Bestec revisited

[everell]

Re: Bestec revisited

Now for scope pictures at 5vsb:
with good cap C1
no cap at C1 to simulate open
100 ohm resistor to simulate open and slightly leaky
47 ohm resistor to simulate open and leaky
6 ohm resistor to simulate very leaky

[everell]

Re: Bestec revisited

When I originally began working on the Bestec ATX-250 12E, the 5vsb was measuring 6.33 volts. Problem was that as I took readings in other areas, and came back to 5vsb, it was slowly increasing. By the time it reached about 12 volts I gave up on trying to fix this particular circuit. The pc board showed heat discoloration, especially around the diode D7. The 620 ohm resistor r23 was smoking, and the 25 ohm resistor in series with the LED in the optocoupler was also smoking. My final measurement on the cathode side of diode D33 was 55 volts.

In my opinion, the caps were crappy, the diode was probably marginal and unable to take the overload, the diode and cap were located in a crowded area near a power resistor, and the circuit is poor design and most likely going into thermal runaway from several components being overstressed.

I had opportunity to work on a Bestec ATX-300 12E which has the IC in the 5vsb circuit. No pc board heat discoloration, but the filter cap for the 5vsb was placed beside a power resistor - great way to cook the capacitor over time. This cap was bulging and was replaced. So I think the designers at this company did not do a good job of component layout and cooling techniques.

[Toasty]

Re: Bestec revisited

The 5vsb never went above 5v in your tests. I thought, from what has been said, if this cap opened, the output would go to 12v or thereabouts...? Instead, you show it goes to zero.

Nice pics btw. What scope is that you're using? Toasty is green with envy.

[everell]

Re: Bestec revisited

The scope I used is the Stingray by USB Instruments Model DS1M12. I bought mine thru Mouser for $220.

Scope probes - have to buy separate. I bought two 40 Mhz probes from Marlin P Jones & Assoc. (www.mpja.com) for $8.95 each.

Measurement of voltage across the C1 capacitor required an "isolated probe". I used one made by EasySync. An added expense but gives important protection to scope.

The software package for the scope included signal generator, frequency counter, digital voltmeter for channel A & B, and spectrum analysis. Downside is that the bandwidth of the scope is low. But for power supplies and audio work it is just fine.

The results of my measurements surprised me also. I think we are dealing with more than just a bad cap. Overall - my vote is "piss poor design" and my solution is pull the two transistor circuit and add DM311 chip circuit. Less time to do the conversion than chasing around trying to figure out this monster. My "converted" Bestec has not given any more trouble.

Of all Bestec psu that have the A6351A chip in the 5vsb circuit, I have found no mobo killers - only some bad caps. Of the three Bestec psu that had the 5vsb problem, all were Bestec ATX-250 12E. Don't assume all 12E are bad - the Bestec ATX-300 12E has the A6351A chip and is not giving 5vsb problem. And finally, don't trust the label. Look inside and make sure what you have is what you think you have!

Tom

[severach]

Re: Bestec revisited

Here's a Bestec ATX-250-12E Rev P5. The only difference I can find is the layout for the diode bridge and some nearby connections. This supply came from a junk dealer so I have no information whether or not it destroyed a motherboard. It won't power on so it destroyed itself so I'll just assume that it fried a mother board. The leads of D7 are mottled and gray. In back the leads for D7 are mottled and gray and the solder joint nearest the diode casing above “R28” shows heat damage. The glue is destroyed since I lifted the trace easily by pressing on D7. The solder joint for the small brown cap C16 above “ZD4” also shows heat damage.

The 220uf C50 ESR is 38 ohm. Though the 10uf C1 is not bulged the ESR is >99 ohm which might explain the heat damage under C1.

my solution is pull the two transistor circuit and add DM311 chip circuit. Less time to do the conversion than chasing around trying to figure out this monster.

It's nifty as a proof of concept but buying a pallet of HiPro seems like a cheaper solution. I only save supplies that are easy to fix and good. These Bestec are neither.

