To overload or not to overload?

[Behemot]

Here's the deal. I have this power supply load tester of mine.

It has been first made for loads up to 520 W (13x40W bulb) on +12 V rail(s). Now I would like to push it forward, first to 650 W (13x50W bulb) and later to 975 W (13x75W bulb).

• I have already increased thicknes of the conductive paths by adding thick copper wire and like a ton of solder
• added some extra wire on the wire junctions (tried to keep it at some 6 A per wire at max)
• added some molex connectors and will try to add at least two 6pin PCIe connectors (so the power will flow through at least 12 pins, equaling in aprox. 6,8 A per pin and wire - I think this needs further reduction by adding at least one extra 6pin PCIe, the problem is, I have to connect it using wires, there is no room left for it

Now I have mostly 3A switches on that. But it's 3 A per 250 V, that means I can pull 6 A at 12 V just fine, right? These switches tend to do 6 A ad 125 V, that's still 10x bigger power

As for bulb sockets, I think they can do it too. Extra airflow will be added as well, holes are already drilled into the board, I will also make some wooden construction to pick it up slightly (so the air can flow through the holes) and to mount the fan on it permanently.

Now, what I am not sure about at all are thermistors. Rated for 4 A (JNR13S100L), they already get hot at 3,33 A. I would suck 6,25 A from them using 75W bulbs. So, should I add another one in parallel with each existing one? I have enough of them and I should be able to squeeze them into it. The question is not so much about them being able to provide 50 % current through them, but more like being able to provide the current over many times of usage. At least couple hundreds cycles. It is more pleasant for me to solder all of them now than replace damaged ones all the time.

[PeteS in CA]

Re: To overload or not to overload?

Re the thermistors, if you parallel them I think you'll be setting up a "race". Whichever one of a pairs happens to heat up faster will have its resistance decrease faster and over a brief time it will basically hog almost all the current. So you'll end up with a 5 or 10 ohm cold thermistor in parallel with a .005 ohm or .01 ohm really hot thermistor ... until the hot one fries or desolders itself and falls out of the circuit (and then the other will do the same in a few minutes).

[Behemot]

Re: To overload or not to overload?

Yeah I understand it, but they'll be in touch with each other (I can even add thermal grease), sharing small pad full of solder so it may take a while, but they will definitelly heat one the other and as soon as some current begins to flow through it, it heats more and more…all in all, nothing is gonna be cold in this board after I will step over first several hudreds watts of dissipation…

[Th3_uN1Qu3]

Re: To overload or not to overload?

Now I have mostly 3A switches on that. But it's 3 A per 250 V, that means I can pull 6 A at 12 V just fine, right? These switches tend to do 6 A ad 125 V, that's still 10x bigger power

No. They may work now, but it's by no means guaranteed that they won't melt, burn or arc in the long run. The contacts are spec'd for a certain amount of current, because they have a certain resistance. Remember Ohm's law: P=I^2*R. Power dissipation in a resistor increases with current squared - double the current and you have 4x the dissipation for the same resistance.

Voltage has no play in this - the voltage rating of the switch is the insulation resistance, aka, how much voltage it can break safely.

[Behemot]

Re: To overload or not to overload?

Check switches on market, most of them state 3 A for 250 V and 6 A for 125 V. I think it is connected with the sparking as for higher voltage you get higher charge on the contacts, and it's the quantum of the charge which can damage something if it is high enough. Now at 12 V, the charge is absolutelly tiny, there will be almost no sparking.

Also remember, everything will be heavily cooled by huge ariflow. Yeah, sometimes the temperature will be higher somewhere, but I don't think it can go so high to create some melting somewhere /it should go at least to 100 °C). Note the thick conduct paths which can conduct heat as well.

[KeriJane]

Re: To overload or not to overload?

Neat tester, but why such high capacity?

It's not necessary to max out a PSU unless you want to compete with JohnnyGuru or blow stuff up.

You can get a good idea of ripple and voltage at 1/2 power, which is where it's likely to be operated at anyway.

Rather than use bigger bulbs, why not use more, smaller ones? One bulb per thermistor, but maybe two bulb/thermistor pairs per branch. This would be gentler on the thermistors.


I like the Thermistor/Bulb pairs that reduce the high warmup current of using bulbs. You don't have a problem with hot de-rating, just with cold over-rating.

