EMRFD Message Archive 13917

Message Date From Subject
13917 2017-05-27 10:13:44 Ashhar Farhan Q of capacitors, SMD et al

while measuring Q of capacitors that i have ordinarily used for my work, I was appalled to know that they are of very low quality. One had assumed that the Q of the ordinary capacitors was high enough to concentrate on inductor Q in a filter. 

this is not the case with at least the selection available in India. 

Q of the capacitors is quite important in even the simple low pass filters that we design as they can dip to two digit figures.

this evening, i tried to measure Q of some SMD caps from the junk box. I started with a T50-6 toroid and measured it to have 1.1 uh with 16 turns. The test setup consisted of a spectrum analyzer with a tracking generator that measured the notch in a series trap made with the this toroid and the capacitor under test. I started with assuming that the toroid Q was 200. it turned out to be an fairly accurate pick. 

EMRFD has an easy way to measure Q. You sweep the series trap, measure the depth of the notch and plug it into the q-measure.exe and out pop the Q. The Q is of the resonator, that is inductor and cap in parallel. With the assumed value of the inductor Q, you can back calculate the capacitor Q. it works a little like resistors in parallel ( 1/Q_res = 1/Q_coil + 1 / Q_cap).  Details are on http://w7zoi.net/twofaces.pdf

This was used to measure a few Qs.

1. Polysterene 220pf caps from FDIM vendor night.
A whopping Q_cap of 1800 at 9.9 MHz

2. Low cost 220pf SMD 1206 from my component box.
Just 30! I had to double check this through the method 1 from the pdf quote above. The capacitor Q was really at 30 

3. Low cost 100 pf SMD 1206 from my component box
Q cap of 852, at 15 MHz

4. Paralleled 100 pf to make 300 pf, Qcap of 1133 at 8.736 MHz

Some observations: 

first, the attenuation notch for a particular capacitor type remains equally deep if you parallel them. the increase in Q is from lowered frequency in Q calculation.

second, it is good to have a second method to confirm unusual results. The values of Q will vary wildly. I will need a VNA at audio frequencies to measure the ubiquitous 0.1uf that we use everywhere. i am not sure if having a bad q is a good thing.

 third, paralleling up smaller values makes for better Q than the large capacitors. I am not sure if this is easier to model in physics or lt spice.

have fun, f.

13919 2017-05-27 10:40:02 Nick Kennedy Re: Q of capacitors, SMD et al
​That's quite interesting. I wrinkle my brow and look at those SMT (and sometimes leaded) caps and *hope* they're OK.  For those around 0.001uF and smaller, I usually go with NP0 or C0G and hope that their temperature stability correlates with good Q. But I have no basis for that hope.

I used to just grab a disk or silver mica and go. But with all these scary dielectrics used to make components smaller, I'm not sure what to think. 

I need to read that paper. I've used my PHSNA device to measure some coil Qs, similar to how I do crystals. I'll use resistive minimum loss pads to put the resonant circuit in a lower resistance environment for a better peak or null.​

I should try some capacitors.


Nick, WA5BDU

13923 2017-05-27 11:35:55 Ashhar Farhan Re: Q of capacitors, SMD et al
i would also add that, consistently, the Q of capacitors 100 pf and below is not bad. it coasts around 800. but it suddenly dips after 100 pf. i guess they change the technology to get more capacitance in.

- f

13932 2017-05-29 06:35:09 k1rf_digital_stev... Re: Q of capacitors, SMD et al
Ashhar, you might want to look at Murata smd cap data sheets.  They are a lot more comprehensive than the data sheets one typically sees from most vendors.  You would have to wade through the various types, but they do offer high Q types designed for RF.  They are a lot pricier than the run of the mill ones though.
-Steve K1RF
13933 2017-05-29 12:06:49 Ashhar Farhan Re: Q of capacitors, SMD et al
I have ordered a few of all types of avx and murata from mouser. I hope they arrive this week.

- f

13936 2017-05-30 08:11:31 Roelof Bakker Re: Q of capacitors, SMD et al
Hello Ashar,

This thread spurred my interest and today I have been looking at
some Yaego brand SMD size 1206 NPO capacitors using the first
procedure as described in 'The Two Faces of Q'.

The series capacitors used are 1 - 9 pF high quality piston
trimmers and were set to provide > 30 dB insertion loss. As signal
generator and detector a Rigol DSA-850-TG spectrum analyser was
used, which has a feature to automatically calculate the - 3dB
bandwidth. This is convenient, but not really necessary. In the past
I have been making this measurement using a home made signal
generator and a home made AD8307 based power meter.

