RIP-Felix
Senior Member
A Retrospective Analysis of the NEC/TOKIN Proadlizer
A proadlizer is an Aluminum "Hybrid" Capacitor, with characteristics between a "wet" electrolytic and "solid" polymer. In Development and Mass-Production of the New "Proadlizer" Decoupling Device here is NEC describing it's construction:Because of it's chemistry, it's performance must fall in between the extremes for those 2 types of capacitors. So lets first define those extreems. Then we can evaluate NEC's promotional product documentation to see if the Proadlizer's performance data falls within the expected extremes. We should also critically evaluate their statements and graphs for spin, which is expected from promotional information.NEC said:
The following is taken from KEMET's published information on the advantage of Aluminum Film Polymer Capacitors over Aluminum Electrolytic Capacitors. Let's first look at change in capacitance over time, also called capacitor aging:
This characteristic happens regardless of temperature or use. It is simply the time since manufacture. Even if the cap is never used, this graph shows it's capacitance will decay. Electrolytics hold for awhile then drop off quickly. This has been called "drying out." That's part of it, but the other part has more to do with the dielectric chemically degrading. "Electrolytic capacitors (aluminum electrolytic, tantalum etc.) like batteries have their functionality based on a chemical reaction. Because of entropy, time will eventually slow, stop or reverse that reaction and the capacitors will become non-functioning. Electrostatic capacitors (ceramic, polymer film, etc.) do not function due to chemical reactions, but ceramic capacitors contain certain base elements and dopants that can radically affect their longevity" (1). Entropy dictates that the dielectric electrolyte cannot maintain maximum stability, regardless of its chemistry. Liquid dielectrics decay the fastest, polymer or solid dielectrics are very stable, but cannot escape the laws of thermodynamics. As you can see they too degrade slowly as well. However, their decay lacks the dramatic exponential drop off, instead following a more predictable linear decay. This makes them far more reliable.
It has been claimed that heating capacitors can restore capacitance. This sounds like the fountain of youth, the holy grail solution. Just bust out the hot air station and heat the capacitors back to their youthful glory! Like most "all or nothing statements," there is a seed of truth, but it doesn't apply to everything. So the question is, what's the seed of truth and which type of capacitor's does it apply to?
On Page 53, @Shinichi Kudo posts that NEC promotional research paper (linked above) on Proadlizer construction and performance. He also quotes another article and speculates as to why ESR might increase. That analysis has to do with sealed aluminum electrolytic capacitor and deformation due to high pressure gas during reflow. 'The plastic deformation of the casing reduces the contract pressure and thus ESR changes.' This doesn't happen in NEC/TOKIN proadlizers. They have a completely different construction. So that's out.
However, @Deus had earlier brought up an explanation on Page 50, one that's not so easily dismissed.
In physics and material science there is a property called the Curie Temperature (TC) or Curie Point. It is defined as the "temperature above which certain materials lose their permanent magnetic properties, which can (in most cases) be replaced by induced magnetism" (2). According to Rick A. Price of Paktron Capacitors, "Ceramic capacitors [MLCC] can be de-aged by subjecting the capacitors to temperatures above the dielectric's Curie point (>120°C) after which the aging process starts anew" (1). While He doesn't elaborate on the physics behind why it works, he does state it works on the solid dielectric used in MLCC capacitors.
Usually, temperature has a negative effect on capacitors. The following graph shows the theoretical lifespan of capacitors exposed to various temperatures. The higher the temperature, the faster they die:
There are three points I want you to note. First, it clearly shows why heat is bad for electronics. The more heat, the faster your console will die. Second, it shows that at higher temperatures polymer caps lose their lifespan advantage. Perhaps they live ten times longer at 65C, but at 90C it's only twice! Lastly, the graph stops at 105C. It doesn't show us the effect of heating the caps beyond the Curie point! The reason is probably because the metrics they use to graph capacitor lifespan no longer provide any meaningful data above a certain temperature. The physics that allow the capacitor to work breakdown. Beyond this point, there is no longer any useful data KEMET cares to provide. That may limit of their product information, but not our curiosity.