I used wire wound resistors and fans on mine. The fans keep the resistors at the bottom of their derating curve.
https://www.badcaps.net/forum/showthread.php?t=6687

It looks like you took the hot road and I took the cold one!

[Behemot]

Re: To overload or not to overload?

Review without loading up to at least 100 % of rated power is not a review, it is a joke and it is telling absolutelly nothing.

Problem is that the sockets for this type of bulbs (GY6,35) are not anywhere around anymore so I have to go with higher power per bulb. Also as you can notice, there is no room left on this board for additional bulbs. Making two huge boards is impractical although it is an option, but once again, I won't find any more GY6,35 sockets I am afraid.

[KeriJane]

Re: To overload or not to overload?

Oh! Reviews!
So you do want to compete with JohnnyGuru!
Sorry, I didn't understand the purpose of your tester.

Why not just use a bigger board?
And use complete lamp assemblies with reflectors, which would direct much of the heat away from the board and components.

Of course it would then be a really effective floodlight too.


How are you attaching your scope? I don't see any BNC connectors.

[Behemot]

Re: To overload or not to overload?

Oh! Reviews!
So you do want to compete with JohnnyGuru!
Sorry, I didn't understand the purpose of your tester.

Not exactly, I am aiming at, say, local market Czech only. I will use it to laod my own PSU's for sure, but now I can do reviews as well, you see (:

Why not just use a bigger board?

This is already the biggest cuprexit I have found…

And use complete lamp assemblies with reflectors, which would direct the heat away from the board and components.

Fan blows from top, and actually even halogens can transfer like 10 % of power into light anyway, the rest is heat.

How are you attaching your scope? I don't see any BNC connectors.

Directly to PSU connectors.

[KeriJane]

Re: To overload or not to overload?

Actually, reflectors reflect heat along with the light.

I came up with an idea for your size problem....

Instead of bulbs on the control board, just have a multi-wire connector like maybe an ATX with fairly heavy wires.

Put all of the bulbs on a separate board or box with their thermistors, attached to the control board with a multi-wire cable.


If you make it a box, you can have bulbs on 5 sides, facing all directions except down. This would allow lots of room for lamp assemblies and dissipate the heat and light over the greatest area.

It would probably earn the nickname "Lighthouse"!

Plus, by building the load into a separate unit, you could switch to a different style of load without starting over.

[Behemot]

Re: To overload or not to overload?

I think I rather wont make an UFO

As for daugther board, I can as well just make a new one which will suck power only from PCIe connectors. Because there is really hardly anything else you can take power from on some 1000W+ ATX PSU.

It seems the bloody shop got re-stocked with new sockets, so I can actually make daughterboard just fine

OK than, I think it's clear, I will buy 50W halogens now (thats 4,17 A, which is just fine for both switches and thermistors I think) and make second board later. There will also not be any problems with additional connectors as 12 pins for 650 W are fine I think (4,5 A per pin and wire on the PSU side).

[mariushm]

Re: To overload or not to overload?

With those light bulbs, you could just use regular screws and a par of large washers to trap the leads of the leds between the washers.

lame ascii drawing

   _  <--- phillips screw
  _|-|_
  ====
  _____ <-- washer 
----------------------- < --- lead from bulb 
  ____ <-- washer 
  |||
  ||| <-- hex style nichel spacer (keeps bulb a bit further away from pcb, a bit of heatsink
 --|||- - pcb
  [nut] <-- locks hex spacer to the pcb. maybe add solder here for perfect connection to pcb copper
  |||

However, if I were you I would seriously consider doing one or several of those constant current loads, they're more suitable for testing power supplies, as you can finely tweak the current.

I've made the one here for my own needs: http://www.sleepyrobot.com/?p=136

My version came out like this, really ugly on vero but it does the job. I tested it up to about 16v and 3A with my linear power supply and the heatsink went up to 40-45c.
It can probably do 6-10 amps with this type of heatsink with a fan chilling it.

But the beauty of the design is that is that you can add mosfets and increase its current ability, as long as the heatsinks are properly sized

The most expensive part is the fets, about 2$ a piece, the 2 channel rail to rail opamps are cheap and the rest is 5-6v linear regulators (i used a plain boring 7805) that can be shared between several loads a and resistors and capacitors. (later edit: and that heatsink was about 40 cents, but you can put two fets on it one on each, it's tall, it was this one : http://uk.farnell.com/jsp/search/pro...jsp?SKU=265275 - picture doesnt match, but datasheet shows it correctly)

[Behemot]

Re: To overload or not to overload?

