Post by thread:[EE] 1.0 uF Decoupling Cap

As far as I remember couple of guys here bet on the 1.0uF ceramic ones (or other low ESR) - Olin was talking about that a while ago I think. Couple of days ago someone else here (sorry, could not recall his name) said 1.0uF is not good for low drop regulators. I guess all depends on your application, power source etc. Tamas On Thu, Apr 16, 2009 at 1:51 AM, solarwind <.....x.solarwind.xKILLspam@spam@gmail.com> wrote: {Quote hidden}> On Wed, Apr 15, 2009 at 8:43 PM, <davidcouKILLspamaol.com> wrote: > > My advise is to use a 0.1uF in parallel with a 10uF (or 22uF) if you have > them. > > > > The goal is to provide a local current path as close to the Maxim IC for > the internal switching system. > > > > The 1uF is a minimum recommendation but I always recommend a 0.1uF > additionally as close to the IC. > > Thanks. I'll do the 10 uF + 0.1 uF. > -

solarwind wrote: > I just sampled some MAX233A chips (TTL <---> RS-232 drivers). These > are the newer ones that require no external caps for the charge pumps. > However, the data sheet says to use a 1 uF capacitor from VCC to VDD. > Why 1 uF? Unless the datasheet decides to explain the reasoning behind it or you can talk to a Maxxim engineer envolved in the design, don't expect to ever know. What you need to know is that's the spec. My guess is that the internal charge pump draws current in bursts and can't guarantee the output voltage and current specs if the supply is allowed to droop. They are probably assuming a barely good enough impedence of the power supply especially once it reaches the chip, and they found 100nF insufficient to cover their butts, so specified 1uF. > I have none of these. Would a 0.1 uF cermic cap work just as > well or is there some weird requirement that it has to be 1 uf? There is a requirement since the datasheet clearly says so. 100nF would probably work most of the time if you have a good power supply, but I would use the 1uF. The days of 100nF bypass caps are long gone for "normal" ICs anyway. Yes I know lots of people still tell you to use 100nF caps, but that's a sign they haven't kept up and should be a warning they probably don't know other things they are talking about either. Today's 0805 multilayer ceramic 1uF caps have lower inductance and higher resonant frequency than the 100nF caps of a few years ago, even though those were considered adequate at the time. I can only recall using less than 1uF decoupling cap in one circuit in the last 8 or so years where 1uF 0805 ceramic caps have been cheaply and widely avaialable. That was a RF circuit where the higher resonant frequency of a 100pF cap of a specific model was important. Each power lead to the chip had its own 10ohms in series follwed by this specific 100pF to ground. The common supply point before all this still had the 1uF to ground though. So the point is if you don't have 1uF 0805 ceramic caps you should get some. If you have a larger than 1uF ceramic cap you can use that instead. Anything ceramic of 1uF of higher should be good enough as the power supply cap for a MAX232 chip. Do not substitute electrolytic though. These don't have the frequency response to work in this situation. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Tamas Rudnai wrote: > Couple of days ago someone else here (sorry, could not recall his name) > said 1.0uF is not good for low drop regulators. Huh? There seems to be a lot of dead fish waving here today. All LDOs that I recall reading the datasheets of require some output capacitance for stability. Most also specify some minimum input capacitance because they need low input impedence. The gotcha is some LDOs, particularly but not limited to the cheap asian ones, require a minimum ESR of the output cap for stability. Unless you are doing a high volume design where the extra $.10 matters, you should stick to a LDO that is stable down to 0 ESR output cap. You can get yourself into trouble with the ones that require a minimum ESR if you don't know exactly what you're doing. Anyone that has to ask about decoupling here or uses rules of thumb and lore as a substitute for real understanding need not apply. This is one thing I like about the Microchip MCP1700 and MCP1702 LDOs. They are guaranteed stable with 0 ESR output caps. This means you can (and should) put a SMD ceramic cap right accross the output pins as close to the LDO as possible. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

For one thing...if an app note says..use a low ESR 1uF cap across these pins, its a pretty safe bet that you need to, since it is an app note from the manufacturer. Read and understand the datasheet Solar..it will really get you pretty far. Also look for other app notes dealing with the chip..any chip you use. Its the same as the (pick your favorite car maker) you should add oil to the crankcase. Sure, you might get pretty far without it, but at some point it will come back to you, for not following the recomendations. --- On Thu, 4/16/09, Olin Lathrop <.....olin_piclistKILLspam.....embedinc.com> wrote: {Quote hidden}> From: Olin Lathrop <EraseMEolin_piclistspam_OUTTakeThisOuTembedinc.com> > Subject: Re: [EE] 1.0 uF Decoupling Cap - A Must? > To: "Microcontroller discussion list - Public." <piclistspam_OUTmit.edu> > Date: Thursday, April 16, 2009, 5:53 AM > @spam@davidcouKILLspamaol.com wrote: > > The 1uF is a minimum recommendation but I always > recommend a 0.1uF > > additionally as close to the IC. > > Always!!? That's just silly. Do you specifiy the > species of dead fish to > wave over it too? > > > ******************************************************************** > Embed Inc, Littleton Massachusetts, > http://www.embedinc.com/products > (978) 742-9014. Gold level PIC consultants since 2000. > --