It's certainly possible that Rick A. Price is right, that the capacitors with solid dielectrics could be "deaged" somewhat by molecular changes that occur in the dielectric above the Curie temperature (TC). How and if that applies to the NEC/TOKIN is completely unknown, but a critical question to answer. So let's dig into it!
First, Rick talking about Multi Layerd Ceramic Ccapacitors, not Aluminum Polymer. Both use a solid dielectric, but the specific chemistries and deposition techniques are different. It's possible that heat can have the opposite effect and damaged the NEC/TOKIN proadlizer beyond repair. Who knows? However, this cuts both ways. Since it requires heat above 218°C to remove the NEC/TOKIN proadlizer intact, it's probable the Curie point of the capacitor was exceeded. So maybe they were deaged by the process. Proponents of the "heat test" say that it restores capacitance allowing your console to boot. Could this be the underlying mechanism?
Hoping to answer this question once and for all on Page 31, @squeept does testing on a console without a YLOD, BGA or Tokin problems. He was able to remove 5/8 tokins before encountering any display glitches. The console was still able to boot! Unfortunately, the YouTube videos are missing because he later deleted his channel, saying it was because he was getting too many requests to fix consoles. Regardless the console was able to boot without a YLOD with an astounding 60% less capacitance and at least twice the ESR! Since the videos are down and no one mentions it, I would like to know which chip has only one tokin remaining - RSX or CPU? Because one of the chips was receiving 50% of it's capacitance and twice the ESR, while the other was only receiving 25% capacitance and 3 times the ESR. The fact the console worked at all is a credit to how well SONY's engineers designed the PS3!
So if the filter caps can fall so far out of specification before failing, surely we could test them to see if they are bad or not! Right?
Well, yes. You can easily test for bad tokins using an oscilloscope. @squeept artificially created "bad tokins" in his workshop by baking new proadlizers in an drying oven @120C for a month! He installed them on a working console and use an oscilloscope to compare the waveforms. Sadly these images are not recorded on the forum. However on Page 56, He found his first bad tokins in the wild! He posted Oscilloscope images of the waveforms they produced, which are completely different than a working Proadlizer or Tantalum array. He confirmed these real bad tokins looked "almost exactly" the same as the ones he baked. So that's definitely what bad tokins look like on the oscilloscope. He added, it 'took him 40 consoles to find one set of bad tokins.' Fast forward a year later and he has only found 3 consoles that match these results. @squeept does that sound about right?
What about the claim that capacitance can be temporarily "deaged" by reballing temperatures or the "heat test?" Most people don't have an oscilloscope and are not willing to spend the money to buy one, just for one use. Also, if you believe the myth, were you to try to remove the tokins intact, it would require hot air and temperatures that would "deage" them. So you are not able to then measure electrical characteristics such as capacitance and ESR ex situ. Again, if the myth is true. So you couldn't be sure your measurements apply to the proadlizer as it was when in circuit. You would have to electrically isolate them in situ to accurately measure the Capacitance and ESR. Preferably using a network analyzer to obtain a frequency spectrum, or S21 attenuation curve. Then we could see the actual health of the NEC/TOKIN array, without any doubt to what effect the minimal amount of time above the Tc had on the dielectric. To solve this problem @squeept "showed everyone how to isolate the capacitors on the motherboard for capacitance and ESR measurement" in situ. (Pg58)
I wonder if this console was baked in the oven? I remember that was something people used try and reflow the chips using! Super bad idea, but it might explain those tokins. Also, he said, "I tried the "heat test" out on these AFTER reballing and BEFORE replacing the caps - it does nothing. I'm 100% confident now that if the "heat test" works, you have a chipset failure or a BGA issue." I want to know what temperature he used and for how long. But still, the idea that a reball would reset the aging of the caps didn't prove to true in this one case.
