Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711 E-mail: Jack.Smith@cliftonlaboratories.com     To search within the Clifton Laboratories site, enter your search term below.   @import url(http://www.google.com/cse/api/overlay.css);

Home
Up
Current Products
Prior Products - no longer available
Documents
Book
Softrock Lite 6.2
Elecraft K2 and K3 Transceivers

 For digital circuits, and particularly PICs, it's hard to beat the classic solderless plug board for circuit simulation.  The photo shows a Basic Micro 2840 Development Board. The board has a small prototyping plugboard area and the parts necessary to support common 28 and 40 pin PICs.  Basic Micro's programmer, the ISP Pro, plugs into the Development Board and allows you to easily enter your new code into the PIC. For 16F series PICs, I use MBasic Professional from Basic Micro, and you can find out more about MBasic in my book, Programming the PIC Microcontroller with MBasic For 18F series PICs, I use the Swordfish programming language and MikroElektronica's PICFlash 2 programmer. The programmer is incompatible with the 2840 Development Board, so I made a simple adapter board that lets me seamlessly program 18F series PICs in the 2840 Development Board. The 10-pin header J503 in the Z90 is for the PICFlash 2 programmer but since it's only necessary if developing firmware, it may be omitted in production versions.

 As convenient as solderless plugboards are, they leave a lot to be desired for RF prototyping and, for that matter, high speed digital logic as well. Fortunately, there's an excellent and simple RF prototyping solution—Manhattan style construction. I don't know who invented Manhattan-style construction, but for a good summary of the technique it's hard to beat K7QO's article at http://www.k7qo.net/manhattan.pdf. K8IQY has also made good use of Manhattan-style construction in its projects as illustrated at his web site http://www.k8iqy.com/index.html. Here's a typical Manhattan-style construction project. The concept is simple—small pads made from bits of printed circuit board are glued to a piece of printed circuit board stock with ethyl cyanoacrylate ("super glue") glue. Components are wired between the pads. This technique is called "Manhattan style" because it's profile resembles the skyline of a city with tall buildings.I punched the round pads in the illustration from a scrap piece of PCB stock, but more recently, I've used random rectangular pieces of PCB stock that I bought from Dan's Small Parts. Here's a selection of broadband RF amplifiers I built up for an article I wrote that was published in 73 Amateur Radio magazine a few years ago. These are all built with Manhattan-style construction. Building on small scraps of PCB can be difficult mechanically, so I made up a small  test fixture, to hold small Manhattan-technique modules. Here's the first version test fixture. The PCB stock is about 2" x 1-3/8" and is held in place with four 4-40 machine screws. The fixture has BNC connectors on each end. As convenient as these small PCB holders go, there's a limit to how much you can jam into a couple of square inches, particularly when you wish to spread parts out for easy measurement and rework. There's no limit on how large a piece of PCB stock you use, of course, but it's inconvenient to have RF connections made via soldered coax or other jury-rigged techniques. I've made a few BNC mounting brackets to solve that problem. 10 BNC breadboard brackets I designed and made. The brackets are tapped for 4-40 screws and threaded 3/8-32 for a single-hole BNC connector. To use them on a larger PCB breadboard, punch two 1/8" diameter holes 0.800" center-to-center and mount the bracket with two 4-40 x 1/4" machine screws. BNC bracket with connector installed on a scrap PCB sliver. Of course, the idea is that the brackets will be used around the edges of a larger PCB sheet, with Manhattan-style component mounting. Bottom view. Manhattan-style construction has one persistent problem—heat from soldering breaks the superglue bond. I've found 5-minute epoxy is a partial solution, but soldering heat eventually separates even epoxy if you work on a pad enough. One alternative developed by the New Jersey QRP Club was the "Islander Pad Cutter," no longer offered for sale, but the documentation is still available at http://www.njqrp.org/islanderpadcutter/index.html. The Islander Pad Cutter is a diamond "core drill" that, when used with a drill press and a light touch will cut an annular ring, leaving an isolated round pad attached to the PCB. Diamond core drills are widely used by stained glass and lapidary hobbyists and, for the size one would want for pads, are not very expensive. For example, http://www.diamond-drill-bit-and-tool.com/Diamond-Drill/MAIN.htm carries suitable drills  in the $14-$16 range.  I've seen new diamond core sets on E-bay for a fraction of that price, so shop around if you decide to purchase one. I made an island pad cutter from a short length of 0.25" diameter "W1 water hardening drill rod" on the lathe in a few minutes. Face one end of the drill rod, and center drill it. Then, with a No. 4 bit, drill approximately 0.5" into the rod. With a center drill, carefully drill the hole to a thin edge.  With a suitable stone, smooth the outside and inside. Heat to orange and quench in water. Stone again if necessary. It is not necessary for bit's cutting edge to have teeth or to be rough. In fact, when I tried filing notches in the cutting edge, I found that 95% of the pads I tried to cut ripped the copper right off the PCB surface. Chuck the rod in your milling machine or drill press and set the depth stop so that the bit will just cut below the copper layer of your PCB. The speed should be set high, 2000 RPM or more. Although I have not tried it, I suspect it will be somewhere between difficult and impossible to consistently cut pads using a hand held drill. Use a drill press and adjust the depth stop so that you don't have to guess. Note—when you use a tool such as the island cutter with a piece of thin stock, THINK SAFETY. You don't want to hold the PCB stock with your fingers and have the cutter dig in, spinning the stock and sending you to the emergency room to have your fingers stitched back together! Use a drill press vise or C clamps or other tooling to hold the PCB in place. DO NOT USE YOUR FINGERS! First attempts at cutting islands. The pads on the left were cut too deep. The ones on the right are at the correct depth. Close up of properly cut pads. I mark the approximate pad location with a felt tip pen before cutting. It's difficult to cut pads after the board is half-built, so plan ahead and cut all pads at once, before starting construction. PCB with pads cut, ready to add components. The four holes are for BNC mounting brackets I designed and made. I tinned the pads before starting construction. Also, it's a good idea to check all pads for shorts to ground before starting your work. Small copper slivers can easily short across the annular gap, but can be removed by running a sharp object, such as a scribe, around the ring gap. The BNC is mounted in one my brackets. Completed test board, three 2-pole monolithic 21.4 MHz crystals, L network matching at either end and trimmer caps for filter shape tweaking. The pad construction resulted in one of the nicest looking prototypes I've ever built.
I purchased three diamond core drills from Lasco Diamond Products. Nominal sizes are:

5 mm or 3/16"
7 mm or 1/4"
8 mm or 5/16"

I found these drills yield approximate pad sizes as follows:

 Drill Description Approximate Copper Pad Island Diameter (inches) 5/16" Core Drill (8mm) 0.30" 1/4" Core Drill (7mm) 0.20" 3/16" Core Drill (5mm) 0.15"

The pad island diameter will vary to some degree depending on how deep you press the drill. You can see this effect in the photo below, with the two pads cut with a 1/4" drill. The pad at the far right has a slightly larger island than the center pad.

Pads cut with 5 mm (3/16"), 7 mm (1/4") and 8 mm (5/16") diamond core drills.

The amount of copper removed with the two 1/4" drills differs slightly with the depth of cut. It's important not to cut too deeply as there's a risk of removing all the copper if the disk is snagged by the drill.

Typical drill. The drill tip is enlarged and has embedded diamond chips.

The drill is hollow, so it cuts an annular ring, leaving an isolated pad.

The diamond drill will grab the copper so a light touch is important in achieving good results. In addition, the PCB should be flat and not warped or it will be difficult to impossible to obtain satisfactory results.

And, as I've said before, please use good safety practices—clamp the PCB to the milling machine or drill press  table and don't hold it with your fingers. The diamond bit will grab the PCB and spin it around without notice, placing your fingers at risk. Don't do it -- use a clamp!

Geoff, GM4ESD, has asked how much capacitance exists between an island pad and ground, assuming the pad is cut on a piece of double-sided PCB stock and that the top and bottom copper planes are tied together, as is normally the case. This question also applies equally to a standard Manhattan-style construction with PCB pads glued to PCB stock.

