1. 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.
    Dismiss Notice

Dummy Load II

Discussion in 'Electronic Projects Design/Ideas/Reviews' started by jocanon, Jan 12, 2013.

  1. tvtech

    tvtech Well-Known Member Most Helpful Member

    Joined:
    Mar 27, 2010
    Messages:
    3,166
    Likes:
    348
    Location:
    South Africa
    Hi Buddy

    May the wind be at your back and success forever at your side. Good luck with the move.

    Regards,
    tvtech
     
  2. jocanon

    jocanon Member

    Joined:
    Jul 10, 2012
    Messages:
    703
    Likes:
    2
    Location:
    Phoenix, AZ
    Thanks!
     
  3. ronv

    ronv Well-Known Member Most Helpful Member

    Joined:
    May 30, 2010
    Messages:
    4,725
    Likes:
    488
    Location:
    Tucson, AZ. USA
    A 2500 watt fet! unbelievable. Wait till Mr. RB sees this one. :D With 2 or 3 of those babies we wouldn't need to worry about lowering the current at 50 volts (just the house circuit breaker). $13 isn't cheap, but it would sure save a lot of circuitry.
    Of course the water in Phoenix comes out of the tap at 35C - but hey! :rolleyes:
    Maybe mounted to a commercial water cooled heatsink?

    Good luck with the move!
     
  4. dave

    Dave New Member

    Joined:
    Jan 12, 1997
    Messages:
    -
    Likes:
    0


     
  5. Mr RB

    Mr RB Well-Known Member

    Joined:
    Jul 22, 2008
    Messages:
    4,716
    Likes:
    194
    Location:
    Out there

    2500W?? Bah! That's IF it is submerged in liquid nitrogen.

    I just grabbed my pro transistor substitution book, with literally tens of thousands of transistors listed from many decades. Scanning through lots of pages for that package (TO-264) shows it's a 100W package, ie all the highest TO-264 devices are rated at 100W max.

    A TO-220 is about a 60W package, a TO-264 (or TOP-3) is about 100W and the massive old TO-3 is about 200W-250W at best.

    I'm beginning to think the cheaper manufacturers are starting to rate their parts in "whats?" instead of "watts" and it's going the way of the audio amplifier industry.

    You can't expect to get 2500W out of a package that has had decades of use from many manufacturers at a known max safe dissipation of 100W. :(

    (edit) I just googled for a REAL high power semi, this Hitachi IGBT (top middle in photo);

    [​IMG]

    Is 130mm * 140m in size, and is rated for 1200 amps DC at around a typical 2v Vce (ie rated at a 2400W continuous dissipation).
     
    Last edited: Feb 28, 2013
    • Like Like x 1
  6. Dusey52

    Dusey52 New Member

    Joined:
    Oct 2, 2012
    Messages:
    50
    Likes:
    2
    Location:
    Oregon
    Yeah, 2500W is pretty unbelievable. The thermal specs suggest something closer to 800W (still a lot). If they're good for 400-500 on the water pipe they're still worth considering.

    This one http://ixapps.ixys.com/DataSheet/DS100238(IXTN200N10L2).pdf looks like it would actually be good for 500 watts, but I doubt there's an easy way to solder it to a water pipe (it's also ~$35, ouch).
     
  7. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    For what it's worth a couple of us are getting 700 Watts from two FDL100N50F's on a single 5 heatpipe CPU cooler with two 120mm fans at 30C ambient. Rock solid for 15 minutes.

    A single FDL100 will do about 450 Watts on the same cooler. Just on 19Amps at 24Volts at 30C ambient. Using a max. allowable die temp of 150C and an Rcs of 0.15

    You might push it a bit higher but the die temp tends to go up pretty quickly and that's about the safe limit I suspect unless you go for very exotic cooling. It may be that these modern heat pipe coolers are even more effective than straight water cooling because of the very large amount of heat that gets absorbed by the phase change. We also suspect they may have a limit where the Rsa suddenly goes up. By the way, this is all practical testing, not theoretical. :)

    The huge electrical power ratings of these things are meaningless since that is for switching applications. In linear application it is a thermal, not electrical, problem. You still have to watch the DC SOA but it is getting the heat away that is the limiting factor.