[everell]

Re: Bestec revisited

Hi severach - I agree that a pallet of replacement HiPro psu to replace Bestec is a better business solution for a computer repair shop. My objective for this thread was only to show that ALL Bestec psu were not the mobo killers. The Bestec ATX-250 12E is bad for sure, and maybe some others. But several we have discussed in this thread do not have the same problem.

As a hobbyist, buying a pallet of HiPro psu is not an option! I would have little use for that many. So I will have to claim "hobbyist" status. I simply wanted to explore what the problem is for this mobo killer.

Your photos were EXCELLENT. They look just like my Bestec when I first began. The worst heat damage seems to be at diodes D7 and D33. Next is capacitors C1 and C220, which are connected to the diodes. There is also heat damage to pc board at R102 which indicates that transistor Q17 was running hot. Look at the middle optoisolator - it is discolored compared to the other two. Now look at the foil side of the pc board - resistor R75 (surface mount resistor) - it also shows discoloring from overheating (mine was smoking!). Your photos show the same story my psu was telling.

The tests I ran on capacitor C1 with the scope photos did NOT show a significant increase in 5vsb voltage. But when this psu goes bad, the 5vsb does go up to about 12 volts. If the 5vsb goes to 12 volts, the voltage at C1 and C50 has to go up also. I remember measuring about 55 volts on the C50 circuit. My conclusion - several things going bad due to poor design and poor cooling. Therefore, the Bestec ATX-250 12E is like a time bomb waiting for the opportunity. Replace it immediately.

My repair solution for this psu was not a business solution, merely a hobbyist way of making it workable again. With all of these pictures and discussion, I feel that I have found a way to control the 5vsb circuit, but I am NOT convinced that it solved all of the problem. It may still be a time bomb. Time will tell.

Final conclusion: Bestec ATX250-12E bad psu - pull it. Any other Bestec, look for the A6351A chip. If there, it probably has bad caps and is repairable. Anything else, toss it or better yet, post some pictures of it on this forum and we will talk about it.

[severach]

Re: Bestec revisited

First I have a Bestec ATX-250-12E Rev P7 0435 that won't start with a fried T2893 motherboard. The pictures show how damaged the solder joints get from the heat.

Next I have a Bestec ATX-250-12E Rev P6 0245 with a fried T2042 motherboard. The power supply works enough to do some tests. There is a 60 Hz hum in the 5vsb section with the power on and the supply off or on. The hum stops immediately when the power is cut even though the input caps are still fully charged. The hum audibly changes when the supply is started and when C50 is moved. The hum sounds like an arc but is not producing the smell of a corona and the arc can't be seen in the dark. If it is an arc then it is inside some part.

The CoolMax power supply tester 5vsb display alternates between 5.0, 5.1, 5.2, and HH which means high voltage. A DMM shows the 5vsb running between 6.02v and 6.18v. After about 5 minutes it was running at 6.39v. After about 10 minutes it was running about 6.66v. The longer I leave it the higher it goes. At 6.81v the Coolmax power supply tester displays HH only. Turning the power supply on drops the voltage by 0.65v. After 25 minutes of arcing the voltage is at 7.25v. The voltage is stable as it increases. At 30 minutes I see 8.05v and now the power on voltage drop is 1.05v. If I switch the power supply off and on the voltage returns to exactly where it was.

At 35 minutes the voltage went wild. The buzzing disappeared and the voltage rapidly went from 8.65v to 15v and now the power supply is dead. It will still produce 15v on the 5vsb but it will no longer start. I pull and test parts with a meter hoping for a short or burn out.

C50: 128uf, 3.0 ohms
C1: 3.7uf, >20 ohms
D7 1N4005: tests good
Q2 2SC1815: tests good NPN
Q17 2SA1015: tests good PNP
D33 PR1002 horizontal: tests good
D8 1N4007: tests good
PC1 PC2 PC3 ET1109: good
D6 PR1002 vertical: good
D24 RW34 or SB340: good

One of these parts should be burned out yet the meter shows them all good. I've burned out the P6 supply so I can't test there so I'll transfer parts into a good P7 supply until it fails.

