Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Dummy Load II

Status
Not open for further replies.
Not really convenient, sorry.

They are all now mounted on the heat sinks and clamped up with a solid Al angle......

I was wondering how you drilled/tapped a heat pipe sink. They usually don't leave any room for that sort of thing.:)
 
jj, from your description it sounds like we might get by with three FDL100s @ 500W each if the water cooler has sufficiently low thermal resistance (and from observed data it appears to be around 0.025) and with four FDL100s @ 375W each we should be in pretty comfortable territory. Is that correct?
 
Very carefully. :)

There was just enough clearance between the Cu pipes on my particular cooler to get a 2mm drill through and the thermocouple bead just made it.
The Cu tubes were reasonably thick walled and I could see there was probably a bit of slack from the flats ground on the base. In the end no Cu came out of the drill hole so i think I got it exactly down the middle of the gap.
I was wondering how you drilled/tapped a heat pipe sink. They usually don't leave any room for that sort of thing.:)
 

Attachments

  • Screen Shot 2013-03-02 at 4.31.07 PM.png
    Screen Shot 2013-03-02 at 4.31.07 PM.png
    34.9 KB · Views: 180
That's what I'm thinking but I am still testing this rig and, as usual, not everything is plain sailing. I have popped a couple of FETs and we don't know whether they were killed by stupidity (likely) or are in fact under-spec parts (entirely possible as these did not come from a traceable supplier).

Also yesterday after a few tidy ups on the control board the same FETs that did over 700 Watts were looking like they were uncomfortable at 600. There is a way to go as I need to properly isolate my three loads and test them separately. We may have some FET weirdness going on in the unpowered loads.

In short I think yes, but I'm not 100% sure at this time. I would hate for people to burn out these expensive FETs without checking carefully for themselves.

The advantage of the water cooled solution is that if you get it sorted for one FET I think you can just add more and keep the same rating because the heat sink itself is so effective.

John
jj, from your description it sounds like we might get by with three FDL100s @ 500W each if the water cooler has sufficiently low thermal resistance (and from observed data it appears to be around 0.025) and with four FDL100s @ 375W each we should be in pretty comfortable territory. Is that correct?
 
Good industrial design means running a semiconductor continuously at far less than it's absolute max die temperature.

OK, in this case the dissipation is relatively constant and slow to change so you don't have to worry about thermal shock like in some switching apps, but you would still want to keep the die temp well below 125'C, at maybe 100'C.

I'm not sure of the design philosophy that says "OK it seems we can run 3 or 4 of these and JUST get away with it". Is this a price issue? Is it good to build a time bomb that is hard to repair (parts soldered on heatsink) simply to save $20 or $30? What are the priorities?

jj604 said:
...
So at what seems a reasonable value of Rcs for this package, it is just over the allowable die temperature (estimated 155C) at 501 Watts on an air cooled load with a very good Rsa of 0.065. ...

Running at estimated 155'C die temp? How many hours did you run it for at that dissipation? Or was that one of those tests where because it worked for 30 seconds of testing we say "see, it can do this"?
 
I can't speak to the original design goals but the resultant design of 10 TO-220 packages soldered to a copper water pipe suggests maximum cooling at minimum cost. Two goals of the redesign effort are to increase reliability and bump up the power a bit. To that end, it has been suggested that we lower the dissipation in each FET by increasing the number of FETs to 15. Soldering 15 FETs to a pipe seems troublesome so the last couple of pages of posts have been discussing the possibility of using fewer, higher power parts (~ 8 times higher power). We are trying to understand just how much power the new part can safely tolerate so we can dial it back to a reliable level. No one is suggesting we design something that "JUST gets away with it". With what we know, it appears that using four of these will result in a junction temperature around 85 degrees but that number is not set in concrete; it’s just a point for discussion. At this point, there hasn’t even been a decision to use the part, let alone how many to use. That will likely come down to a cost/benefit choice.

Would a water block or exotic air cooler be better than soldering the FETs to a water pipe? Perhaps, but unless the OP is willing to absorb the added cost there’s not much point in discussing it. I believe that was all thrashed out in the original project and the current design was settled on for good reason. If the people involved in the original project want to re-visit it, I’m content to leave it to them to bring up.
 