- I wish he would have given us the history of that board, however. If it was from a previous workshop they may have left it in the drying oven too long. Was there a bunch of other work done on it or was it sealed. Also, was the reball successful? Was the tokin replacement successful? Was he able to sell that console? Or was it DOA never to be reanimated? As far as I know he did not use this technique on the console he confirmed had bad tokins.
- On page 53, Shinichi Kudo makes another interesting post about the importance of ESR and speculates about it changing in the prodilizer depending on heat because of the "silver paste in between as electrolyte." He was onto the trail I explaind above. @squeept responded to this, "I already showed video of a system running TLOU and GT6 just fine with half of the capacitors removed." So that answers that, right?
ESR is one of the most important factors for choosing capacitors. Especially in this application because a High ESR can greatly reduce the efficiency and therefore lifespan of the filter. "ESR determines the I2R heating losses for the capacitor, which in turn establishes the efficiency, pulse handling and indirectly the reliability of the circuit" (1). The whole point of replacing the NEC/TOKIN proadlizer with Tantalum is to make it last longer. All other capacitors are disadvantaged by ESR performance comparatively.
I totally get why SONY went with these Proadlizers. The other options require a combination of Tantalum and/or Aluminum and Multi Layered Ceramic Capacitors. And even then, the frequency response curve is not quite as good. This means noise in the higher and lower frequencies will see a higher ESR, which induces more ripple current and heat in the TaPol cap (dissipation factor). That will shorten it's lifespan. Evidently though the industry has decided it's worth it, since proadlizers have been abandoned in favor of the more expensive TaPol caps. They must have decided Tantalum was more reliable despite the apparent disadvantage. Or did they? We have all been operating under that assumption, but we still need evidence to support the calim.
Comparing ESR vs Time it's clear we don't have to worry about polymer:
This next graph is from NEC/TOKINs promotional research paper (4). Notice that it only takes the Time out to 1000 hours. Now look above at how long it took for the electrolytic caps to begin showing significant increases in ESR. They would have needed to show this graph going out to 5000 hours to be meaningful. As it is, we have basically not idea if the ESR of the NEC/TOKINs ages or not.
Assuming the tokins are like other polymer caps, ESR shouldn't age. However, Proadlizers are actually a Hybrid device. They have that silver paste inside that can degrade, along with other materials that are subject to the laws of thermodynamic (entropy always wins!). It's possible they chose this graph to mislead us into thinking they are as reliable as other polymer caps! Basically, the graph they provide is useless. This is why you need to be carful when reading promotional materials. So I'm still skeptical!
I have previously stated that ESR decreases with temperature, but when I was questioned on it by @squeept, I could not find what I read that gave me that idea. Yes, it's documented in Aluminum electrolytic capacitors that ESR decreases with temperature, and I had seen the above showing they decay with time. Perhaps I made the jump in logic, that it applied to polymer also. That was @squeepts criticism, which is fair. I did some research and couldn't find anything that said polymer cap's ESR decreased with temperature and quickly concluded I must have been mistaken. That was until I found this:
While it's three orders of magnitude smaller, it does show the ESR of Solid Polymer Aluminum Capacitors, which is what NEC/TOKINs are, decreases with temperature. This information combined with the possibility that the capacitance could also be increased by temperature have me re-evaluating claims made in this thread!
Anecdotally, people have been reporting that heating the caps has allowed their console to boot. There are two plausible, yet competing hypotheses for why this is occurring:
- Thermal warping of the PCB caused a BGA defect to mechanically reconnect.
- Heating the capacitors temporarily restored enough capacitance to allow the filter pass boot checks.
Trying to put this to rest @squeept previously conducted an experiment. He heated the NEC/TOKINs while probing them with his oscilloscope. He stated it had no effect on the noise. If there were a lack of capacitance, and it could be restored by heating, then the noise should have decreased. He reported no change. That is a damming result, but in Science we need replication to rule out procedural errors and one-off events.
