I'm currently designing a PCB that is intended to fan out the 38 pins of a MICTOR connector to every other conductor of an 80-conductor ribbon cable (The PATA type one, that is AWG30 wires with a 0.063mm / 25mil spacing). Just as PATA, I intend to use every other conductor as ground wire, which should result in a signal-to-ground impedance of around 80 ohms.

I am having a hard timing getting the characteristic impedance up high enough to hit 80 ohms: The board will be a 4-layer PCB, with "prototypes" produced at JLCPCB. As this is a hobby project, the prototype will be the product, there will be no "production run". The intended stackup is sig-gnd-gnd-sig. The minimum permitted trace width and the minimum trace spacing at JLCPCB is just below 0.1mm, so I am considering 0.1mm traces for the signal. To reduce crosstalk, adding ground traces or ground fill on the signal planes seems like a good idea.

If I understand it correctly, the resulting configuration is either "microstrip" (just looking at a single trace), or "coplanar waveguide with ground" (looking at a trace between grounded shielding elements). The default 4-layer stackup offered by JLCPCB uses a 7628 prepreg layer between my signal and ground layers, which has a relative dielectric constant of 4.6 and a thickness of 0.21mm (8.3mil). Entering er=4.6, track width=0.1mm, track thickness=0.035mm, dielectric thickness = 0.21mm into an online impedance calculator like https://chemandy.com/calculators/microstrip-transmission-line-calculator-hartley27.htm results in 95.3 ohms for the microstrip configurations, and 80 ohms for the coplanar waveguide configuration with a gap width of 0.1mm.

On the other hand, using the JLCPCB impedance calculation tool (which always solves for track width), I get a width of 0.1mm when requesting an impedance of 82.4 ohms (microstrip) or 61.5 ohms (coplanar with ground, gap size 0.1mm). I assume the JLCPCB calculator results better match the actual impedance on the PCB produced by JLCPCB, but I wonder why the difference between the simple standard model and the JLCPCB calculator is that big. One idea that comes to mind is the solder mask, which is another dielectric layer that is not considered in the simple onlince calculators, another idea is that I violate some approximation assumption (e.g. the very low track width) used in the simple formula-based calculators.

So, to reach 80 the target ohms impedance, is it recommended to reduce screening to a minimum (thin traces that stay away from the signal lines as far as possible to get close to the "microstrip" configuration, as suggested by JLPCB to get 80 ohms), or is it recommended to ground pour the signal layers as much as possible to get to the "coplanar waveguide" configuration, as suggested by the simple models)?

  • \$\begingroup\$ Are you asking why one web-tool gives different results to another web-tool? \$\endgroup\$
    – Andy aka
    Commented Jun 4, 2023 at 17:13
  • 3
    \$\begingroup\$ @Andyaka I am mainly asking which web tool is to be considered "more trustworthy" in my scenario. Knowing why the results are different (i.e. what kind of assumptions went into the most likely well-known formulas) would help me understand the (un)reliability of the different web tools. \$\endgroup\$ Commented Jun 4, 2023 at 17:23
  • \$\begingroup\$ That isn't a valid question for this site because in one month or an hour or a decade, things might change and invalidate any answer given. \$\endgroup\$
    – Andy aka
    Commented Jun 4, 2023 at 19:05
  • 1
    \$\begingroup\$ Is a two-layer board with one GND plane on the back an option? Or even a single layer board with coplanar ground traces? These options will easily generate higher impedance without tiny trace widths. \$\endgroup\$
    – tobalt
    Commented Jun 5, 2023 at 4:46
  • 2
    \$\begingroup\$ @tobalt or just relieve inner ground under the traces, to the same end (giving three options for substrate height on a 4-layer board) \$\endgroup\$ Commented Jun 5, 2023 at 6:31

3 Answers 3



Let's first consider the case of JLCPCB.

The last time I investigated the issue, JLCPCB's calculator was based on a lookup table using the model and results from Polar Instruments Si9000. Si9000 is a widely used field solver in the industry. In comparison with simpler model, the Si9000 model features: (1) using trapezoidal traces instead of rectangular trace due to etching effect (The top width is slightly shorter than the bottom), and (2) modeling solder mask as another layer of dielectric.