On Thu, 16 Apr 2009 09:02:28 -0400, you wrote: {Quote hidden}>Tamas Rudnai wrote: >> Couple of days ago someone else here (sorry, could not recall his name) >> said 1.0uF is not good for low drop regulators. > >Huh? There seems to be a lot of dead fish waving here today. > >All LDOs that I recall reading the datasheets of require some output >capacitance for stability. Most also specify some minimum input capacitance >because they need low input impedence. The gotcha is some LDOs, >particularly but not limited to the cheap asian ones, require a minimum ESR >of the output cap for stability. Unless you are doing a high volume design >where the extra $.10 matters, you should stick to a LDO that is stable down >to 0 ESR output cap. You can get yourself into trouble with the ones that >require a minimum ESR if you don't know exactly what you're doing. Anyone >that has to ask about decoupling here or uses rules of thumb and lore as a >substitute for real understanding need not apply. > >This is one thing I like about the Microchip MCP1700 and MCP1702 LDOs. They >are guaranteed stable with 0 ESR output caps. This means you can (and >should) put a SMD ceramic cap right accross the output pins as close to the >LDO as possible. Whearas the MCP1701 does specify a min ESR of 0R1 - min ESR requirements are not all that common but something you do need to watch for, and not something that you'd necessarily expect if you'd not previously read a lot of regulator datasheets..!

Mike Harrison wrote: >> This is one thing I like about the Microchip MCP1700 and MCP1702 LDOs. >> They are guaranteed stable with 0 ESR output caps. > > Whearas the MCP1701 does specify a min ESR Which is exactly why I didn't mention it. I don't like the MCP1701 much. It looks like it was a intermediate product on the way to the MCP1702. > min ESR requirements are not all that common Really? Over half of the LDOs I've looked at don't allow 0 ESR output caps. I'd say this issue is common. Last time I looked into it was maybe a couple of years ago, and at that time not a single one of the cheap asian LDOs were specified to work with 0 ESR output cap. Some ignored the issue which is even worse. Can you show me one that specifically says it works with a 0 ESR output cap? ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Michael Hagen wrote: > I bought some a few months ago, they were not available, Then you couldn't have bought any. > but Microchip Direct had them. So they were available. > Shipped from the factory, I was presently suppressed I guess I'd be "suppressed" too if I were shipped from a factory. Did you at least get your own crate and pallet, or did you have to share it? Hmm. How many people can you fit in a standard shipping crate nowadays, and how much cheaper is it? > and with the price? I have no idea what you are trying to ask here. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Funny. That was supposed to be Surprised! Good old spelling checker got me. I was trying to state that they were out of stock at my usual sources (Mouser and Digikey) so I directly ordered them. The price was competitive and delivery from the factory was less than a week. So, I had a good experience with Microchip direct. Anyway I was answering the question about ESR, the data sheet definitely states 0-2 ohms on this device, making me able to take a resistor in series with the output cap off my board. Nice part! It seemed like some asking if any LDO's could use 0 or near 0 ESR caps? I checked today, Mouser had the 5V ones. Mike {Original Message removed}

On Thu, Apr 16, 2009 at 8:53 AM, Olin Lathrop <KILLspamolin_piclistKILLspamembedinc.com> wrote: > RemoveMEdavidcouTakeThisOuTaol.com wrote: >> The 1uF is a minimum recommendation but I always recommend a 0.1uF >> additionally as close to the IC. > > Always!!? That's just silly. Do you specifiy the species of dead fish to > wave over it too? I knew someone who, as a regular part of their engineering, specified "Sprinkle caps." He'd add a few dozen capacitors across the power supply on one sheet of the schematic, and told the layout guy to just sprinkle them all over the board randomly (which is to say, the layout guy inevitibly put them where it was convenient for him). The boards were 4+ layer, with power and ground layers, chips had their own (correctly rated) bypass capacitors, etc. Turns out he failed an EMI test once, and ever since simply placed a bunch of caps all over to avoid later failures, "Cheaper than retesting!" ... except for the most popular products ... Engineering by the seat of your pants. Also, regarding the dead fish, I usually specify fresh salmon, and have a grill standing by for the conclusion of the test. -Adam

Olin Lathrop wrote: > If you have a larger than 1uF ceramic cap you can use that instead. > Anything ceramic of 1uF of higher should be good enough as the power > supply > cap for a MAX232 chip. Do not substitute electrolytic though. These > don't > have the frequency response to work in this situation. Olin, you haven't read the datasheet yourself, have you? It says nothing about not using electrolytics. alan smith wrote: > For one thing...if an app note says..use a low ESR 1uF cap across these > pins, its a pretty safe bet that you need to, It says nothing about "low ESR", either. Any cap will do. You guys seem to forget that these are level shifter circuits, not microcontrollers. The cap is there simply to keep the voltage on RS232 from going too far below 10V, for the duration of the bit. Vitaliy

Haha, thanks everyone for the detailed info! For now, I just stuck in a 10 uF + 0.1 uF until I buy some 1 uF. The thing also works perfectly without any capacitor at all, but that's probably because I'm powering it from USB. I never intended to question the datasheet. I just wanted to know what was the reasoning behind the 1 uF choice since everyone keeps telling me to use 0.1 uF. I also wanted to know if there was an alternative since I did not have any 1 uF caps available. Also, I wonder why my very simple topics always get so long, deep and sometimes violent :P Anyway, thanks again to all! My MAX233A works beautifully! I love being able to program and test while viewing output without touching my device. Development goes far more rapidly!

