The "Discuss this article in our Forums" link does not point to this thread.

Another congratulation for improving the rig.
I am always surprised by your determination to deliver the highest quality tests possible. The amount of work you are putting into the testing and it's preparation shows that SPCR is my favourite website with good reason. Keep on the good working!

Oops, sorry about that, link corrected.

Thanks for the vote of confidence, jojo4u.

It's certainly impressive to see someone own up to a mistake/problem with testing methodology! IMO, that give you more credibility, not less. Good update!

Greetings Mike,

As always, thanks for the thoroughness and the honest transparency of SPCR articles. It is quite refreshing, and so much more helpful than the over-hyped and self-promoting approach taken by some other web sites; congratulations!

I would also point out that the set up is likely to be a good simulation for a real computer -- the heat produced by the load resistors will heat up the interior of the case in proportional to the load; so, the noise levels for a given load level will be a good approximation of a real system.

The Dale RH25 0.01 ohm shunt resistor has a 100PPM/C temperature coefficient.
Datasheet: www.sal.wisc.edu/PFIS/docs/archive/ public/Product%20Manuals/vishay/rhnh.pdf

The interior of the case can get 30C warmer than ambient as shown the example Seasonics chart.

Where is the shunt resistor located? If it's inside the case (especially if it's near the hot load resistors), it might be heated to 30C or more beyond ambient. This will cause its resistance to increase by 30*0.001 = about 1/3 of 1%. Yes, it's a small error, but you're measuring efficiency to a precision of 1/10 of 1% so this error is within measurable limits.

Probably this error wouldn't be hard to remove by simply calibrating with your specific resistor's R vs C response curve (very likely simple a linear relationship) and making a simple formula for current based on both voltage drop and resistor temperature. Annoying though... and probably not worth it for improving accuracy by 0.3%.

spworley --

Intersting point. I had not considered that, really, tho I did take care to try and keep the shunt resistors cool.

Actually, for the Seasonic PSU, the ambient was 21C and the max seen in the box was 46C, so the temp rise was 25C. I don't think it has ever gone higher. Higher power PSUs have more airflow, while lower airflow PSUs don't put out as much heat.

3 of the shunt resistors are bolted tightly to underside of the front panel of the DBS-2100 PSU load tester. In other words, the top panel, which is made of fairly sturdy steel, is being used as a heatsink. On the right side where they are located, there are no load resistors directly beneath them. One of them is actually bolted right up against the front edge vent intake of the DBS-2100. The grill is visible in this photo:



This grill is the only way intake air gets to the 4 fans blowing out on the other side. I've described in the previous PSU test rig article that these are mid-speed 80mm fans undervolted to 5V; I'd guess they each push maybe 10-12 cfm in free air; inside this tightly packed PSU loader, it's probably half of that -- maybe 20~25 cfm in total. All three shunt resistors are quite close to the air intake path, and you can feel a fair amount of air being pulled in at these vents. In other words, even though the shunt resistors are inside, I doubt we'd see 30C above ambient in their location.

As for the 4th shunt resistor, it is bolted to the aluminum inset panel for the 5 switches for the +12V2 load bank. The screws for this resistor can be see in this photo (the 2 in the center):



Because of the heatsink action of this aluminuim panel, one side of which is on the outside of the thermal simulation box, the temp of this resistor is considerably lower than that of the internal PSU "intake temp" sensor that is positioned in the box about 3 inches behind it. When you put your fingers on that panel, even when the PSU is at >400W load, it never feels more than warm.

So I suspect inaccuracy due to shunt resistor temp rise is not really a cause for concern, espcailly as the potential error is so tiny compared to potential errors elsewhere...

But thanks for bringing my attn to this.

It's probably a very small effect anyway.. and that's still worst case scenario.

I thought of it only because of a magnetic field physics experiment I did in college involved precise current measurement. It was important to do temperature calibration. It was terribly hard to get accuracy better than 1%! We also used a 4-wire measuring rig, which gave us a 5X accuracy boost. You can buy 4-contact low resistance shunt resistors designed for current measurements. We got to about 0.1% accuracy but with a lot of work.

Some info about 4 wire measurement (Thanks Google!)
http://www.allaboutcircuits.com/vol_1/chpt_8/9.html
Top half of the page is about resistance measurement. Current measurement in the second half.

My hat's off to your continued dedication in effort and time to get the best results possible, Mike. Surely there are no other review sites that can touch SPCR in this area now.

My mind boggles as to what continues to drive you, and for what personal reward to yourself?

Those questions arise from time to time... I suppose the main thing is to try and do it right. By continuing to raise review standards higher, SPCR increasingly becomes the standard bearer. It's a slow and hard way to build traffic... revenue on the web always boils down to traffic.

MikeC writes "I suppose the main thing is to try and do it right. By continuing to raise review standards higher"

Talk about raising the standard! You're already the gold standard in my book on PSUs, and _the_ standard in regards to noise review of any hardware - and you're still raising the bar higher!!

Trying to do it right is something not enough do - congratulations on doing it right and then some.

Whenever someone asks a PSU question on the other main forums I participate in (whirlpool.net.au and overclockers.com.au) SPCR inevitably comes up as the definitive reference - something which I can see is going to continue.

Er. hm. this is a massive undertaking. really big effort. I am certain that whatever review comes out of this, it will be THE review.

my only hope is that mikeC gets to test 3-4 various versions of the same psu's. That always is important it seems when people have odd claims to what their psu is doing or is not doing.

How about something new??

Like, if your psu is really crappy but shouldnt be, you send it to mikeC?

Thanks for explaining the whole process of discovering the problem and deciding on the solution. It helps me understand how much work went into improving the PSU test system. I would have been tempted to just throw out the PSU efficiency measurement instead of going through all that effort.

It's the attention to detail, both in the work and in the write-ups, that makes SPCR the best hardware review site on the Web, and even one of the best sources of information about new PC hardware being published or broadcasted anywhere. You don't publish as frequently or cover the market as broadly as other major review websites, but the level of trust people have in SPCR is amazing. Keep it up!

wow!! excellent job. the commitment to accuracy is amazing and i am glad you are setting such a great standard

Wow!

MikeC, What sort of LCD display did you use for measuring voltage? i.e. what brand/model?

Way to go man! Excellent writeup. Ace!

I knew that Fluke 36 would come in handy! Thank you ebay!

Good work, Mike!

Using a shunt to measure current draw is genious. It must be a pretty common practice in the industry, because the multimeters we're building in one of my labs also uses a shunt to measure current.

Although there is more than 1 year from this artical. I must admit you did very good job. Thanks for this info cause it prooved more than useful.

Ummm... did you know that we upgraded our test system yet again recently?
SPCR Power Supply Test Rig, v.4

Hi editorial team, I just noticed that the Recommended PSU article is starting to look a bit out of date, for instance no mention of Seasonic M12 and S12 Energy Plus models, or Corsair; a quick touch-up to reflect recent reviews would be most welcome.