'EMCFiltr' EMC Filter simulation Software written by C. Bateman. see 'about the software section' or 'E-Mail' cyrilb@ibm.net Installation. ------------- Due to the restrictions imposed by Font sizes etc. two versions of this software are available, optimised for either VGA 640*480 screen or SVGA 800*600 screen, each in a separate directory on the disc. While both versions will install and run on either Windows screen resolution, some on screen features will be incorrectly displayed. The SVGA version is also suitable for higher screen resolutions but the presentation will not be optimal. If you have purchased the 'Demo' disc then all files required have been included and are uncompressed. If you have downloaded the 'Demo' then for economy of download, If VBRUN300.DLL is not already in your Windows/system directory, this file must downloaded separately and copied to Windows/system BEFORE commencing with the 'EMCFiltr' installation. To minimize download times, since the VBRUN300.DLL file is quite large, most ftp sites request this be excluded from the download and if not already on your harddisc, it be obtained separately. It is readily available from most MS Windows archives. The current version of VBRUN300.DLL used to write this software was 3.00.0538 12/5/93 extracted size 398,416 bytes. This version or newer is required. Before commencing with a VGA installation, if this software has been downloaded, it is necessary to copy certain files from the SVGA directory into the VGA directory:- DDEML.DL_ SETUP.EXE SETUPKIT.DL_ VER.DL_ To reduce download times, since these files are common to both SVGA and VGA installations, only one set is included. If you have the normal distribution disc, they have already been duplicated so no action is needed. To install follow the usual Windows practise of File Run 'A:\SVGA\SETUP' or File Run 'A:\VGA\SETUP' as appropriate. When installation is complete the program body 'EMCFiltr.exe' will have been installed in the directory named EMCFiltr and Setup1.exe in Windows. If not already present, three files needed to run a compiled Visual Basic program will have been added to your Windows/System directory. VBRUN300.DLL DDEML.DLL VER.DLL You need take no action, the file Setup.exe checks for their existence and actions accordingly (see setup.lst on the disc). This install makes no alterations to your existing 'Ini' files or autoexec.bat or config.sys hence can be easily uninstalled by simply deleting the program group while running Windows, then deleting the program its directory and Setup1.exe. The three files in Windows/System are standard Microsoft files required by many other programs and can be left in place should you uninstall the main software.. VBRUN300.DLL DDEML.DLL VER.DLL Having completed the installation, double 'click' on the EMCFiltr Icon to launch the program. The full program provides analysis of nine pre-set Filter configurations. The Demo version is restricted to analysis only of the Two-Tee style, however all other screens are still available and can be actioned for value change etc., but will not proceed to the analysis stage. Familiarisation. ---------------- The quickest way to check program operation is to repeatedly push Enter which will cycle the program through each default screen in turn, to demonstrate program action. Having ascertained the program is functioning at all four screens, try choosing a filter style - say Two-Tee, and run the default values, its not a very good Filter. Try setting silly component values e.g. 0.0 etc. To avoid 'divide by zero' errors the program automatically applies a safe minimum value. This minimum for the Demo has been increased, the full program however works down to extremely small values. Let us now try redesigning this Two-Tee Filter to a target specification. Lets assume as a target, good transfer of audio to 12 kHz, with interference rejection better than 40 dB at 40 kHz and better than 70 dB at 100 kHz and above. Exit the program then re-run to restore the original defaults. Run the default Two-Tee values, its an awful Filter in every respect.- Terrible Return Loss - Inadequate Pass Band Transmission and poor EMC attenuation. 1) Try halving the capacitors to 0.5 uF Looks better 2) Make all inductors 100 uH. Better again. 3) Increase inductors to 500, 1000 , 500 uH Very much better but Return Loss still poor. 4) Reduce two outboard inductors to 300 uH Not quite there. 5) Try reduced capacitance say 0.33 uF. Almost to spec. 6) Try 330 uH, 920 uH, 330 uH Looks good. 7) Make DC resistances attainable. Try 0.5 ohm, 1.0 ohm, 0.5 ohms Still good. 8) Reduce inductor Self Capacitances to 2.5 pF Specification surpassed. The above few steps give a good indication of the affect on Return Loss and Insertion Loss of various component changes. Dependant on the transmission requirements further fine-tuning may be needed to improve the Group Delay response. The default capacitors in the Demo version are X7R Ceramic Surface Mount chips, these typically have 0.5 to 0.8 nH of Self Inductance, depending on their physical size. Other styles have differing Self Inductances, with wire leads the inductance will be much higher, for the so called 'Discoidal' types the inductance will be almost