On Wed, 2009-04-15 at 20:03 -0400, solarwind wrote: > I just sampled some MAX233A chips (TTL <---> RS-232 drivers). These > are the newer ones that require no external caps for the charge pumps. > However, the data sheet says to use a 1 uF capacitor from VCC to VDD. > Why 1 uF? I have none of these. Would a 0.1 uF cermic cap work just as > well or is there some weird requirement that it has to be 1 uf? Generally, if the datasheet says use something, just use it. There is often a reason for these things (the reason of which may be very surprising) and it's best to just follow the directions. That said, and this is pure speculation: those chips have charge pumps. Charge pumps can sink relatively large amounts for currents for very short periods of time. Specing such a large bypass is probably also to ensure the droop on Vcc isn't beyond and acceptable level. So, with that said, it would probably be OK to use two caps, a 0.1uF for "normal" bypass reasons (i.e. noise on Vcc) and say a 10uF to minimize the droop. Note that this recommendation is based on PURE speculation as to the reason for the spec. TTYL

Vitaliy wrote: > Olin, you haven't read the datasheet yourself, have you? It says > nothing about not using electrolytics. We're talking about power supply "bypass" or "decoupling" caps. Basic electronics should be enough to assume that electrolytics are not suited for this application unless the datasheet specifically says so, and I expect it doesn't. There is really no place left today for a 1uF electrolytic capacitor. > It says nothing about "low ESR", either. Any cap will do. It may not say anything about ESR, but this gets to the level of obvious unless the datasheet specifically says otherwise. At some point they have to rely on the person reading the datasheet to know at least a little about electronics. Keep in mind that even though a electrolytic would not be violating the datasheet if taken literally, neither would a ceramic, and there are plenty of good reasons to use a ceramic instead of a electrolytic in this case. > You guys seem to forget that these are level shifter circuits, not > microcontrollers. The cap is there simply to keep the voltage on > RS232 from going too far below 10V, for the duration of the bit. It's a digital circuit that has switching transients and two or three charge pumps. Unless the datasheet explicity says otherwise, the assumption should be that good power supply bypass is required. The fact that they explicitly call for 1uF should be a big clue that something there matters more than usual. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

solarwind wrote: > The thing also works perfectly > without any capacitor at all, As far as you can tell over the limited range of test conditions you have applied to it. Running it without a power supply cap is asking for trouble. The likely symptom is that it works 99.9% of the time. > but that's probably because I'm powering > it from USB. Probably not. > since everyone keeps telling > me to use 0.1 uF. No they haven't. I certainly haven't. As I said before, 100nF bypass caps are so 1990s. If someone tells you that should be the standard value, you should be careful about what other bad or outdated advice they may be dispensing. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Olin Lathrop wrote: > Vitaliy wrote: >> Olin, you haven't read the datasheet yourself, have you? It says >> nothing about not using electrolytics. > > We're talking about power supply "bypass" or "decoupling" caps. The OP was specifically asking about a decoupling cap for a MAX233A. > Basic > electronics should be enough to assume that electrolytics are not suited > for > this application unless the datasheet specifically says so, and I expect > it > doesn't. There is really no place left today for a 1uF electrolytic > capacitor. This is completely backwards. If Maxim doesn't want us to use electrolytics, they should explicitly say so. >> It says nothing about "low ESR", either. Any cap will do. > > It may not say anything about ESR, but this gets to the level of obvious > unless the datasheet specifically says otherwise. Most datasheets would specifically tell you whether a low-ESR cap is required. For instance, Microchip goes to great pains to explain this when it matters (I recently read an entire app note devoted to this issue). > At some point they have > to rely on the person reading the datasheet to know at least a little > about > electronics. Keep in mind that even though a electrolytic would not be > violating the datasheet if taken literally, neither would a ceramic, and > there are plenty of good reasons to use a ceramic instead of a > electrolytic > in this case. If it ain't forbidden, it is permitted. If a non-electrolytic low-ESR cap is required, it is the manufacturer's responsibility to tell me so. This has nothing to do with my level of knowledge of electronics. The IC in most cases is a black box, and I trust the mfr to tell me what to hook up to it. {Quote hidden}>> You guys seem to forget that these are level shifter circuits, not >> microcontrollers. The cap is there simply to keep the voltage on >> RS232 from going too far below 10V, for the duration of the bit. > > It's a digital circuit that has switching transients and two or three > charge > pumps. Unless the datasheet explicity says otherwise, the assumption > should > be that good power supply bypass is required. The fact that they > explicitly > call for 1uF should be a big clue that something there matters more than > usual. I would have no problem if you said that SW should just buy a bunch of ceramic 1uF caps, and forget about the whole thing. But woe unto you, for ye lade newbies with burdens grievous to be borne unnecessarily. Vitaliy

> > We're talking about power supply "bypass" or "decoupling" caps. Basic > electronics should be enough to assume that electrolytics are not suited for > this application unless the datasheet specifically says so, and I expect it > doesn't. It doesn't actually seem (btw I hate the way maxim datasheets cover so many bloody parts that is is hard to tell what stuff is intended for your chip) to explicitly specify the cap is required it just shows it in the typical operating circuit. In that circuit it is shown as a polarised capacitor which (absent further information) would seem to indicate to me that an electrolytic is acceptable.

On Thu, 16 Apr 2009 18:30:46 -0400, "Olin Lathrop" <TakeThisOuTolin_piclistEraseMEspam_OUTembedinc.com> said: {Quote hidden}> > You guys seem to forget that these are level shifter circuits, not > > microcontrollers. The cap is there simply to keep the voltage on > > RS232 from going too far below 10V, for the duration of the bit. > > It's a digital circuit that has switching transients and two or three > charge > pumps. Unless the datasheet explicity says otherwise, the assumption > should > be that good power supply bypass is required. The fact that they > explicitly > call for 1uF should be a big clue that something there matters more than > usual. I'm going with Vitaliy here. The 1uF requirement is for bulk capacitance, not low ESR. Any cap will work perfectly fine. The resistance in the switching elements is higher than the ESR of the worst electrolytic. Cheerful regards, Bob -- http://www.fastmail.fm - mmm... Fastmail...

On Fri, Apr 17, 2009 at 7:56 AM, Bob Blick <RemoveMEbobblickTakeThisOuTftml.net> wrote: > > I'm going with Vitaliy here. The 1uF requirement is for bulk > capacitance, not low ESR. Any cap will work perfectly fine. > The resistance in the switching elements is higher than the > ESR of the worst electrolytic. Sorry I do not quite understand you. What do you refer by "switching elements" here? The discussion topic here is the power supply "bypass" or "decoupling" caps and I agree with Olin that in general electrolytic is not recommended in this application. It might work in some cases. The Maxim datasheet shows a polarized capacitor. I would assume that they mean tantalum capacitor or similar. Actually I would think a ceramic capacitor is more suitable nowadays. -- Xiaofan http://mcuee.blogspot.com

On Fri, Apr 17, 2009 at 9:29 AM, Xiaofan Chen <xiaofancEraseME.....gmail.com> wrote: > The discussion topic here is the power supply "bypass" > or "decoupling" caps and I agree with Olin that in general > electrolytic is not recommended in this application. It > might work in some cases. > Interestingly, Planet Analog just published a three part article on this topic. http://www.planetanalog.com/features/showArticle.jhtml?articleID=199905522 I still use mostly 100nF and 10nF ceramic (0603 or 0402) capacitors for bypass. For some more power hungry ICs, I will typically put 1uF (0603 or 0805), or 10uF (1206) in parallel with the 100nF/10nF. Even for bulk capacitor after the linear regulators, ceramic capacitors are replacing tantalum capacitors and electrolytic capacitors due to space constraint. The main problem is the high voltage (say 35V or 50V) bulk capacitors for DC/DC converters, that is still the kingdom of electrolytic capacitors most of the time. Even there I will put several ceramic capacitors in parallel. -- Xiaofan http://mcuee.blogspot.com

Xiaofan Chen wrote: > On Fri, Apr 17, 2009 at 7:56 AM, Bob Blick <RemoveMEbobblickEraseMEEraseMEftml.net> wrote: >> I'm going with Vitaliy here. The 1uF requirement is for bulk >> capacitance, not low ESR. Any cap will work perfectly fine. >> The resistance in the switching elements is higher than the >> ESR of the worst electrolytic. > > Sorry I do not quite understand you. What do you refer > by "switching elements" here? The on resistance of the charge pump transistors. It defines the peak current. > The discussion topic here is the power supply "bypass" > or "decoupling" caps and I agree with Olin that in general > electrolytic is not recommended in this application. It > might work in some cases. Why shouldn't it work in any case? It's a low speed device and needs bulk capacitance. > The Maxim datasheet shows a polarized capacitor. > I would assume that they mean tantalum capacitor > or similar. Actually I would think a ceramic capacitor > is more suitable nowadays. Why assume it's a tantalum capacitor? Why not assume it's an electrolytic? I realize that we're talking about Maxim, but if you look at the TI datasheet for the MAX232, it says on page 7, "Nonpolarized ceramic caps are acceptable. If polarized tantalum or electrolytic capacitors are used, they should be connected as shown" with the diagram showing polarity marks on the appropriate ends. So as far as TI is concerned, any of the three types of capacitor is OK with them. And I say the choice of 1uF is for bulk capacitance, and that their allowance of any type of capacitor backs that statement up. Cheerful regards, Bob

Speaking on Decoupling Caps (this is like an "Which is the BEST Oil Thread on my Motorcycle Groups", just keeps going), In the late 80's I worked for Philips Ultasound Corp. We had the best HP Impedance analyzer money could buy! I could not believe how good those cheap (OK , inexpensieve) 1uf Electrolytics were as far a ESR! Yes the ones from Digikey and Mouser, like 1uF/50V. I bitched about buying standard run of the mill capacitors for a machine that had high power drivers for the Ultrasound Tranducers. Everything was very critical, to image, the bandwidth of our array drivers was from about 700Khz all the way to 14 Mhz. These driver modules ended up being ceramic hybrid modules made in Israel. Lots of individual FET drivers, because of the focusing of the ultrasound transducer. Lots of noise! Hey, the best part was when we had demo in the lab at looking at babies in mom, facinating! The .1uf ceramics were going down in impedance until about 100Mhz, would dip (bottom out) and start up because of lead and self inductance. We soldered everything to copper pcb with no lead length to clamp in the analyser. I wish I could remember the exact numbers, but in the mili ohms, often a cheap electrolytic would really supprise. So a .1uf in parallel with the electrolytic did well. Of couse, lots of Spicing, and a really good Analog Engineer was the key. Not me, another retiree! I never got to work for many places that had the instruments and gave you the time to do such tests. I think PCB layout was much more important than the problems we had with components? But that is just another opinion from a real old fart! Mikey, remembering the HP Impedance Analyzer with fond affection!

