Products > Test Equipment
DL24 Atorch type electronic loads: considerations on how to choose the best mosf
(1/6) > >>
Filippo52:
Mine is meant to be only a contribution from an enthusiast.
If we look in the forums and on Youtube there are many cases of Mosfets that sometimes explode in a pyrotechnic and dangerous way.
We neglect the case of FAKE components and limit the discussion to the original ones.
In products such as the DL24, in its various versions; powerful but cheap Mosfets are used: currently comes with IRPF264.
Enthusiasts who go to replace the mosfet often rely on the most visible parameters such as Amps and Watts supported. The voltage is related to the type of source to be tested.
If we want to give an example, the IRFP90N20D 200v 94A and 580W type il Mosfet is certainly considered better than the IRFP264 currently mounted in the DL24pcb and which has the following data 250v 38A and 280W.
In reality both of these Mosfets are intended for common use for MOSFETs, i.e. as switches and therefore in dynamic regime and not to act as a static resistor in an electronic load and this, wanting to simplify a lot, means that currents and dissipations are possible only for very short instants in which the MOSFET works in switching therefore in conditions where in most of the time the Mosfet is either closed or open and therefore dissipates very little.
In an electronic load this thing is not respected and the MOSFET is asked to behave like a resistor, returning to a state of partial conduction and therefore of dissipation for even very long times.
To evaluate these things in the Mosfet Datasheets, the manufacturer puts a graph called briefly "SOA" or in full "Forward bias Safe Operating Area" or even "Maximum Safe Operating Area" where there are curves remaining "under" which the manufacturer guarantees "survival" of the Mosfet under those indicated conditions.
Below is an image where there are both MOsfets we just talked about

The first IMPORTANT, indeed very important, thing to note is that the curve with the longest time is 10 msec, ie the manufacturer does not supply any data for continuous operation in "DC"; this makes it clear that we will not be able to have guarantees, quite the opposite.
I have traced some curves in red imagining a deterioration which generally exists with the increase of time, but it is only my hypothesis: no guarantee from the manufacturer.

We can make comparisons with the curve at 10msec but perhaps it is better to use my curve to get values which, even if not guaranteed, are more compatible with continuous use.

let's take a trivial voltage of 20 v the IRFP90n20D at the crossing on the red line will give about 8 A while the "less powerful" IRFP264 will give 20A; the same difference can be found using the manufacturer's official curves at 10 msec.
This means that the IRFP264 is more suitable for use, always improper let us remember, what we want to do with it even if it has lower general data.
But it's not exactly always true; for example at 4 volts (just the voltage of a lithium element) we see that the IRFP90n20D will carry 70A while the IRFP264 only 50 and if we use the official curves at 10 msec at 7v the IRFP90n20D will carry 200A while the IRFP264 will it will stop just above 100 A.
In short, the comparison is not as successful as looking at the main current and wattage specifications where the IRFP90n20D clearly doubles the IRFP264.
In practice, who can say with which voltage and current pairs it will be used and since the "DC" curve is invented, the result will be ...any. But we have found that the general data "alone" is not enough to find the best MOsfet and it is necessary to look at the SOA graph.


Let's take a step forward, armed with these considerations, let's look for Mosfets that have large currents, large wattages but also a nice "DC" curve in the SOA graph and we exaggerate to be clearer.
Here is a good example IXFB100N50P, a real monster with 500v 100A and 1890W of dissipation, with a large plus264 container and with an impressive weight of 10.28gr (IRFP90n20D or IRFP264 weigh just under 6g and incidentally their fake 4,4gr)
This "monster" as seen from the SOA graph also has the DC curve



and at 20 v we read 100A, in short we arrive almost at 2000W and we will be limited to 1890W by the dissipation.
let's take into account that the initial mosfets are still available and cost from 5 to 7 €, I'm always talking about the original ones; this monster is instead more difficult to find in this period (so beware of fakes) and costs between €25 and €30.
So is everything okay? do we have the solution? just pay?
Yes and no
There's still one thing that can escape and instead it's VERY IMPORTANT.
If you look at the SOA graph of the IXFB100N50P, if it is true that there is a DC curve BUT it is also true that the whole SOA graph refers to these two temperature conditions
Tj of 150°C and Tc of 25°C.
If you think about it for a moment in an electronic load and in a situation of continuous dissipation how do you keep the case at 25 °C?
It is practically impossible except by liquid cooling, and not like water but strong things like nitrogen and the like.
This happens because even the IXFB100N50P monster has not been designed to work in a linear way and therefore the manufacturer is not able to give guarantees if you put it in an electronic load with the case at 50°C for example, a rather normal situation instead.
Here is the last point.
There is a category of MOSFET called linear, born to do the work like of the electronic load.
Let's take an example here too
the IXTK110N20L2 less powerful than the previous one with "only" 200V 110A and 960W and weighing only 6.8 gr but which costs at least another 10€ more and is even less easy to find in this period.
here is its curve



