Fullscreen HTML5Med Res (??MB)HTML4

Digital waveform generation using a computer, Pt.4 This month, we have produced a PC board for an upgraded D-A converter & filter circuit. We also have a 5-¼in floppy disc full of upgraded software so that you can use your computer as a function generator or tone burst generator. By STEVE PAYOR This series started out as a simple example of computer interfacing; a practical demonstration of how a little bit of electronics could enhance your PC, and vice-versa. On the way, we created a log/linear frequency sweep generator, a tone burst generator, and a sine/triangle/square wave function generator. That's not bad value, really, for a handful of resistors, a 64 SILICON CHIP 36-pin socket and some optional filtering components. PCB and floppy disc In this final article, we will put it all together with the introduction of a PC board (to make the filter easy to build) and a floppy disc with fully executable programs. But first, let's recap the story so far : The February 1990 issue, pages 68-71, described the parallel printer port and showed how to access the various bits directly. We also showed how a binary-weighted resistor network could turn the 8 data bits into an analog voltage. In the March 1990 issue, pages 80-84, we showed how a low-pass filter could remove the "steps" from the digitally generated waveform. On page 83, we gave a listing of a log/linear frequency sweep program, which proved very useful for adjusting the "flatness" of the filter passband response. The frequency sweep program is written in TURBO BASIC and has data rate of 20k samples/sec. When combined with the simple third-order LC filter , it can generate high quali- a HXACT VALUES ALL RESISTORS 1% OUTPUT S1b--------, S1a * 2. 769k : FIL TEA .._ LOAD 68[1 2.7k J. ---, (9) 07 * 5.583k 5.6k 50pF. 3.3k (8) 06 -!- C1 * 11.21k 5.6k 5.6k (7) 05 22k OUTPUT tRIM VR1 2.2k <5kf .,. 470!1 (6) 04 FILTER TYPE 180k (4) 02 3RD ORDER CHEBYSHEV : -0.1d8 AT 4.4kHz I 5TH ORDER CHEBYSHEV -0.1dB AT 36kHz C1b L2 C3 .0047 .01 38.3mH .033 .001 .001326 5.943mH .006308 7.618mH .005537 + 50pF LOAD 100k SAMPLES/ SECOND .001 297pF 4.415mH .005782 8.335mH .006424 + 50pF LOAD 100k SAMPLES/ SECOND L4 C5 SUITABLE DATA RATE C1a 20k SAMPLES/SECOND 180k 180k (3) 01 (16,19) TO (30,33) GND 5TH ORDER BUTTERWORTH -0.1dB AT 22kHz o---~Mr--- 7 (1) - S T R O B E 0 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0 CRO TRIGGER CENTRONICS PIN NUMBERS ty waveforms up to 4.5kHz. In the May 1990 issue, pages 92-97, we presented two more programming examples. The program on page 93 is for generating IHF standard audio tone bursts. It is also written in TURBO BASIC and has the same data rate and frequency limitation as the frequency sweep program. The program on pages 94-96 is a wide range (20kHz plus) sine/triangle/square wave function generator. The program itself is in GWBASIC but it uses a high speed subroutine, written in 8088 machine code, to achieve a data rate of 100k samples/sec. OUTPUT D-A CONVERTER/FILTER Fig.1: this D-A converter & filter circuit plugs into the Centronics port of your computer. Note that you have a choice of filters: 3rd or 5th order Chebyshev, or 5th order Butterworth (see text). In most cases, the 3rd order Chebyshev filter will be adequate (see text). programs which can be run directly from DOS. The executable versions of these programs have been considerably enlarged, to make them more "user friendly". For example, they will automatically locate the 1/0 ad- dress of your parallel printer port. The data rate compromise You may well ask, why not use the highest possible data rate in all the software? Well, for short, repetitive waveforms, like sine, No compiler needed All these programs [and more) are on the floppy disc. The source listings of these programs are exactly as published. However, the source listings of the TURBO BASIC programs are mainly for information only. Since most readers will not have a copy of the TURBO BASIC compiler, we are including compiled and executable versions of these The 5th order Chebyshev filter is suitable for the function generator program but not for the frequency sweep & tone burst programs. The 36-pin Centronics socket is soldered directly to the PC board. ]ULY 1990 65 data rate, the filter components can be calculated to suit. Filter choices The 3rd order filter is easier to build than the 5th order types. Note that the Centronics socket fitted to this board is the "solder bucket" type which has a pin spacing of .085-inch. triangle and square waves, the highest data rate of took samples/ sec is fine. However, for the long waveforms produced by the frequency sweep and tone burst programs, the number of bytes required becomes excessive. For example, a 0.5s tone burst would require 50,000 bytes at a data rate of 100k bytes/sec. Not only is this getting close to the maximum allowable array size (64k bytes) but you would also have to wait an awfully long time for the program to set up this waveform. A data rate of 20k samples/sec is much more realistic for this task. Execution speed If you have an 8087 mathematics co-processor, the compiled programs will use it, resulting in a considerable saving in the time re- quired to set up the waveform. For example, setting up a 0.5s tone burst takes 4 seconds with the coprocessor and 41 seconds without it. Note: the data rate during waveform output remains unaffected by the presence of the coprocessor. What will affect the data rate is the speed of your CPU and/or 1/0 card. The programs allow you to enter sample times other than 50µs. So, for example, if you have a toMHz Turbo XT, you may find that a tkHz waveform comes out as 2. tkHz. To correct this, try entering a sample time of 23.85µs at the start of the program. The best way to determine the actual sample time on your machine is by experiment, so the first thing to do is get the D-A converter working. Once you have determined the Where to buy the PC board & software An etched PC board plus a 360K 5 ¼-inch floppy disc of the software are available from M. Radvanyi, PO Box 49, Kingswood, NSW 2750. The price is $39.95 (includes instructions on running the programs plus packaging & postage) Note: does not include on-board components. Please make payment by cheque (to M. Radvanyi) or by mail order. The floppy disc includes three separate executable programs which can be run from DOS: frequency sweep, tone burst generator and function generator (sine, triangle & square waves). Also included are some GWBASIC programs for those who wish to create their own waveforms. The software is easy to use and all frequencies and functions for each program are simply entered into the opening menu. Note: copyright of the PC pattern and the software is retained by the author. 66 SILICON CHIP The low-pass filter is only required if you wish to remove any visible traces of the D-A steps from the waveforms. With the machine code software, you certainly won't hear any difference because the lowest alias frequency will be 50kHz. So, if you don't have a CRO, you can leave the filter section of the PCB empty, and replace the switch with a wire link. Capacitor CtA can still be fitted, if you wish, to absorb the D-A glitches. You have a choice of filters from 1st order to 5th order in fact. We have listed the component values for suitable 3rd and 5th order filters on the circuit diagram. In most cases, the 3rd order Chebyshev filter will be adequate. This only requires one inductor and is easy to align (see the CRO photo on page 81, March 1990}. Unless you are an experienced constructor, we suggest you stick to the 3rd order filter, even for the higher data rate. (In this case, simply get out your calculator and divide the values for Cl, 12 and C3 by 5). Don't forget to fit a link in place of 14. The main advantage of the 5th order filters is that their steeper cutoff allows a passband which extends closer to the theoretical limit of half the sampling frequency. For example, with a took samples/sec data rate, the 5th order Chebyshev filter allows a frequency response up to 36kHz ( - 0.tdB). However, the very steep cutoff results in some ringing on square and sawtooth waveforms. The 5th order Butterworth (maximally flat) filter will give slightly cleaner waveforms. However, the - 0.tdB bandwidth is reduced to only 22kHz. Its principal advantage is that it can be adjusted to be dead flat up to 20kHz. In practice however, such precise alignment is rather difficult. Changing the parameters You can scale the 1 & C values, to take into account different source impedances or to move the cutoff frequency. PARTS LIST 36 . WAVEFORM OUTPUT GNO GND 19 CRO TRIGGER OUTPUT 36-PIN CENTRONICS SOCKET Fig.2: parts layout for the 3rd order filter. C3 consists of three capacitors connected in parallel to give the required value. The circuit values shown are for a data rate of 20k samples/sec but the filter can also be built for the higher data rate by dividing the L & C values by 5. 1 PC board, code PC-Driven Waveform Gen. 1 5.25-inch 360K floppy disc, "Silicon Chip Digital Waveform Generation Software" 1 36-way IEEE-488 type Centronics socket 1 2k0 or 1 OkO miniature horizontal trimpot (see text) 4 PC stakes Resistors (0.25W, 1 %) 5 1 1 1 3 1 1 1 1 180k0 27k0 22k0 18k0 5.6k0 (one selected for 55830 ± 1 00, two selected for 11,2100 ±500) 3.3k0 2.7k0 (selected for 2,7010 ±50) 4700 680 Optional .components for low-pass filter CRO TRIGGER OUTPUT PINS 1 TO 18 SOLDERED TO COPPER SIDE OF PCB 36 PIN CENTRONICS SOCKET Fig.3: parts layout for the 5th order filters. The winding details for the inductors are given in an accompanying panel. You wHI need a CRO to accurately align the filters (by adjusting the inductors) to give a flat response across the passband. For example, to change the 3rd order filter to accommodate a 1.200k0 source impedance, multiply all the L values by 1.200/1.083 and divide all the C values by this same amount. This gives ClA + ClB = 13.lnF, 12 = 42.44mH and C3 = 29.8nF. Now let's say you have a compiled "C" program which has a data rate of 60k samples/sec. Dividing the above L & C values by 3 will make the filter suitable for this data rate. Assembling the PCB The PCB is designed to be soldered directly to a 36-pin "Centronics'' socket. There are two types of socket available. The "solder bucket" type has a pin-topin spacing of .085-inch, while the JDC type has a 0.1-inch pin spacing. If you have the latter type of socket, cut the finer connector pattern off the end of the PCB. Solder pins 1-18 directly to the PCB pads and connect the signal ground pins (19-30 & 33) using three tinned wire links on the component side of the board, as shown in the wiring diagram (and photograph). Sorting the resistors Before fitting the resistors, it is desirable to check them with a digital multimeter. The most important ones are those at the top of the ladder - in particular, the resistor for the MSB should be within 0.2% for an accuracy of 1/2 LSB. Therefore, try to get the resistors for bit D7 to add up to 2.769k0. Of the three 5.6k0 resistors, put the one closest to 5.583k0 in the D6 position and try to get a pair which add up to 11.2 lkO in the D5 position. Normal 1 % accuracy will do for all the lower bit positions. By the way, the resistor values 1 DPDT slide switch 1 or 2 RM 1 0 ferrite core assemblies (each consisting of two matched core halves, inductance adjuster, two mounting clips and coil former.) AL values to suit required inductance (see text) length of 0.25mm, 0.315mm or 0.4mm grade 2 enamelled copper wire (see text) Miscellaneous Assorted polystyrene, silver mica, or high-stability foil/film capacitors for C1 a, C1 b, C3 & CS (see Fig.1 ). have been optimised for standard TTL output source and sink impedances. If your printer port is driven directly from a custom LSI chip, some modification of the resistor network may be necessary. The most likely scenario is insufficient drive to the high-order bits. If you have access to a CRO, run the 0-255 test ramp program and you will probably see small "backward" steps 1/4 and 3/4 of the way up the ramp and a larger backward step at the half-way JULY 1990 67 Pins 19-30 of the Centronics socket are all joined together using a tinned wire link. These pins, along with pin 33, are then connected to earth on the PC board. This is the opening menu for the tone burst generator program. The default parameters give the IHF standard but you can change any parameter simply by typing in a new value and hitting RETURN. point. Take a large value resistor (eg. 330kn) and temporarily shunt the resistor for the second-mostsignificant bit (D6). Note the effect and keep trying different shunt resistors until the waveform is perfectly straight at the 1/4 and 3/4 points. Below: the 5th order filter also uses parallel connected capacitors to give the required values. Note how the pot cores are mounted. 68 SILICON CHIP Now trim the MSB (D7) until the gap at the half-way point disappears, and that's it. The secret is to start trimming from the lower bits up. Output level trimming The D-A output can be trimmed to lOmV per step with the 3.3kQ resistor and the Zkn trimpot. This gives an output voltage range of 0.1 V to 2.65V for a digital input of 0 to 255. For audio applications. you may wish to fit a lOkQ trimpot. which will allow you to trim the output to 2.828V peak to-peak (ie. exactly 1V RMS for sine waves). The output trimming circuit may seem a little unusual but it performs a vital role in maintaining the linearity of the D-A converter. The heavy DC load to ground helps to stabilise the TTL "high'· level output voltages. The output impedance is approximately l.083kQ when trimmed for 2.55V p-p and l.200kQ when trimmed for 2.8V p-p. (You will need this information later when calculating the filter