, but I may just go with a much larger to220 package and slap a heatsink on it. Also the shunt I'm using is 50W rated. So if I change out the diodes I'll bet I could push 30A+ safety 8n a failure.

[C]

If you want good time for loging and are using wifi then why not use NTP

[Bootalito]

If you want good time for loging and are using wifi then why not use NTP

If I want to turn this into a product, the end user may not want to enable the WiFi for security concerns or whatever.  At the end of the day I can always choose not to populate the PCB.  I've got a TON of room on the PCB.

[Phoenix]

70W is a fair bit on a TO-220 package. Considered a TO-247? E.g. IRFP3703 similar spec to yours.

Also you're stating 30V input - this is Vds rating of the MOSFET. If you want to go up to 30V you will want a higher rated MOSFET.

[C]

When it comes to WIFi, I would want good easy simple control.

Think of button  "WIFI Time Sink"
Leads to option question of changing WIFI Modes if needed.
If WIFI not in mode for NTP then if enable WIFI needed then enable WIFI, Time sink using NTP then WIFI back to previous state.

Best option is have ESP32 ask questions when needed.
Many ways to do something.

Might think about video for fast changing information, see some ESP32's on youtube doing this.

[Bootalito]

70W is a fair bit on a TO-220 package. Considered a TO-247? E.g. IRFP3703 similar spec to yours.

Also you're stating 30V input - this is Vds rating of the MOSFET. If you want to go up to 30V you will want a higher rated MOSFET.

Thanks, I know most TO-220 Fet's would be strained at 70W, but this is why I choose the IRL3818.
The IRL3818 has a 0.64 C/W j-c terminal resistance with a 150C max temperature.  So that's 234Wmax (@ 25C).  I want a max of 85C absolute, so this is a 60C delta from ambient. To calculate the Wmax with derating:  235Wmax - 60Cdelta *1.6 W/C derating factor (from data sheet) = 138Wmax(@ 85C) with derating.  I'm going to software clamp the temperature at 85C.  With a 60C delta this should be only 94W (60C delta / 0.64 C/W).  94W is below the 138W.  And as a good measure I'm leaving 20% margin.  94W *0.8 = 70W (rounded down).  I'll also be software clamping the power to 70W.  I also have to keep in mind that I'm using a CPU cooler that can probably only dissapate 90W anyway

Using only one TO-220 FET is not the greatest design, but I'm trying to keep it simple.  Most commercial designs use 6-10 TO-247 fets and are able to limit their temp to a cool 50C which obvisouly helps with longevity and reliability.

My next design will most likely feature a single fet that also, but a fet that is designed for the linear region such as a any of the hulking beasts that IXYS produces.  ref. IXFK320N17T2-ND (160V Vdss, 320A cont Id, a whopping 1670W dissapation...).  I'm guessing commercial products don't use these because they are expensive at $24 a piece.  But one of these'll do the business.

As far as the 30V goes, yeah I was thinkinga bout that.  Perhaps I'll limit it to 20V or something.  Or just spring for the IXYS part now since I only need one of them.

[Phoenix]

I think your thermal calculations are a bit off. Junction to case 0.64, but you missed typical case to heatsink (flat greased surface) the datasheet estimates 0.5. Then you assume your heatsink is actually able to keep the thermal connection point at ambient.Then the heatsink itself will have it's own thermal resistance to ambient characteristics I can't estimate.

Your junction at 70W is now +80degC from the heatsink temperature. Plus whatever the heatsink is at, plus ambient 25degC.

Sent from my G8441 using Tapatalk

[Phoenix]

Oh if you want a real whopping thermal package find an isotop. I've seen some with a square SOA!

Sent from my G8441 using Tapatalk

[blackdog]

Hi Bootalito,


First off all, read the datasheet!
Look at the SOA grafics of the IRL3813, There is no DC Line.
Do not Use this Mosfet for DC applications!