As for the first part, I have no idea what are you talking about sir

As for the active load, I have some schematics and partly working prototype here. The thing is, I can make load with bigger increments through the bulbs and than have just one small linear load for fine tweaking. You have to see the magic

[mariushm]

Re: To overload or not to overload?

Ok for the first part.

Drill larger hole in PCB.

Get some of these :



It's the kind of spacers that are put between motherboard and computer case.

There's also larger sizes :



Put it through the hole on the pcb, use a nut to tighten it on the other side and use some solder to really lock it in place and make contact.

Now you can use a regular computer style screws on top. They're M3, M4 or 6/32 depending on what spacer you get.



You can also use a pair of washers (or fasteners?) between the spacer and the screw to catch the leads of the light bulb easier.



And that's it, you don't need a light bulb socket anymore.

[Behemot]

Re: To overload or not to overload?

OK, this is how to make easy things complicated, right? I have a name for such things…большая техника

One thing which bothers me more and which is gonna be really challenging is how to measure high currents really accuratelly (at levels under +-0,5 % instead of over +-2,5 % as now).

Very precise shunts maybe? But what voltmeters accurate enough to measure that tiny voltage difference? Analog maybe?

ADD// or just use milivoltmeters? Cause I can measure voltage directly accross the shunt, not difference in voltage before and after it

[mariushm]

Re: To overload or not to overload?

You said you weren't able to find those sockets for light bulbs anymore so I gave you an alternate method of connecting the light bulbs in the circuit.

The method is more complicated, but it would also be arguably a bit better, as everything up to the light bulb leads is solid metal with minimal resistance. When doing 1000w of load, improper connections in the connectors of the psu can be quite a problem.

As least in theory, in your design your wires are thin, there's some small voltage drop there, wires heat up a bit, there's less perfect connection between the socket and the light bulbs and so on.

I'm also not so found of using light bulbs to do measurements, their consumption will slightly change as they warm up, and they're also going to be affected by the surrounding light bulbs and the heat coming from them.
You would be better off using something with a larger surface to dissipate the heat of, in the worst case even those 20w resistors would do. But I agree light bulbs are really cheap.

It sucks however cause you'd have a hard time doing crossload tests when you can only jump up a few Amps at a time.

[Behemot]

Re: To overload or not to overload?

As I said later, the shop here got another sockets on stock…

I measure current and voltage so I don't really care the load varies. After some time, it always settles. Than I measure the actual numbers.

Bit of a problem I was dealing with was about accurate measurement of current, but I think I can solve that by using industrial shunts and measure just the voltage over them. There are shunts for 50 and 100 A with up to 60 mV drop on them. Even the cheapest voltmeters have +-0,5 % accuracy which is way better than clamp meter with +-2,5 %

[KeriJane]

Re: To overload or not to overload?

Большая техника???

LOLOLOL!


I like your phrase for the overly complex and have adopted it as my own.

Yes, a millivolt reading across a shunt will display Amps fairly accurately.
In the automotive trade we are used to making large current measurements and this is how our ammeters are or rather, were constructed.
Millivoltmeter across heavy shunt.

Nowadays, we generally use clamp-on.

The shunt will be most accurate within a range of current and voltage. The ammeter that works great on a 200A starter isn't the best at measuring ma.

I'm not all that sure of the accuracy though. When dealing with tens or hundreds of Amps, we have no need of high precision.
Maybe use different shunts for different load ranges, like a multimeter has?

Спасибо Большое и С Новым годом

[Behemot]

Re: To overload or not to overload?

I think as for first and cheap generation, I will leave it with 650 W (+12 V) at most, measuring with clamp meter and so. No major changes on this one. You convinced me I gave it a thorough thought and I won't be able to transfer this one into much better load and accuracy anyway.

As for next generation, I will jump to totally different level, similar to more professional instruments I think I am thinking about separate boards for different voltages. Load will be achieved by high-power rated resistors in TO packages mounted on heatsinks. Some linear load for fine adjustments too.

On every board there will be one (or more) shunts and panel voltmeters - I found different types of digital voltmeters in 200 mV range with accuracy of 0,3 % and 2 digits, which is great I think. Yeah

[tom66]

Re: To overload or not to overload?

I hope that you're going to use a fan (or three) to keep that array cool.