The model looks similar to this:


JLCPCB also gave parameters used for the calculations. It's worth noting that I found the parameters given in the JLCPCB's Si9000 model (such as dielectric thickness) are sometimes different from the official "stackup" possibly to "fit" the results.

JLCPCB's Si9000 parameters

If you have access to Si9000, one could reproduce the result almost exactly. Since Si9000 has a license fee of many thousand dollars, I also compared it with a free and open source field solver TNT-MMTL. After modeling the solder mask, both agreed to within 1 Ω for microstrip. One can see an impedance reduction around 2 Ω by turning the solder mask on and off. Once it's off, they also agree with the common Wheeler or Hammerstad-Jensen microstrip formulas found in EDA calculators.

Since the last time I've checked, apparently now the situation has changed. In the past, only microstrip was supported, but now new geometries like coplanar waveguide have been added, so I cannot comment on how does JLCPCB obtain their numbers or their accuracy. All the previous comment only apply to the historical situation. Since I haven't reinvestigate the issue, I can't make any claim with confidence.

But... I thought it would be interesting to do a re-comparison right now. So, for the purpose of answering this question, here, I compare the result from JLCPCB calculator, Chemandy Electronics website calculator, TransCalc calculator (as seen in Qucs and KiCad), TNT-MMTL field solver and Si9000 field solver.


  1. 50 Ω microstrip:

    • JLCPCB: 0.34925 mm (100%)
    • Chemandy Electronics: 0.39 mm (111%)
    • TransCalc: 0.367 mm (105%)
    • TNT-MMTL (mask): 0.353 mm (101%)
    • Si9000 (mask): 0.3539 mm bottom, 0.3285 mm top (97.7%)
  2. 0.1 mm microstrip

    • JLCPCB: 82.4 Ω (100%)
    • Chemandy Electronics: 95.3 Ω (115%)
    • TransCalc: 88.67 Ω (107%)
    • TNT-MMTL (mask): 84.46 Ω (102%)
    • Si9000 (mask): 83.6 Ω (101%)
  3. 0.1 mm coplanar waveguide, 0.1 mm gap to ground

    • JLCPCB: 61.5 Ω (100%)
    • Chemandy Electronics: 80.48 Ω (130%)
    • TransCalc: 77.02 Ω (125%)
    • TNT-MMTL (no mask): 66.78 Ω (109%)
    • TNT-MMTL (mask): 58.52 Ω (95%)
    • Si9000 (no mask): 69.79 Ω (113%)
    • Si9000 (mask): 61.71 Ω (100%)


  • For microstrip transmission lines, almost all calculators give roughly the same results, within 5%-10%.

  • For coplanar waveguide with ground, impedance calculators start to break down with diverging results. But both the free field solver TNT-MMTL and the commercial field solver Si9000 agree with each other. So I'd consider field solvers more trustworthy.

  • Solder mask has a significant effect on the impedance of coplanar waveguide, and simple calculators don't model them. This can be a major source of error. In my experience, for microstrip, it's only 2 Ω or so. But for coplanar waveguide it can be as high as 5 Ω, possibly because much more electromagnetic field exists above the board, so it's very sensitive to an extra dielectric layer.

Calculators Only Give Guidance

Don't always trust calculators. There are multiple tools for calculating characteristic impedances, including different closed-form approximation formulas, 2D field solvers, and full-wave solvers. Each model is different, all of them (but in particular simple formulas) should only be seen as a guidance, and ultimately one should verify them by measurement.

In fact, the dielectric constant given by manufacturers is sometimes not physical but mathematical - a "fitted" value to allow the use of common formulas. This is even inevitable in a sense, because it's frequency-dependent. For the purpose of digital design, its value at 1 GHz is commonly used.

As a result, if it matters in volume production, the target impedance should be specified in the manufacturing requirements and be routinely checked in the quality control process using "test coupons". As a hobbyist, if you're doing it only for experiments, and if the impedance is critical, my suggestion is to use the board vendor's calculator as a starting point, then create a custom impedance test fixture with varying line geometry. Measure it with a TDR or VNA, tweak it until you get the optimal result. Even then, there's still a variation due to manufacturing tolerance.