    John

    PS: jocanon did you see this post?

    http://www.rcgroups.com/forums/showthread.php?t=1841615#post24275721
     
    Last edited: Feb 28, 2013
  8. ()blivion

    ()blivion Active Member

    Joined:
    Nov 24, 2010
    Messages:
    857
    Likes:
    210
    Location:
    Level 5
    I have to agree with Mr RB, 2500 watts isn't going to happen. That rating is almost certainly what it is calculated to switch, and even that is a dubious claim. At some current, even if the die can take it, the leads will melt. It would have to get most of it's watts from voltage to be able to switch that.

    For dissipating power, you have to remember that the actual part making all the heat is a square of what basically a gram or two of glass the size of your pinky nail. Well, the several feet and several grams of Ni-Cr resistance wire in my space heater gets red hot at 1300 watt. Trying to get that much heat to come out of that tiny piece of glass would require it to be several times over it's melting temp, let alone above it's max junction temp. Simply not happening.

    Even with amazing heat-sinking, the material the chip is made out of will have enough thermal resistance to hold the heat in and destroy the thing.

    Here is what the inside of a (larger) power transistor looks like...

    dual-mosfet.jpg

    And here is the size of the full case...

    NTE1912.jpg


    It's just as jj604 said, virtually every power transistor that we have looked at was intended to be used as a power switch. And the datasheet power rating is very often a reflection that application, sans marketing.
     
    Last edited: Mar 1, 2013
    • Like Like x 1
  9. tvtech

    tvtech Well-Known Member Most Helpful Member

    Joined:
    Mar 27, 2010
    Messages:
    3,166
    Likes:
    348
    Location:
    South Africa
    That Roman is the chirp of the decade. I agree with it 100%. I wish I could rep you to Gold Status here. Not merely more green.

    Regards,
    tvtech
     
  10. Dusey52

    Dusey52 New Member

    Joined:
    Oct 2, 2012
    Messages:
    50
    Likes:
    2
    Location:
    Oregon
    I don't think anyone has suggested that the FDL100N50 can operate anywhere near the 2500W spec in linear mode. But it does have a Tjc that is 10 times better than the FET used in Dummy Load I so it's much better at transferring heat. I think everyone agrees this is a thermal management issue more than anything else and the thermal specs along with jj604's real world experience suggest that 3 or 4 of the FDL100N50s could do the job of 15 or so of the original FETs. When it comes time to build the thing, soldering ~4 of these to a copper pipe and building the support circuitry on a perf board is going to be a lot easier for Jeremy. It's also a 500V part, so protection against line voltage is less of an issue. Just my 2 cents...
     
  11. tvtech

    tvtech Well-Known Member Most Helpful Member

    Joined:
    Mar 27, 2010
    Messages:
    3,166
    Likes:
    348
    Location:
    South Africa
    To add to my post above.....a lot of us "older people" here see Standards that were adhered to strictly in past going to pieces since China started the mass production of rubbish. For the masses.

    Sure, there are Chinese factories out there that produce goods for the likes of Apple etc. But they are well run by a proper organization..There is strict quality control.

    Here where I live, I would hate to be a buyer of Electronic spares right now for an Electronics importer. There is soooo much crap floating around right now.

    As a tvtech, my job is already difficult. If you cannot trust the new spares you are using to fix your fault...you have to rely on a sixth sense to warn you that the new part you fitted has taken the job backwards.......

    For example: Set had Frame Collapse. Line was running (obviously). But Frame collapse. Frame output pulse is missing from Jungle IC. Checked with scope. Yew tells me that Voltages are correct to Frame oscillator supply. Jungle has failed.

    New IC ordered. Arrives. Fit an IC socket in case something changes....unexpectedly...I have learn't by now.

    Fit socket, plug Mr Brand New IC in......set is dead. No line running or anything. Dead.
    Take out Mr Brand New and put old one back and back to Frame Collapse.

    Job goes forward. And so on. Customer got his set back after the 2nd supplier supplied me/us with a original SUSOC IC.

    Google SUSOC if you don't know. Made by Sanyo. Not the first time this has happened.

    I feel I need to swear again. But I won't. New faulty spares can drive a Tech crazy. Unless you know what you are doing. And don't trust new spares anyway.....

    And you are methodical with your approach to repairing stuff ;)

    That is the Secret. And that is what gets the Job done.

    On time. Everytime. In spite of obstacles.