Perfectly sound (if completely negative) comments:

In answer:

1) What has good industrial design got to do with it? My interest is for use in a hobby involving toy planes. As others have pointed out many times on other forums, we do with our toys what would not be acceptable in a commercial environment where you very properly run things at well under their specified ratings and build in proper factors of safety. I'm pretty sure this project was to see if a decent load could be built with reasonable chance of working reliably at a low cost. That is certainly the aim with mine. Only I will use it and I know what its limitations are. It is not commercial quality with all that implies. If I want industrial reliability from a 2000W+ variable load I would go out and pay several $k for one.

2) On design philosophy, I can't answer for this project, but for ours I can. All three people involved are trained Engineers, two have extensive experience in designing Mil Spec power supply equipment (not me - I'm the light relief).

We all understand exactly what are setting out to do.

3) Regarding

Or was that one of those tests where because it worked for 30 seconds of testing we say "see, it can do this"?

An excellent point but, No.

I have run an aircooled load at the power specified for 15 minutes on several occasions. That is longer than any battery test I will want to perform on this load which is designed to test at 20C and above rates (<3 minutes). There is no design requirement for it to run for "several hours" although that would be a reassuring test to have.

Anyway it's not my thread or project and I will butt out. Jocanon can state his own position. :)

Good industrial design means running a semiconductor continuously at far less than it's absolute max die temperature.

OK, in this case the dissipation is relatively constant and slow to change so you don't have to worry about thermal shock like in some switching apps, but you would still want to keep the die temp well below 125'C, at maybe 100'C.

I'm not sure of the design philosophy that says "OK it seems we can run 3 or 4 of these and JUST get away with it". Is this a price issue? Is it good to build a time bomb that is hard to repair (parts soldered on heatsink) simply to save $20 or $30? What are the priorities?



Running at estimated 155'C die temp? How many hours did you run it for at that dissipation? Or was that one of those tests where because it worked for 30 seconds of testing we say "see, it can do this"?
 
Here is my 2 cents:
I worry because we don't have real good data on the thermal resistance of case to pipe or pipe to water. :D
My thinking has been case to heatsink is 0 because it is soldered. But that may not be true because tin lead is nowhere near copper.
The case for the heatsink is probably not quite as big of problem.
So looking at the Key parameters for both FETs:




Current New
Qty 15 $42 4 $68
W/FET 100 375
T max. 175 150
Jc = .5 .05
Cs= .1 .1
Sa= .025 .025


Tot. .625 .175

C rise 62.5 65.6 Maybe not to bad. But, if we double the unknowns.

Total .75C/W .3C/W

Crise 75 112.5

Tj 100 - 57% 137.5 - 92%

Seems a little risky to me unless we use more of the big guys, but then we kind of defeat the purpose.
 
Could have sworn they were $17 a few days ago. Maybe they know we are looking.:rolleyes:

Anybody see a water cooled heat sink we could use?
 
Dusey52 said:
I believe [the heat sinking] was all thrashed out in the original project and the current design was settled on for good reason. If the people involved in the original project want to re-visit it, I’m content to leave it to them to bring up.

I'm one of the originators, and I really don't see a reason to change the cooling part of the project. Cooling doesn't appear to even remotely be our problem. The water pipe method is cheep, it's effective, and the OP has all the parts, knowledge, and tools needed to do it... nuff said really.

Dusey52 said:
[less, but more powerful FETs] also elminate several $5-$6 op amps...

When the heck did the Op-Amps get all the way up to $5-$6 each? The 1$ TLV272IP is perfectly acceptable IMHO. Or did I miss something??? Not upset obviously, just confused...

https://www.digikey.com/product-detail/en/TLV272IP/296-12671-5-ND/454247

https://www.mouser.com/ProductDetai...=sGAEpiMZZMtCHixnSjNA6Dh9CQmOfcFc0hzMUdzRuvg=

jj604 said:
1) What has good industrial design got to do with it? My interest is for use in a hobby involving toy planes. As others have pointed out many times on other forums, we do with our toys what would not be acceptable in a commercial environment where you very properly run things at well under their specified ratings and build in proper factors of safety. I'm pretty sure this project was to see if a decent load could be built with reasonable chance of working reliably at a low cost. That is certainly the aim with mine. Only I will use it and I know what its limitations are. It is not commercial quality with all that implies. If I want industrial reliability from a 2000W+ variable load I would go out and pay several $k for one.