For the first series of measurement I used a T68-6 toroid with 16
turns and an inductance of 1.4 uH. Using a good quality air variable
capacitor the following Q of the parallel tuned circuit was found:

15 MHz - Q = 238
10 MHz - Q = 200
7 MHz - Q = 211

The variable capacitor was replaced by a SMD capacitor and the
following Q of the parallel tuned circuit was found:

68 pF - Q = 250 at 15.7 MHz
100 pF - Q = 254 at 13.1 MHz
220 pF - Q = 275 at 8.95 MHz

A Philips wire ended 100 pF NPO capacitor gave a circuit Q of 266 at
13.15 MHz.

Calculating the Q of the 68 pF capacitor gave a Q of about 1000 at
15.7 MHz.

A second test was done using a T68-2 toroid with 25 turns and an
inductance of 4.3 uH. The following circuit Q's were found:

220 pf - Q = 255 at 5.2 MHz
470 pF - Q = 220 at 3516 kHz

An air variable capacitor gave a circuit Q of 226 at 3516 kHz.

I have been using these type of capacitor in low pass and band pass
filters, mainly for long wave and it looks like there is not much
wrong with them, which is reassuring!

Roelof Bakker, pa0rdt
13937 2017-05-30 10:15:27 robert.hair Re: Q of capacitors, SMD et al
Ashhar, I have some experience with Q issues concerning SMD chip capacitors.  While now retired, I worked for many years in RF circuit design of automotive remote keyless entry transmitters and receivers for a major supplier of automotive electronics.  (BTW, the detailed analysis of a Colpitts oscillator in Wes's "Intro to RF Circuit Design" was an invaluable help to me as a beginner in this field many years ago).
Our products operated in the low UHF range (mostly 315 to 433 MHz) and used lots of small value ceramic chip caps.  We did have sporadic problems with some suppliers of these caps related to Q and finally settled on Murata as primary supplier.  Some of our problems were with US suppliers, which we used for many other non-RF applications without problems.
The problem we had was that although most samples exhibited an ESR of under 1 Ohm, occasionally caps from certain suppliers would show an ESR of well above that value - sometimes 10 to 25 Ohms.  This was unacceptable to the proper operation of our circuits. 
Generally capacitor suppliers specify Q (or more commonly they specify dissipation factor, 1/Q) but measured at usually 1 MHz or even lower.  For the very low values (most of what we used in our RF sections were under 100 pF) this can give a very poor indication of Q or ESR, since capacitive reactance at 1 MHz is very high and X/ESR can give a high value of Q even when the ESR is far too high for the cap to be useful for our purposes.  I think the problem was related to the bond of the end conductors to the capacitor plates.
To resolve our problems we had to add a requirement to our component specification calling for the ESR to be less than 1.0 Ohm when measured at 300 MHz.  Murata was willing to accept this requirement and I think the caps we purchased were from their GRM series.  Most were in the 0805 and 0603 package sizes.
Having network analyzers, we were able to characterize these capacitors at the frequencies of their use.  Now that is even relatively easy for amateurs to do, since there are several low cost vector network analyzers available to us.  The VNWA by SDR-Kits is one good example. 
Rob Hair
13938 2017-05-30 10:20:31 Ashhar Farhan Re: Q of capacitors, SMD et al

The other method of using a series trap uses a single measure of the notch depth. it is pretty accurate and simple. It is easy to 'guesstimate' the Q values of inductor and capacitors.

If one starts with inductor Q of 250 for the T68-6, you can back calculate the capacitor Q. The Qs add up like resistors in parallel. 1/Qres = 1/Q-cap + 1/Q-inductor. The values look pretty good, probably too good. Can you recheck with the trap method too?

I don't have any NP0 in the box, i have ordered a few from mouser. i am awaiting the delivery. let met get them and i will post more.

- f

13939 2017-05-30 10:26:43 Ashhar Farhan Re: Q of capacitors, SMD et al

the big takeway from your note is for experimenters is to test the Q at the frequencies of interest. we tend to measure q as something that is constant across frequencies and voltages. this is simply not true and has no theory to back it. (the theory is contrary to this, Q is defined as reactance / resistance).

suprisingly, the smaller value caps, less than 100pf did well in these measurements. even at VHF. I have to construct a better jig to test them north of 100 MHz as the stray leads can create problems.  Soon, though...

- f 

13944 2017-05-31 06:26:20 Roelof Bakker Re: Q of capacitors, SMD et al
Hello Ashar,

Thank you for your comments.

I have repeated yesterday tests using the series trap method.

A new test jig was made on a 25 x 25 mm scrap of single sided PCB
laminate. Using a hack saw three traces of 8 mm wide each were made
and two BNC connectors were mounted.

This is what I found. The circuit Q found yesterday is noted within

1.39 uH 68 pF 33.73 dB 15.91 MHz 264 [250]

1.39 uH 100 pF 35.75 dB 13.25 MHz 279 [254]

1.39 uH 220 pF 40.16 dB 9.066 MHz 319 [275]

4.305 uH 220 pF 34.47 dB 5.250 MHz 295 [255]

4.305 uH 470 pF 36.83 dB 3.555 MHz 263 [220]

Using this method, the circuit Q appears to be a little higher,
albeit the difference trend is the same. I used the same inductors
and SMD capacitors as yesterday.

Roelof Bakker, pa0rdt
13945 2017-05-31 14:02:13 Ashhar Farhan Re: Q of capacitors, SMD et al
It is a resonatore Q that must be offest from the Q of the capacitor. I guess if you reduce it to the inductor Q, the results would be almost identical. The precise value of the Q is not as important as the magnitude of Q is.
The Q has two implications : it sets the loss in a filter and as a consequence of it, it also sets the noise floor of the oscillator. as broader resonators will allow greater noise bandwidth.

13949 2017-06-01 08:33:49 Roelof Bakker Re: Q of capacitors, SMD et al
Hello Ashar,

Some final comments. I believe that the differences I found are due
to the use of a different test jig and probably also due the two
series capacitors used with the 3 dB method. As you said, the series
resonant method is straightforward.

Today I have tried a number of good air variable capacitors and none
gave better results than the SMD parts, albeit some were close. The
test layout used short wires, about 10 mm.

A leaded NPO 100 pF and a 100 pF Styroflex capacitor were tested and
these provided slightly better results. The notch at about 13 MHz
was 0.3 dB deeper, so the difference is very small.

It looks like the Q of this type of SMD capacitors is excellent.
They are also cheap. Temperature stability is an other issue,

Roelof Bakker, pa0rdt
13950 2017-06-01 10:23:16 Ashhar Farhan Re: Q of capacitors, SMD et al
That's excellent. Mouser is listing them.
Here is a bizarre incident. I ordered 100 of 100pf, 220pf and 470pf NP0 caps from mouser.com and sat back expecting the postman to deliver them in a week as per their excellent service. 

However, this morning i got a mail from them, saying these are export restricted items from the USA. The funny thing is that I have purchased the caps made by TDK (a japanese company). I have purchased Si5351s, ferrites, many other ICs from mouser without a problem. Now, i ask for a few capacitors and I have to sign up all sorts of forms to get them here. Ugh!! 

Looks like a case of 'you can check out caps anytime you like, but you can never leave!!!', is joe walsh on this list?

- f

13985 2017-06-09 09:44:23 ashhar_farhan Re: Q of capacitors, SMD et al

Just an FYI to this thread, I bought some NP0 capacitors from mouser.com . These are listed as 12101A71JAT2A (for the 470pf) . The Q as measured by the notch method in EMRFD and calculated with q_measure.exe (also supplied with EMRFD) is above 1000 for all these caps.

These are pretty low cost caps. So, i guess, it is time to purchase strips of 100s of each of the regular values of these caps for filters.

- f
p.s. there are some styrene caps available as well, not very expensive. Q is north of 3000!

13987 2017-06-09 13:07:52 Andy Re: Q of capacitors, SMD et al
Catching up ...

My first reaction is that one might be naive to think that an "SMT" capacitor adequately describes it.  Caps come in all types.  Many SMTs are ceramics, and even those come in a wide variety, with significantly different electrical characteristics.  You have to know what you're using.  If it's a ceramic, the ceramic composition (usually a 3-character code) should be the first thing to look at.  Treat all parts with a different code as if they are a totally new component.

Many years ago I took some flack from my boss when I suggested changing from (I think) X7R to Y5V bypass capacitors in digital circuits.  The Y5V's were a little smaller and cheaper, and at first glance looked like they should be similar, but there is a world of difference.  At room temperature and no bias voltage they seem the same.  But -- they had MUCH wider tolerances, MUCH greater temperature dependency, and MUCH greater capacitance variation with voltage.  With a PC board having a couple hundred of these capacitors on it, the cost savings add up.  However, by running the numbers, you get a much smaller capacitance than the number printed on the cap, when operated at-voltage and at typical (elevated) operating temperature.  The drop-off can be quite dramatic.

While we were not looking specifically at Q (or DF), they tend to go hand-in-hand with the other characteristics.  Caps with higher capacitance density tend to have wider tolerances (part to part, over T, and over V), and much higher DF and lower Q.

NP0 (or NPO) formulations are among the better ceramics, but usually limited to smaller values.  This might be why there is good Q for the caps up to 100pF.  Again, you need to know what is the ceramic formulation -- not just the capacitance value!  The manufacturer isn't what matters, except for quality control.  Most manufacturers make multiple lines of ceramic formulations.

Higher dielectric constant ceramic caps like X7R and Y5V are usually meant to be bypass caps, where high density is important and Q is not.

In fact, very low ESR (high Q) in a bypass capacitor can hurt, by increasing the Q of parasitic resonances in the power distribution network on a board, which is a bad thing.  Good power designers often choose capacitors that don't have the lowest ESR, and therefore not a very high Q, for this reason.

Junk box SMT caps are a problem, since we usually don't know what they are.