    Regards,
    tvtech
     
    Last edited: Mar 2, 2013
  12. ronv

    ronv Well-Known Member Most Helpful Member

    Joined:
    May 30, 2010
    Messages:
    4,725
    Likes:
    488
    Location:
    Tucson, AZ. USA
    Thanks JJ.
    Jeremy is using water cooling at about 20C. I can't find where he did the measurements but as I recall he only had a case temperature rise of about 4C on the last FET in line so it is almost an infinite heat sink. He also measured the top of the TO220 and got a number very close to the calculated junction temperature so I almost belive it.:rolleyes:
    Not sure what the junction to heatsink might be from soldering the FET to the copper pipe, but it must be very low.
    () The big difference in this FET is the thermal resistance junction to case for this guy is only 0.05C per watt whereas a TO3 like the 3055 has a junction to case of 1.5C/watt. So it is 30 times better getting the heat from the die to the case. I'm not sure how they do that but Fairchild has a pretty good reputation.
     
  13. ()blivion

    ()blivion Active Member

    Joined:
    Nov 24, 2010
    Messages:
    857
    Likes:
    210
    Location:
    Level 5
    I don't want to dispute anyone's work, but I can't really explain that other than if some kind of mistake was made somewhere along the way. I don't recall the exact specifics, but when I was doing my preliminary testing with ~6 transistors and ~100 Watts on virtually the same heat sinking apparatus, I was getting a higher indicated Watts to temp ratio. The backs of the FETS were getting pretty hot too. My best guess would be that he is likely not putting all the Watts into the device that he thinks he is.

    If that's not it, and there are no other errors, then this data alone would make it seem perfectly sound to go ahead with four, three, or even two transistors at 2000 Watts. And that just seems completely unrealistic to me. Even with resoundingly positive data, I can not recommend going forward with something like that. It goes against what I would consider transistor common sense.

    Yeah, if we are just talking whats better or worse when compared to the other parts, then I'm with you guys. Those are better performers on paper. I just want to make sure we are not using the numbers to assume we can get 2~4 transistors to dump 2000 Watts of heat. Doing that will almost certainly destroy them no matter what the heat sinking is, I'm fairly sure of it.

    I'm thinking larger die most likely. Spreading the heated area out will lower the effective thermal resistance. It costs much more to make the larger die though. If it isn't that, then they are putting a thermal bridge on the top and sides of the die that transferred heat around to the tab. This would have the effect of more surface area without increasing the die area. I don't even know if they do that or not though, but it seems reasonable to accomplish. The only other thing they could possibly be doing would be to not use silicon. They would have to be using something else with a higher thermal conductivity. Crystalline silicon is already fairly high thermal conductor though, so to be a material with even better thermal performance it would have to be some crazy exotic semiconductor made with like diamond, carbon nano tubes, or silver-oxide or something. Even so, if they were doing that, then the limiting factor would just get shifted over to the copper tab.

    Maybe we should de-encapsulate some power transistors some time, and see what is really going on. I'm fairly sure they will not look much different from the images I posted though.



    Final thoughts
    I hate to drag things out for a long time, and I hate to be a naysayer on something I helped come up with. However, I would hate to have this be dangerous and unreliable even more so. As far as I am concerned, using the package as an indicator of wattage is the only guarantied reliable method for choosing the transistor, as it is a manufacturer independent metric. Anything else is dubious and would need to be thoroughly field tested before it could be deemed reliable. Using the datasheet numbers on something like this depends solely on the manufacturer both knowing what they are doing, and them not pushing marketing into the equation. Both of these things are far from a guarantee, even with a good manufacturer.
     
  14. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    Those FDL100s are pretty impressive.

    I knocked up a spreadsheet for calculating the SOA for a FET as a DC load if that would be useful. I can post it if anyone wants it.

    If you plug in the T j max, Rjc and Rcs from the spec sheet, the number of FETs you are using on the cooler, and know the Rsa of the cooler it will show the max power for an ambient temp of 15, 20, 25 and 30C. In our case I plot it as the max. allowable current for any particular pack voltage. You can play around with varying the various parameters and see what difference they make to the power rating which can be illuminating.

    Finding Rsa is the trick. Only practical way we found was to embed a TypeK just under the surface of the cooler at the centre of the FET tab. Smallest possible hole, good thermal epoxy etc. I found very consistent values for Rsa over a range of test loads and close agreement with the manufacturers quoted value. Not sure how it would go with a water cooled load though.