If I may intervene, lets not forget that this project and thread is jocanon's. Having your own motivations and project is wonderful, and you are more than welcome to learn off of this project and thread, as well as add your input if you think it will help. But you should also try to be extra careful not to change or confuse our objectives with your own. "Thread jacking" is quite against the rules here, and the above quoted comments seem dangerously close to crossing that line.

Jocanon wants something that can load up a server power supply for several hours to prove that they are still working and are reliable at the power level he will sell them for. He wants to do this with the best bang for the buck, as anyone should.

As for what industrial design has to do with it, robust design is common sense for power electronics and for tools as far as I know. We may not need to go overboard, he may be able to get away with things that are consumer level. But there is also no compelling reason why we can't go the extra mile. Cost? The cost is a non issue. The power handling per unit is going to be offset by the number of parts we will need to have to meet total power, this has the effect of making the total cost be within 10~20$ of each other no matter what we do. If anything, we get better cost per Watt ratios if we buy MORE transistors, because we get the better price breaks for buying in bulk.

The real issue is, at any cost, what is going to handle the most power? More low power parts, or fewer parts with better power per unit?

Everyone seems to be divided on this because some of us think the datasheet power rating is the limiting factor, and some of us think the power handling of the case is the limiting factor. I am personally on the fence. I have seen a transistor destroy it's lead and still be otherwise functional. So at some power level, you do have to take the packaging into account. And I think we are in just such an area to be sure. But I also know that the first dummy load incarnation worked fine at what should be higher than the package rated power. And I can't deny all the rest of the data every one else has provided. Yours included.

When I'm on the fence, I always try to take the safest route. And severely underestimating part power handling is always safer than even slightly overestimating it.

ronv said:
I worry because we don't have real good data on the thermal resistance of case to pipe or pipe to water. My thinking has been case to heatsink is 0 because it is soldered. But that may not be true because tin lead is nowhere near copper

I wouldn't worry about that, It is really really low. I would call it near 0 also. At the very least it is not the limiting factor. The solder may be lower than copper, but it is still a metal, so should be much higher than the crystalline silicon the die is made of.

ronv said:
Seems a little risky to me unless we use more of the big guys, but then we kind of defeat the purpose.

I agree, just because the bigger parts are better, doesn't mean they are better per dollar. And in the end, I am more able to trust the power handling of of a larger number of smaller parts, as it spreads the heat out. As it has long been decided on, the main enemy of the transistor is heat at the parts junction.


== Page Break ==


Not that everyone's input isn't welcome, but I have had enough of the transistor debate. It's gone on far too long. So here is my ultimate decision.

I personally think the actual parts can take more than what the case would typically be rated for. I think this, because the cooling is unprecedented, and the first incarnation worked fine while running with just such higher power levels. However, I don't think they can run all the way up to some of the numbers we are seeing in the datasheets. 2000 watts per FET is an an absolutely ridiculous exaggeration.

Not that it makes me the boss or anything, but as one of the originators of the project, I'm putting my foot down on the issue so we can move forward. For this project, the parts will be rated at 2X whatever the package is rated to handle, no more, no less (provided we don't exceed SOA and claimed power). I think this is an excellent compromise that satisfies both parties positions. Moderators, Ronv, and jocanon are the only people that can change my mind at this point. As Ronv is the one that is doing the legwork on designing everything else, so takes a large part of the responsibility if it should fail. And jocanon is the one who has to use the thing, and as such takes all the risk. And moderators are... of course moderators... and have the ultimate say in everything.

Part ratings should thus be...
TO-220 = 50 X 2 = 100 Watts
TO-3P = 125 X 2 = 250 Watts
TO-247 = 125 x 2 = 250 Watts
TO3 = 200 X 2 = 400 Watts

If anyone disagrees, and wants to debate the issue farther, you can always start a new thread. Debating it more in this thread is only going to be prohibitive to the project, and helps no one. We can deal with the consequences of all this if and when they present themselves. But for now, I want to move forward with the project rather than be stuck on the transistors.