Use Mosfet who are designed for DC applications, you wil kill the IRL3813, even with water cooling!

Kind regarts,
Blackdog

[fcb]

I'd concentrate on getting your basic circuit working, get the thermals right, get the performance right.  Then I'd design the casing and do all the fun secondary stuff.

As others have pointed out, 70W from a TO220 is big ask - and you'll need to make sure you have superb cooling and test the s**t out of it.  Probably split the power across 2 or 3 FET's, simpler to cool as more surface area.

Even if your main control loop is digital, probably look at some sort of supervisor circuit as belt & braces.  It'll save you a few $ during development.

[gfolin]

It seems that there is a short between VIN and OUT on LDO1.

[Bootalito]

It seems that there is a short between VIN and OUT on LDO1.

Yeah I found that last night and correct. Thanks!

[duak]

If you're concerned about the current capacity of some PCB traces it can be better to lay some copper wire on top rather than a thick layer of solder.  Tin is the major constituent of solder and its resistivity is about 7X that of copper.  The AWG 14 wire in typical house wiring is a cheap source.

TO-220 packages are notoriously bad for meeting their thermal specs.  It's hard to get the clamping force except by pushing down right over the die.  Using just the mounting hole doesn't quite cut it.  If you look at processor packaging, you'll see they're quite a bit larger than the TO-220 tab for similar amounts of power.

Best o' luck

[Bootalito]

If you're concerned about the current capacity of some PCB traces it can be better to lay some copper wire on top rather than a thick layer of solder.  Tin is the major constituent of solder and its resistivity is about 7X that of copper.  The AWG 14 wire in typical house wiring is a cheap source.

TO-220 packages are notoriously bad for meeting their thermal specs.  It's hard to get the clamping force except by pushing down right over the die.  Using just the mounting hole doesn't quite cut it.  If you look at processor packaging, you'll see they're quite a bit larger than the TO-220 tab for similar amounts of power.

Best o' luck

Those are good points. I started laying out the Pcb last night and there is a big distance from the front to the mosfet. Perhaps some copper wiring harness would be better.
The irl3813 has specs similar to a larger TO247 packages, that's why I was going to use it. But there is a large difference between a cpu and a to220 package as it's being sandwiched. I think I'll still use the to220 and just deal with the limitations, even if it can only dissapate 40 or 50w. The next version I'll use larger heat sinks, larger multiple ic packages with proper current sharing circuits. Probably each mosfet will have its own op amp with a 5w 1ohm ballast resistor or something, with a single control op amp providing the feedback.

[Bootalito]

I'm thinking I should really use a larger package, and multiple like may of you have suggested, but I'm having trouble trying to source the type of heat sinks that are used in professional electronic loads such as the BK Precision and rigol, etc?  The type that seem to have the fins on the inside?
I've found these, which should work if I stand two of them on their side:
https://www.digikey.com/product-detail/en/wakefield-vette/511-12M/345-1203-ND/5068127

I've found this thing on banggood which shows the type of heat sink I'm looking for:
https://www.banggood.com/12V-60W-Thermoelectric-Peltier-Refrigeration-Cooling-Cooler-Fan-System-Heatsink-Kit-p-1203806.html?rmmds=search&cur_warehouse=CN

I've also found plenty of these which I could stack back to back:
A Really Huge AliExpress Link
But 69mm is too tall for my choosen enclosure (which ultimately could be my problem...

But I havn't found anything on ebay/AliExpress/Banggood that is really close to what I've seen in tear down videos

[schmitt trigger]

The very large heatsinks found in commercial products are generally custom-made.

If your project is one-off, the way that I would suggest is to pick up a defunct power supply from Ebay. The cost of the heatsink, case and power transformer alone justify its price. Plus there will be some extra goodies like meters and buss bars.
The only downside is that these tend to be quite heavy, and shipping costs may exceed the actual unit cost.

