Datasheet Specs vs Average Conditions In DIY Use Cases

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I’ve been looking through a lot of data sheets while sourcing components for a number of projects and had some questions regarding how this information translates into use cases of DIY where there seems to be more emphasis on derating than in a production environment where cost cutting is a major driver.

I read Morgan Jones book in which he details overrating electrolytics for increased performance and reliability which got me thinking.

Im also curious about some new technologies and how maybe their reliability and performance benefits haven’t trickled down as much into the DIY community but I’ll start with derating.

A resistor example:

Let’s say say theres 1/20th of a watt passing through.

An RN50 might be rated at 100ppm at 1/8th watt, however RN50s designated as 50ppm are rated at 1/20th of a watt.

Usually the 50ppm example is at a higher cost for a given resistance.

These numbers are at rated power, and are temperature dependent. So, my assumption would be using an RN60 in this position might provide comparable or even increased performance to the 50ppm RN50 example at less cost. It would stay cooler, drift less and provide less noise is my assumption?

So, let’s say the 1% RNs are 25 cents each, and a Vishay .1% S102 is $20. Would it not be sensible to just get 40 RNs at twice the rated power of the foil type for $10 and select the best matched pair (R and L channel) rather than spend $40 on the S102 or other foil? Wouldn’t the performance be rather comparable unless let’s say if you require it to survive on a lunar rover under massive environmental stress with no servicing ability?

Also, it seems as a general rule of thumb the lower wattage / miniature precision resistor seem to be more costly than larger cousins. Is this due to difficulties in production?

So, one could reduce cost and increase performance simply by overrating resistors by a certain factor?

Are there rules of thumb I’m unaware of?

How might a 5ppm resistor running at it’s rated power compare to a 50ppm resistor running at one half it’s rated power?

For capacitors:

Given Vishay’s RN derating practice, I wonder if long life examples (10k hours or more) are simply derated versions of other ranges?

How does reduced temperature effect other parameters?

For various capacitor use cases, what’s your general practice of selection and derating?


Disclaimer:

I don’t really want to get into the audibility of such things, but speaking purely from a quoted figures vs diyaudio in-situ and from a long term reliability standpoint.


Other capacitors and their measurements:

With the wide availability now of low cost MLCC ceramics and Organic Polymer capacitors, is it more tradition to continue to use standard electrolytics, tantalums, micas and so on? (Except in high voltage applications)

Of course I’m discussing primarily for diyaudio applications, where a small price per unit hike isn’t a deal breaker necessarily.

Given voltages, values and footprints where both are available would it not make sense to:

Replace all tantalums with organic polymer electrolytics?

Replace all standard electrolytics with Polymer types, except perhaps when one needs a bi-polar or other special attribute?

All micas with C0G ceramic?

Polystyrene with C0G?

Are some of these still being used simply because those designing these circuits come from industry backgrounds in which there’s high cost per unit awareness?

What do these older types have in operational parameters over these newer examples?

Again, whether it’s audible I imagine it’s circuit dependent.

Or perhaps there’s functional or audible parameters in some of these types I’m not considering.

Again, speaking purely from reliability and theoretical improved on-paper performance from using more modern replacements. I don’t want to compound confusion by discussing subjective experiences before sorting out the objective.

X7Rs:

X7Rs have a bad rap from their past lives, but from what I gather that’s largely not as relevant these days except in certain circumstances.

For digital circuits for example, or noise absorption in power supplies (use cases outside of the signal path) is there anything to be gained from C0G/NP0? The cost difference is getting smaller every day and C0Gs are getting higher values than before. You can also use multiples to get the values you need.

Is there any valid reason to replace the standard .1uf X7R be by a lower (for size or less cost) or same value C0G? It’s about a $1 difference in cost for same value.

From someone else’s comments in another thread my understanding is that it’s precisely their poor operating behavior that in some way makes them superior for noise absorption applications?

Do the new MLCC X7Rs still have the same beneficial “issues”?

Curious to hear everyone’s thoughts / benefit from their experience.
 
Last edited:
I don't know the answers to most of your questions, but here's my two pennies' worth:

As far as I know, polystyrene capacitors have much less dielectric absorption than NP0 / C0G. Whether that's relevant for audio is a controversial subject. Otherwise I can't think of any particular advantage of polystyrene over NP0.

