By critical I mean in active filters, equalizers etc.. a fairly high voltage X7R, say 50v rated, looks as though it might suffer a fraction of a % capacitance change with 1V bias across it.
So I reckon for the larger value capacitors slinging a few high value X7R's in series should produce a really stable capacitor in a fraction of the space & price of an 'audio grade' capacitor. By soldering flexible leads on the ends or spacing adequately/orthogonally etc on the PCB the microphonics could be mitigated...
For my application the physical size is crucial, as I'm miniaturising some analogue audio processing gadgets, so obviously surface mount MLCC's would be a great advantage.
What experience or measurements have readers here had/made? Also any information on which product lines are most suitable would be really helpful.
Cheers,
Robert.
So I reckon for the larger value capacitors slinging a few high value X7R's in series should produce a really stable capacitor in a fraction of the space & price of an 'audio grade' capacitor. By soldering flexible leads on the ends or spacing adequately/orthogonally etc on the PCB the microphonics could be mitigated...
For my application the physical size is crucial, as I'm miniaturising some analogue audio processing gadgets, so obviously surface mount MLCC's would be a great advantage.
What experience or measurements have readers here had/made? Also any information on which product lines are most suitable would be really helpful.
Cheers,
Robert.
For Filtering capacitors read the reports (Jung. Marsh etc).
They tested the distortion of RC and report that electrolytics are much poorer than film.
But even of the plastic film type they found a significant difference between the better and the not so good.
Teflon, polystyrene & polypropylene generally perform better than polyester, polycarbonate, polysulfone and others (https://en.wikipedia.org/wiki/Film_capacitor).
NP0/C0G if the values suit seem to get close to the better films.
They tested the distortion of RC and report that electrolytics are much poorer than film.
But even of the plastic film type they found a significant difference between the better and the not so good.
Teflon, polystyrene & polypropylene generally perform better than polyester, polycarbonate, polysulfone and others (https://en.wikipedia.org/wiki/Film_capacitor).
NP0/C0G if the values suit seem to get close to the better films.
It also depends on the overall linearity of the circuit you are using. TI do use mlcc x7r caps in places on their lower budget chips (of lower performance) for coupling and filtering. Cost and size do play an important role in many applications and going with film caps simply doesn't make sense.
Also not all mlccs will be created equal. If you can always measure to see if your performance criteria are being met and perhaps try different mlccs until you find one that could perhaps work.
Also not all mlccs will be created equal. If you can always measure to see if your performance criteria are being met and perhaps try different mlccs until you find one that could perhaps work.
range of np0/c0g vs film smt: http://www.kemet.com/Lists/Technica... Class-I C0G MLCC and SMD Film Capacitors.pdf
some film smt parts can be destroyed with hand soldering poor temp control - especially if you use lead free
some film smt parts can be destroyed with hand soldering poor temp control - especially if you use lead free
Happily using X7Rs and X5Rs as audio coupling caps in many of my circuits and have yet to encounter them being a problem. Yes I do 'oversize' them to minimize the voltage drop they might encounter. That's barely a problem though as 0603 4.7uFs are so cheap and miniscule, just parallel them to suit.
Where I have found ceramics to be a problem (in that they're mechanically noisy) is in switched capacitor voltage converters. They'll sing there so microphony is definitely a potential issue.
Modern SMD NP0/C0G caps are really good and really cheap. The problem is getting very large capacitances, but today's technology uses calcium zirconate and nickel electrodes, which are very inexpensive and extremely reliable. Even though calcium zirconate is a perovskite, it has none of the nonlinearities of even some of the ceramics used in NP0/C0G ceramics with precious metal electrodes from a while back.
These are very high quality caps, and since the electrode spacing is getting smaller and smaller, you can get larger capacitances in pretty small packages with extreme reliability for not much money today. Oddly enough, the least expensive NP0/C0G caps are the ones you want - they use the good dielectric with modern processing to achieve thin electrode spacing.