I'm temporarily out of 0.062" double-sided PCB stock, so I ran tests on thinner material I've recently used. This PCB stock is 0.0299 inches (0.759 mm) thick, measured with a digital micrometer accurate to 0.0001.". My original assumption was that the stock has 1 oz. copper, a very common value, is 0.0014" (0.036 mm) thick. Graham, KE9H, wrote to correct this assumption, saying that it's almost certainly 0.5 oz foil  in its present form, as the additional copper for a 1 oz delivered finish is supplied during the electroplating process (traces and holes) as the raw stock is made into a completed PCB. Hence, instead of 1.4 mils copper thickness, it's 0.7 mils. (0.036 mm and 0.018 mm, respectively.) I've accordingly corrected the calculations on this page. (I also re-measured the PCB stock  thickness at five locations and use the average, 0.0299" in the calculations.)

If so, the fiberglass dielectric thickness is thus 0.0285 inches (0.724 mm) .

The material appears to be FR-4 fiberglass, the dielectric constant of which ranges from 4.2 to 5.2, with 4.7 used as a nominal center value. However, the dielectric constant is a function of humidity, temperature and frequency and a study by Dupont shows measured dielectric constant for FR-4 material as high as 5.82 under conditions of high humidity and elevated temperature, and no values below 4.52.

Measured Dielectric Constant

To determine the board's dielectric constant, I measured the capacitance between the top and bottom foils using a DCM-601 digital capacitance meter.  The board sample is rectangular, with a total area of 2.681 square inches, or 1730 mm2 if I've done my metric conversion correctly, and measured 110.4 pF.

The formula for a parallel plate capacitor is:

$C = \epsilon \frac{A}{d}$

where
C is the capacitance in Farads
ε is the effective permittivity consisting of ε0 x εR
ε0 is the permittivity of free space, 8.85x10-12 F/meter
εR is the relative permittivity (dielectric constant) of the PCB material
A is the plate area in square meters
d is the separation between plates in meters

I've recast this formula into Imperial engineering units as:

Cpf = 222 * A * εR / d
where
A is the area in square inches
d is the separation or PCB fiberglass thickness in mils (0.001" units)
εR is the dielectric constant.

If  εR is 4.7, the "average" value for FR-4 PCB stock, then this equation can be further simplified as:

Cpf = 1041 A/d

For back-of-the-envelope engineering purposes, the capacitance in pF for FR-4 material can be simplified to an easier-to-remember number of 1000 pF * area in square inches / thickness in mils.

Using the exact parallel plate capacitance formula, working backwards, from the measured capacitance and board area, I computed  εR as 5.22 for my PCB stock.  This is at the edge of the nominal 4.2 to 5.2 range, but within the range measured by Dupont under conditions of high humidity. Remember that  εR varies between different board samples, and also changes with temperature, humidity and frequency, as well as the ratio between glass fibers and epoxy in the material.

This calculation assumes the "fringing" capacitance is negligible. Fringing capacitance is from the electrostatic lines of force at the perimeter of the board that are in air. I don't have a good estimate of the fringing capacitance, but expect it to be at most a couple of pF for this size board. If fringing capacitance is considered, then the true parallel plate capacitance is less than measured and hence the computed dielectric constant will correspondingly decrease. (I am aware of free finite-element electrostatic simulators available on the Internet, and will leave it to the interested reader to duplicate the  test setup with such simulation software. The simulation software will consider and compute the fringing capacitance.)

We can easily calculate the capacitance of an island pad, using the equations above. In this case, with εR =  5.22, and with a board thickness of 0.0285" we calculate C = 41.18 pF / in2.  (This number could, of course, be more simply derived by dividing the measured sample capacitance, 110.4 pF, by the sample area, 2.681 square inches.)

A single pad, cut with the 8 mm (5/16") diamond core drill leaves a pad diameter of about 0.29." The exact pad diameter varies with how deep the drill is pushed and if the PCB is not perfectly flat, the annular insulating area will not be of uniform width.

For a 0.29" diameter circular pad, we compute the capacitance as 2.72 pF. To this value should be added an allowance for fringing capacitance, i.e., the capacitance from the pad to the top surface foil across the annular insulating area.  Again, I don't have a good estimate for this fringing capacitance, but there is reason to believe it's rather small, as will be seen shortly.

I've measured the capacitance of a 0.29" diameter pad two ways. First, using eight pads in parallel, and second, a single pad.

Eight 0.29" diameter pad arrangement. The pads are connected together with a cage arrangement. The idea is that the cage to foil capacitance will be small compared with the eight pads. This assumption may or may not be the case, of course.