Ronv, are you at a point that you can build a BOM now with the above assumed part ratings? I still think 10 X TO-3P's or 10 X TO-247's is going to be the best all around choice. But I will leave that decision in your hands.

Jocanon, be aware that the device could mysteriously fail prematurely, so take the proper safety precautions when operating it, and be prepared to need spare parts.
 
Last edited:
Absolutely, I said I would butt out, and I will, but just to make it clear - I wasn't trying to take over this thread although in hindsight I can see how you could interpret it this way. :)

Jocanon and I had some correspondence off line earlier and he was really interested in our project but it was early days and I wasn't confident of the data. I originally posted here to add to the information pool but I got a bit taken aback by the negative posting about "you guys don't know what you are doing" when clearly this had been a productive and cooperative thread of the best kind.

Apologies - it wasn't my business and I should have stopped one post earlier and not reacted.
If I may intervene, lets not forget that this project and thread is jocanon's. Having your own motivations and project is wonderful, and you are more than welcome to learn off of this project and thread, as well as add your input if you think it will help. But you should also try to be extra careful not to change or confuse our objectives with your own. "Thread jacking" is quite against the rules here, and the above quoted comments seem dangerously close to crossing that line. .
 
I got a bit taken aback by the negative posting about "you guys don't know what you are doing".

That's perfectly understandable.

I have seen and made some harsh comments with people on this forum before myself. All I have to say about this particular incident is that Mr RB certainly knows what he is doing more so than most. And that I am sure he wasn't trying to rub you the wrong way on purpose.

It seems to me more that he feels there is a real safety and reliability issue, and he feels that no one is taking his point seriously. One can understand how this would be frustrating and might lead to a negative sounding comment or two from him. If that is the case, I would ask that you try and understand his frustration and forgive him.

Even so, I'm reminded of the quote from HowToAskQuestionsTheSmartWay...

"Much of what looks like rudeness in hacker circles is not intended to give offence."

And although I strongly disagree with the working attitude of HowToAskQuestionsTheSmartWay, I think that particular quote applies nicely to this situation.
 
When the heck did the Op-Amps get all the way up to $5-$6 each? The 1$ TLV272IP is perfectly acceptable IMHO. Or did I miss something??? Not upset obviously, just confused...

I'm just working off the latest schematic. There was a desire to use an op amp with much better input offset voltage. The dual version of the OPA234 (OPA2234) is ~ $5-$6 (unless I miss read something). The TLV272IP appears to be similar to the LM358 in that regard.
 
Ah yes, that was probably my doing actually... :)/)


In any case, the input offset of the TLV272IP is lower than the LM358, but higher than the OPA234.

LM358 = 2000μV
TLV272IP = 500μV
OPA234 = 100μV
 
I kinda like the idea of the bigger FETs because it would be easier to build, but I think unless we can get data everyone trusts about the case to heatsink and heatsink to ambient we should use about 6 of them. I agree running right up against the absolute temperature rating is not a good idea.
Some things to think about if we go this way:
We would now have a 10 amp fuse. Do you think they will still blow before the FET bond wires?
Would need to find new sense resistors. Maybe .1 ohm?
Will the new case solder well to the water pipe?
 
... I have seen and made some harsh comments with people on this forum before myself. All I have to say about this particular incident is that Mr RB certainly knows what he is doing more so than most. And that I am sure he wasn't trying to rub you the wrong way on purpose.
...

I apologise to everyone if my posts(s) were seen as an attack, that was never the intention. However I was negative, not so much towards this dummy load build but more toward the idea that a TO-264 can be run at 2500W, or even at >500W with 155'C die temp.

Maybe I'm becoming one of those "grumpy old men"? :) I do think it's important that the forum sets a standard regarding things like package dissipations, heatsinking die temps etc so that's why I tend to jump on those type of things.