components). At this stage. the D-A converter is complete and you can test it using a digital voltmeter and the DC How To Wind The Inductors The inductors are wound on PCB-mounting RM 1 0 ferrite core assemblies. Use AL 400 cores for inductors greater than 1 0mH and AL 250 cores for inductors less than 10mH. If you are using a different core type, you will need to know its AL value in order to calculate the number of turns required . The calculation is quite l:limple: (1 ) . Express the desired inductance in nH; eg, 38.3mH = 38,300,000nH (2). Divide this by the AL value of the core; eg, 38 ,300,000 -+- 400 = 95,750 . (3) . Take the square root; eg, (93,250)½ = 309.4 = the required number of turns (309 and a half will be near enough) . If the core has an adjuster, find out how much it can increase the inductance and subtract half this amount before doing the above calculation. For example, the AL 400 core has an adjustment range of 0 to +20% while the AL 250 voltage software, or a pair of headphones and the signal generation software. Note that the D-A network will produce a current output (eg. 0 to 2.55mA) when used with a short circuit load, such as a pair of headphones. However, with the filter in circuit, the load should be higher than 2k0 to maintain a flat frequency response. Winding the inductors Full instructions for calculating the number of turns are given in the accompanying panel. Note that RM10 cores are recommended. For preference, use A1 400 cores for inductors greater than 10mH, and A1 250 .:ores for inductors less than 10mH. The coil formers for both cores will accommodate up to 400 turns of 0.25mm grade 2 enamelled copper wire, 250 turns of 0.315mm wire, or 180 turns of 0.4mm wire. The A1 400 core has an adjustment range of Oto + 20%, while the A1 250 core has an adjustment core has an adjustment range of 0 to + 1 7 % . Thus, when calculating the number of turns, reduce the desired inductance by 10% and 8.5% respectively. This will put you in the middle of the adjustment range . Thus, to obtain 38.3mH on an AL 400 core, we have to calculate the number of turns for an inductance of 38.3mH less 10% = 34.47mH. This gives 293.5 turns. We used 0.25mm enamelled copper wire, for which there is ample room within the core window. Alternatively, if you don't want to bother with aligning the filter, leave out the adjusters and the inductance will be accurate to within ±2% as is, provided you count the turns correctly. Suitable cores, formers, mounting clips and adjusters are available in boxed sets from Radiospares Components for $6 .83 each, plus sales tax. The AL 400 set is stock number 228-258 and the AL 2 50 set is stock number 228-242. range of O to + 17%. So, when calculating the number of turns, reduce the desired inductance by 10% and 8.5% respectively. This will put you in the middle of the adjustment range. Alternatively, if you don't want to bother with aligning the filter , leave out the adjusters and the inductance will be accurate to within ± 2% as is, provided you count the turns correctly. On the subject of accuracy, it is well worth selecting the exact capacitor values with the aid of a digital capacitance meter, especially if you intend to use fixed inductors. With adjustable inductors, the filter "flatness" can easily be restored, even if the capacitors are 5% off. Note that the PCB will accommodate up to three capacitors in each location, so that standard values can be combined to give the exact value required. For best results use polystyrene, silver mica or similar low temperature coefficient capacitors. ~ Problems? ... and you don't have our NEW 1990/91 148 page electronic parts and accessories catalogue ... Its our latest TRADE catalogue for the consumer ARISTA ... Your one-stop problem solver. ... Stylus ... ... Plugs, Jacks and Sockets... .. .Batteries .. . ...Cable .. . ...Tools and Technical Aids .. . ...Plug and Power Packs .. . ...Car/Auto Accessories .. . ... Boxed Hi Fi Speakers .. . ... Raw Replacement Speakers ... ...Speaker Accessories ... ... Telephones and Intercom ... ...Public Address Accesso ries .. . ... Security and Alarm Accessories .. . ... TV/Video/ Antenna Accesories .. . ... Videocam Accessories .. . ...Audio Accessories ... ... Headphones... ... Computer Accessories .. . ...Microphone Accessories .. . ...Mixers, Amplifiers, Equalizers ... Just about anything you want ... Get your catalogue complete with "Recommended Retail Prices" free from your local ARISTA dealer or send $2.50 P & H and your return address to: ARIST~ ELECTRONICS PTY LTD PO BOX 191, LIDCOMBE, NSW, 2141 JULY 1990 69