Dummy Load II

[jocanon]

Hi all, I just wanted to check in and let you all know I am still here! It has been pretty crazy moving across country and all...we have arrived safely in Arizona and now I am trying to get settled in my new job and get our house all unpack and organized. I need to finish my new solid hobby website and get my power supply side business up and running as again as soon as possible because everyday that passes I am missing sales. I have already had a lot of people ask when they can buy a power supply...anyway, I am just telling you all this so you know that I am still serious about the project even though I have been silent for a while.

As far as priorities, the way I see it, there really wasn't anything wrong with the original design. I think it would still be functioning and serving the purposes I needed it for today had I not shorted the op-amp and ruined it by my own human error. Now that being said, I thought if I have to build it again, I might as well see if we could improve it by making it not oscillate as much and improve the amount of power it could handle. In actuality, if that turns out to be more trouble than it's worth, I would just as soon simply rebuild the original design ... don't get me wrong, I am appreciative of all the input and hopefully the end result is we get a better more robust dummy load out of it.

To put a little bit of perspective on things as well, I may only need this to work long enough to test about 150 more power supplies, probably about 6 moths time. I have about that many in current inventory and I am not sure I am going to keep going once I sell them all. I haven't decide for sure yet...it will depend on a couple factors, one, if I can find more power supplies to purchase in bulk for cheap (I don't see any on eBay right now). And two, if I feel like taking the time it takes to do it. It takes about 3.5 hours to make 1 power supply. Anyway, point being, I tend to agree with the point () was hinting at that is that if we keep the debate going we may not be done debating it before I completely sell out of power supplies and then it will all just be academic. I like the idea of bigger FETs. All in all, I would like to make it as robust and solid as possible for around $200 or less.

Jeremy

P.S.
I don't know how much of a difference it will make, but the water here in AZ does come out of the tap hotter than in NY, so we may want to throw a few more FETs in there for good measure

 

[ronv]

Welcome to the desert. If you have a cat make sure it stays inside. If you have a small dog make sure you have a very high fence. And watch out for the snakes in a couple of months. There I've done my civic duty.
Hope the move went well!
The difficulty with the big FETs is that their improved performance depends on two parameters that we don't have bullet proof data on. That is the temperature rise due to soldering them to the case and the other exactly how hot the case is. Using the smaller ones makes those 2 parameters less important. My gut feeling is we could do 6 of the big guys or the 15 smaller ones and be safe either way at 1500 watts. Makes no difference to me. I suspect cost will turn out to be about the same, but of course assembly would be easier.
What does everone else think?

 

[jocanon]

Thanks for the warnings ...I grew up here in the desert so I pretty much know what to expect...124F summers! But it's a dry heat and let me tell you, it beats shoveling snow and scraping your car everyday.

 

[()blivion]

I personally don't care too much any more about which FET's are better for the wattage. If it breaks, it breaks. I'm thinking we can deal with that bridge when we burn it.

As for other physical aspects, less FET's means they could be easier to solder, but it's hard to say. It really depends on whether it is hard because of space, or because of wetting the metal, or if it's even hard at all. If the space is an issue, then the larger ones are best, but you could allso just use the other side, or do some other creative hacking to make the cooling work. If it's about getting the solder to wet the metal, well then that's a whole other issue. The more metal there is, the more heat will be needed to get the solder to bond. You also need to be careful about only partially soldering the FET down. Larger FET's will be more likely to trap an air pocket under them, or end up with an entire half of the FET left unsoldered. It's critical that you ensure that the solder blob starts from the center of the FET, and has properly wet the part. Then you need to make sure when putting it on solder fully pushes out from under the part on all sides. The only good news here is that you only have to repeat soldering success a few number of times, a lower number than with the other FET's.

As for the circuitry, some parts get simpler, some parts get harder.

For starters, the current sense resistors will have to be either much higher power, or much lower resistance. This creates problems no matter what we do. As they were, they were already bordering on needing active cooling. If they are made to be even higher power, they will make more heat, then they will *require* cooling. The problem is, this will likely be with something other than our super fantastic water cooling setup. I consider this far from optimal. If they are lower value however, heat becomes a non issue, but then we need to pay particular attention to noise and resistor tolerance. And the thermal drift will start to come into play.

Given only these options, I still say we can't go wrong with the before suggested tabbed resistors. They are more precise, and they will be heatsinked to our almost constant temp cooling system. And, as luck would have it, we already have a thermal sensor on the cooling. so if thermal drift does still happen to be a problem, we will have the framework necessary to deal with it in our circuit. Though with the tabbed resistors, I doubt it will com into play.

The other problem with the circuitry change is redundancy and fail over. Lower number of power parts controlling more power makes it harder to control the whole system in the event of catastrophic single part failure. Also, if one part has issues, like oscillation, the noise will be amplified more as that one part is a greater percentage of the whole. Of course, proper design can work around all this, so it's mostly academic. Just realize this is going down hill from what we had, and we don't even know that what we had was rock solid.


Anyway, my point is we need to make sure we keep in mind there are things we sacrifice and have to change when we are looking at doing this. Changing the power parts creates chain effects in the whole system, it's could lead to radical redesign.

 

[jocanon]

I think I may just keep it simple and go with the original design, since it worked...but just add 15 mosfets instead of 10. Also, I will put a watt meter in there to get current readings...this one:

**broken link removed**

 

[ronv]

Hi Jeremy. Welcome back.
I hope the move went well and the new job is fun.
I couldn't find the spec sheet on the watt meter but it seems like it should work and eliminate some circuits at the same time.
Do you still have a schematic for dummy load one? I have changed computers since then.........
If you stay with DLI we might want to at least change the op amp and make the resistors higher power to cool them down a bit. Whatca think?
Did you find a source for more supplies?.

