PC PSU DIY load tester, looking for the right (cheap) components

Thread Starter

adelage

Joined Aug 22, 2014
38
Hi all
Since I often have to deal with crappy bog standard pc power supplies, I am trying to build a 12v load tester (single rail for now) recycling some stuff I have laying around, but I am not sure about some of the components, and in general, the final design:

The load tester should be able to load max 80A on the 12v rail in 4A steps.
For the female connectors I am gonna desolder the 24pin and 4+4 pin from a motherboard, and for the 4pin molex and 6/8pin pcie I will get them from some spare cable adapters.
For the load itself I am gonna use an heating element from an old fan heater, cutting it at the proper length so to get 20x3ohm pieces (I already tested one and a 20cm piece of resistance gets just a bit hot with 12v taking 4A, and I will place them so that they don’t touch anything unnecessary with a fan blowing fresh air, so hopefully there shouldn’t be any fire hazard) and for the box I was planning to use some mdf boards, painted with fireproof heat reflective paint.
Now, what I am missing are the switches, I was planning to use some 10A lever switches for each load increase, but I would need 20 of them, so I was looking for some kind of rotary switch with 20 positions, but it should withstand a load of 80A when fully turned on, and I don’t think such a thing exists..any idea?

P.s.: I know I should test also at least the 5v and 3.3v rails, but for the sake of simplicity I am only focusing on the 12v single rail part, adding the other sections later if it proves successful.
 
Last edited:

Dodgydave

Joined Jun 22, 2012
9,666
I would use either high Wattage resistors or an Heating element/ nichrome wire broken down, fan assistance of course, as for switches use Auto relays as these can handle 30A+ easily.
 

bob91343

Joined May 29, 2019
15
Hi all
Since I often have to deal with crappy bog standard pc power supplies, I am trying to build a 12v load tester (single rail for now) recycling some stuff I have laying around, but I am not sure about some of the components, and in general, the final design:

The load tester should be able to load max 80A on the 12v rail in 4A steps.
For the female connectors I am gonna desolder the 24pin and 4+4 pin from a motherboard, and for the 4pin molex and 6/8pin pcie I will get them from some spare cable adapters.
For the load itself I am gonna use a resistance from an old fan heater, cutting it at the proper length so to get 20x3ohm pieces (I already tested one and a 20cm piece of resistance gets just a bit hot with 12v taking 4A, and I will place them so that they don’t touch anything unnecessary with a fan blowing fresh air, so hopefully there shouldn’t be any fire hazard) and for the box I was planning to use some mdf boards, painted with fireproof heat reflective paint.
Now, what I am missing are the switches, I was planning to use some 10A lever switches for each load increase, but I would need 20 of them, so I was looking for some kind of rotary switch with 20 positions, but it should withstand a load of 80A when fully turned on, and I don’t think such a thing exists..any idea?

P.s.: I know I should test also at least the 5v and 3.3v rails, but for the sake of simplicity I am only focusing on the 12v single rail part, adding the other sections later if it proves successful.
An automotive relay would work but to save money, why not use banana plugs and jacks. You can get them cheaply enough and they can easily handle the 4 Amperes.
 

Thread Starter

adelage

Joined Aug 22, 2014
38
I was tempted to use the relays with some kind of rotary selector (yep, I found one with 20 positions!), it would be neat, but the overall cost is just too much (and also I should have to figure out a way to keep the previous relays closed every time I increase the load) so I am going for this kind of switch
B32DEF09-DE7F-406B-BFB3-B94692B4DBF1.jpeg
which I managed to get for a good price.
Now, from what I could understand watching some videos on youtube, it seems that measuring the ripple is not an easy task even for someone who knows electronics very well (which I am not for sure), but do you think it would be possible to use something like this thing to get a rough ripple measurement?
 
Last edited:

bob91343

Joined May 29, 2019
15
I was tempted to use the relays with some kind of rotary selector (yep, I found one with 20 positions!), it would be neat, but the overall cost is just too much (and also I should have to figure out a way to keep the previous relays closed every time I increase the load) so I am going for this kind of switch
View attachment 216117
which I managed to get for a good price.
Now, I know that measuring the ripple is not an easy task even for someone who knows electronics very well (which I am not for sure), but do you think it would be possible to use something like this thing to have a rough ripple measurement?
It seems that it should work. The bandwidth is limited but for ripple, probably ok. As for the switch, make sure it's rated for the current.
 

Thread Starter

adelage

Joined Aug 22, 2014
38
The switches are rated for 6A, I would have preferred 10A, but they didn’t have them..as for the toy digital oscilloscope, I cannot understand if I can use it to measure 12v
 

bob91343

Joined May 29, 2019
15
The switches are rated for 6A, I would have preferred 10A, but they didn’t have them..as for the toy digital oscilloscope, I cannot understand if I can use it to measure 12v
Well it seems it will do the job but there is no clue as to its power supply. Well there is a clue, that it seems connected to USB so that means 5V source.