X7R and X5R capacitors are quite non-linear and inaccurate, and piezoelectric as well. Their values change with temperature, applied voltage and time since they were last heated above their Curie temperature. The drop of capacitance with applied voltage is far worse for small-size SMD X7R and X5R capacitors than it used to be for old-fashioned single-layer X7R ceramic disc capacitors. Many models have only 20 % of their capacitance left when you apply their nominal voltage.

Because of their non-linearity (and piezoelectric behaviour) they are not particularly suitable as filter or coupling capacitors. Nonetheless they are often used as such in consumer equipment because they are cheap.

The non-linearity doesn't matter much for power supply decoupling and their relatively high losses help to damp unintended resonances (not much though, as they typically still have a Q of about 50). In that sense they are better for decoupling than NP0 with its very high Q. Also keep in mind that for decoupling of high-frequency signals, like in fast digital circuits, the physical size of the capacitors and of the wiring matters a lot: the smaller the capacitor and wire length, the less parasitic inductance.
 
Last edited:
Hi Marcel thanks for your input.

Regarding SMD I am not sure, I’m a through hole kind of guy.

I will stay away from X7Rs in the audio path. My thoughts were more-so the trade offs in avoiding them entirely.

As I said price drops in C0G/NP0 caps and higher capacitances make them attractive as an X7R replacement up to around .1uf.

Specifically talking about the “obligatory .1uf bypass capacitor” which across the board seems to be X7R.

Let’s say a .1uf 20% X7R is 20 cents

And a .068uf 5% C0G/NP0 is 75 cents.

That X7R could be .08uf anyhow, .068uf isn’t much of a stretch.

For a DIY project 4-5 75 cent caps won’t break the bank. Lord knows that’s a drop in the bucket compared to some of my other epic blunders....

It’s less than the difference between an orange drop and a Mundorf certainly.

At the very least X7Rs don’t seem to hold up well over time and in a nice piece of audio kit it is a nice feeling to know you have a few thousand extra hours before worrying about it. Less possibility of poking around and more listening to music seems like a nice proposition to me.

Plus since the overall capacitance here isn’t critical, why not go for a lesser value more stable c0g?

I don’t expect magical unveiling of hitherto unknown audio bliss....

Just minimizing desoldering / feel good move not unlike overrating an electrolytic capacitor for example.... and perhaps gain some smidgen of reliability.

Mostly I’m asking as a general sanity check.

https://www.murata.com/~/media/webrenewal/support/library/catalog/products/emc/emifil/c39e.ashx

In some pieces of the article it seems to suggest that attributes found in C0G would provide increased reliability and performance in areas which traditionally used X7R. However, I’m no expert so perhaps I’m missing something.

Again, not really discussed probably due to the “acceptability” of X7R and the cost considerations in production environments.

I also wonder if the C0G benefits are moot when off on the pcb in comparison to an X7R right on the pins of a DIP. Some interesting layout info in that doc that leaves me with more questions than answers.... funny how that works.... the more you know; the dumber you feel.

Regarding resistors...

I know that there was a linear audio article on resistors and I would imagine it would provide at least some insights related to the resistor Q.
 
Last edited:
The non-linearity doesn't matter much for power supply decoupling and their relatively high losses help to damp unintended resonances (not much though, as they typically still have a Q of about 50). In that sense they are better for decoupling than NP0 with its very high Q.

This is what I was trying to remember / understand.

That the inherent limitations of the X7R turn into assets in certain use cases... the losses help to damp unintended resonances.

That’s a bit beyond my comprehension level.

The article however also went into some benefits of capacitors which share the attributes of c0g in this same use case.

I’m just uncertain what the relative weight of these pluses and minuses are.

I wonder if anyone out there has tested both and measured the practical differences.
 