Edit: I see that you have already selected a case.
I'll join the chorus of other posters mentioning that you are attempting to drink the milk before milking the cow.
You might have to build a "dumb" prototype without bells and whistles to get the current and thermal handling correct.

Oh! and don't forget to check the closed loop response.

[Bootalito]

The very large heatsinks found in commercial products are generally custom-made.

If your project is one-off, the way that I would suggest is to pick up a defunct power supply from Ebay. The cost of the heatsink, case and power transformer alone justify its price. Plus there will be some extra goodies like meters and buss bars.
The only downside is that these tend to be quite heavy, and shipping costs may exceed the actual unit cost.

Edit: I see that you have already selected a case.
I'll join the chorus of other posters mentioning that you are attempting to drink the milk before milking the cow.
You might have to build a "dumb" prototype without bells and whistles to get the current and thermal handling correct.

Oh! and don't forget to check the closed loop response.

Thanks that's pretty much what 8 had thought... That they were custom made.

I often find it useful to select a case early in a project as it can assist in design decisions. Dave has alluded to this on many occasions, and most of the time you generally know the size you need. But 8j this case I probably should have started with the thermals, selected a heat sink and dissapation method, then select the case. I kind of did this by selecting a cpu cooler, it's just that for right now I can only fit one mosfet under it

I'm kind of thinking forward to the future to make this project diy able by others by using off the shelf products. Even if I have to lower the ratings a bit to 50w or something

[Kleinstein]

The  irl3813 is not at all suitable for a electronic load. It is not only that the DC curve is usually missing, but the SOA curve looks really bad - indications of thermal instability starting at 4 V.

For an electronic load one should more look at MOSFETs for a much higher voltage (like > 400 V). A possible candidate would be an IRFP250 - still not specified for DC from most manufacturers, but at least for the Fairchild version. You may not need the high voltage rating for normal operation, but the higher voltage FETs tend to have less trouble with thermal instability.

[Bootalito]

Unfortunately all of you were right .  I went back tot he drawing board and replaced my puny single IRL3813 with 6 IRFP250N with 5W 0.1R ballast resistors.  I found a suitable heatsink http://bit.ly/2nIC3ir.  I changed the to be similar to Kerry Wongs project where a separate op amp controls each mosfet via a 5W 0.1R ballast resistor. 
I was able to up the specs to 150V, 20A, and 200W.
Using an Rth junction to ambiant of 1.14 (0.7Rjc, 0.24Rcs, and 0.2 estimated thermal grease) I now have each mosfet dissipating 33.3W.  With a max of 7Amps/fet limited by the 5W 0.1R ballast resistor (@ 2.5W for safety margin)

Here is the updated shematic
ElectronicLoadSchematicV0.2pdf

If I have a max input voltage of 150V, what is the best way to select different voltage dividors?  I'm assumingg through an analog switch, but how does one switch fast enough before you blow an input?  I'm using zener diodes to clamp the voltage, but do I just switch as fast as the uC can sense that the voltage is maxed, then just hunt through a couple of ranges and just hope that the zeners can take the abuse for a couple seconds (or perhaps it will only take 100ms or so...)?

[viperidae]

Do you have enough gate drive voltage?
With 10A flooring through your main shunt it will drop 1V
The current sharing resistors will drop another 180mA, leaving the source pins of your MOSFETs at 1.18V. the max Vgs you can give them is 5 - 1.18 = 3.72V
That's below the max gate threshold.

Even worse for your max spec you want.

Why not use an opamp with a higher voltage rating. You have 12v supply.

[Bootalito]

Do you have enough gate drive voltage?
With 10A flooring through your main shunt it will drop 1V
The current sharing resistors will drop another 180mA, leaving the source pins of your MOSFETs at 1.18V. the max Vgs you can give them is 5 - 1.18 = 3.72V
That's below the max gate threshold.

Even worse for your max spec you want.

Why not use an opamp with a higher voltage rating. You have 12v supply.