To break a lot of glue and make other parts easier to desolder I start with transformer T2. Voltages are good.

D7 is our biggest suspect. Voltages good.

Q2, Q17, D33, D8, PC3, PC2, PC1, D6, D24, voltages good.

None of the suspect parts have failed. I burned something out hitting 15v so I'll swap the remaining parts until I find it. I switch the primary transistors Q24 2SK2749 Enh N-Ch Mosfet and Q80 2SC5027-R NPN found in Rev P7. The Rev P6 supply has 2SK2850 and 2SC5353. The Rev P7 board supply works fine with the Rev P6 parts.

Next up is C16 a tan 10nf ceramic disc capacitor in the middle of the heat, heavily abused but produces good voltages.

IC1 UC3842BN, IC4 TPS3510, IC7 LM358N, all work fine.

D19 H48, D24 H48, good.

Output diode assembly: D100 STPS30-45CW, D103 BYQ30E, D131 STPS20-45CT, U125 7905CV; all good.

Q16 2SC1815, Q13 2SC1815, IC3 TL431AC, IC5 PJ431 or TL431AC, IC2 PJ431 or TL431AC. Changing IC2 changed the 5vsb voltage a bit and it is part of the circuit.

Q14 2SC1815, Q10 2SC1815, Q11 2SC1815, Q12 2SC1815, Q9 V-B772 PNP, Q6 V-B772, D30 H48, D25 H48, D26 H48, D15 UF3003, D16 UF3003, D10 PR1002, D29 PR1002, D21 PR1002, D22 PR1002, D31 PR1002, D11 PR1002, D1 PR1007.

All the semiconductors on the top work. Now for the bottom.

ZD2 green green red with ZD2 yellow, D3, ZD4 green red brown, ZD1 green green red with ZD2 yellow, Q4-LG NPN was destroyed on removal, D50, D5, D4, D18 exploded from the heat on removal, D51, T1. Damaged parts were not swapped.

No semiconductors damaged except possibly D18 or Q4 so I switch all the large output caps, still works.

With almost all of the parts out I go around and check all the resistors. Only 1 is out of spec: R65-221 should measure 220 ohms but measures 744 ohms.

Frustrated I transfer the bad caps starting with the bad C50, voltages are good. Then I put in the bad C1 and the hum is back and 5vsb jumps to 5.75. Then I put the good C50 back in and the hum is quieter but still there so C1 alone can cause hum. Then I put my good C1 in parallel with the bad C1 and the circuit works without a hum so C1 and C50 are only arcing because the bad C1 is out of spec.

So now I need to duplicate the failed capacitor C1 with good components that I can measure. I rig up a good 10uf 50v for C1 with a 400 ohm potentiometer. Testing shows that my meters need to be hooked in up to 5 places. I measure the DC voltage of 5vsb, the DC voltage on C1, the DC voltage on C50, the AC voltage across my artificial ESR, and the resistance of my artificial ESR. The resistance and voltage will tell us the power dissipated by the artificial ESR.

C1 10uf 50v; C50 220uf 25v
ESR Ω  5vsb  C1-DCV C50-DCV ESR-ACV ESR-Watt
0    5.0v  11.5v  17.7v  0v    0
12.2   6.09v 12.08v 21.1v  0.04v  ~0
20.7   6.66v 12.80v 23.5v  0.30v  4mW
29.6   7.61v 13.40v 27.2v  3.15v  250mW
40.2   8.70v 13.8v  29.7v  7.9v   1.5W
51.4   9.45v 14.00v 31.5v  10.6v  2.2W
62.8   9.57v     33.5v  13.7v  3W
72    9.93v     35.2v  14.8v  3W
85.3   10.4v     37.6v  16.0v  3W
101.8  11.26v     39.7v  16.5v  2.7W
200   ?   ?    BOOM!  ?    ?