You can get 5V from your 12V by adding a cheap regulator, say LM317 or LM7805. You will have to figure out how to program it or just connect to a computer via the USB.
 

Reloadron

Joined Jan 15, 2015
5,667
When I did this years ago for a hardware test site I used mechanical switches but if I were to do it again today I would use MOSFETs and to be exact Logic Level MOSFETs.

For the female connectors I am gonna desolder the 24pin and 4+4 pin from a motherboard, and for the 4pin molex and 6/8pin pcie I will get them from some spare cable adapters.
I just used a simple 24 pin / 20 pin adapter cable and hacked it.

I used resistors mounted on a 1/8" aluminum sheet. The resistors were wire wound. I used 100 Amp current shunts and amplified their outputs. I also used an ADC software driven solution.

Before I built the full unit I fabricated a demo for the hardware site owners. The remains of the original web page can be found here. The resistors in the proto type were Ohmite Stackohm series and the switches were I think Carlton manufacture rated for 10 or 20 amps DC. Remember you are switching DC. Not shown is a fan I used for heat removal.

Anyway, you might get some ideas from all of it. This may have been as long as 20 years ago. The finished final unit included some high power adjustable resistors so the current could be trimmed. We also did some test with case temperatures including PSU inlet and outlet temperatures under different loads. What images still remain can be found here.

Anyway today I would have done things differently including MOSFET switching rather than screwing with relays or mechanical switches. Think about how many steps you will be switching. I can buy adequate MOSFETs for a buck apiece. :) Before I forget the data acquisition was dumped into Excel so there was a record of loads and voltages. Additionally when placing the data in a spreadsheet anything out of tolerance printed in red.

Ron
 

Thread Starter

adelage

Joined Aug 22, 2014
38
Well it seems it will do the job but there is no clue as to its power supply. Well there is a clue, that it seems connected to USB so that means 5V source.

You can get 5V from your 12V by adding a cheap regulator, say LM317 or LM7805. You will have to figure out how to program it or just connect to a computer via the USB.
It looks like it gets powered either by a 9v battery, or any 8-12v psu, I found some videos on it on youtube, and apparently there’s is a small community behind this toy, since it is conceived mainly for educational use

When I did this years ago for a hardware test site I used mechanical switches but if I were to do it again today I would use MOSFETs and to be exact Logic Level MOSFETs.


I just used a simple 24 pin / 20 pin adapter cable and hacked it.

I used resistors mounted on a 1/8" aluminum sheet. The resistors were wire wound. I used 100 Amp current shunts and amplified their outputs. I also used an ADC software driven solution.

Before I built the full unit I fabricated a demo for the hardware site owners. The remains of the original web page can be found here. The resistors in the proto type were Ohmite Stackohm series and the switches were I think Carlton manufacture rated for 10 or 20 amps DC. Remember you are switching DC. Not shown is a fan I used for heat removal.

Anyway, you might get some ideas from all of it. This may have been as long as 20 years ago. The finished final unit included some high power adjustable resistors so the current could be trimmed. We also did some test with case temperatures including PSU inlet and outlet temperatures under different loads. What images still remain can be found here.

Anyway today I would have done things differently including MOSFET switching rather than screwing with relays or mechanical switches. Think about how many steps you will be switching. I can buy adequate MOSFETs for a buck apiece. :) Before I forget the data acquisition was dumped into Excel so there was a record of loads and voltages. Additionally when placing the data in a spreadsheet anything out of tolerance printed in red.

Ron
wow, it looks like a proper tool, even after 20 years it looks amazing, it even had software control!
I don’t have such high expectation, or better, I don’t have the budget/time/knowledge to make something like this, or anything similar with a more modern design (like using mosfets as you suggested), consider that I am spending something like 25€ for the switches, maybe 10€ for the fireproof paint, and eventually in the final design I might add a small digital amperometer/voltmeter display and a “toy” oscilloscope like the dso138, everything for about ~25€, the rest is gonna be scavenged from broken stuff I have laying around in my garage: it’s gonna be bulky, rough, and with a WW1 look, but it will be cheap and very easy to build, and to diagnose in case something doesn’t work.
I appreciate your suggestion, I will come back to it in case in the future I want to make a 2.0 digital version.
 

andrewmm

Joined Feb 25, 2011
658
Ive done this two ways in the past, I now use a commercial tester,

two solutoins,
a) resistors on a heatsink
b) transistors ( 2N3055 I think ) on a heat sink, with variable resistor on the gate to vary the current / effective resistance.

I ended up never varying the resitance , just giving the PSU a heavy load.
I use a comercial unit now for the same reason.