Also my real-world electronics experience comes from two areas; 1; Maintaining and designing industrial/automotive equipment etc where it is designed with big safety margins, NEVER to fail. 2; Decades in electronic goods servicing, repairing everyone else's designs that DID fail, and were designed with totally inadequate safety margins. You can probably see why my viewpoint is slanted towards "good industrial design". ;)

Sorry for any argument and thread derailing my posts have caused, and I won't comment on heatsinking again. :)

RonV said:
... We would now have a 10 amp fuse. Do you think they will still blow before the FET bond wires?

Very unlikely. The FETs are probably rated for 60A continuous, and a 10A fast blow fuse would blow long before any semi device rated for such high continuous current. Even de-rating for a high die temp of 120'C that 120'C will have little effect on the low PTC resistance curve of the 60A bond wires.
 
Mr RB, my apologies also for the rather clumsy way I made my comments - and I am very conscious this is not my thread and I will go way immediately if asked- but what you say raises a very important point.

I was rather taken aback by the absolute nature of your statements and I really am sorry if what I said came across as a criticism - it was not intended as such.

1) Let's be clear. I totally agree with your position. For commercial products.

2) But this is not what this project is about, I thought Jocanon is trying to build a useful controlled load at a reasonable cost and with minimum complexity. It has to run reliably for a decent length of time (1hr or so) but it does NOT have to do so under every conceivable condition imposed on it by idiot users.

3) Given that, then it is reasonable to make some compromises. That after all is what practical engineering is all about.

4) One of the sensible compromises might be universal reliability sacrificed for reduced cost and complexity.

5) If (and it is an important if) we can show that a particular FET combo can run consistently up to a certain load then PROVIDED Jocanon understands that and makes sure those conditions are not abused, that is a satisfactory solution. Not the optimum or safest or most reliable solution, but satisfactory. It is a decent cost benefit outcome.

6) The simple fact is that for a reasonable period of time (15min +) I have run a single TO-264 FDL100N50F FET at >500 Watts with 130C die temp at an ambient 22C. As far as my experimental method allows me to determine that die temperature. Certainly the >500 Watts is not in doubt; 20.8Amps at 24.1V. Whether it will continue to be the case I do not know without further testing. It seemed entirely relevant to the discussion people were having about suitable components.

So the decision is not whether it is "good industrial design" that this should not happen but whether for the purposes of the task it is an acceptable risk (suitably verified by further measurements) and for how long testing needs to be carried out to be reasonably confident of future performance given that deterioration can be expected operating at the thermal limits.

It is not whether one of us is right or wrong - I freely acknowledge you are right from the position you start from. The key point is whether for the purposes of this task it is an acceptable risk to run these FETs at the borderline.

This is a judgment not a fact. As such it is something that involves balancing the benefits and risks. If a load failure would be catastrophic to Jocanon's task then substantial over engineering, conservative factors of safety and redundant protection would be the order of the day. If he is trying to build a reliable load that will be expected to perform consistently day in day out but a failure is not catastrophic, then "commercial industrial quality" as you recommend is required. If he is trying to build a DIY loud at minimum cost that will do the job sufficiently well, then different standards apply.

That's all.

We need to be careful what assumptions we bring to the table when we say something is right or wrong.

Again not trying to start an argument, nor questioning anyone's expertise or experience, just asking people to think about what the assumptions are before they come to a conclusion.

Sorry to everyone for taking up your time again but It does seem to be an important point in this kind of design problem where there are so many unknowns.

Hope this is constructive and helpful.

John
 
... For commercial products.
2) But this is not what this project is about, I thought Jocanon is trying to build a useful controlled load at a reasonable cost and with minimum complexity. It has to run reliably for a decent length of time (1hr or so) but it does NOT have to do so under every conceivable condition imposed on it by idiot users.
...

I agree completely and your interpretation of the main design priorities were much closer than my interpretation. :) Remember I did ask in my post "What are the priorities?".

... I was rather taken aback by the absolute nature of your statements and I really am sorry if what I said came across as a criticism - it was not intended as such. ...

No offense was taken JJ604 (John), and again the fault was mine for being too heavily focused on reliability issues when the thread had obviously pushed priorities in a different direction. I was bucking the trend and causing problems.

As for absolute statments, well I think like most people my posts are 99% opinion, even though it seems like we state our viewpoints as absolute. Some of that is from hurried typing (or lazyness) and some is just conversational style. :)
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top