On Fri, Apr 17, 2009 at 11:10 AM, Bob Blick <RemoveMEbobblickTakeThisOuTspamftml.net> wrote: > Why assume it's a tantalum capacitor? Why not assume it's an electrolytic? > > I realize that we're talking about Maxim, but if you look at the TI > datasheet for the MAX232, it says on page 7, "Nonpolarized ceramic caps > are acceptable. If polarized tantalum or electrolytic capacitors are > used, they should be connected as shown" with the diagram showing > polarity marks on the appropriate ends. > > So as far as TI is concerned, any of the three types of capacitor is OK > with them. And I say the choice of 1uF is for bulk capacitance, and that > their allowance of any type of capacitor backs that statement up. > Even if you are right here that electrolytic can be used, in any case, I have not seen any design using a low voltage low value electrolytic (1uF is low value for electrolytic) in the past 10 years as electronics design engineer. I have only used 1uF electrolytic once when I was designing some universal input sensors (level and photoelectric). The voltage input is 12-240V DC and 20-240V AC. Other than that, for low voltage application, you typically use ceramic and tantalum or similar (if ceramic is not good). -- Xiaofan http://mcuee.blogspot.com

On Apr 16, 2009, at 10:13 PM, Xiaofan Chen wrote: > I have not seen any design using a low voltage low value electrolytic > (1uF is low value for electrolytic) in the past 10 years as > electronics > design engineer. I see them all the time. Mostly attached to MAX232-type rs232 drivers on hobbyist devices (through-hole construction still being seen as an advantage, and high-value ceramic (1uF is relatively high value for a ceramic) being mostly SMT.) (Although I have some 100uF 10V ceramics caps in 2022 smt package; a bit mind-blowing...) http://www.sparkfun.com/commerce/images/products/28PinPicBoard-01-L_i_ma.jpg BillW

On Fri, Apr 17, 2009 at 2:06 PM, William "Chops" Westfield <EraseMEwestfwspamspamBeGonemac.com> wrote: > > On Apr 16, 2009, at 10:13 PM, Xiaofan Chen wrote: > >> I have not seen any design using a low voltage low value electrolytic >> (1uF is low value for electrolytic) in the past 10 years as >> electronics design engineer. > > I see them all the time. Mostly attached to MAX232-type rs232 drivers > on hobbyist devices (through-hole construction still being seen as an > advantage, and high-value ceramic (1uF is relatively high value for a > ceramic) being mostly SMT.) (Although I have some 100uF 10V ceramics > caps in 2022 smt package; a bit mind-blowing...) > > http://www.sparkfun.com/commerce/images/products/28PinPicBoard-01-L_i_ma.jpg You are right, there may be a lot of hobbyist design which is of different requirement than what I have seen. I should have put "professional design" or something similar. -- Xiaofan http://mcuee.blogspot.com

Vitaliy wrote: > Most datasheets would specifically tell you whether a low-ESR cap is > required. For instance, Microchip goes to great pains to explain this > when it matters (I recently read an entire app note devoted to this > issue). This doesn't mean everyone does. The manufacturer can't possibly put everything into a datasheet. At some point they have to rely on common sense and a little basic knowledge from the reader. > If it ain't forbidden, it is permitted. Then go ahead and use electroytics for bypassing. When units fail and customers complain, just tell them the datasheet didn't say anything against using electrolytics. I'm sure they'll understand. > If a non-electrolytic low-ESR > cap is required, it is the manufacturer's responsibility to tell me so. Should they also tell you not to bend the pins too often, not to chew the corners of the chip, not to put a 10ohm resistor in series with the power supply? > This has nothing to do with my level of knowledge of electronics. Of course it does. If you know nothing about electronics any PIC datasheet is going to be over your head. It's intended for a certain level of audience. Some knowledge has to be assumed. {Quote hidden}>> It's a digital circuit that has switching transients and two or three >> charge >> pumps. Unless the datasheet explicity says otherwise, the assumption >> should >> be that good power supply bypass is required. The fact that they >> explicitly >> call for 1uF should be a big clue that something there matters more >> than usual. > > I would have no problem if you said that SW should just buy a bunch of > ceramic 1uF caps, and forget about the whole thing. Actually I did tell him to get a bunch of 0805 1uF ceramic caps. > But woe unto you, > for ye lade newbies with burdens grievous to be borne unnecessarily. Huh? In this case I also explained why, partly because I don't like to just give a answer, and partly because he asked why. I don't understand your beef here. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