Removing solder mask can improve impedance control. For a simple microstrip, my experience is that it can reduce its characteristic impedance by 1 to 2 Ω, as shown in both my simulations (via field solver) and experiments.

In the book Right the First Time, a Practical Handbook on High-Speed PCB and System Design, its author Lee W. Ritchey says,

Comparison of Field Solver to Equations

Comparison of calculated impedances using different methods

Figure 24.3 is a plot of impedance versus trace width for the three most common types of transmission lines in PCBs-- surface microstrip (SMS), buried microstrip (BMS) and centered stripline (CSL). All transmission lines are 1 ounce thick (1.4 mils) and are 5 mils above the nearest plane with a dielectric whose relative dielectric constant is 4. The only variable is trace width.

The three curves with the square markers are the impedances predicted by the three impedance equations presented earlier. The three curves with the diamond markers are the impedances predicted by a 2D field solver. As can be seen for the centered stripline case, the 2D field solver and the equation results (Equation 24.2) are very close to each other. For the two microstrip cases, there is a significant difference.

Which is right? It has been shown in many papers that 2D field solvers predict impedance that agrees within the measurement accuracy of the test equipment used to measure impedance. It has also been shown that equations commonly predict the wrong impedance. Because of this equation error problem, the users of equations have often resorted to iterative, trial-and-error methods for adjusting the equations to allow them to predict the correct impedance. Some PCB fabricators have gotten quite good at “fudging” the values of er used in their calculations in order to get good answers from the equations. The problem with this “fudging” is it requires considerable trial and error with new materials until they are “dialed in.”

Even if there is time to “dial in” a new material by manipulating the er values to arrive at accurate impedance calculations, it leaves the PCB engineer not knowing the true er of the material. This is needed in order to accurately calculate velocity for the purpose of predicting the time of flight along traces.

Even with a good impedance predicting tool, such as a 2D field solver, impedances are still often calculated wrong. This is because the er value given for a particular laminate may be incorrect. The most common published values for er are measured at 1 MHz. Figure 16.2 shows that the relative dielectric constant of all common PCB materials decreases with frequency. It also shows that the relative dielectric constant varies with the ratio of glass to resin in the material. Therefore, getting the impedance calculation correct requires using a good tool; knowing the glass to resin ratio of the laminate and knowing the frequency at which the transmission line will be used. Once all three of these factors have been taken into account, the results are predictable and repeatable.

Raw Data

Chemandy Electronics

  • Microstrip formula: Rick Hartley's simple formula
  • Coplanar Waveguide formula: Transmission Line Design Handbook by Brian C Wadell, Artech House 1991 page 79.

I'm pretty sure the simple microstrip formula here was not invented by Rick Hartley, other books like High-Speed Digital Design gave similar formulas. These formulas are less sophisticated than more complex formulas such as Hammerstad-Jensen.


  1. er = 4.6
  2. track width = 0.1 mm
  3. track thickness = 0.035 mm
  4. dielectric thickness = 0.21 mm
  5. coplaner waveguide spacing to ground: 0.1 mm


  1. Microstrip: 95.3 Ω
  2. Coplanar Waveguide: 80.48 Ω

TransCalc (or KiCad, or Qucs)

  • Microstrip formula: Wheeler's formula for single microstrip synthesis
  • Coplanar Waveguide formula: ?

TransCalc is an open-source impedance calculator of various transmission lines based on common closed-form approximations. This is the same calculator integrated in other open-source tools like the microwave circuit simulator Qucs, or the circuit board design tool KiCad.

My impression is that the tool is pretty robust (within the limitation of closed-form formulas), based on tested-and-proven formulas such as Wheeler's formula microstrip impedance formula.


  1. er = 4.6

  2. track width = 0.1 mm

  3. track thickness = 0.035 mm

  4. dielectric thickness = 0.21 mm

  5. microstrip target impedance: 50 Ω.