One problem is that he didn't provide a very clear description of his procedure for anyone else to follow, nor did he say if he has repeated the experiment since. I would also like to know the details of the YLOD on that console to see if it was a likely candidate for bad filter caps as the OP originally described (instability in games, a delayed YLOD, not instant).
If the console had confirmed bad tokins, that's one thing, but no one has defined the Vpp for unacceptable noise. I have read that (+/-)50mV is generally good enough for processors, but looking at the datasheets for the RSX's onboard DDR memory the acceptable noise is (+/-)20mV (VDDR, VDDQ, VDDA and VDDIO). Granted those voltages are supplied by linear regulators driving N-channel MOSFETs, not the VDDC that feeds the GPU. VDDC is coming off the second stage filter. The NEC/TOKINs are bulk filter caps for that area of the second stage filter. The reason for them is to filter noise generated by the switching regulators converting 12v to the ~1.3v the GPU needs. Switching VRM are noisy, whereas linear VRM generally isn't. However, if more than 20mV of noise is enough to interfere with the normal operation of onboard memory it's possible that the GPU could also be that picky. To be safe, I would therefore say (+/-) 20mV PP of AC coupled noise is the limit.
@squeepts probing technique was inadequate to distinguish waveforms that varied by as little as (+/-) 20mV pp. Now He knows that's what ground springs are for! That was my constructive criticism for him. That's how it goes.
I think this experiment needs to be repeated to put this question to bed!
To Research:
@squeept artificially destroyed new old stock NEC/TOKIN proadlizers in a 120C oven for a month to obtain a set of known bad caps to see the effect they would have on a working console. They did not fail in a shorting condition, but He never specified the other characteristics (capacitance, ESR, frequency spectrum). It gave an instant YLOD, no surprise.
@squeept, what happened to those and can you make more for testing purposes? Also, were they what you based the "bad tokin" waveform oscilloscope image on? Because if so, that's not necessarily the same type of failure mode we might see in real tokins. It could be a subtle increase in noise that begins to slowly cause instability in demanind games that progresses to normal games, then the XMB, and finally to a full blown instant YLOD. I would't expect it to suddenly come on though. That still reeks of BGA.
Lastly, and getting back to @squeept's question about why a bad tokin spoils the bunch, I thint that there is more to the way proadlizers work than we've taken into account. In this research I've come across a statement that tokins work based on "transmission lines." I have been meaning to look into this further as that RLC calculation I have been talking about all along is optimized using frquency bands, pass/stop filtering frequencies, basically transmission line EE. I haven't done much digging here, but suspect that if one is performing badly it's resonating and "chaffing" the RLC. It could have a narrow band where the noise suppression all of a sudden drops dramatically, like tuning into a radio frequency. Anyway, that's where I currently am with my research - The point where it devolves into speculation.
Sources:
(1) Rick A. Price, "Polymer Film Capacitors Provide Needed Performance," Paktron Capacitors 1205 McConville Road Lynchburg, VA 24502. URL = https://paktron.com/wp-content/uplo...ilm-Capacitors-Provide-Needed-Performance.pdf
(2) https://en.wikipedia.org/wiki/Curie_temperature
(3) https://paktron.com/wp-content/uplo...ilm-Capacitors-Provide-Needed-Performance.pdf
(4) NEC Promotional Paper: https://www.nec.com/en/global/techrep/journal/g06/n05/pdf/t060514.pdf
...take 2, so i re-did the Tantalum + MLCCs combination (just on the bridge wire only), this time i doubled checked the (+) and (-) lol, so far the PS3 is running great!
Again thanks so much for spotting my silly mistake on my last build. I'm gonna aim to test this latest working set up and update you guys soon all being well, just need to run some intensive games. Unfortunately i don't have an o-scope to measure anything, i wish i did though, maybe soon.
