Spice, smice. You got good answers. Wake up, spice is just an engine to run your component models in circuits of your making. Once one gets beyond the "toy" level of using spice it becomes necessary to vet and complete the models of all your critical components, including parasitic circuit elements. Want a good RF capacitor model, you do well to make it yourself, from the manufacturer's data and info, and from your understanding of the part, aided by theory and bench measurements. Then vet your model with more bench measurements. Thinking about component leads and PCB wiring? Hey, you need to explicitly add all these into your circuit. Just don't blame any resulting shortcomings on "spice."
I didn't realize that this question was addressed to all "'spice' folks", but was directed only to the OP. That's probably why the "'spice' folks" haven't leapt to respond. Anyone who uses SPICE professionally, and indeed a great number of amateur users should be able to answer the question easily.
At 150 MHz, a capacitor lead can generally be handled as a single lumped inductance. The amount depends on the wire's diameter and environment, but around 7 nH is a good working number for a half inch lead. The inductance of the capacitor body itself might have to be included in the model if high accuracy is important. Capacitance to ground, from both the leads and the capacitor body, might also be important if the impedance of the circuit to ground is high. If so, it can be included. The way I'd approach inclusion of the C would be to calculate the impedance and length of the transmission line comprised of the lead and ground plane or capacitor body and ground plane, then convert those values to a single series L and shunt C rather than just directly using a transmission line model -- very short transmission lines in a model can greatly slow SPICE calculations unless there are also other very short time constants involved.
On the other hand, if the "leads" are PC traces over a ground plane, shunt capacitance will be higher, and the approach I mentioned with the transmission line is the way I'd always do it. The model for the leads would include both L and C. As an example, if the "lead" is a .010" trace on .032" FR4 material (er ~ 5) over a ground plane, it makes a transmission line of Z0 = 105 ohms, velocity factor 0.55. The equivalent L and C of a half inch of this line are 8 nH and 0.73 pF respectively.
Other characteristics of the capacitor such as ESR might have to be included in the model depending on the application.
You gave a capacitor value in your question -- an ideal capacitor of that value would of course be the other part of the model.
SPICE is used daily, as it has been for decades, by professionals and produces strikingly good results in the hands of someone who is skilled at modeling and has a good understanding of the circuitry being modeled. I've personally used SPICE for modeling linear to highly nonlinear circuits up to 50 GHz, where even tiny SMT components were often modeled as transmission lines and every pad and solder blob is significant and included. The results were used in the design of products which have been successfully produced by the thousands and sold for years.
I am not an expert, but can you not just add a bit of series L to approximate this? Or even transmission line for higher frequencies / longer tracks.
I have had surprisingly good results simulating a single transistor UHF oscillator at ~500MHz. The results agreed with reality quite well, even *without* explicitly modelling all the parasitics. Certainly spice was better than me at predicting what would happen (although that is not saying much!) I guess perhaps the reason that I could get away with this was because of the small size and high performance of SMT parts these days.
But that wasn't the question. THe question was to simulate a 1000 pf (1 nf) capacitor at 150 MHz. with an inch of lead or an inch of normal pcb trace and tell me what you get.
Such a question is stupid and incomplete. First of all, in cases where impedance matters, one wouldn't dare use an inch of lead at 150MHz, we'd cut that short, 0.1-inch max. And we certainly wouldn't use an inch of pcb trace unless it was field-controlled with a ground plane. This is true whether a 1nF cap is involved or not. If you were to insist on analyzing an inch of lead, we'd insist on knowing *all* about the ground scene. Since you aggressively put your question without any relevant information about what the ground is like, and where it is, the question is intrinsically-stupid and incomplete.
I simulated your question in a first trial. Here is the result: approx. 70MegHz resonance for 100pF, 30mOhm ESR and 5nH for the trace. Surely I would prefer smd chips having better results.
BTW: I cannot understand why you're so aggressive and I would prefer that the thread will go in direction of my FIRST posting. I'm doing electronics since I was 12 years old and that is 25 years back. I have for example a patent application made and worked in the communication industry. So you can surely think I know what I'm doing. Thanks!
- Henry
"RST Engineering (jw)" schrieb im Newsbeitrag news: snipped-for-privacy@corp.supernews.com...
