VOGONS


Tantalum recapping project

Topic actions

First post, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Project goal: replace wet aluminum capacitors with tantalums wherever possible.

For decades tantalum capacitors have been a premium choice for high-end products, providing extremely long life and very stable characteristics. Some people hold fears about them blowing up at random, but this isn't supported by data.

1. Noname S3 ViRGE/DX (below)
1.1. Revisit
2. STB Velocity 128 (RIVA 128)
3. Noname serial ball mouse
4. Sapphire Radeon 7000
5. ATI Rage Pro Turbo
6. SiS 315L-128
7. Matrox G100
8. Diamond Fire GL 1000 Pro (Permedia 2)
9. Edimax EN-9130 (Realtek RTL8139)
10. Datapath VisionRGB-PRO2
11. S3 Trio 3D/2X
12. Matrox G450 LP
13. Nvidia Vanta-16
14. Nvidia Quadro4 980 XGL
15. PC Chips M577
16. ATI Radeon 9600
17. Matrox G550

First experimental subject: cheap noname S3 ViRGE/DX graphics card. The board comes with about a dozen small alus, mainly NOVA branded. Nine 10 uF 25 V, one 22 uF 16 V, and one 1 uF 50 V, all through-hole and about 3 mm lead spacing. All appear to have 5 V across them, so 10 V tantalums should be enough. My small grab bag has some 1990s 3 mm 10 uF 10 V anyway so that's what it's getting.

The attachment 1.jpg is no longer available

SMD on through-hole, not great but I don't have any th tantalums. The card wouldn't output anything with tantalums for the 1 uF and 22 uF, replacing the 1 uF made no difference. I didn't bother measuring the originals or replacements, besides for capacitance, so who knows. With two placeholder vintage ELNA alus the card works. 2D image quality is about as bad as it always was on this unit.

The attachment 2.jpg is no longer available

Maybe in the future I'll return to this to give it proper through-hole tantalums instead of random reused SMD ones.

Last edited by vvbee on 2025-04-29, 04:45. Edited 15 times in total.

Reply 1 of 51, by StriderTR

User metadata
Rank Member
Rank
Member

Well, at least it's working. 😀

I've got a good stock of a variety of common caps on hand for situations like this, though, I have more TH than SMD. Need to go through them one day and see what I have vs what I need.

On a side note, I see a lot of people talk about their S3 cards bad 2D quality, perhaps I'm just lucky, but the few I've used have looked and worked great. Honestly, it's probably one of my favorite DOS cards for sheer compatibility. Kinda wish I had a Voodoo 1 to use with mine in my DOS rig, but not at current prices.

Last edited by StriderTR on 2025-03-30, 19:13. Edited 1 time in total.

Retro Blog & Builds: https://theclassicgeek.blogspot.com/
3D Things: https://www.thingiverse.com/classicgeek/collections
Wallpapers & Art: https://www.deviantart.com/theclassicgeek

Reply 2 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

I have a regular ViRGE with crisp image quality, so it's possible for sure. Only uses ceramic capacitors as well.

Reply 3 of 51, by melbar

User metadata
Rank Oldbie
Rank
Oldbie

I have replaced lot's of electrolyt capacitors....
In general, i have recapped on mainboards and PSU's.

But i think, there are two VGA cards in my collection, which are equipped with these tantalum caps.
A Riva TNT2 card and a Geforce 2 Ti.
Never tested any SMD caps until now...

And furthermore, i have an old socket 7 board which i've not used yet.
It has lot's of these tantalum capacitors....

https://theretroweb.com/motherboards/s/tyan-s1572-turbo-atx

The bad issue with this board:
There are several tantalums, where are installed quite close to the PCI slot, the CPU and ram socket.

#1 K6-2/500, #2 Athlon1200, #3 Celeron1000A, #4 A64-3700, #5 P4HT-3200, #6 P4-2800, #7 Am486DX2-66

Reply 4 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Very rare to hear of post-1980s graphics cards with failed SMD tantalums. They seem to often have extreme voltage derating, seen 35 V tantalums on a 3-volt rail on one RIVA 128. Not surprised if they outlive about everything else on the board.

All five wet alus on my GeForce2 Ti leaked in about a year or two back then, not very durable. They may have been Sanyo so not even a bad brand (now that I think about it, they may also have been GSC). But these cards start to have larger 470+ uF capacitors, not really feasible to replace with tantalums unless they only see low voltage.

Last edited by vvbee on 2025-03-31, 18:49. Edited 1 time in total.

Reply 5 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Second subject: STB Velocity 128 (RIVA 128 AGP). Again about a dozen through-hole wet alus, most 10 uF 16 V, one 22 uF 16 V, and a 47 uF 10 V through-hole tantalum. The tantalum I left in place. All removed alus measured 3-Ohm ESR and a capacitance as rated. Most Nichicon, a few T.I. My parts bag had 1990s 16 V tantalums for the 10 uF and a more modern 10 V for the 22 uF. Based on my measurements of a PCI version of the RIVA 128, some capacitors get 3 and some 5 volts across them.

The attachment 1.jpg is no longer available

Again SMD on through-hole vias, since that was god's plan.

The attachment 2.jpg is no longer available

Card works and I think 2D image quality is better.

Reply 6 of 51, by robertmo3

User metadata
Rank Oldbie
Rank
Oldbie

vvbee want me to send you nv1 card?

Reply 7 of 51, by pvlst

User metadata
Rank Newbie
Rank
Newbie

Have you considered using polymer electrolytic caps instead of tantalum ones? Just being curious.

I successfully recapped some ISA riser boards with polymer caps and it works very well. I mean those are decoupling/filtering caps, so polymer ones were a perfect fit.

Reply 8 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie
robertmo3 wrote on 2025-03-31, 13:47:

vvbee want me to send you nv1 card?

No thanks.

pvlst wrote on 2025-03-31, 14:19:

Have you considered using polymer electrolytic caps instead of tantalum ones? Just being curious.

I successfully recapped some ISA riser boards with polymer caps and it works very well. I mean those are decoupling/filtering caps, so polymer ones were a perfect fit.

Specs not as close to the originals, shorter lifetime, and availability not as good. I've considered them for 100+ uF replacements where tantalum availability starts dropping off, but might just keep using regular alus.

Reply 9 of 51, by auron

User metadata
Rank Oldbie
Rank
Oldbie

i have a pci elsa victory erazor riva 128 that uses black 10uf/16v tantalums, however comparing with images of the agp version, that one has much larger yellow tantalums, same value though. i wonder if the yellow ones are supposed to be higher grade parts?

there are more differences between these boards - the pci one uses some wendell 220uf/16v thru-holes for the tv in/out, the agp one has japanese 220uf/4v smds there. also, the finish on the edge connector of my pci board has a noticeably paler color than the agp one, which looks more standard for a card from the time. perhaps these differences are due to the agp version being considered the flagship card or something.

Reply 10 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Color varies by manufacturer, you can find good ones in black. I doubt package size says anything about quality for the same specs, maybe physical robustness.

Reply 11 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Third subject: noname serial ball mouse, apparently from 2004. It has one wet through-hole alu, 10 uF 25 V, which measures the right capacitance and an ESR of 4 Ohms. Looks to have 5 volts across it in use.

The attachment 1.jpg is no longer available

The pads are on the back. The white wire has partially lifted its pad, cord been yanked probably and all the stress went through that wire.

The attachment 2.jpg is no longer available

A 10 uF 16 V SMD tantalum went onto the pads of the capacitor. The wires had to be moved to make room, and later adjusted some more for a better angle. You could probably think about rerouting the whole cable so it exits the PCB from the side rather than the top so the white wire or one of the other ones isn't so likely to rip out entirely, but maybe some other time.

The attachment 3.jpg is no longer available

It's not a great mouse, but probably works the same now as it did before. Does work anyway.

Reply 12 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Fourth subject: Sapphire Radeon 7000. Three wet through-hole alus, 100 uF 16 V, Licon LRM. Measured the right capacitance and about 0.6-Ohm ESR. A couple of them ran with about 2 volts across them, the third one I didn't bother checking.

The attachment 1.jpg is no longer available

Got replaced with new through-hole 100 uF 16 V tantalums. Same ESR as the originals.

The attachment 2.jpg is no longer available

Card works. Image quality is unchanged, but can now be overclocked from the anemic 133/133 default. Used to artefact above 140 or so, now can do the more usual 150/150 without issue. Though this is in a different system with different software.

Reply 13 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Revisit of the noname S3 ViRGE/DX to get rid of the two remaining wet alus that wouldn't work with tantalums. The 1 uF cap was replaced with a through-hole 1 uF 35 V tantalum, and the 22 uF one with an SMD 22 uF 16 V tantalum.

The attachment 1.jpg is no longer available

A pad was lifted and rerouted. I doubt this PCB can take many recappings, so you might as well try and do it once with tantalums.

The attachment 2.jpg is no longer available

Card works this time. 2D image quality may be better.

Reply 14 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Fifth subject: ATI Rage Pro Turbo. Five SMD wet alus, 22 uF 16 V. Capacitance measured per spec, ESR of 5 Ohms. Some have 3 volts across them, some have 5, one has 12.

The attachment 1.jpg is no longer available

Replaced with four 22 uF 16 V tantalums and a 22 uF 35 V tantalum. Two of them were donated by a kind Voodoo Banshee, the rest were new.

The attachment 2.jpg is no longer available

Card works. The alus were likely not doing anything critical anyway.

Reply 15 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

Sixth subject: SiS 315L-128. Ten wet alus: 6 47 uF (Tocon), 2 10 uF (Tocon), 1 470 uF (Tocon), 1 1000 uF (Sanyo). I didn't measure voltages across them, but the 12-volt rail looked to only go into an unpopulated fan header and maybe some other unpopulated parts. ESR on the bigger cans was about 0.3, on the smaller ones about 5 or something like that.

The attachment 1.jpg is no longer available

The <= 47 uF alus were replaced with through-hole tantalums, 47 uF 16 V and 10 uF 35 V. The two bigger cans got Panasonic FR. All unused. The Panasonics had about 0.2 ESR, the tantalums I didn't measure but probably normal.

The attachment 2.jpg is no longer available

Card works.

Reply 16 of 51, by akimmet

User metadata
Rank Member
Rank
Member

Using non polymer SMT tantalums is ill advised. When they fail and explode, they ruin the pcb. Any blackened fiberglass will become conductive. At least with leaked electrolyte, there is a chance you can repair damaged traces.

If you insist on using tantalums, ALWAYS derate the voltage specification. Tantalums do not tolerate overvoltage spikes. I aim for at least 3x on voltage rating.

Polymer electrolytics or polymer nobium/tantalum capacitors at least won't violently self destruct.

Electrolytic and polymer electrolytic lifespan is calculated at maximum rated temperature and ripple current. While a 1000 hour lifespan sounds scarily short, it in reality will be much higher at more reasonable temperatures. A 1000 hour 105°C capacitor will have a longer lifespan than a 5000 hour 85°C capacitor inside a PC case.
Since most polymer electrolytics are rated at 125°C they will likely outlast any standard tantalum.

Reply 17 of 51, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie
akimmet wrote on 2025-04-09, 22:05:
Using non polymer SMT tantalums is ill advised. When they fail and explode, they ruin the pcb. Any blackened fiberglass will bec […]
Show full quote

Using non polymer SMT tantalums is ill advised. When they fail and explode, they ruin the pcb. Any blackened fiberglass will become conductive. At least with leaked electrolyte, there is a chance you can repair damaged traces.

If you insist on using tantalums, ALWAYS derate the voltage specification. Tantalums do not tolerate overvoltage spikes. I aim for at least 3x on voltage rating.

Polymer electrolytics or polymer nobium/tantalum capacitors at least won't violently self destruct.

Electrolytic and polymer electrolytic lifespan is calculated at maximum rated temperature and ripple current. While a 1000 hour lifespan sounds scarily short, it in reality will be much higher at more reasonable temperatures. A 1000 hour 105°C capacitor will have a longer lifespan than a 5000 hour 85°C capacitor inside a PC case.
Since most polymer electrolytics are rated at 125°C they will likely outlast any standard tantalum.

https://www.kyocera-avx.com/news/solid-tantal … e-applications/

Due to known wear-out mechanisms and instability in AC or DC performance across electrical and environmental conditions, tantalum capacitors with polymer cathode systems are not recommended for long-term use in all high-reliability applications. Tantalum capacitors with MnO2 cathode systems, on the other hand, have a long history of extremely stable performance over decades of employment in demanding, harsh-environment military and space applications, which squarely establish it as the preferred technology for long-term use in high-reliability, mission- critical military- and space-level applications.

One critical benefit of MnO2 tantalum capacitors, and one of the primary reasons that they’re ideal for use in high-reliability and long-life applications, is that there are no known wear-out mechanisms for this technology. In fact, MnO2 tantalum capacitors that have been operating in applications for 30 years or more — since before the publication of the MIL-PRF-55365 specification even — continue to improve in leakage performance and reliability with age. Conductive polymer capacitors employ the same dielectric as MnO2 tantalum capacitors with the same reliability and long life characteristics, but the conductive polymer that replaces the inorganic manganese dioxide is sensitive to humidity and oxidation, which can cause long-term parametric changes.

The probability of capacitor failure goes down as you increase the percentage of tantalums, and for retro gear you also want parts that don't degrade in storage. Reports of exploding tantalums in old machines never come with data exploring what might've caused it, and the cause is very likely them being driven out of spec by some other poorly functioning part in the decades-old machine, which for the tantalum is preventable by derating.

Reply 18 of 51, by analog_programmer

User metadata
Rank Oldbie
Rank
Oldbie

There is an "urban legend" according to which tantalum capacitors (especially the "old ones" - whatever it means) that stay long time out of operational usage tend to explode more often than the ones in regular usage.

I've never seen an exploding tantalum capacitor in live, but I often see leaky standard electrolytic capacitors. Each person decides for himself what to use when replacing old electronic components. I don't use tantalum capacitors as replacements for electrolytic ones just because they're not as common and often are more expensive.

The word Idiot refers to a person with many ideas, especially stupid and harmful ideas.
This world goes south since everything's run by financiers and economists.
This isn't voice chat, yet some people overusing online communications talk and hear voices.

Reply 19 of 51, by akimmet

User metadata
Rank Member
Rank
Member

I have had hundreds of tantalums fail short over the years. It is a big enough problem that anything over 30 years old I automatically replace them. Examples from the 1970's and earlier are especially failure prone. The failure mode is well known and its causes.
https://www.kyocera-avx.com/docs/techinfo/Fie … tallization.pdf
https://www.edn.com/what-a-cap-astrophe/

Tantalums do degrade in storage, almost always from moisture ingress. I have had to throw out entire reels of them because of improper storage in a wet environment.