The faults I found using the fixed loads, or even variable loads, were,

a) Totaly dead power supplies
b) Dead rail of the PSU
c) PSU fails under load,

Which were most of the faults,

the ones I did not find, are the transient load problems,
GPUs in particular have a very very irregular power need,
the average might be 150 Watts, but they need 50 watts a lot, and 250 watts occasionally for a few milli second,
New PSU's cope with this, cheap / old ones don't.

so its good to make a load,
and it will detect a lot of faults,
but I'd not waste time with a variable load, have at most two settings ( say resistors in parallel or series )

and be aware of what the load will not help you detect.
 

Reloadron

Joined Jan 15, 2015
5,667
It looks like it gets powered either by a 9v battery, or any 8-12v psu, I found some videos on it on youtube, and apparently there’s is a small community behind this toy, since it is conceived mainly for educational use


wow, it looks like a proper tool, even after 20 years it looks amazing, it even had software control!
I don’t have such high expectation, or better, I don’t have the budget/time/knowledge to make something like this, or anything similar with a more modern design (like using mosfets as you suggested), consider that I am spending something like 25€ for the switches, maybe 10€ for the fireproof paint, and eventually in the final design I might add a small digital amperometer/voltmeter display and a “toy” oscilloscope like the dso138, everything for about ~25€, the rest is gonna be scavenged from broken stuff I have laying around in my garage: it’s gonna be bulky, rough, and with a WW1 look, but it will be cheap and very easy to build, and to diagnose in case something doesn’t work.
I appreciate your suggestion, I will come back to it in case in the future I want to make a 2.0 digital version.
Thanks and it was done around the needs of a business testing about 10 ~ 12 power supplies a week. For decades Intel published a Desktop Platform Design Guide covering ATX PSU form factors. You may want to give it a read with a focus on testing parameters you have an interest in. A PSU test bench can be as simple or complex as you wish it to be as well as as inexpensive as you want it to be. Start like you have already began and list the various parameters you want to test and to what uncertainty you wish to test. Then do as you are now and determine what hardware to use. I have used load banks ranging from old carbon piles to nichrome wire heating elements, they all got their intended jobs done. :)

When setting up your loads consider the wire gauges used.
The load tester should be able to load max 80A on the 12v rail in 4A steps.
A 12 volt 80 amp load is just under a 1.0 KW load and 80 amps in 4 amp step changes is 20 step changes. Additionally figure a single 12 volt 4 amp load is 36 watts so allow overhead in your loads.

Ron
 

Thread Starter

adelage

Joined Aug 22, 2014
38
Thanks and it was done around the needs of a business testing about 10 ~ 12 power supplies a week. For decades Intel published a Desktop Platform Design Guide covering ATX PSU form factors. You may want to give it a read with a focus on testing parameters you have an interest in. A PSU test bench can be as simple or complex as you wish it to be as well as as inexpensive as you want it to be. Start like you have already began and list the various parameters you want to test and to what uncertainty you wish to test. Then do as you are now and determine what hardware to use. I have used load banks ranging from old carbon piles to nichrome wire heating elements, they all got their intended jobs done. :)

When setting up your loads consider the wire gauges used.


A 12 volt 80 amp load is just under a 1.0 KW load and 80 amps in 4 amp step changes is 20 step changes. Additionally figure a single 12 volt 4 amp load is 36 watts so allow overhead in your loads.

Ron
For the gauge I was planning to use awg 14 for each step, thanks for the Intel guide link, will definitely come in handy..I would like to be able to test the nominal currents written on the psu labels, simulate the sustained full load of a pc before installing a psu that is not “branded”, things like that..it would be a nice plus to be able to get a measurement for the ripple voltage when under load, but I am still trying to figure that out

Ive done this two ways in the past, I now use a commercial tester,

two solutoins,
a) resistors on a heatsink
b) transistors ( 2N3055 I think ) on a heat sink, with variable resistor on the gate to vary the current / effective resistance.

I ended up never varying the resitance , just giving the PSU a heavy load.
I use a comercial unit now for the same reason.

The faults I found using the fixed loads, or even variable loads, were,

a) Totaly dead power supplies
b) Dead rail of the PSU
c) PSU fails under load,

Which were most of the faults,

the ones I did not find, are the transient load problems,
GPUs in particular have a very very irregular power need,
the average might be 150 Watts, but they need 50 watts a lot, and 250 watts occasionally for a few milli second,
New PSU's cope with this, cheap / old ones don't.

so its good to make a load,
and it will detect a lot of faults,
but I'd not waste time with a variable load, have at most two settings ( say resistors in parallel or series )

and be aware of what the load will not help you detect.
I believe that modern cpus have the same problem, they change their load very quickly and very often, maybe I should add some button that only when pressed applies a big load (say 150/200w).
I want to have the option to vary the load so that I can test the nominal currents written on the labels for each rail, which varies for each psu.
BTW, are you able to find the transient load problems with a commercial unit?
 