No, Win, the question is well formed and quite complete. Sleep off the ten shots of Old Rammycackle and let's have the discussion when you are sober.
First of all, in
No, Win, neither you nor I would do such a thing. But somebody who is (as the OP posted) new to the RF world would do so without a second thought. You and I have been playing this game all our lives and take self-resonance into account without even thinking about it. However, a student new to the field (as my freshman engineering students are) makes the mistake repeatedly, even when using a decent text called ... um ... The Art Of Something Or Other. When their RF amplifier starts squeeging or motorboating, I tell them that the power supply isn't bypassed well enough, and I'll be damned if the first thing they do is put a BIGGER capacitor on the supply line.
I'll then ask them what they think the bypass impedance is and get the stock answer "1/(2*pi*f*c)". Hm, says I, how about the three inches of wire between the capacitor and the supply line. Oh, says them, that's a direct short. Straight wire doesn't have a reactive component. Hm, says I, let's see what the network analyzer says about that. Hm, says student, it says 60 nanohenries. How can that be? Mm, 20 nanohenries per inch for #20 wire sounds about right, so what does that series circuit look like? Hm. Inductive at the frequency of interest. Now, grasshopper, tell me about self-resonance of capacitors with long leads.
And
That's not always an option in commercial gear, Win.
This is true whether a
If I didn't say what the ground is, then we can assume that I formulated the question without ground plane. 99% of the commercial products run this way.
The question is unclear. Any component must have a return path, either when you measure it or when you use it in a circuit. The entire loop determines the "lead" inductance. If I solder an axial cap, with 1" of extra leads, onto the end of a hunk of coax, and analyze it with a VNA or TDR, I can bend the cap leads into various fat/flat loops and push the L all over the place.
Maybe I'm newer to rf as you but where is the difference between a microprocessor decoupling from the power supply at 100MegHz and a rf stage at the same frequency? Truly the cpu is more challenging because of the broad used spectrum above 100MegHz.
It is not a complete question. You could get all sorts of different parasitic inductance values by putting the component near or far from other metallic objects, which usually occurs to some extent in any practical situation. Many other things will affect the answer.
You have not yet supplied anything other than a single dimension measurement for the component. If you want the inductance, skin effect, etc. then you would have to give me a dimensioned drawing showing the placement of the wires, the plating material, plating thickness and the internal construction of the capacitor. (You would also have to pay me enough to make it worth me bothering to simulate it.) Your question is incomplete.
Well as you have not specified whether this is "commercial gear" and what type of PCB material, dielectric thickness, trace width etc. of course we can't tell you the answer. Neither could a guy who was going to answer your question by building one and measuring it. He could find one possible answer but there are lots of possible answers which differ because you have not given us a complete problem to solve.
You still gave insufficient information on the wire geometry. I would refute your claim that 99% of commercial products don't use a ground plane. The cell-phone market is in the high hundreds of millions of units this year, and is likely to reach 1 billion units per year next year, and I guarantee you that every one will contain a multi-layer PCB with ground planes and microstrip traces etc. every one of them designed using field simulators and some version of SPICE to model the integrated circuit packages and bondwires, as well as the antenna. I don't believe that this one billion units would fit into the 1% minority of products that you think have ground planes!
Anyhow, until you tell me how long a piece of string is, (to the nearest micron or micro-inch whichever you prefer), I have had enough of this thread.
Am Thu, 24 Nov 2005 09:55:13 -0800 schrieb RST Engineering (jw) :
Like some of us could lough nicely, when a new guy at a customer (semiconductor fab/asics) of the last company I worked for had to test an RF Chip. Another engineer came to look what the guy is doing, saw he didn't bypass the power supply and recommended to do so. The young engineer put a nice 2200µF electrolytic at the terminals of the lab-power-supply.
Model making is driven by need. 1nF capacitors don't exist on-chip (at least not very often :), so I don't have a model. But I have built a model of a wirebond that is good up into the 3GHz range.
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
Some of us have done better than 5% prediction BC (before CAD)... like for close to 20 years before I ever saw a computer... and 25 years before simulation software.
...Jim Thompson
-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |
formatting link
| 1962 | I love to cook with wine. Sometimes I even put it in the food.
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.