And from Kemet’s official documentation, seeming to be putting forth C0G as a good decoupling / bypass choice:

(And seemingly a great choice for everything, where you don’t need high capacitance or voltage ratings)

C0G
Ceramic capacitors with a C0G dielectric are produced with a temperature compensating Class-I material. This material set makes them suited for applications where the Q and stability of capacitance characteristics are required, e.g. filtering circuits. KEMET’s automotive-qualified C0G’s capacitance over temperature changes a small 30 ppm/°C from -55°C to +125°C. With such high stability, C0G devices are suitable for use in critical timing or tuning circuits, high-current or pulse applications, and circuits where low losses are critical, as well as for decoupling, bypass, filtering, transient voltage suppression, DC-blocking, and energy storage.


Technical >> Understanding Ceramic’s Capacitance Over Temperature Performance | Engineering Center
 
Mostly I’m asking as a general sanity check.

https://www.murata.com/~/media/webrenewal/support/library/catalog/products/emc/emifil/c39e.ashx

In some pieces of the article it seems to suggest that attributes found in C0G would provide increased reliability and performance in areas which traditionally used X7R. However, I’m no expert so perhaps I’m missing something.

Given similar sizes and capacitances, the one advantage of using a dielectric with relatively high losses is that the antiresonance peak in Murata's figure 3-12 would be a bit lower. When you have a board with many digital ICs all with their own decoupling capacitor, you have a large network of wire inductances and decoupling capacitors with many series and parallel resonances.

Like I wrote, the losses of X7R or X5R would help a bit, but not more than that. The ESR of an X7R or X5R capacitor is usually frequency dependent and about 1/50 of the capacitive reactance. To damp the parallel resonances, it would be better to have far higher losses. Just adding a resistor in series with the capacitor would help, but would also reduce its efficacy at all frequencies where there is no parallel resonance. What you can do is add extra decoupling capacitors with series resistors here and there, or put in some aluminium electrolytic capacitors (no low-ESR types, but run-of-the-mill electrolytics since you use them as dampers), or put high-loss ferrite beads in the supply lines.
 
Would lesser classes of ceramic capacitor have the higher losses you suggest?

I take it even if they did, these are depreciating returns?

I gather from what you are saying the .1uf cap is often added in an automatic manner, but the real answers as always aren’t as simple or direct.

How many ICs / parallel resonances before this begins to be problematic?

I would very much appreciate if it wasn’t a further burden on your time a simplified example/schematic in an audio application in which the additional resistors and capacitors you suggest would be fruitful.

My building has been mostly confined to tube circuits... almost exclusively films, electros and micas. So, as I begin to experiment in solid state there’s a whole wealth of parts and standard practices which are a bit foreign to me.
 
Are there rules of thumb I’m unaware of?
Yes. Read the datasheet very carefully. A good data sheet will answer many of your questions.

X7Rs have a bad rap from their past lives, but from what I gather that’s largely not as relevant these days except in certain circumstances.
Like all capacitors, there are things they do well and things they do badly. The big mistake made by many audiophiles is to regard dielectric X as 'good for audio' or 'bad for audio'; these are meaningless judgements. By "thing" I do not mean 'audio' but 'coupling cap' or 'decoupling cap' etc.
 
Thanks DF. I’m just formulating an understanding here not looking for magical audio runes.

Do you have any thoughts on the resistor derating scheme / it’s correlation with noise that I suggested in the Op?

I read the datasheets. However sometimes they read more like advertisements.

https://www.wima.de/wp-content/uploads/media/WIMA-Audio.pdf

And sometimes as with the Murata doc the more technical discussions are lost on me.

If I could already grasp it all there would be no reason to ask here, and I could join you in correcting and clarifying for everyone else.
 
Would lesser classes of ceramic capacitor have the higher losses you suggest?

No, I just looked up the dissipation factors of a couple of Z5U capacitors and they are similar to those of X5R and X7R.

I take it even if they did, these are depreciating returns?

I gather from what you are saying the .1uf cap is often added in an automatic manner, but the real answers as always aren’t as simple or direct.

How many ICs / parallel resonances before this begins to be problematic?

Two capacitors and one parallel resonance if the resonance happens to be at a very inconvenient frequency. I've seen a situation like that at work once, in an RF circuit.