Reloadron

Joined Jan 15, 2015
5,667
Below is a few pictures of part of the final load banks. These were done using 1/8" sheet aluminum. The resistors happen to be Dale now I believe Dale Vishay but matters not as everyone makes them. They are 5 ohm 1% parallel pairs. I used AWG 12 simply because I had a bunch of AWG 12/2 Romex which I easily stripped and which happen to fit the hold size perfectly on the resistor terminals. The blue lead you see was one of many and was some fine stranded AWG 12 made by Teledyne and was silver plated. Years ago I ended up with a 500' spool which I would hate to buy today. Back then the resistors were about $2.00 USD each and today they are pushing $5.00 USD. I also used some high wattage variable resistors on each series of loads so I could tweak the current. One problem is as the resistors warm up even a little they change and it doesn't take much for the current to really start drifting. In the attached images I paralleled 5 Ohm resistors for a 2.5 Ohm total which was 2.0 amp steps on a 5 volt rail. The resistors are 50 watt so each pair was good for 100 watts and I only needed to dissipate 10 watts so tremendous overkill.

The sheets were stacked and force air cooled.

Load Bank 1.png

Load Bank 2.png

There was no need for thermal compound.

Next, as I mentioned I used switches, a whole panel of them. Today I would use Logic Level MOSFETs like for example the FQP30N06L which cost about $1.00 USD each and is a 60 Volt 32 Amp MOSFET with a very low On resistance. I would use a simple D flip flop with a push on and push off to move loads in and out.

Today there are dozens of inexpensive data acquisition units available which include charting software. A nice feature is the ability to actually see and chart what actually happens in milli-seconds what happens on a rail when a large load is placed on it. Then too it all depends on what you want and your budgetary constraints. :)

Ron
 

Thread Starter

adelage

Joined Aug 22, 2014
38
unfortunately where I live (Italy) those kind of resistors are a bit more expensive, while an heating element can be scavenged for free from any broken heater, even though it will not look so tidy and compact...I am interested in the data acquisition unit, can you tell me some model name I can look for on ebay? Also, about the ripple, is it true that it can be measured just with a good multimeter set on ac voltage?
 

Reloadron

Joined Jan 15, 2015
5,667
unfortunately where I live (Italy) those kind of resistors are a bit more expensive, while an heating element can be scavenged for free from any broken heater, even though it will not look so tidy and compact...I am interested in the data acquisition unit, can you tell me some model name I can look for on ebay? Also, about the ripple, is it true that it can be measured just with a good multimeter set on ac voltage?
Ah yes, Italia. I lived in the Napoli area for several years and my son was born there. This was early 80s and I recall we had a guy who would deal with a local parts house for us. I worked out of Capodichino. Actually lived in Ischitella on the beach. Fond memories. :) Understand how getting parts can be difficult and costly. :(

Ron
 

Thread Starter

adelage

Joined Aug 22, 2014
38
Ah yes, Italia. I lived in the Napoli area for several years and my son was born there. This was early 80s and I recall we had a guy who would deal with a local parts house for us. I worked out of Capodichino. Actually lived in Ischitella on the beach. Fond memories. :) Understand how getting parts can be difficult and costly. :(

Ron
:D:Dyep, great area if you want to enjoy some good holiday time, not so much if you plan to live there forever..and still Napoli is a big town with a lot of chinese import companies, I live in a small town down in Sicily, which being an island has very limited electronics/tech supply, I often have to order stuff from Germany/UK/China, with long delivery times, expensive shipping costs and sometimes annoying custom fees (anything above 23€ coming from outside EU is subject to vat+import fees)..also, I mainly deal with IT support/repair, only recently I am getting into electronics, as a hobby for now, so my budget for this kind of experiments is often pretty low
 

Reloadron

Joined Jan 15, 2015
5,667
During the early 80s I got down to Sigonella quite a bit. Had a good friend married to an Italian National names Bill (William) Walker. His wife was Conchetta (not sure about the spelling). Eventually he and his wife ended up in Sicily and as far as I know opened a restaurant called Conchettas's Kitchen (cucina). Not sure it still exist. Guessing today their daughter Jose would be in her mid 40s. :)

My mother's mother was Napolitan descent and her father the north of Italy. I really enjoyed all my time in Italy and Europe in general.

Ron
 

Thread Starter

adelage

Joined Aug 22, 2014
38
Funny how small the world is, Sigonella is not too far from where I live! Back in the 80s life was much easier in Italy, now aside from the sun and the historical monuments there’s not much left, especially the south is going down, unemployment rate is close to 40% among young people, and they are (rightfully) all running away. Anyway thanks for all the precious suggestions, I think now I got a better view of the setup I am gonna build!
 
Top