Before you tell me the monster was better, look at the test conditions: Tj equals 150°C but Tc is 75°C. Here's the BIG difference is this; the "survival" of this Mosfet is guaranteed by the manufacturer in the real conditions in which an electronic load works and therefore this Mosfet if cooled well but with feasible things will be able to give a real 600W electronic load coupled with a heatsink of those for very high-end PCs powerful. We must remeber that il DL24 Atorch mosfet is not the only limit for 600W, many other things must be enforced like diode schottky , shunt resistors and also some parts of PCB

I don't know how clear and useful it was but I'll stop here, any insights on this question.

I apologize to those who already knew these things.



elecdonia:
Thank you for explaining this.

1)   It is critically important to select MOSFETs which are intended for “linear operation” where both current and voltage are >0 simultaneously. Vendors like Digikey have a “linear” parameter which can be selected when searching their inventory.

2)   The “safe operating area” (SOA) graph is an essential guide for selecting the type and quantity of MOSFET devices required to handle the maximum expected load current/voltage.

I’m beginning to think of alternative electronic load circuit designs which dissipate most of the power in a bank of resistors. This could take the form of a “high power” DAC with binary weighted load resistors. Each load resistor will have a MOSFET in series to select it. But these MOSFETs will always be entirely on or entirely off. Therefore standard low-cost switching MOSFETs can be used here.
There could also be one linear MOSFET used as a “vernier” for the LSB. This MOSFET won’t have a series resistor.
If the binary-weighted load resistor bank is properly scaled then this LSB linear MOSFET will dissipate only 10% (or less) of the total power.
Filippo52:
I welcome you to share my thoughts on SOA and linear MOSFETs.

Your idea of a resistor bank is very suggestive; but I fear that in the end it would be much more expensive than even the most expensive linear mosfet that solves all the problems with only one component
Silicium81:
Thank you for this informative article.
My DL24 electronic load has just seen one of its mosfets drop today in full use!
I was discharging a 48V/20Ah lithium-ion battery with a modest power of 300W (that's a 600W load made up of four mosfets/heatsinks). The mosfet of the main board is shorted after 2 hours of operation, the voltage was then 46V... The measured temperature of the heatsink was 58°C.
The mofets present were IRFP260M.
I'm thinking of replacing the mosfets with IRFP264s
The boxing:
Filippo52:
Dear Silicon81

Congratulations for the particular assembly of the 4 DL24 modules.
From what I see they are last year's models, now each module has 4 Mosfets and in theory they are more resistant, I mean in theory because they put 4 tiny heatsinks for each Mosfet.

The differences between the IRFP260M that you have on your modules and the IRFP264 that you would like to use are not relevant, there is an improvement but it is not very big and if we don't try to understand what happened you may not solve it.

Let's try together if you like.
I understood that you were discharging a 48V battery with a load of 300W (half of what the 4 DL24 modules should bear), if I'm not mistaken the current should have been 300W/48V= 6.25A.
These divided among the 4 modules are 1.56 A for each module and therefore for each mosfet.
The mosfets should have held.
I ask you
1 - you checked the temperature but I think your model has a sensor only on the main module; so the 58°C is only on the main module? Can you confirm?
2 - you know that the sensor is only close to the Mosfet and does not detect the true temperature of the component.
3 - have you had the opportunity to check if the current is divided equally between the four modules?
4 - have you checked the contact between the mosfet and the dissipator, the chinese guys sometimes make mistakes

Let's start discussing these things, let's try to understand what could have happened and then if you want to solve it by changing the Mosfet, do you have where to buy guaranteed original ones?

Since you're intervening, perhaps it's time to think of something a little better like a Mosfet so you don't think about it anymore.

But first it's important to understand today if your mosfets were all carrying the same load and if they were all cooled properly.

a greeting
Navigation
Message Index
Next page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod