Electrolytic vs Tantalum for Supply ByPass?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Just changed some 70's electrolytics in a compressor today

There's a noticeable degradation in sound when dealing with old electrolytic caps.

Manufacturers admit that shelf life is not eternal.

Tantalum does not go soft and fuzzy, by comparison.

If you are only concerned with power rail filtering then maybe tantalum is a bit of a risk, because they react badly to high current peaks. Some manufacturers install current limiting resistors as a safeguard against this weakness.

Electrolytic caps with film caps in parallel sound better IMO, generally speaking, but some of the sound of a lot of 70's discrete gear is due to Tantalum caps.

As far as distortion caused by polarized caps, SSL was certainly aware of that aspect when designing their stuff. Everything distorts. Perfection is more of a Marketing thing than if is reality, IMO
There are objective bench measurements that quantify non-ideal capacitor behavior. Electrolytic caps are a mature technology but have improved since the 1970s (mostly lower impedance improvements for switching supply use).

Coincidentally back in the 70s I spent some time on my bench quantifying the impact of paralleling electrolytic caps with smaller film caps. In my tests the film cap needed to be at least 10% of the total capacitance to dominate behavior. We routinely see small ceramic caps in parallel with electrolytic caps on PS rails but that is a different very specific above audio frequency application.

JR
 
There are objective bench measurements that quantify non-ideal capacitor behavior. Electrolytic caps are a mature technology but have improved since the 1970s (mostly lower impedance improvements for switching supply use).

Coincidentally back in the 70s I spent some time on my bench quantifying the impact of paralleling electrolytic caps with smaller film caps. In my tests the film cap needed to be at least 10% of the total capacitance to dominate behavior. We routinely see small ceramic caps in parallel with electrolytic caps on PS rails but that is a different very specific above audio frequency application.

JR

I like to use relatively large values, so that the film cap is doing most of the work into the mid-range frequencies.
 
Yes - Bateman's stuff still seems to be the best read wrt audio use.
I don't want to diss Bateman's work, but the conditions where he measures the effects are often not pertinent to typical audio use, in particular electrolytics used in AC coupling, where the voltage across the cap must be very low. OTOH using electrolytics in speaker x-overs is not good practice, although common place.
IMO the weak point in his papers is the absence of warning that some of the tests are not typical of a product's performance if it is designed with good practice.
His papers are not for just anybody.
 
I don't want to diss Bateman's work, but the conditions where he measures the effects are often not pertinent to typical audio use, in particular electrolytics used in AC coupling, where the voltage across the cap must be very low. OTOH using electrolytics in speaker x-overs is not good practice, although common place.
IMO the weak point in his papers is the absence of warning that some of the tests are not typical of a product's performance if it is designed with good practice.
His papers are not for just anybody.

It looks like the tests did use 0.5V pulses which is a lot better than the 70 VRMS in the last link. Also the articles are kinda old. His "Ultra low distortion test oscillator" THD is "under 0.01%" which is just not that great these days and easily beaten with a mediocre USB audio interface. [wr-wr-wrong, see #48]. So lets just appreciate that he did that work almost 20 years ago.

But I think most of us understand where an when to use various caps. G0G and film (non-metal) are great everywhere, electrolytics are fine for bypass and coupling as long as there's no voltage across them (meaning not used to filter), tantalums are generally to be avoided but I'm not going to drag anyone down for replacing tant with tant and MLCCs shouldn't be used for anything except bypass and selected with care to account for bias loss (X7R, higher V and uF).

One more point about MLCCs. Like it or not, that's all there is dudes. Even audiophools don't use film for bypass. Good luck finding old-school ceramic disc caps. MLCCs are high Z at low frequencies but presumably they're used at IC pins and that is where you need > 1kHz performance where MLCCs are perfectly suitable. Maybe use 220n instead of the usual 100n to account for bias loss but I have a huge tape of 100n that I'll be using happily for sure.
 
Last edited:
there seems to be some confusion about importance of capacitor non ideal performance. Capacitors used across power supplies are a completely different application than capacitors used in audio filters (where terminal voltage changes with signal), or to a lesser extent DC blocking/coupling applications.

Different applications with different sensitivities to non-ideal behavior, not one question to evaluate but three or more..

JR
 
Also the articles are kinda old. His "Ultra low distortion test oscillator" THD is "under 0.01%" which is just not that great these days and easily beaten with a mediocre USB audio interface. So lets just appreciate that he did that work almost 20 years ago.

But I think most of us understand where an when to use various caps.
As I said, I don't want to criticize Bateman's work; my gripe is with people who refer to this work and conclude that electrolytics are distortion generators and advocate expensive and bulky film caps when they don't understand that it's useless.
 
It's not "every day" but I'm seeing more and more tant capacitor failures in power rail bypass applications as found in decades-old gear. The culprits look like a small, colorful bit of candy in a "blob" shape.

Most all fail as a dead short from a rail to 0V. Cap catches on fire, burns the PCB. If there was a series R that also burns. A few weeks ago I had to salvage an ADM console module (from a rack up) where a pair of tants and associated series resistors had done some serious PCB damage. I used "low impedance" 105C AL lytics and flame-proof resistors as replacements.

One recent oddity with tants I also encountered recently was on an audio card in an Ampex ATR-102. The card would function properly for a few minutes, then became "stupid". Tracked it down to the bipolar rails for the opamps that would drop down to a low value. Ampex used a dual-tracking regulator on each audio card in a TO-66 style package (with extra legs)...I forget the part number.

One bad tant across one output rail caused the regulator to go into current limiting after a few minutes and thus cause the other rail to follow. The TO-66 reg would get quite warm. I found the bad tant with my fingertips...it also was getting quite warm.

After removal, I measured no resistance across that bad tant with my Fluke DVM. Capacitance, however, gave wonky readings. IOW, it didn't fail in the typical fire and glory. It was one of those "piece of candy" tants. I hate them as they are crapping out too often.

Bri
 
Last edited:
IMO the weak point in his papers is the absence of warning that some of the tests are not typical of a product's performance if it is designed with good practice.
Agree re atypical performance / practice.
But I'd tend to look at it 'from the other end' and say that those tests illustrate the importance of good design practice. Same thing really.
But - yes - they shouldn't be taken in isolation and without being put into an up to date context re limits of measurement etc.
 
One more point about MLCCs. Like it or not, that's all there is dudes. Even audiophools don't use film for bypass. Good luck finding old-school ceramic disc caps. MLCCs are high Z at low frequencies but presumably they're used at IC pins and that is where you need > 1kHz performance where MLCCs are perfectly suitable. Maybe use 220n instead of the usual 100n to account for bias loss but I have a huge tape of 100n that I'll be using happily for sure.

I've no problem with X7R as stated but Polyester is a realistic option for DIY and through hole.
Link is to quick search and filter on RS. Scroll down for options to buy 25 or 50 etc rather than 1000s.

Polyester Film Capacitors | RS Components
 
Is it summer already? Time for the yearly rehash of all the same talking points we've only had 10 times before and everybody is fully aware of.
 
I've no problem with X7R as stated but Polyester is a realistic option for DIY and through hole.
Link is to quick search and filter on RS. Scroll down for options to buy 25 or 50 etc rather than 1000s.

Well I stand corrected. That is certainly a valid option. The 7.2x2.5 footprint requires quite a bit more board space but at $0.15 @10 for a 100n 63V film does beg the question, "why not?". If I had a lot of space, I might just use those for bypass of select amps.
 
Well I stand corrected. That is certainly a valid option. The 7.2x2.5 footprint requires quite a bit more board space but at $0.15 @10 for a 100n 63V film does beg the question, "why not?". If I had a lot of space, I might just use those for bypass of select amps.
Well much depends on whether you are working SMT or thru hole. Then you can get onto lead inductance and disruption of ground planes etc...it goes on and on if you let it !
On effective decoupling strategies - have you used/seen the X2Y configuration type caps ?
 
Well much depends on whether you are working SMT or thru hole. Then you can get onto lead inductance and disruption of ground planes etc...it goes on and on if you let it !
On effective decoupling strategies - have you used/seen the X2Y configuration type caps ?

Lead inductance? Come on now.
 
Then you can get onto lead inductance and disruption of ground planes etc...
It's a concern for RF and high-speed logic. I've never could note a difference between film and ceramic by-pass caps in audio applications.
On effective decoupling strategies - have you used/seen the X2Y configuration type caps ?
Do you mean using these for rail decoupling? I use them for mains inlet; they're quite enormous.
 
Do you mean using these for rail decoupling? I use them for mains inlet; they're quite enormous.

No. He means a particular series called "x2y" that has nothing to do with mains protection. They're SMD MLCCs that have two caps in one package so that inductance cancels and so on. We talked about them in one of the CMRR threads. They're good for common mode shunt on entry for signal.
 
Some interesting discoveries about inductance, mostly made worse by our friend, Eddie Current. He can be a real asshole when it comes to hum, at any frequency. I had an issue one time with a 4 channel mic pre (the 3124 was my first product back in the olden days) where channel 4 always had a higher level of hum. I was closer to the power supply. After removing and moving all the transformers on the audio side of the PCB, I finally discovered that the cutout for the transformer had a ground plane around it, and after all attempts, I cut a slot in it, breaking the circle, their was a large spark and the hum went away.

This was caused bu the field of the AC transformer (a toroid) and it was inducing current into the loop around the transformer, basically placing hum on the whole ground plane. This is the same concept as the Weller soldering iron. Wrap 1000 wordings around a single loop and get enough current to heat up a copper wire.

Just remember that when laying out PCBs, make sure that no traces are on a loop and no cutouts have loops. Even with an external PSU, you may have another piece of equipment above or below it that can cause problems.

And lead inductance, well, maybe if you have a G-Hz bandwidth circuit. Like the skin effect, it's out of your range... Now, if your product sits under a cell tower, you might get some calls, and these won't be coming from the 5G bot in your vaccine.
 
And lead inductance, well, maybe if you have a G-Hz bandwidth circuit. Like the skin effect, it's out of your range...

Even SMD caps start becoming inductive around a couple hundred MHz, leaded caps would have inductance in the 5 nH range for relatively small leaded ceramic caps. For the 7.2x2.5 mm footprint discussed above it will probably be higher, inductance is approximately 1 nH per mm of lead length and capacitor body length combined, so depending on whether the 2.5mm or 7.5mm is the length, it could be in the 8 or 10 nH range.
The TDK film cap data shows the self resonance frequency of a 0.1uF film cap around the 1MHz to couple of MHz range, well within the gain bandwidth product of most modern op-amps.
The cap impedance only begins to increase at that point, it still isn't very large (maybe 1 Ohm or so) around 10MHz, but I don't think you would want to use one of those caps for bypassing any op-amp that was flagged as a "high speed" or "video" amp and might still have gain into the high MHz range.
And of course don't use them for bypassing modern logic family devices which have really fast edge rates, that is just designing in lots of power supply noise.
 
Back
Top