  6. coplaner waveguide spacing to ground: 0.1 mm


  • 50 Ω microstrip: 0.367 mm
  • 0.1 mm Microstrip: 88.67 Ω
  • 0.1 mm CPW: 77.02 Ω

JLCPCB calculator


  1. Stackup: JLC04161H-7628 (Standard)
  2. Microstrip: L1 to L2
  3. Target impedance: 50 Ω
  4. Target impedance: 82.4 Ω


  1. 50 Ω: 13.75 mils (0.34925 mm)
  2. 82.4 Ω: 3.94 mils (0.1 mm)

However, when I attempted to calculate the 82.4 Ω microstrip, the result is shown in a red, not green color. I also got an error message that says:

Failed to calculate the value, please check your impedance configure

So its accuracy at high impedance is possibly not optimal, possibly due to limited range of its precalculated lookup table.


TNT-MMTL is a 2D electromagnetic field solver that solves transmission lines with an arbitrary number of dielectric layers using the Boundary Element Method (a.k.a. Method of Moments) according to the laws of electrostatics.

It was developed by Mayo Clinic’s Special Purpose Processor Development Group (SPPDG) from the 1980s to the 1990s. In the early 2000s, it was released as free software under the GPLv2+ license. Development was then discontinued, and the project fell into obscurity and is mostly forgotten.

Self-promotion: I'm currently working on reviving its code and create a Web calculator based on this field solver. I already successfully made its Boundary Element Method to run inside the Web browser.

It's possible to model trapezoid traces and different dielectric layers, but since it's based on creating geometry shapes manually it's too cumbersome. So for now I use rectangular traces.

Input (microstrip):

  1. Substrate 1 Height: 8.28 mils
  2. Substrate 1 Dielectric: 4.4
  3. Coating Above Substrate: 0.6 mils
  4. Coating Above Trace: 1.2 mils
  5. Coating Above Both Edges of the Trace: 1.6 mils high, 1 mil wide
  6. Target impedance: 50 Ω
  7. or trace width: 0.1 mm


  1. 50 Ω: 13.9 mils.
  2. 0.1 mm: 84.46 Ω

Input (grounded coplanar waveguide):

  1. Substrate 1 Height: 8.28 mils
  2. Substrate 1 Dielectric: 4.4
  3. Coating Above Substrate: 0.6 mils
  4. Coating Above Trace: 1.2 mils
  5. Ground Strip Width: 1.27 mm
  6. Ground Strip Separation: 0.1 mm
  7. Trace Width: 0.1 mm
  8. Trace Thickness: 1.6 mils
  9. Coating Above Substrate: 0.6 mils
  10. Coating Above Trace: 1.2 mils
  11. Coating Between Traces: 0.6 mils
  12. Coating Dielectric: 3.8


Input (microstrip):

  1. Substrate 1 Height: 8.28 mils
  2. Substrate 1 Dielectric: 4.4
  3. Trace Thickness: 1.6 mils
  4. Coating Above Substrate: 0.6 mils
  5. Coating Above Trace: 1.2 mils
  6. Coating Dielectric: 3.8
  7. Target Impedance: 50 Ω


  1. Lower width: 0.3539 mm (13.9341 mils)
  2. Upper Width: 0.3285 mm (12.9341 mils)


  1. Lower width: 0.1 mm (3.9370 mils)
  2. Upper width: 0.0746 mm (2.9370 mils)


  1. 83.6 Ω

Input (Coplanar Waveguide):

  1. Substrate 1 Height: 8.28 mils
  2. Substrate 1 Dielectric: 4.4
  3. Lower Trace Width: 0.1 mm (3.9370 mils)
  4. Upper Trace Width: 0.0746 (2.9370 mils)
  5. Lower/Upper Ground Strip Width: 10 mm
  6. Ground Strip Separation: 0.1 mm
  7. Trace Thickness: 1.6 mils
  8. Coating Above Substrate: 0.6 mils
  9. Coating Above Trace: 1.2 mils
  10. Coating Between Traces: 0.6 mils
  11. Coating Dielectric: 3.8