There may be specific applications that can use X7R, for example, but those caps are pretty annoying IMHO. I spent last week in a futile search for a thin X7R that would not lose 80% of its capacitance when biased to ~20V. If you must use those, you need thick 100V parts. My suspicion is that 'X7R' is not one dielectric, but a vendor specific family of dielectrics of various degrees of nonlinearity, applied to various parts according to how much breakdown voltage they can handle, how easy it is to make a stable and thin dielectric layer, and lord knows what other odd parameters that conspire to prevent me from finding a thin but usable X7R for 20V supplies!!
Re-think the X7R - they're pretty dirty, and my guess is that there are a number of specific dielectrics that get used for different voltages and thicknesses. After all, how could one X7R lose only 20% of its value at 20V when another loses 50% and a third loses 80%? That sounds like a mess to me… maybe it's just geometry and dielectric thickness that causes this, but you really have to start talking about specific parts and not generic families like an X7R tempco. Some parts may be good enough, but some also seem like they'd be a disaster.
Stay with the thick high voltage parts if you must, and share your results if you can!
These are very high quality caps, and since the electrode spacing is getting smaller and smaller, you can get larger capacitances in pretty small packages with extreme reliability for not much money today. Oddly enough, the least expensive NP0/C0G caps are the ones you want - they use the good dielectric with modern processing to achieve thin electrode spacing.
There may be specific applications that can use X7R, for example, but those caps are pretty annoying IMHO. I spent last week in a futile search for a thin X7R that would not lose 80% of its capacitance when biased to ~20V. If you must use those, you need thick 100V parts. My suspicion is that 'X7R' is not one dielectric, but a vendor specific family of dielectrics of various degrees of nonlinearity, applied to various parts according to how much breakdown voltage they can handle, how easy it is to make a stable and thin dielectric layer, and lord knows what other odd parameters that conspire to prevent me from finding a thin but usable X7R for 20V supplies!!
Re-think the X7R - they're pretty dirty, and my guess is that there are a number of specific dielectrics that get used for different voltages and thicknesses. After all, how could one X7R lose only 20% of its value at 20V when another loses 50% and a third loses 80%? That sounds like a mess to me… maybe it's just geometry and dielectric thickness that causes this, but you really have to start talking about specific parts and not generic families like an X7R tempco. Some parts may be good enough, but some also seem like they'd be a disaster.
Stay with the thick high voltage parts if you must, and share your results if you can!
I use an LCR meter and a bench power supply to provide the bias. Apply the bias voltage through a resistor with a value at least 100X the reactance of the capacitor at the measurement frequency. Use a series electrolytic capacitor between the meter and the cap itself so the meter's not exposed to DC - it should be at least 100X the value of the cap being measured.
Another way to do this, without breadboarding, is to use a vendor simulator to give you a basic indication of what can be expected from a part. For example, Murata has a nice simulator that seems to show a pretty accurate estimation of the behavior of specific capacitors under a variety of conditions, including DC bias. Their simulator is located at:
SimSurfing
Other companies also have similar standalone software simulators, such as AVX's Spicap, and I think Kemet has one (forgot the name) that handles all of this as well. In my experience, they are good tools to put some boundaries on the behavior of a part. The benefit is that it's possible to get a lot more detailed information about a part than what could be obtained from a data sheet. Yes, breadboarding is always a good idea, but being able to see whether a part is basically suitable or clearly not suitable without breadboarding it is very useful.
SimSurfing
Other companies also have similar standalone software simulators, such as AVX's Spicap, and I think Kemet has one (forgot the name) that handles all of this as well. In my experience, they are good tools to put some boundaries on the behavior of a part. The benefit is that it's possible to get a lot more detailed information about a part than what could be obtained from a data sheet. Yes, breadboarding is always a good idea, but being able to see whether a part is basically suitable or clearly not suitable without breadboarding it is very useful.
TDK leads the field as far as I'm aware - their webpage 'search by characteristic' feature allows inputting the desired capacitance at the projected bias and it'll return all the capacitors which meet or exceed the target. Pretty neat.