peter green wrote: > It doesn't actually seem (btw I hate the way maxim datasheets cover so > many bloody parts that is is hard to tell what stuff is intended for > your chip) to explicitly specify the cap is required it just shows it in > the typical operating circuit. In that circuit it is shown as a > polarised capacitor which (absent further information) would seem to > indicate to me that an electrolytic is acceptable. I was taking Solarwind's word for it that the datasheet specified 1uF. No, I'm not going to chase down the datasheet just to answer a basic question. Of course if he gave us incorrect information, then the resulting responses could also be incorrect. Note also that a polarized capacitor could mean tantalum, which has rather different properties for that application than electrolytic. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Bob Blick wrote: >> Sorry I do not quite understand you. What do you refer >> by "switching elements" here? > > The on resistance of the charge pump transistors. It defines the peak > current. It defines the peak current of the charge pump perhaps, but not of the whole chip. It's a digital chip, so there can be other switching transients that could be quite fast. > Why shouldn't it work in any case? It's a low speed device and needs > bulk capacitance. You don't know that. I agree part of the requirement is likely for bulk capacitance. Good engineering principles should indicate that a ceramic cap is the right answer here. At 1uF is satisfies the bulk capacitance requirement because the datasheet says so, but it also will provide power supply bypass for the high frequency transients if there are any. Since we can't tell for sure from the datasheet whether there are any, and basic electronics tells it's at least reasonably possible, using a ceramic cap is just good engineering. Let's keep this in perspective. I'm not advocating for a large or expensive solution. In fact, a ceramic 1uF SMD cap will be smaller, cheaper, more reliable, and more rugged than any tantalum or electrolytic of suitable rating. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

Olin wrote about what Bob said: >> But woe unto you, >> for ye lade newbies with burdens grievous to be borne unnecessarily. > > Huh? In this case I also explained why, partly because I don't like to just > give a answer, and partly because he asked why. I don't understand your > beef here. I think he is referring (loosely and jokingly) to Matthew 23:4 "They tie up heavy burdens 3 (hard to carry) and lay them on people's shoulders, but they will not lift a finger to move them.", where Jesus refers to the way the Pharisees wrote long lists of regulations for people to follow, then ignored the more fundamental truths of God's law (mercy, justice, helping others, etc.) http://www.usccb.org/nab/bible/matthew/matthew23.htm Sean

I do not know about the ESR of tantalum vs Al electrolytic, but I did recently learn that high value SMT ceramic caps often have a large change in capacitance with voltage. For example, I have some 2.2uF 100V 1210-size SMT caps which are only around 1.6uF at 50V bias. So, caution must be exercised when using these "magical" newer high value ceramic caps that you provide enough of them to get the minimum capacitance you need. It seems to be especially a problem for higher voltage ones, but I think it happens to a lesser degree even with the 16V and 25V ones. Sean On Fri, Apr 17, 2009 at 8:23 AM, peter green <plugwashSTOPspamspam_OUTp10link.net> wrote: {Quote hidden}> >> Interestingly, Planet Analog just published a three part >> article on this topic. >> www.planetanalog.com/features/showArticle.jhtml?articleID=199905522 >> > Hmm, I was under the impression that tantalum was lower ESR than > electrolytic and i've seen datasheets specify "ceramic or tantalum" > which reinforces this view. But the table in that article puts tantalum > as a higher ESR than electrolytic. > > Can anyone clarify? was I wrong? is this something that has changed with > recent technical developments? >

At 08:01 AM 4/17/2009, Olin Lathrop wrote: >William Chops" Westfield" wrote: > > 1uF is relatively high value for a ceramic > >Not for a few years now, at least for normal logic voltages. Take a look at >DigiKey, Mouser, etc. You will see 1uF 10V readily and cheaply available. >In fact 1uF 10V 0805 ceramic caps will be cheaper than any 1uF electrolytic >or tantalum, and also smaller and more reliable. So very true. For anything up to about 10uF at low voltage I now use ceramic because of the size and the high frequency performance. Up to 5V is can pretty much always use an 0603 package which helps a lot. On things like CPU's and FPGA's, especially with 1.2, 1.5, 1.8 or 3.3V supplies it is common to need a mix of bypass caps from 1nF to 1uF for each supply rail. For the switch mode regulators it is often necessary or desirable to use ceramics up to 22uF or more. In my current designs I don't think I have an electrolytic below 39uF or 47uF. For bulk energy tantalums have become less attractive as prices have risen and reliability has not. A combination of high value ceramics and electrolytics is likely to be equal in performance, lower in cost and have equal or better reliability. John

On Fri, 17 Apr 2009 08:34:42 -0400, you wrote: >I do not know about the ESR of tantalum vs Al electrolytic, but I did >recently learn that high value SMT ceramic caps often have a large >change in capacitance with voltage. For example, I have some 2.2uF >100V 1210-size SMT caps which are only around 1.6uF at 50V bias. So, >caution must be exercised when using these "magical" newer high value >ceramic caps that you provide enough of them to get the minimum >capacitance you need. It seems to be especially a problem for higher >voltage ones, but I think it happens to a lesser degree even with the >16V and 25V ones. I think this is moslty an issue with caps with '5' in the dielectric type, usually Y5V, which also have poor stability with temperature, and I've also found they can be slighly piezoelectric, emitting audible noise under high DI/DT situations. '5' series dielectric ceramic caps are best avoided in anything vaguely critical (like if you care about a 50% capacitance change), but are probably OK for non-critical decoupling at normal temperatures.