Nonetheless, I've built digital and audio hobby circuits with dozens of ICs with no apparent problems without thinking very much about the resonances. Besides the ceramic decoupling caps, I did always include a few cheap aluminium electrolytic capacitors here and there. They are good dampers because they have quite high losses. If the power grid looked regular (similar-sized ICs placed at a fixed pitch), I usually used several different values of ceramic capacitor to spread the resonant frequencies a bit, but again without thoroughly analysing anything.

I would very much appreciate if it wasn’t a further burden on your time a simplified example/schematic in an audio application in which the additional resistors and capacitors you suggest would be fruitful.

My building has been mostly confined to tube circuits... almost exclusively films, electros and micas. So, as I begin to experiment in solid state there’s a whole wealth of parts and standard practices which are a bit foreign to me.

With a possible exception for oversampled data converters, I don't think it is likely to be a big problem in audio circuits. The resonances are usually way above the audio band, so as long as nothing starts to oscillate, there shouldn't be any problem in audio amplifiers. As long as you don't need to pass EMC tests and the resonance doesn't occur right at the clock frequency, digital circuits are also likely to work fine.
 
I would recommend you spend time and effort on specifying your operating conditions, and trying to define what if any change in a circuit's performance behaviour would occur due to different part choices.

Eg. some simplistic operating conditions would include your operating conditions, such as in the range 10C to 35C temp for the actual parts in operation. How many hours operation per day and per your diy lifetime will you be using the equipment, and an expectation that you won't stop using it next year/decade as your interests change.

At least then you would get some value from your effort, as imho if you are starting at each part's datasheet without any tangible practical goal to use as a target then your musings have no relevance.
 
I first ran into capacitor resonance with a video amplifier I was developing for the BBC. The combination of capacitors caused an early video opamp to have an odd frequency response wobble at about 6 MHz. The root cause was a two layer pcb with through hole capacitors and long tracks. These days you would use a ground plane and smd capacitors, which shifts the problems up into the VHF/UHF
 
I would recommend you spend time and effort on specifying your operating conditions, and trying to define what if any change in a circuit's performance behaviour would occur due to different part choices.

Eg. some simplistic operating conditions would include your operating conditions, such as in the range 10C to 35C temp for the actual parts in operation. How many hours operation per day and per your diy lifetime will you be using the equipment, and an expectation that you won't stop using it next year/decade as your interests change.

At least then you would get some value from your effort, as imho if you are starting at each part's datasheet without any tangible practical goal to use as a target then your musings have no relevance.

This is simply not so. I’m happy to share with you some of my practical goals.

I did not share them in the interest of understanding some of the larger themes involved and for being of use for all parties, but here you go:

I always assume greater than 30C operating, 7 days a week 4 hours per day average. So let’s say 30 hours / week x 10 years = 15,600 hours at 40C would be my minimum requirement.

I will avoid at all costs a capacitor rated under 20k.

For me long term reliability is a goal only secondary to its original measured performance. If I build something I want it to be rock solid for a long time to come.

I don’t want to build and build cheaply, accumulating and reselling and so on. I’d rather focus my energy on very few designs and use them indefinitely. Maybe this is uncommon, but I am the slow and intentional type.

The c0g vs x7r was really a secondary question, as I was wondering if the use of X7R was mostly based on cost considerations and if the c0g, now more reasonably priced than a decade ago, might be considered a reasonable substitution. There was an older thread on this topic which was abandoned before any consensus or recommendation was put forth.

The resistor issue was my main curiosity and is related to tube applications or other scenarios with higher heat, but required low ppm.

I’ll try to provide some specific info in relationship to my question with regards to resistor derating, temperature and tempco:

From digikey:

Resistance is usually specified at the ambient temperature of 20°C while most manufacturers specify the power rating at 70°C and free airflow conditions.

It seems to suggest the tempco is per degree Celsius over 20 degrees Celsius to their power rating temp at 70C.

They go on to say:

The temperature coefficient of resistance (TCR) is a constant that represents the resistance change per degree Celsius of temperature change over a specific temperature range; it’s expressed as ppm/°C (parts per million per degree centigrade).

Where I live the temperature is frequently 30C (tropical, close ocean proximity environment). People in the middle of Canada who built the same circuit as me will likely achieve better performance over a longer lifetime.

I feel it’s reasonable to compensate for my own harsh climate.