  1. (no-coating) 69.79 Ω
  2. (coated) 61.71 Ω
  • \$\begingroup\$ So basically your answer shows that the JLCPCB impedance calculator results are generally in the same ballpark as the proven field simulators. The JLCPCB results are more reliable than (some of) the simple formula-based approaches. The main takeaway for me: To get to around 80 ohms, I should not pull ground on the signal layer very close to the traces, because that would yield 60 to 70 ohms. \$\endgroup\$ Commented Jun 5, 2023 at 8:34
  • 3
    \$\begingroup\$ FYI: The error message you observed at JLCPCB is most likely confusing. Currently, JLCPCB obviously does server-side impedance calculation, and they seem to calculate the parameters for their whole set of available stackups when you press the "calculate" button, with a separater request per stackup. If you put in parameters that are less sensible (like a target impedance of 150 ohms), you do not get the error message once or twice, but around 8 to 10 times. In this case, 80 ohms seems to be out of range for one or two of the stackups only, but not the standard one. \$\endgroup\$ Commented Jun 5, 2023 at 8:38
  • \$\begingroup\$ Hey, is the tnt-mmtl continuation still an ongoing effort? I couldn't find much about it, aside from the original source code; can't read your blog, your TLS certificate expired :( \$\endgroup\$ Commented Apr 16 at 10:39
  • 1
    \$\begingroup\$ @MarcusMüller The project is put on hold indefinitely because there's an extremely tricky Web GUI problem and I currently do not have the energy or attention required to finish it, as I'm currently working on another project. I may return to this project if the other project is finished. But the kernel is fully functional, you can read more about it at here. My MMTL fork is available at here. \$\endgroup\$ Commented Apr 16 at 17:06
  • \$\begingroup\$ Thanks so much, @比尔盖子!! WOW, this is a lot of work you put in! \$\endgroup\$ Commented Apr 16 at 17:19

Check the higher-up (calculators) page; Chemandy has two other microstrip formulas as well.

The Wadell variant gives 75.25 ohms for the given dimensions, and IPC-2141, 83.93 ohms. It seems likely JLC is using the latter formula. (If JLC is using clientside calculation, you can likely inspect the Javascript and confirm this exactly.)

All of these formulas are approximations of either a much more complicated analytical solution, or an open-form solution (infinite series, insoluble integral, etc.), which is impractical to specify/formulate and compute. The approximation constraints are given, but the corresponding numerical error is not.

The discrepancy isn't a big deal anyway, as manufacturing variation is comparable. That is, even if you order an impedance controlled process, you might only get 10% or so. If these are carrying digital signals, likely you won't notice the difference (BER < 1e-9?), short of actually measuring it with a VNA.

  • \$\begingroup\$ I can reproduce the IPC-2141 result, but I can't reproduce your Wadell redult. I get 83.93 from the Wadell model, which is only insignificantly different from the 84.84 result from the IPC-2141 model, and also in the same ballpark as the JLCPCB calculation, so it seems likely I happened to pick the most unfit model for the setup I tried to calculate. \$\endgroup\$ Commented Jun 4, 2023 at 17:29
  • \$\begingroup\$ Oh, that was a typo. \$\endgroup\$ Commented Jun 4, 2023 at 17:31
  • \$\begingroup\$ Huh? I can confirm that IPC-2141 does yield 84.84, as your answer said originally. I suspect that the Wadell result in your answer is incorrect, though, because I get 83.93 on chemandy.com/calculators/… \$\endgroup\$ Commented Jun 4, 2023 at 17:36
  • \$\begingroup\$ Huh. Well now I don't remember if I saw 84.84 originally, or if I did, how that happened, if I entered values incorrectly, or the result is somehow nondeterministic. Well, in any case, it's in the right ballpark. \$\endgroup\$ Commented Jun 4, 2023 at 18:47

I dont know which calculator is correct or how incorrect each one is respectively. But in general do these things:

  1. Use a Sig/Gnd////Gnd/Sig stackup for more consistent impedance at top and bottom

  2. For high impedance, make the outer dieletric as thick as possible.

  3. Leave enough space between traces and ask the manufacturer for controlled impedance traces. They will trim the trace width based on their experience with their stackup, accordingly.

The last point in particular should give you the most accurate result, but might cost extra. Just ask the Chat support. I found they are rather knowledge about the tech or at least they can inquire with the engineers within a literal minute.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.