The eight-pad arrangement measures 24.4 pF total, or 3.05 pF per pad. This is 0.33 pF above the calculated value. This discrepancy can be attributed to a combination of errors in actual pad diameter, stray capacitance of the cage to the foil and the individual pad fringing capacitance.

To reduce the effect of the cage, I tried a different approach, as illustrated in the photo below. I cut three pads into the PCB as seen in the photo. I mounted the board on a Boonton RX Meter, model 250A, set for 1 MHz measurement. I moved the wire until is was just a hair short of contacting the pad and nulled the 250A RX meter. The null adjustment therefore considered and cancelled the capacitance of the wire contact to the PCB and the instrument. I then moved the wire contact a tiny amount to contact the pad and adjusted the 250A's X dial to restore bridge balance. I measured 2.80 pF. The rated accuracy at this frequency and dial reading is ±0.165 pF. The RX 250A meter is calibrated in 0.01 pF increments and easily meets this accuracy.

Test circuit for Boonton RX Meter 250A. By nulling the bridge with the contact wire just above the pad, stray wire capacitance is eliminated from the measurement.
Comparing the RX 250A meter data with predicted we find:

Calculated: 2.72 pF
Measured: 2.80 pF ±0.16 pF.

This is excellent agreement and suggests that the fringing capacitance for this size pad and board thickness is small, at most being 0.24 pF, and a mean value of 0.08 pF. And, my pad diameter measurement is only an estimate, made with a digital caliper held against the pad. A small error in pad diameter will have a 2x larger effect in capacitance error, since capacitance is proportional to the area (diameter or radius squared).

The remaining question is "is 2.8 pF capacitance too much?"  There's obviously no single answer to that question, as it depends on the point in the circuit and the frequency. Moreover, if 2.8 pF shunt capacitance is too large, there are several strategies to mitigate it:

1. Use a smaller pad. If cut with a 5 mm (3/16") core drill, the pad diameter is 0.15" and hence the shunt capacitance to ground will be on the order of 0.7 pF.

2. Use 0.062" thick PCB stock. In fact, I prefer the 0.062" board, but happen to have used all I had on hand at the moment. 0.062" PCB stock will have only about 40% of capacitance I measured with the 0.0299" material. A 0.15" diameter pad on 0.062" stock will thus have about 0.28 pF shunt capacitance. In this case, fringing capacitance may become an important factor and should be considered.

3. If your circuit otherwise permits, use single sided PCB stock. This will reduce the capacitance to nearly zero, but may or may not cause other problems.

4. Mount the sensitive portions of the circuit in air, spaced above the PCB. In some cases, high value resistors may be used as improvised stand-offs. Or, the purpose-made Teflon insulated terminal pins, such as those from Keystone Electronics, 11065 - 11074 series, with pin heights up to 1." The Teflon base press-fits into a hole in the PCB and the pin provides an insulated structure, above the PCB base. Or, one might glue an inverted DIP socket to the PCB and use the pin leads. (I've seen this done with octal sockets years ago.)

I should also add that using conventional Manhattan-style construction has exactly the same issue—there is shunt capacitance from the pad mesa to the PCB surface, which will be similar to the shunt capacitance of an island pad arrangement with the same dimensions and thickness.

One problem with either island pads or glue-on Manhattan pads is capacitance to the ground plane. For many circuits, the small additional capacitance is not a problem. But, it can be for some circuits. If you are building an oscillator, even if 1 or 2 pF of pad capacitance isn't a problem by itself, changes in the pad capacitance can result in frequency instability. Standard FR-4/G10 PCB material is not known for high dielectric constant stability, after all, as it changes with both temperature and humidity.

One alternative mentioned in the the text above this topic, is to build sensitive portions of the circuit on Teflon stand-off pins. I purchased a couple hundred pins from a surplus dealer on E-bay recently and recently used them for the critical parts of a crystal oscillator circuit.

Teflon push-in pin. The pin diameter is 0.142" and fits nicely into a hole made with a #27 drill.

As suggested by Geoff, GM4ESD, I made an installation tool. It's a short piece of brass rod, 3/16" diameter with a hole drilled in one end to fit the metal pin. The handle is a short length of 1/2" diameter Delrin (acetal) rod, with a tight press fit onto the brass rod.