 

[jocanon]

Thanks Ron. The move went well. I did find a good source for more power supplies. I agree, let's at least change the current sense resistors to not get so hot and the op-amps. The final schematic for dummy load I is attached to post 1. I don't remember if I have the actual LTsplice file. It seems like you may have sent it to me at some point. Let me know if you want me to search around for it or if the PDF on post 1 of this thread will be enough.

 

[jocanon]

I wonder also, how much would we have to change, other than the TVS to make it 48v capable?

 

[ronv]

The biggest problem with the higher voltage is that without sensing the voltage if you hook up a 48 volt supply at 50 amps it's toast. We could do a switch, but if you ever forgot . If it is just 24 and 48 and nothing in between sensing it and lowering the current might not be to bad.
I'm having some trouble finding the right op amp that isn't in a surface mount package, but I'm sure there are some out there - just a matter of finding one.

 

[jocanon]

That makes sense. It would just be 24v and 48v most of the time, occasionally 12v too.

 

[Mr RB]

The biggest problem with the higher voltage is that without sensing the voltage if you hook up a 48 volt supply at 50 amps it's toast. We could do a switch, but if you ever forgot . If it is just 24 and 48 and nothing in between sensing it and lowering the current might not be to bad.
I'm having some trouble finding the right op amp that isn't in a surface mount package, but I'm sure there are some out there - just a matter of finding one.

When I make simple linear dummy loads, they are either current controlled (with the pot biasing the emitter resistors in reference to a fixed 5v) or are roughly "resistive" (so the pot biases the emitter resistors in reference to the input voltage).

The good thing about those types of operation is that the main voltage voltage can be anything, the only thing that happens with a higher Vin is that it will make more dissipation (for the same pot position).

 

[()blivion]

To allow the system to transition from 24v -> 48v safely and "automagically" would simply require that you have a comparator/op-amp that senses the input line and de-bias the current control pot by half whenever Vin passed a certain threshold. Such an implementation is fairly straightforward.

Something that I was saying about this before though is that it should be just as equally straightforward to solve this problem with a fully linear solution and almost the same circuit if we wanted to. We simply need to find the correct offset coefficient and apply that to the current pot via an inverting op amp circuit. Should be as easy as pie.

The benefits of doing this are debatable however. All it would mean is the circuit could take mostly arbitrary voltages between 0 and 48 volts and would never allow you to set itself to draw more than the wattage that it could safely dissipate. Of course, even with this setup, one could always apply more voltage than some critical part can handle and the system would then still break down.

The final solution is to "simply" make the system able to safely handle 48v * 50A = ~2,500 watts. Which means... it would need to be quite beastly.

 

[jocanon]

Heck, at this point, I'm not opposed to putting 30 MOSFETs on the thing if that is what it takes...I mean why not, when it comes right down to it, it's really not that expensive it's just more labor. Granted, at some point I am going to be tripping my household circuit, but I could always run it off 220/240 dryer circuit too. I know guys in the hobby that do that with some of these beastly chargers. In fact I've heard that some houses are coming with a couple 240v outlets in the US now. Point is, if I put double the MOSFETs in there it would only help make it safer under normal operation and then allow me the option to test a larger power supply when needed. One of the manufacturers came out with a charger that is capable of a ton of power and needs 48 v to get all its capable of. I think it was designed more with other countries in mind that have 240v or 250v on the AC input side but there are still some guys in the US that want to push that limit.

 

[()blivion]

I digress... but in an elitist PC world, there is sound reason to modify your house for 240 volts for the sole purpose of running a single PC that breaks out of what a 120v outlet can do. This kind of excessive power consumption can happen when you are running four 150+ watt graphics cards on a dual 16 core cpu server motherboard with 64GB of RAM, 8 HDDs, water cooling, and so on and some such nonsense. Quite overkill really, and for not a whole lot more extra power in all reality. The machines will be eclipsed by newer hardware in a few short years. Reminds me of the P4 Extreme(ly expensive) PC's few would build and others would drool over not that long ago. Those things are absolute junk now.

But that's computers we are talking about. A power supply/dummy load is not really susceptible to obsoletion do to moore's law.

 

[ronv]

How about this? You could test each supply until the over current trips. 55 amps? then build them into 48 volt supplies and test them at 30 amps. This would have the added benefit that they might run on a 120 volt outlet. We could auto adjust for the lower current, as ()blivion suggested at say 28 volts.

 

[jocanon]

Sounds good

 

[ronv]

Dl2

I kinda like this one. I think the 75 ohms in the FET gate will also make it stable. This op amp has 4 per package so I made it for 16 FETs.

Here is a link to a 10 watt .15 ohm resistor.

https://www.electro-tech-online.com/custompdfs/2013/06/cp-205274.pdf

 

[()blivion]

Do note that, all shunt based amp meters have an internal resistance needed to develop across it a measurable voltage relative to the current. So... having the amp meter between the ground and the sense resistor could potentially de-calibrate the actual current from the intended current setting. With this in mind, it would be best to put the meter above the MOSFET rather than below, if at all humanly possible.

That being said, that resistance is itself an amp measuring device. So regardless of where it is, we should still be able to know the correct amps.

 

[jocanon]

Thanks ronv and duly noted ()blivion. So could I run 48v through that circuit? Of course being careful not to exceed the rated wattage.

 

[()blivion]

So could I run 48v through that circuit? Of course being careful not to exceed the rated wattage.

Yes the MOSFET's are the part that would fail if anything, and they look to be able to handle 100v/2. With 16 of them at 150 watts each (playing it safe-ish), 2400 watts should be the upper limit for power.

I don't think the proposed circuit has any kind of auto current limit feature as of yet though. Ronv may get back to us on that yet still. So may not want to heat up the soldering iron just yet.