Standard DMM do not measure ACV well on high frequency circuits so ESR-AC may be inaccurate. The power dissipated seems consistent with how warm my artificial ESR gets for the short time I allow the circuit to operate.

The numbers vary a lot from run to run.

The voltage rise resembles 1v per 10 ohms. ESR in C1 is unintentionally interacting with the rest of the circuit to call for increased voltage.

So long as C50 is in the circuit it is hard to get 5vsb to exceed 12v.

When the ESR of C1 passes 25 ohms the voltage blows out C50 which causes the wet spot in many dead Bestec supplies. C1 is radiating 1W of heat and will soon radiate 3W of heat so it rapidly breaks down. After the blown C50 has baked its voltage resistance away it arcs under the excess voltage caused by the high ESR of C1.

C50 is not required. 5Vsb will produce the proper voltage with or without C50 so long as the ESR of C1 stays low.

Once C50 is out of the circuit it is easy to get 5vsb to exceed 15v.

10 seconds at 200 ohms ESR blows C50.

A 1uf cap increased 5vsb a bit probably from the extra ESR. A 100uf cap ran 5vsb at the same voltage as the 10uf cap.

I know that the 5vsb can go as high as 15v and I recall that around 15v the power supply will not start even though I found no damaged semiconductors. To get 5vsb that high I disconnect C50. Now I can dial the 5vsb voltage anywhere I want and C50 can't blow out. The power supply will not start if 5vsb is above 14.5v. The power supply tester doesn't even blip so the supply knows it should not start under this condition.

Hoping to find some semiconductor damage I go back and repair all the Rev P5-P7 power supplies that would not start. A new C1 and C50 and they all run with normal voltages. The bad caps had already been pulled so I couldn't tell if the 5vsb was still running.

So I know that C1 dies first which quickly destroys C50 and the motherboard but I still don't know what causes C1 to go bad so much sooner than the other capacitors. I don't have a meter that will test ripple current with ~0 ESR to see if it is too close to the max value 46mA for a Jamicon 10uf 50v 105*C TK. I also don't have an infrared camera to see if any nearby parts produce excessive heat during normal operation.

Now that I've made the Bestec ATX-250-12E Rev P5, P6, and P7 perform from best to worst, let's see if I can get a Bestec ATX-250-12Z Rev D2 to produce a damaging 5vsb. C12 2.2uf 50v, C13 0.1uf 63v, and C50 220uf 25v are the same as in the other supply. C60 10uf 50v matches the capacitor that fails in the Rev P5-P7 supplies. I add 0-20 ohms of ESR to C60 and the 5vsb voltage doesn't change. I do the same with the remaining 3 caps and the 5vsb doesn't change.

Other failed caps will be caught by the TPS3510. The A6351 equipped Bestec look good to me.

[everell]

Re: Bestec revisited

Hi severach - just wanted to thank you for the time spent analysing the Bestec ATX-250 12E. That report was fantastic, well written, and easy (for me) to understand. wizzard, i4004, and severach all said that the experiment I did was incorrect, that instead of open capacitor (no capacitor) and a resistor, what should be done is to use good capacitor and series resistor to simulate esr. Your method duplicated the problem just as I experienced it. Now I want to duplicate your experiment and look at some scope pictures!
It is so easy to have an opinion, make dogmatic statements, then find out how wrong you are. So I admit I was wrong on this one, and have egg on my face! But thanks to the efforts and contributions you have made, and the patience others have shown to us newbies, I think we have all learned a lot. Most important is our concept of capacitance versus esr, and how to correctly simulate a problem.
So again thanks. Your report was the best analysis I have ever seen. I think the college kids could learn a lot by reading your report and going to the same thoroughness in reporting to their professors.
I really enjoy this forum. Just didn't know I would stir up so much controversy over a power supply I was just playing with.