So, a 50ppm 1/4W resistor operating at 30C baseline would have 500ppm, a 100ppm resistor would have 1000ppm.

Let’s say the resistor is dissipating 3/16th of a watt.

The more power dissipating, the warmer the resistor. At 75% of rated power with 30C ambient I’m unsure of the change in tempco but I imagine that a 50% stress factor or less would be much wiser for a more reliable environment.

Stress Ratio = Operating Power / Rated power

To use an example of a Dale PTF series with 5ppm vs a CMF55 with 50ppm, the PTF at 30C has the same exact ppm as the CMF55 at 20C if I understand correctly.

The cost differential is 10-30 cents vs $5 USD. For a quantity of 50 resistors this becomes a dramatic cost differential which may be mitigated simply through some sensible practices which are available to the diyer but maybe not so much a commercial enterprise.

It’s my understanding that reducing tempco by means of derating, improved heat dissipation, cooling methods, fans etc can provide large cost benefits in diy applications.

Also larger resistors can dissipate heat more effectively and thus reduce collective ambient temperature rises from let’s say a whole board of components.

Tightly coupling all resistors to a PCB seems to be a recipe for higher tempcos across all components.

That seems to mean: derate and provide airflow between component and pcb for min. Ppm.

My assumption would be given the above is that a larger 50ppm CMF60 (1/2W) or 65 (3/4W) resistor with a small board clearance for airflow would provide improved tempco, increased reliability and greater precision over a 1/4W PTF 5ppm resistor given that both begin with the same precision (1% or .1%) at less cost.

This is especially so in tube amplifiers where ambient temperatures are much higher.

At other temperatures the calculation goes:

R= Rref[1+α(T -Tref)]
Where
R = resistance at temperature, T
Rref = resistance at reference temperature Tref
α = temperature coefficient of resistance
T = material temperature in °C
Tref = reference temperature at which the temperature coefficient is specified


A lot of ink gets spilled on costly ultra precision foil resistors.

At $25 each or more for those Texas components, I wonder if a much more cost effective behavior in more critical locations for the audio hobbyist is to heavily derate precision low ppm .1% resistors that can be had for around 25 cents.... especially when they need not be retrofitted to an existing design but simply altering the through hole spacing in an eagle file or mounting them in a “tombstone” configuration.

So, is this another case of hype where people don’t understand the real nature of behavior at the level of the resistor?

Or is there something I’m missing? If so, what is it?

Of course the foil resistor is a sound engineering achievement, but we aren’t going to space here where every kilo or square inch is precious and incredibly expensive to launch and opportunities for service are nil. Their image is one of undeniable superiority where the best is required but I wonder in our applications by how much in practice.

In addition:

It’s clear large component manufacturers have caught on to the financial benefits of catering to the audio community. Their accounting departments have wet dreams about their components getting labeled as the new best sound. While it’s certainly not “snake oil” I wonder if there is something missing from the picture.

For example it seems at point since NP gave some love to Elna Silmic II in a thread here quite a while back they’re now labeled as audio caps in the mouser catalogue.

Anyhow, here are some datasheets for comparison;

TE YR1 and YR2 15ppm .1%, ~30cents:
https://www.mouser.com/datasheet/2/418/NG_DS_1773265_A-732410.pdf

Dale RN/CMF 25-100ppm, 1-0.1%, ~10c-$1
https://www.mouser.com/datasheet/2/427/cmfmil-223788.pdf

Dale PTF 1-0.1%, 15ppm ~$5
https://www.mouser.com/datasheet/2/427/ptf-239718.pdf

Vishay S102 1-.01%, 2ppm, ~$10-$40:
https://www.mouser.com/datasheet/2/428/63001-4478.pdf

TX2575/2352 .1%, 1-2ppm, ~$10-$55:
http://umwxd.xdqqt.servertrust.com/v/TxCC_Graphics_Files/DataSheets/TX2352_Data_Sheet.pdf

http://www.texascomponents.com/pdf/tx2575.pdf

I’ll do some more research on the topic of drift and see if the tempco relationship can be extended there as well.
 
Last edited:
On the topic of stability/ drift, since it is detailed as a function of hours at its rated power at 70C, one could assume the derating practice would also effect drift.