    Agree with the comments about the ability of a particular package to dump heat. Getting transfer internally from the die to the package is probably proprietary magic but the package itself is fairly constant - and in the case of the FDL100 the key limiting factor. I would be very cautious about assuming an infinite heat sink. At these power levels even a small value of Rsa makes a big difference.

    Here's a sample output for a single FDL100 on an air cooler with an Rsa of 0.065.

    Just out of interest the max. power for a single FDL at 20C ambient on my air cooler is predicted to be 491 Watts. If you change Rsa to 0 (i.e. an infinite heat sink) the value rises for a single FDL100 to 650 Watts. You can't do better than that under any circumstances without reducing the ambient temp by cooling with iced water or immersing the whole assembly.


    John

    PS: This is not just theoretical. I am actually running these FDL100s at close to these loads. Haven't pushed it to the absolute limit as I am conscious of the cost of these things. :)
     

    Attached Files:

    Last edited: Mar 1, 2013
  15. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    Updated comment

    Of course all of the below is for the THERMAL SOA.

    You still have to comply with the ELECTRICAL SOA for each individual FET as well for the DC line. In almost all cases however the thermal limit is the critical one.
     

    Attached Files:

  16. Dusey52

    Dusey52 New Member

    Joined:
    Oct 2, 2012
    Messages:
    50
    Likes:
    2
    Location:
    Oregon
    If I remember correctly, the current Dummy Load II design is targeting 1500 Watts using 15 of the STMicroelectronics FETs. The specs suggest (and jj's experience supports) that 4 of these could comfortably handle 1500 watts using the water cooling method employed in the first project. I agree with () that expecting a To-264 package to handle 400 or more watts is counter intuitive and just ‘feels’ wrong but the specs say it will work and jj has done it on air no less. Using 5 would drop dissipation to 300 watts per part adding a little more wiggle room while still being 1/3 the parts count.
     
  17. Dusey52

    Dusey52 New Member

    Joined:
    Oct 2, 2012
    Messages:
    50
    Likes:
    2
    Location:
    Oregon
    Last edited: Mar 1, 2013
  18. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    Sounds about right to me.

    Just to be completely clear, I ran a single FDL100 at a measured 501Watts (20.8A at 24.1V) on an air cooler with twin 120mm fans. The measured sink temperature was 54.7C at 22C ambient. At that power if the specs are correct, the junction/case power dissipation is 25 watts and the case/sink dissipation 50 Watts. That last figure is slightly doubtful as the spec sheet gives a value of Rcs as a Minimum of 0.1 which seems meaningless. From then on I used a figure of 0.15 as being more in line with other numbers published for this package. In that case the dissipation case/sink would be 75 Watts.

    A 54.7C sink temperature gives a value of 0.065 for the Rsa for this cooler and predicts an allowable sink temp of 75C at that power using Rcs = 0.1 and an allowable sink temp of 50C at that power using Rcs = 0.15.

    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. A well constructed water cooled load should do slightly better so I think 500Watts/FET is an achievable value for this project using FDL100N50Fs based on my preliminary work to date.

    This all assumes the FETs are genuine and that the published specs are as well. :)

    John
     
  19. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    Don't forget if you have more than one FET that the heat paths are in parallel for the FETs but are then in series with that for the single heat sink.
     
  20. ronv

    ronv Well-Known Member Most Helpful Member

    Joined:
    May 30, 2010
    Messages:
    4,725
    Likes:
    488
    Location:
    Tucson, AZ. USA
    If you believe the data sheet (and I do) we have 2 unknowns in this application:

    1- Thermal resistance case to heat-sink. This could be very low in this case since the FETs are soldered to the water pipe. - IC engineer turns over in his grave. :rolleyes:

    2- Thermal resistance of the heat-sink - estimated at .025C/Watt or better based on ()blivion and Jeremy data.

    JJ, If you still have 1 running would it be possible for you to measure the hottest spot on the plastic (opposite the metal tab)? I think we can get a good idea of the junction temperature from this and maybe refine Jeremy's or ()blivions data.
     
  21. jj604

    jj604 New Member

    Joined:
    Oct 3, 2012
    Messages:
    11
    Likes:
    0
    Not really convenient, sorry.

    They are all now mounted on the heat sinks and clamped up with a solid Al angle. I did try measuring the FET tab temperature directly with an embedded thermocouple but I think I screwed up the thermocouple when I soldered it in. Did it again with thermal epoxy but the readings are lower than the cooler temp so I am regarding that as a fail.
     

Share This Page