We have an RF Device which GHz RF Signal Reception, Demodulation and Decoding Circuits.
The Power Rails for different sub circuits have 5V, 3.3V, 2.7V respectively.
And those power supply rails have 1nF, 100nF and 10nF capacitors with tolearance 5%.for each of them.
My colleague asked me whether it is possible to use 10% capacitors of the same value in this Design.
I doubt the RF Device would work in the long run, however, when we tested the RF Module on the functional tester (used in production) with some or all of the above mentioned capacitor removed, it does not show significant impact on the performance and functionality of the module.
Why would you want to use 10% capacitors? Under the terms of the purchasing agreement, 5% caps may be cheaper than stocking both 5% and 10% caps. Unless you have very high volume, saving a penny or two will be eaten up by documentation, qualification, and all the other overhead of making a change. How long would it take to recoup the engineering expense already incurred by taking out caps and testing the module.
It's tempting to conclude that if the tolerance were critical, there'd be values other than 1,100,1000...and it shouldn't matter. That's the thinking that makes so many cost reductions go so horribly wrong. I can't count the times a production shutdown or high failure rate was traced to a purchasing agent saving 2-cents and asking the wrong person for permission to change.
There are other electrical, mechanical, reliability specifications for a cap. It may be that, to get the other parameters you need,
5% is/was the available tolerance.
A "Functional Tester" is designed to test the function of the module at the lowest possible test cost. For a well-designed test, you get high probability that it works as designed. But when you change the design, you need to re-evaluate the functional tester. Just 'cause it works at the test conditions doesn't mean that it will work at the customer's use frequencies. Taking out bypass caps and verifying the FUNCTIONAL test is VERY RISKY BUSINESS.
Until you verify with the designer which material to use, dont make any changes, a cheaper often comes with nasty changes in temperature or a self resonance, and unless you can measure both your transmit spectrum and receive sensitivity, changes to a transceiver are foolhardy.
so now you have two against and one for, lets see how this pans out.
Unless we are talking about tens or hundreds or thousands, etc. piece counts, in a single one off situation, it is not worth switching unless you are having a problem sourcing the proposed ejectee. For the lack or knowing the effect alone. Though it would likely be none.
This is really a do whatever you want situation, and the chance for it to throw off the compliance or operation or such is probably only like
If you do not have access to any kind of design documentation that might explain why certain things have been done, look at the layout, if some of the components are in unexpected places.
If components are in strange places, it might be that the prototype did not work properly (e.g. failed EMC tests) and some experimentation was needed to make it work. In such cases it would make sense to use the components as specified.
However, if the layout is conventional and especially when only 1, 10 or 100 nF components are used for bypassing it is very unlikely that the design would be sensitive to tolerances.
However, the type of capacitor should not be substituted with something with very different self-inductance or different losses at the frequency of interest, since the series self-resonance frequency and the depth of the notch would be radically different and this might alter the operation of the whole circuit.
Those are typical generic values picked as bypass caps, so they are almost certainly not anywhere near critical, so the 5% figure doesn't matter a rats.
Almost certainly, and that's even without your test below...
Not surprising. That goes to show the values are not critical, they were just designed in as "best practice", or simply taken from an app note etc.
Your own practical test answered your own question.
Dave.
--
================================================
Check out my Electronics Engineering Video Blog & Podcast:
http://www.alternatezone.com/eevblog/
It is not common to offer smd ceramic capacitors with dielectrics other than NP0/C0G in 5% tolerance, but all values including 100nF parts are available in X7R, body size permitting.
It's possible that stocking both 10% and 5% values has presented difficulties in the past that are avoided simply by using only the 5% parts. This decision, having been made, should not be countermanded out of ignorance, for an apparent short-term benefit.
There is no electrical reason why parts with the same body size, dielectric and voltage rating should require a 5% tolerance for supply decoupling alone. At 1GHz, however, I wouldn't make any changes - even as to the supplier brand name or model type, without physical testing for suitability. This information may not be included in a simple schematic or a BOM.
The purchasing professionals will show their expertise by ensuring the supply of the specified component on time and under budget. Enquiries re substitution are long-term issues that should not be allowed to affect immediate scheduling.
While I tend to agree with what you said you are also not taking into account the situation. These are bypass caps. If they are electrolytic then they will have a huge variance over time. In fact most electrolytic's loose 80% of their capacitance over 20 years when not in use. Hence a variation of 5% is not necessarily significant.
In fact, since the caps are for bypass and in parallel the mean is exactly the same as that of 5%'s or 1% if one assumes a symmetric distribution. In layman's terms you'll have some capacitors with larger capacitance that will make up for those that were low. You might get unlucky and get an overall very low capacitance but this is unlikely and being bypass caps may only result in spurious operations which one might call a defective unit. Cost wise it might all cancel but if there are other compelling reasons to switch then it might be worth it. Also, bypass caps tend not to be chosen at the critical threshold for operation so a 5% drop will still probably be just fine.