Thanks for point that out, I didn't think about the voltage drop accross the ballast resistors raising the Vgs.  I'm actually going to a 120V design with a center tap transformer to get some good low noise split rails.  I've been toying around the idea of using different op amps: OP07 (cheaper) or NE5532 (cheap and high slew rate) to drive the mosfets

I've also changed the shunt to a 0.001R (CSS4J-4026) which necessitated adding an in amp which will feed the ADC (upped to an 18bit from 16bit) as well as feeding the feedback on the main driver op amp.  I've also increase the number of mosfets to 10 and added a secondary heat sink.  I've added a bridge rectifier so that the input voltage polarity can be ambiguous and changed the AFE for the voltage measurement.

What do you think of the following voltage and amperage measurement scheme.  Noting that the output of the in amp is also input to the main op amp controlling the current throught he now 0.001R main shunt:

[void_error]

Unfortunately all of you were right .  I went back tot he drawing board and replaced my puny single IRL3813 with 6 IRFP250N with 5W 0.1R ballast resistors.  I found a suitable heatsink http://bit.ly/2nIC3ir.  I changed the to be similar to Kerry Wongs project where a separate op amp controls each mosfet via a 5W 0.1R ballast resistor. 
I was able to up the specs to 150V, 20A, and 200W.
Using an Rth junction to ambiant of 1.14 (0.7Rjc, 0.24Rcs, and 0.2 estimated thermal grease) I now have each mosfet dissipating 33.3W.  With a max of 7Amps/fet limited by the 5W 0.1R ballast resistor (@ 2.5W for safety margin)

Have you applied the linear derating factor? The higher the junction temperature the less power your MOSFETs can dissipate.

[Bootalito]

Unfortunately all of you were right .  I went back tot he drawing board and replaced my puny single IRL3813 with 6 IRFP250N with 5W 0.1R ballast resistors.  I found a suitable heatsink http://bit.ly/2nIC3ir.  I changed the to be similar to Kerry Wongs project where a separate op amp controls each mosfet via a 5W 0.1R ballast resistor. 
I was able to up the specs to 150V, 20A, and 200W.
Using an Rth junction to ambiant of 1.14 (0.7Rjc, 0.24Rcs, and 0.2 estimated thermal grease) I now have each mosfet dissipating 33.3W.  With a max of 7Amps/fet limited by the 5W 0.1R ballast resistor (@ 2.5W for safety margin)

Have you applied the linear derating factor? The higher the junction temperature the less power your MOSFETs can dissipate.

Yup, all accounted for.  According to Kerry Wongs build the IRFP150's started to fail when they reached a dissapation between 45 and 60W (http://www.kerrywong.com/2016/10/08/linear-mosfet-and-its-use-in-electronic-load/).  However I'm using 10 fets for 200W.  My limiting factor is actually the heat sinks I choose.  2 X these:AliExpress listing.  they are 40mm x 40mm x 150mm (I choose a short case so I have that design constraint).  They are roughly 1/4 the volume of a single 80mmx80mm heatsinks used in rigol's, and bk precisions lower models.  I'll have two, so I should be able to just get away with 200W, albeit they may run a bit warmer than the nice cool 55C at full load I saw from one of Dave's teardowns (rigol or bk precision, can't remember).  Also basing this on my gut knowledge of how much mass and air flow goes through modern 60W, 90W, or 120W CPUs.

The FETs should run around 65C at full load which puts them at a 26A max per FET.  But each one will see no more than 7A.

Here is my ultra rough in on the PCB just to start seeing where I'll generally place things, in order to assist me with the design.  Note the power traces are just "ideas" of where and how I'll place them, but you can see the board output, the two ling heat sinks, and all of the mosfets. I've added a bridge rectifier on the input (using 4 x 60A TO-247 diodes), those aren't shown yet


Again..this is just me moving stuff around the foot-print seeing what goes where, mainly for the FETs and the shunt.  Just seeing where things fall in relationship to my front panel stuff