Search Result | Multilayer Ceramic Chip Capacitors | TDK GLOBAL
Search Result | Multilayer Ceramic Chip Capacitors | TDK GLOBAL
Hello,
tiny up for this thread, is a bit old but even more up-to-date in '21.
In '21, we have on the market some X5R MLCC at 330µF in form factor 1210 SMD (4V rating voltage).
These SMD can be stacked, thus, use 4 of them and you have 1.3mF X5R MLCC.
In 2021, why still can't we see 1.3mF of stacked X5R on PCB of digital audio ?
I'm not saying to put them anywhere.
I'm pointing the chips (CPLD, FPGA etc... XMOS...) <= those chips have a big steady-state, but on top of that, like any chip, there is a transient state.
By using 1.3mF X-type MLCC in front of each face of chips, you match their transient response.
Why still using electrolytics as bulk caps to match the transient responses of these chips ?
ESR of electrolytics (os-con is roughly equal) is so high that they can't match the transient reponses. And adding a decoupling MLCC of some nF before the chips is a true joke vs the transient response.
So, what's the up-to-date design to power the chips ?
Thanks for advice
Rgds
tiny up for this thread, is a bit old but even more up-to-date in '21.
In '21, we have on the market some X5R MLCC at 330µF in form factor 1210 SMD (4V rating voltage).
These SMD can be stacked, thus, use 4 of them and you have 1.3mF X5R MLCC.
In 2021, why still can't we see 1.3mF of stacked X5R on PCB of digital audio ?
I'm not saying to put them anywhere.
I'm pointing the chips (CPLD, FPGA etc... XMOS...) <= those chips have a big steady-state, but on top of that, like any chip, there is a transient state.
By using 1.3mF X-type MLCC in front of each face of chips, you match their transient response.
Why still using electrolytics as bulk caps to match the transient responses of these chips ?
ESR of electrolytics (os-con is roughly equal) is so high that they can't match the transient reponses. And adding a decoupling MLCC of some nF before the chips is a true joke vs the transient response.
So, what's the up-to-date design to power the chips ?
Thanks for advice
Rgds
If the capacitors are not in the audio signal path and are just to power/decoupling an IC even an audio IC (not only the DAC buy anything else, even an OPA), best solution is of course MLCC because ESR is very low (so ripple and transient will be better). Just be aware of voltage derating (a 100µ 6.3V will have an effective value of only 30-40µF @ 5V). Note that MLCC is more expensive than electro too (not important in some cases but it can be. On some of my boards at work I have for more than 100-200€ of capacitors because electro caps are not as reliable than MLCC). SOme Cpas in MLCC can have a cost of 3-4-5€ and equivalent in electro is 1€.
For audio path, only NP0/COG ceramic capacitors should be used --> if it's not possible because of cut frequency (need a big value and it is difficult witj NP0/COG caps), try to increase the resistor or use a X7R with a much more voltage rating than necessary in a big size package (X7R MLCC with a package of 1812 has much less THD than in 0603 package. If you have an audio signal of 1Vrms, use a X7R/X5R/X8R capacitor for coupling of at least 16V, 25V, 35V...).
For audio path, only NP0/COG ceramic capacitors should be used --> if it's not possible because of cut frequency (need a big value and it is difficult witj NP0/COG caps), try to increase the resistor or use a X7R with a much more voltage rating than necessary in a big size package (X7R MLCC with a package of 1812 has much less THD than in 0603 package. If you have an audio signal of 1Vrms, use a X7R/X5R/X8R capacitor for coupling of at least 16V, 25V, 35V...).
Antoher remark: the nF cap for decoupling is not for transient manangement but for HF "cutting". This is because the caps in µF range are no more a capacitor at HF frequencies. Only a small cap with a small value can "decouple" the signal @ HF frequency (for example bluetooth or wi-fi signal). So this is the reason why the decoupling of power IC is done by 2 caps : one 1µF or 10µF in MLCC or Tantal (less expensive) for transient/low frequency filter and a small 100nF MLCC in small package (in 0603, voltage rating is at least 50V)