He says they removed all the capacitors, not replaced, and the circuit still worked. If this is the case then I doubt there will be an issue.
An analogy to your rant is that changing the hubs on a car is going to change it's performance drastically. This is not true. In general, removing or replacing an arbitrary part in a car can drastically change it's performance but it's not necessary true.
I think it's YOU who are not taking into account the situation. Details were sparse, but I interpreted the situation to be that there was a working product in successful production. Somebody wanted to change the caps for no apparent reason. I guessed cost, but no details were given.
I never said it wasn't likely to work. I'll paraphrase what I said as, "only an idiot would sign off on the change." But there are plenty of idiots with pens who listen to engineers with tunnel vision.
"Hey boss, sign here. There's only a 5% chance of a product recall that will bankrupt the company."
That's a whole other topic. Don't know how they do it these days, but in the old days, if you specified a 10% resistor, about the only thing you could count on was that there'd be almost none within 5%. Those all got marked 5% and sold at a higher price. A symmetric distribution at the production site doesn't imply a symmetric distribution in this month's shipment to you. Counting on distributions within the spec is a rookie mistake. But I digress...
Quoting the original posting, "We have an RF Device which GHz RF Signal Reception, Demodulation and Decoding Circuits."
In "layman's terms", at RF, there ain't no such thing as caps in parallel. It's a distributed network in at least two dimensions. If you want to have some fun, stick a spectrum analyzer on one end of a board from power to ground. Put a tracking generator on the other end of the board and sweep it.
And you're assuming that the 5% number had anything to do with the reason that cap was chosen. There are a parameters at RF that matter. The tolerance number may be just what was available or convenient in a cap that had the proper RF, mechanical, environmental specifications.
In layman's terms you'll have some capacitors with
Your carefree attitude about spurious operations that cause a defective unit boggles my mind. Of greater concern is that you have a LOT of company.
Cost wise it might all cancel
That's NOT what he said at all. He said it "passed functional test". Try this experiment. Take most any product. Remove a heat sink somewhere. Run it thru functional test. I'll bet MOST of the time it will pass. (for you nitpickers, no, you can't take the heat sink off your
100W microprocessor, pick something else) Would you bet your company that the heat sink should be removed based on that experiment???
And what do we do about the possibility that one of those caps really isn't a bypass cap?
A better car analogy would be to put different diameter tires on front and back of your 4-wheel-drive vehicle without worrying about the additional stresses on the drive train and tires...and send the change order off to production. It'll drive around the block just fine. Come talk to me after 100,000 miles...if you can catch a ride.
The major point of my RANT is, IF IT AIN'T BROKE, DON'T FIX IT.
The relevant corollary is, Don't mess with it if you don't know what you're doing.
A person with a clue wouldn't have asked that question here. Taking advice from a newsgroup to modify a PRODUCTION product is insane. We have not nearly enough information to make an informed proposal. Unfortunately, we have plenty of hip-shooters willing to tell you to "go for it".
If the question had been, "can I use 10% bypass caps in my new design?" you wouldn't have heard a peep outa me. Of course, you can.
The question asked in this thread was more like, "I don't know enough to know if bypass cap tolerance matters, but I'm gonna "fix" a working production product by replacing the caps with crappier ones...and ship it based on functional tests. That OK? Oh, by the way it's a GHz. receiver/demodulator."
NO, IT'S NOT OK!!! IT'S INSANE!!!!! And the reasons it's insane have NOTHING to do with whether the tolerance is 5% or 10%.
As you point out it's most likely to do with dielectric rather than tolerance. Without information on the dielectric only an idiot would change anything.
I got caught by Z5U back in the 80s when I specified a value and not the dielectric. Loads of returns due to malfunctioning PLL. I wonder why anyone would use Z5U or similar.
It gets more annoying when, after carefully multi-spec'ing a part through easily-obtained triple sources, you get inquiries from purchasing for substitutes for the same part - every time it's scheduled into a production run over the next two years.
And those are the times when there is actually an inquiry - you have to assume there are others when you're either not there or there is no checking. There is apparently no learning curve being surmounted.
Thank you so much for your comments and discussion on this topic.
Finally, we have decided not to use 10% capacitors for the product.
My personal interest is mainly on the capacitor selection for power supplies in RF Devices, especially those GHz range, how the tolerance, dielectric property, and other ratings affect on the performance of this kind of circuits.
Is there any link which could provide me this info in details?
I would really like to start learning on this topic so that I would have more in-depth knowledge on this.
We have Agilent 8648C Signal Generator, and E4402B Spectrum Analyzer, which I believe, I could use them to learn on this issue, however, I am still curious to know whether there are the other things that I could try out on this.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.