John Day wrote: > In my current designs I don't think I have an electrolytic below 39uF > or 47uF. For bulk energy tantalums have become less attractive as > prices have risen and reliability has not. A combination of high > value ceramics and electrolytics is likely to be equal in > performance, lower in cost and have equal or better reliability. I agree. I use electrolytics when the combination of capacitance and voltage is out of range of ceramics. Since switching power supplies are the norm nowadays and it seems every board has a few, the demands on large bulk storage caps has changed. Several times I've had to put electrolytic caps in parallel not to get the extra capacitance, but to not exceed the ripple current spec. Ripple current is a serious stress on capacitors used in switching power supplies, and all electrolytics are definitely not created equal. This is one of the few places I often specify particular models. I used to use the Panasonic FC series in these applications extensively, but the new FK series has even better specs for ESR and ripple current tolerance. (ESR and ripple current tolerance generally go together since it's typically not the ripple current itself that destroys a cap, but the heating it causes) ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

At 09:22 AM 4/17/2009, Mike Harrison wrote: {Quote hidden}>On Fri, 17 Apr 2009 08:34:42 -0400, you wrote: > > >I do not know about the ESR of tantalum vs Al electrolytic, but I did > >recently learn that high value SMT ceramic caps often have a large > >change in capacitance with voltage. For example, I have some 2.2uF > >100V 1210-size SMT caps which are only around 1.6uF at 50V bias. So, > >caution must be exercised when using these "magical" newer high value > >ceramic caps that you provide enough of them to get the minimum > >capacitance you need. It seems to be especially a problem for higher > >voltage ones, but I think it happens to a lesser degree even with the > >16V and 25V ones. > >I think this is moslty an issue with caps with '5' in the dielectric >type, usually Y5V, which also >have poor stability with temperature, and I've also found they can >be slighly piezoelectric, >emitting audible noise under high DI/DT situations. > '5' series dielectric ceramic caps are best avoided in anything > vaguely critical (like if you care >about a 50% capacitance change), but are probably OK for >non-critical decoupling at normal >temperatures. Good points. You really need to determine what is the minimum capacitance you require in the high value parts and then study the manufacturers data carefully to make sure you get what you need. There is really nothing magical about these parts, you still have to carefully read the data sheet to understand what you are getting! This WikiPedia page gives an excellent summary of what you can expect:" http://en.wikipedia.org/wiki/EIA_Class_2_dielectric from various dielectrics. It is not the 5 is the problem, it is the T, U or V on the end! You should only go for things like Y5V where space is an issue and you can be sure to get what capacitance you need. X7R, or even X5R is generally the better choice, as long as you have room for the higher values which do get big. Most of the class 2 dielectrics are based on Barium Titanate and have very high dielectric constants, hence why the can become piezo transducers under those high di/dt conditions. They can also introduce microhpnic noise under ihgh vibration conditions too. I once saw an audio system which used that class of capacitor and the engineer who designed it couldn't understand why he could get acoustic feedback when he moved the board so that it was exposed to the audio. John >

I recommend following the datasheet closely, especially for these components. I tried to use those parts a number oif years ago and was unable to get the product to pass FCC radiation tests. I had to drop back to a part that used 4 0.1uF for switching. --Bob On Wed, Apr 15, 2009 at 5:03 PM, solarwind <spamBeGonex.solarwind.xSTOPspamEraseMEgmail.com> wrote: > I just sampled some MAX233A chips (TTL <---> RS-232 drivers). These > are the newer ones that require no external caps for the charge pumps. > However, the data sheet says to use a 1 uF capacitor from VCC to VDD. > Why 1 uF? I have none of these. Would a 0.1 uF cermic cap work just as > well or is there some weird requirement that it has to be 1 uf? > > -- [ solarwind ] -- > -

At 09:39 AM 4/17/2009, Olin Lathrop wrote: >John Day wrote: > > In my current designs I don't think I have an electrolytic below 39uF > > or 47uF. For bulk energy tantalums have become less attractive as > > prices have risen and reliability has not. A combination of high > > value ceramics and electrolytics is likely to be equal in > > performance, lower in cost and have equal or better reliability. > >I agree. I use electrolytics when the combination of capacitance and >voltage is out of range of ceramics. > >Since switching power supplies are the norm nowadays and it seems every >board has a few, the demands on large bulk storage caps has changed. Indeed. My current design is a pair of boards that comprise essentially a PowerPC & FPGA based CPU board and a rather complex interface and analog control board. There are a total of about 15 regulators, and only a couple of those are LDO's where we need low current at low voltage. Especially as SM regulator frequencies increase we are naturally using more ceramic and less tantalum. It was only a few years ago that most switch mode regulators ran at 25 or 50kHz. Now even high current regulators run at 250kHz or more and for the lower currents it is not uncommon to go to 1.6 or even 3.2MHz. Oh how delightfully small they have gotten. >Several times I've had to put electrolytic caps in parallel not to get the >extra capacitance, but to not exceed the ripple current spec. Ripple >current is a serious stress on capacitors used in switching power supplies, >and all electrolytics are definitely not created equal. Have you ever noticed that many regulators on PC motherboards use three or four electrolytic capacitors in parallel in the output. Often they will have 4 caps, series inductor, and another 4 caps. Just proves that MORE can be BETTER! > This is one of the >few places I often specify particular models. I used to use the Panasonic >FC series in these applications extensively, but the new FK series has even >better specs for ESR and ripple current tolerance. (ESR and ripple current >tolerance generally go together since it's typically not the ripple current >itself that destroys a cap, but the heating it causes) Of all of the time I spend choosing components I probably spend more on things like power supply capacitors, especially electrolytics, than on anything else. Heating in capacitors is fatal. I agree about specifying particular models. I have gotten to the point now where I have one type form Panasonic, one from United Chemicon and one from NIC that make it onto my approved parts list. But another difficulty right now is specifying electrolytics that are also able to withstand the elevated temperatures required in RoHS / Lead free soldering processes. We use some 1000uF 16V and 330uF 50V SMT electro's and one of our contract manufacturers wants to hand solder them because they are not rated for a 260C reflow cycle. John >******************************************************************** >Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products >(978) 742-9014. Gold level PIC consultants since 2000. >