Cracking of the coated body and subsequent moisture entry leading to drift is likely a function of thermal conditions in the resistor, if not in unusual non-audio applications like a military radio in a helicopter. So, less heat, less power = less drift, at least hypothetically.

See “Characteristics in Figures” on page 5/8 in Vishays resistor basics application note:

https://escies.org/download/webDocumentFile?id=62214

Another interesting finding in this document is that Wirewound resistors show the worst stability at rated power per 1000hrs, ranging from 1% to 10%, alongside that of carbon comp being +4 to -6 with -3% being typical.

In addition, in a high humidity ocean environment, resistor drift seems to be more of an issue.

From the military derating practices (55% rated power) on through hole metal film resistors:

Construction: This resistor is constructed by vaporizing a carbon, metal, or thick cermet film onto a ceramic or glass substrate. The element is spiraled to increase available resistance. Resistance is calibrated through trimming of a helical grove in the resistive layer. Terminations are usually a tinned copper wire welded to nickel-plated steel end caps. **The body is coated to prevent moisture and contaminate penetration, but coating varies with manufacturer and quality level.**

Reliability: The reliability of this style is
considered better than that of other
resistor styles. **The primary failure mode is resistance drift, which is often caused by cracking of the external coating layer and moisture permeation into the resistive element.** This resistor is ESD sensitive. **Heat dissipation is accomplished through the leads.**

Full info here: http://www.navsea.navy.mil/Portals/103/Documents/NSWC_Crane/SD-18/ResistorsDerating.pdf

This would seem to me to indicate that larger sized resistors, which have a larger coating area and thickness would be less susceptible to drift given that they were operated beneath their power ratings.

The heat dissipation through the leads is interesting as it was my understanding it was done via the resistor body.

Would this imply better heat dissipation not only from being raised from the board for airflow but also due to the increased lead length from resistor to through hole?


I hope I’ve made some more practical and concrete suggestions for informed discussion. Looking forward to your thoughts and insights.
 
Last edited:
Regarding the effect of self-heating on resistance: you usually can assume a linear relation between self-heating and power dissipation. Often you see a derating curve indicating the maximum power as a function of ambient temperature. The temperature at which the maximum power becomes zero is the temperature that the resistor element reaches when it dissipates its maximum power.

For example, a typical 1 % through-hole metal film resistor usually has a derating curve that crosses 0 at 155 degrees C and a temperature coefficient between -50 ppm/K and +50 ppm/K. Suppose the resistor is rated 0.6 W at 70 degrees C ambient. That means that 0.6 W causes a temperature increase of 155 degrees C - 70 degrees C = 85 K. The change in resistance due to self-heating is then 85 K * (+/-50 ppm/K) = +/-0.425 %.

When you only let it dissipate half its rated power, the resistance change due to self-heating also halves: +/-0.2125 %.

(On top of that you have the long-term drift, which also gets worse at higher temperatures.)
 
Last edited:
Some comments would be:

Equipment may be in a 30degC ambient, but certain parts can easily be much higher depending on the equipment cooling design - the part temperature is where a good appreciation is needed.

Very few people would look beyond electrolytics for part life, perhaps as they are the 'low hanging fruit' with respect to the temperature and time service life.

Most of the other topics raised, such as resistance and capacitance variation, with temp or voltage or time, are nebulous without a close connection to a particular circuits operational performance. The better commercial designs know very well which parts are of any influence at all, and quite happily pull every other part down to a simple QC level for the lowest cost, most basic spec.
 
My observations re: diyaudio is that aside from power amplifier heat sinks there doesn't seem to be much regarding "cooling design"...

Although I will run a temp probe over some of my equipment when I get a moment on some key parts out of curiosity.

While I genuinely appreciate your time and wisdom, you don't seem to understand that my preference is to be nebulous. A cursory look at my username should've been an a-ha moment for that one.

My line of thought was going toward feedback resistors, loading and riaa resistors and so on in tube designs. So, I would imagine they have relevance. My goal was to explore how one could optimize lifetime, stability and performance for various key areas of interest by derating, lead handling (not damaging coating), placement, airflow, etc. rather than let's say spend $$$ on a Vishay S102... in areas where one might be warranted.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.