Low absorption SMD capacitors

Look for polyphenylene sulphide capacitors. It has worse drift than the polypropylene, but otherwise very good characteristics, and a few types are available in SMD. Also try a company called 'Electronic Concepts Inc.', who specialise in slightly 'upmarket' capacitors, and were helpful to me in the past when looking for something with better characteristics. I think they do a polypropylene type. The reason for the 'lack', is that the film ceased production a little while ago. They actually arranged a run of film specially for these capacitors. They also do PTFE capacitors. :-)

Best Wishes

Some years ago I found that NPO ceramic microwave capacitors worked very well. The ones we used were made by American Technical Ceramics. This was a true 16-bit hybrid track/hold amplifier (AD386 if you still have an old Analog Devices catalog on your shelf). They provided about 20-40ppm of DA error alone IIRC, we were able to get nearly an order of magnitude better with some laser trimming (not the cap, but some other compensating stuff).

We learned some other interesting things during this project as well. They may not apply to your design. When we were breadboarding the circuit prior to designing the substrae for this chip-and-wire hybrid, we found that sometimes the DA was awful, and sometimes not. Ultimately it came down to the op-amp packaging! Metal cans were great, plastic nearly as good, ceramic (a.k.a. CERDIP) was right out. It definitely wasn't the op-amp, (AD744, BTW) we had part from the same wafer lot in all three packages.

Steve - still at ADI after all these years

For direct email replies remove the 3-letter food item from the address

Thanks, I'll look them up.

One thing which has bitten me already is surface leakage on the prototype board (soldering residues etc). I must have got lucky with the first one as it worked perfectly. #2 and #3 were totally unusable at first, I couldn't scrub the residue off as it would rip all the fine wires off the SOT23 transistors I had bodged onto the protoboard. A gentle wash with isopropanol improved things a lot, with a proper PCB I can give it a proper scrub and then conformally coat it. We are trying to measure photocurrents of about 200 pA upwards, so I will have to keep a close eye on leakage paths. Is ordinary FR4 PCB with conformal coating adequate for this job or will I need to look at ceramic hybrids etc?

OK, thanks.

Oh dear. This design will be in production for 5+ years so I may end up having to do a lifetime buy. At least we aren't talking massive volumes here...

OK, thanks.

Oh dear. This design will be in production for 5+ years so I may end up having to do a lifetime buy. At least we aren't talking massive volumes here...

Steve is right. I have taken measurements on three brands of NP0 C0G 0805-size capacitors and found them to be either perfect, or below 0.1% D.A. over a 100Hz to 250kHz range (this is close to my hp 4192's supposed 0.2% measurement accuracy), which is as good as my "pefect" polypropylene capacitor measurements. This compares to 5 to 15% D.A. that I measured for other ceramic-material 0805 capacitors.

Steve, is your concern the D.A. of a ceramic package's capacitance?

You have to clean the board shortly after soldering, then add a conformal layer to a dry board after test. FR4 should be OK in the region you're using, provided you have appropriate guards. In some instances you'll want to add rows of closely-spaced guarded vias to intercept the currents that could flow along fibres inside the PCB.

You will find that companies who are used to working with low leakage boards, will offer low residue cleaning, as a special service. I have some boards used for reading a couple of sensors, where virtual grounds around the inputs are used to make the input 'see' an extremely high characteristic impedance, and the input feeds in through Teflon standoffs and connectors, and with properly specified cleaning proceedures it is suprising how good FR4 can be.

Best Wishes

Others have mentioned NPO caps--AKA C0G. You can get these over the whole range you're looking for, and a decade on either side. It's possible to get them in 1% tolerance, if you ever do need closer than

Interesting to hear from another poster about the polyphenylene sulphide film production. I need to pass that info along to a team member who is planning on using them (in part because of the low DA).

Cheers, Tom

OK, great. These seem to be a lot more readily available in SMD.

By 'guards' I assume you mean a ring around the sensitive nodes connected to GND?

All our boards are made by a subcontractor so I don't have much control over the cleaning - some have been fine, some have come back covered in white detergent residue and other cr@p which has played havoc with even fairly low-impedance circuits. If I scrub them myself with a toothbrush and isopropanol after they've been sitting on a shelf for a week will this be OK or does the contamination leach into the FR4 itself?

I have used the LT1112 (plastic DIP package) for the prototype, it seems to work fine but I note from the data sheet that the worst-case input bias current and offset voltage could be a problem. In one sense it doesn't matter because the photodiode dark current will usually dominate, and I subtract this out by taking a dark measurement at the same integration time immediately afterwards, but if there is a large enough negative bias current the ADC will clip at zero and this method won't work. Another 'oddity' is that the SO8 package of this device (which I'm planning on using for production) has significantly worse specs than the PDIP - any ideas why? I always assumed the die was identical between different packages.

Yes, that was it precisely. Parts from the same wafer lot in different package types showed different amounts of dielectric absorption. We never bothered to figure out if it was the ceramic in the package, or the glass frit seal material, but parts in the CERDIP package showed much worse DA than in plastic or metal cans. Ultimately we didn't really care as the final product was a chip-and-wire hybrid on a single-layer thinfilm substrate with laser-trimmed Nichrome resistors.

We did also see a difference in cap materials. NP0 were generally good, but the best ones were the ATC microwave caps designed for really low ESR at superhigh frequencies. I can't remember if they actually specified DA, but those caps were quite good. At $1+ each they should have been!

steve

snipped-for-privacy@rblack01.plus.com wrote: ...

Guards are not necessarily connected to ground (optimally), but to a potential which is driven from a low impedance and follows the node (trace; vias) that you want to protect. If the node you want to protect is a virtual ground, then a grounded guard is perfectly appropriate. But if you want to guard the + input of an op amp that's used in the non-inverting configuration, you can guard that node with traces and vias that are connected to the - op amp input. You have to beware of capacitance to ground (or to other nodes) from that guard trace, though, because that can affect the response. You can also use an op amp buffer to drive the guard. Using a guard reduces both DC leakage currents and effective capacitance.

Don't have much control over your subcontractor? Ouch! Complain, and if they don't solve the problem, fire them and find one that will listen to you. Your situation may prevent doing something that drastic, but at least give them a good piece of your mind if they deliver cr@p. We've found that our contract manufacturers don't always get it right the first time, but they do want to deliver a product that meets our needs, and they do listen to us.

... Cheers, Tom

greetings, Vasile

snipped-for-privacy@rblack01.plus.com wrote:

My recollection is that glass has poor dielectric absorption characteristics. Do you have other info on it? The link seemed silent on the subject.

Cheers, Tom

I'm asking info directly from the AVX when the parts meet my requirements. They answer quite fast. However AVX is expensive (twice than Panasonic for example). On the application posted, the capacitor's dielectric loss is not the only problem as long the photodiode current is *very* small. Keeping the board clean around the low leackage points is much difficult than founding a good capacitor.

greetings, Vasile

Hi Vasile,

But the question is not about dielectric loss or leakage, it's about dielectric absorption. See, for example, Bob Pease's article at

I have an ongoing experiment involving the room-temperature self-discharge time constant of polyester (mylar) and polypropylene caps. Last time I checked, the polyesters managed between 5 and 10 years, and the polyprops came in at about 50 to 100 years. Indeed, it becomes diffidult to guard traces with a voltage that's close enough to the voltage on the guarded trace to get leakage currents that low. Even a 100uV offset, with 10^10 ohms, comes out 10fA I believe, and it's easy to end up with less than 10^10 ohms, especially if the humidity is high.

Cheers, Tom

Wow, that triggered an old memory. Back in the day ~20+ years ago as a test engineer i had to create fixtures for repeatable mass testing of SMD capacitors from very large (0.250 inch by 0.350 inch) down to about 0603 size. It turned out to be precision construction (less that 0.001 inch tolerances) and air insulated.

I heard there used to be problems long ago when they changed EPROMS from sidebrazed to CERDIP. The sealing glass released quite a lot of water during fritting and the chip passivation did not take that into account.

regards, Gerhard

Some articles on DA are on

regards, Gerhard