I agree that this depends greatly on the type of dielectric used, but I've seen the problem with X7R types, too, at least for ones which have a very high voltage rating for a small package (2.2uF 100V 1210 size in my case). Sean On Fri, Apr 17, 2009 at 10:01 AM, John Day <KILLspamjohn.dayspamBeGonesiliconrailway.com> wrote: {Quote hidden}> Good points. You really need to determine what is the minimum > capacitance you require in the high value parts and then study the > manufacturers data carefully to make sure you get what you need. > > There is really nothing magical about these parts, you still have to > carefully read the data sheet to understand what you are getting! > This WikiPedia page gives an excellent summary of what you can > expect:" http://en.wikipedia.org/wiki/EIA_Class_2_dielectric from > various dielectrics. It is not the 5 is the problem, it is the T, U > or V on the end! You should only go for things like Y5V where space > is an issue and you can be sure to get what capacitance you need. > X7R, or even X5R is generally the better choice, as long as you have > room for the higher values which do get big. > > Most of the class 2 dielectrics are based on Barium Titanate and have > very high dielectric constants, hence why the can become piezo > transducers under those high di/dt conditions. They can also > introduce microhpnic noise under ihgh vibration conditions too. > > I once saw an audio system which used that class of capacitor and the > engineer who designed it couldn't understand why he could get > acoustic feedback when he moved the board so that it was exposed to the audio. > > John > > >

William Chops" Westfield" wrote: > The rest of the paragraph was supposed to clarify "1uf is relatively > high value for a through-hole ceramic cap." So? Thru hole caps haven't been relevant in years. You can stick to point to point wiring using tube socket eyelets too, but the rest of the world has moved on long ago. > Out of the 934 caps that > digikey reports for in-stock 1uF ceramic caps, 44 are TH. That's probably because there are some existing old designs that have been qualified and continue to be produced. Even those are clearly a small fraction of the capacitor market. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

At 02:39 PM 4/17/2009, Olin Lathrop wrote: >William Chops" Westfield" wrote: > > The rest of the paragraph was supposed to clarify "1uf is relatively > > high value for a through-hole ceramic cap." > >So? Thru hole caps haven't been relevant in years. You can stick to point >to point wiring using tube socket eyelets too, but the rest of the world has >moved on long ago. LOL But of course I go into our factory and I see lots and lots of through hole ceramic caps, but they have ratings like 5kV, 15kV and the like. I wish I could find some nice 15-20kV rated ceramics in SMT that would handle immersion in oil. > > Out of the 934 caps that > > digikey reports for in-stock 1uF ceramic caps, 44 are TH. > >That's probably because there are some existing old designs that have been >qualified and continue to be produced. Even those are clearly a small >fraction of the capacitor market. A VERY VERY small fraction of the market. Some products we have been building for 15-20 years are still done through hole, but we are moving them to SMT because we can no longer get parts we need. As time goes on even the average home experimenter will need to adapt to SMT, it isn't hard, I have used pretty much all SMT even for home built hobby stuff for most of the last 10 years. But Olin I htink sometimes that people like you and I are the exception, plenty of the folk on this list are frightened by SMT and lots of them wont even try it. John

solarwind wrote: >> not to put a 10ohm resistor in series with the power >> supply? > > Wait, what's wrong with that? (I'm a total noob to analog electronics.) What does Ohm's law have to say about the supply voltage to the part as the part draws current when there is 10 ohms between it and the power supply? How much current causes the supply to dip below specs? How much current does the part draw? That last question is the zinger. Explain why or take a guess at it. ******************************************************************** Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products (978) 742-9014. Gold level PIC consultants since 2000.

> Hmm, I was under the impression that tantalum was > lower ESR than > electrolytic and i've seen datasheets specify > "ceramic or tantalum" > which reinforces this view. But the table in that > article puts tantalum > as a higher ESR than electrolytic. > > Can anyone clarify? was I wrong? is this something > that has changed with > recent technical developments? Peter, Never rely on a "generalized" table. Tantalum caps can have extremely low ESR, depending on the manufacturer and series. Some have internal fuses in series which raises the ESR (and negates the reason for using tants in the first place). You must read / understand / interpret the spec sheets for all parts in order to choose correctly. Manufacturers typically are poor at rating their caps, especially for items such as temperature rise and ripple current. All of the above being said, E-Caps typically have highest ESR (violent hand waving now in progress). Large value (multi-uF) ceramic caps have gotten much cheaper in recent years, have decent ESR and are often preferred over tantalums for various reasons. Ken