Home made installation tool, with a pin.

Inserting a pin using the tool. The PCB is a piece of scrap used for mechanical experiments.

Completed prototype board, oscillator, buffer and power amplifier (+15 dBm output). The power amplifier is built with glue-on Manhattan pads, whilst the frequency sensitive oscillator and buffer are on Teflon pins.

Close up of pins around the oscillator stage.

Z90 Prototype

My Z90 started out with a combination of plugboard and Manhattan-style construction. The microcontroller and interface to the graphics LCD were constructed on a solderless plugboard, and the rest of the circuit on a series of Manhattan-style modules. Each module was first developed and tested with a test fixture like the one shown above, or one of a similar design I later built.

Here are two early design crystal filters in test fixtures.

Here's the first Z90 breadboard. The module in the lower left is an SA612 mixer (abandoned due to inadequate dynamic range and strong signal performance). The module at the lower right is the log amplifier and the two crystal filters and relay switching are at the top. Each module was developed and tested to work independently.

Rather than go to one large PCB at the outset, I designed and had made small prototype PCBs for selected modules. Here's the crystal filter test PCB. The test PCB has variable capacitors that I tried as an experiment to determine whether or not 5% tolerance standard capacitors would work or whether trimmers would be necessary. It turns out that trimmers are not necessary, and standard capacitor values with 5% tolerance are adequate.

Here's a "two-fer" where I combined the AD9851 DDS prototype PCB layout with the Gali-74 amplifier layout. The only common element between the two modules is a common +12V connection.

This may seem like a lot of extra work. However, by verifying that each module worked independently, and that the interface impedance and voltages were correct, when I designed the master PCB, everything worked perfectly first time, except for a couple of connection traces that I missed.

I've subsequently made changes in that original design, and PCB layout, to make it easier to build (for example, moving the DDS to a small daughter board) and to improve performance. But, I could make those changes with confidence that each module was a functioning unit with stable interfaces.

 I recently designed and made a slightly different style fixture. The concept is similar to my earlier designs, except the holder is sized to fit one half of a Radio Shack DIP Prototype Board, PN 276-159. (Available at most RS retail outlets or via Radio Shack's web page.) The current price is $2.29, so each board half is$1.15. The fixture can, of course, be used with a square of PCB stock, just like the earlier fixtures. But, sizing it to fit the 276-159 board makes it easy to prototype modules with DIP integrated circuits. Not so fast, you're probably thinking--the 276-159 board is upside down.Yep, that's right--install it upside down. Then take a socket that fits the ICs you wish to install and splay the pins outward. Solder them onto the copper traces just as if the socket were a surface mount component, as in the photograph. Here's how it works in a larger arrangement. This particular module is the mixer and VCO test unit from my analog pandapter. The 276-159 board has sockets for two DIP-8 chips, an NE612 mixer/oscillator and a 7555 CMOS version of the 555 timer to generate the sweep.The 276-159 board is mounted on larger scrap PCB piece where more Manhattan-style construction can be seen. Here's the finished analog panadapter. (The CRT is removed from the unit for the photograph.) It's all built Manhattan-style, with extensive use of the 276-159 boards. All analog design, and with 1000 Hz and 200 Hz crystal Gaussian filters.)My design is centered around a surplus Regco 21.4 MHz panadapter made for use with intercept and surveillance receivers. The IF frequency and resolution bandwidth (several hundred KHz) made it impracticable to modify for HF ham use, so I saved the power supply and CRT module and built the rest of the panadapter around those modules. Here's the completed analog panadapter on top of the receiver it normally is used with, a Watkins-Johnson HF-1000. The HF-1000 has a 455 KHz IF output port, with a total available bandwidth of about 20 KHz, as the 1st IF roofing filter is the limiting factor. Here's an AM broadcast signal displayed on the analog panadapter. In many respects, the analog scope display is more pleasing to my eyes than the LCD or computer display of the Z90/Z91. If I could have found a reasonably priced source of small CRTs, I might have designed the Z90 to use an analog CRT display, keeping the rest of the circuitry digital. Alas, small long persistence CRTs are exceedingly difficult to find and are far from inexpensive when found.