[everell]

Re: Bestec revisited

I did an experimant similiar to what severach has done. Using Bestec ATX-250 12E with C50 removed, and simulated esr in series with capacitor C1. The capacitor used for C1 measured 9.70 uF, .046 DF, 1.8 esr. I used 7 different resistor values to simulate the esr. I measured the 5vsb for each resistor value, and made a scope picture across C1 for each resistor value. Here is a table of the voltage readings:

Resistor Volts on 5vsb line
6 ohm 5.62 Volts and increasing
12.5 ohm 6.32 volts and increasing
25 ohm 7.65 volts and increasing
50 ohm 10.37 volts and increasing
100 ohms 13.94 volts and increasing
150 ohms 16.22 volts and decreases
200 ohms 17.0 volts and decreases

Scope photos attached

[everell]

Re: Bestec revisited

scope photos of signal across capacitor C1 during esr testing - put in better format

[everell]

Re: Bestec revisited

Now that we know how to make the Bestec 5vsb line go high and blow mobos, how do other brands compare? I ran the same ESR test on a SeaSonic SS-250FS. Date on the pcb is 2001.05.22. The capacitor in similiar circuit is C24, 47 uF/50 volts. When I forst got this psu the 5vsb measured 12 volts. The capacitor esr measured >99. See attached document for scope pictures and voltage readings.

[everell]

Re: Bestec revisited

The next psu I tested was a Raptor LC-B400 ATX (400 watts), made by L&C. The capacitor in a similiar circuit is C8 22 uF/50 volts. It measured 19.62 uF with esr=1.2. In the test I used a new Nichicon PWM series cap which measured 22.78 uF with esr=0.40. See attachments for scope pictures and 5vsb voltage readings.

[everell]

Re: Bestec revisited

Now lets look at a chart that compares the three brands on 5vsb voltage versus simulated esr resistance.

Comparison Chart – ESR versus 5vsb voltage



R 5vsb
ESR resistor SeaSonic Raptor (L&C) Bestec

6 ohms 5.00 volts 5.07 volts 5.62 volts
12.5 ohms 5.01 volts 5.08 volts 6.32 volts
25 ohms 6.46 volts 5.57 volts 7.65 volts
50 ohms 8.55 volts 6.69 volts 10.37 volts
100 ohms 10.31 volts 8.65 volts 13.94 volts
150 ohms 15.20 volts 9.48volts 16.22 volts
200 ohms 16.97 volts 10.63 volts 17.0 volts

As computers with the two transistor design get older, I think the cheap psu with cheap brand capacitor powering a cheap mobo is somewhat like a time bomb. After all, who bothers to see what the 5vsb is doing til its too late. As for the above comparison test, I agree that the capacitor is the root cause of the Bestec failures, but I think something else is aggravating the situation because the voltages are so much higher than the other two brands tested.

[Toasty]

Re: Bestec revisited

Gosh, I wish you'd go back to the regular pics. Now it's download or open in Word.



Toast

[everell]

Re: Bestec revisited

Toasty - the comparison chart just above your post came out terrible, so I put it in document form so that it will make some sense. As for the documents with the photos, I spent a lot of time making them look pretty .........just for YOU!

[Toasty]

Re: Bestec revisited

>>I spent a lot of time making them look pretty .........just for YOU!<<
-blushes-

Your efforts are appreciated.

I just think of the others who may not have Word or its Open Office counterpart.

If you use the "code" formatting option, the charts come out fine. It's a fixed width font.

The 22uF cap in the Raptor falls in spec going by the chart on the Blue ESR meter. By that it could have an ESR of about 5 or less and still be considered good.

>>In the test I used a new Nichicon PWM series cap<<
Do you mean the PW series? If so, then that cap is (per spec sheet) supposed to be 1.2 ohm, which would match the one that was there.

All this destruction for the cost of a 3¢ zener and a 1¢ resistor... sad.

Toast