zero. The default inductors in the Demo version assume use of medium u ferrite toroids, these provide usable inductance value with little Self Capacitance, although this capacitance is influenced by winding techniques, hence a value must be input. Since the d.c resistance of inductors depends on wire gauge and winding methods, it cannot be pre-determined and is a required input. The software calculates the frequency dependant AC loss resistance which is then added by frequency to the d.c. resistance value. The full program allows the Capacitor and Inductor base material to be chosen for each simulation, since this determines the Frequency Dependant variables needed for realistic simulation. The full software package includes a manual describing Inductor winding techniques together with typical values to be expected, also covers capacitor choice in detail. However mostly this software will be used to simulate an existing Filter when used with source and load impedances other than 50 ohms. Most Filter specifications state the capacitance values used or alternately the total capacitance for Pi (two capacitors) and larger Filter styles, and generally will be stated as a Minimum value, but Inductor information is less common. A telephone call to the makers Technical Support will often fill in the details. If not available, all is not lost. Simply 'reverse engineer' the Filter by inserting the stated capacitances perhaps at +10%, then guess at an inductance value and analyse the Filter, compare results with stated claims. Generally after five or six attempts the Filter will have been reverse engineered with sufficient accuracy. You can then use these derived values to analyse the Filter's behaviour with any desired source and load impedance combination. Remember their catalogue claims will be chosen such that no Filter fails insertion loss testing. It is also common practise to considerably underclaim at the higher frequencies to avoid test failures, since small construction variances have most affect as frequency increases. Significance of Return Loss. ---------------------------- In order to pass low frequencies with little attenuation, the inductors used must have a low DC Resistance, hence unlike an attenuator the Filter cannot attenuate EMC by dissipating energy. Since energy cannot be destroyed the only usable mechanism is to 'return' (Reflect) the unwanted EMC to its source. This is exactly the mechanism by which Low Pass EMC Filters work. Depending on the input impedance of the filter the wanted signals, which should pass through the Filter, can also be reflected. Automatically resulting in loss of wanted signal. The perfectly matched Filter would lose by Reflection less than 1% of the wanted signal, expressed in dB this equates to a Return Loss signal -40dB lower than the incident wanted signal. A Return Loss of -26dB equates to a VSWR of 1.1:1 and would be perfectly acceptable in practise, reflecting 5% of the signal. Return Losses of -20dB and less are not desirable. Obviously Return Loss results from the value of source and load impedances used as well as the Filter's Input impedance, thus making this an important parameter when designing for EMC suppression with good transmission of a wanted signal. With DC Power supplies, Return Loss traditionally has been ignored, resulting in the oft quoted advice to design for maximum mismatch between source and Filter. This is extremely bad practise whenever a wanted signal must be transmitted through the Filter. About the Software. ------------------- Simulators come in three basic forms:- Those based on measured 'S Parameters' and generally used for Microwave design only, since the software is expensive. Measurement of 'S Parameters' for capacitors and inductors is time consuming, also extremely expensive 'characterised' jigging is essential, otherwise the measurement is valueless. While most RF rated semiconductors have 'S Parameter' information available, this is almost non-existent for capacitors and inductors, especially those not specifically intended for RF use. Lower cost 'Frequency Domain' simulators which calculate node by node the 'Kirchoff' solutions, are available but these assume 'idealised' capacitor and inductor models. While it is obviously possible to apply a series loss resistor, since with real components this resistance is frequency dependant, this would only be applicable to a restricted range of frequencies. The most common circuit simulation software derives from the 'Spice' system detailed in Lawrence Nagel's doctoral thesis. This was funded and targeted expressly towards the simulation of Integrated Circuits having amplitude dependant non-linearities, thus the basis of 'Spice' is modelling in the 'Time' domain and not 'Frequency' domain. While it is true that 'Spice' offers 'Small Signal Simulation' by Frequency this does not permit the use of measured frequency dependant variables. The 'EMCFiltr' software models Insertion Loss strictly in accordance with the US Specification MIL-STD-220A, which is recognised world wide. Real world capacitors and inductors, when carefully measured, each comprise three elements, all of which are strongly frequency dependant. Capacitors translate to a 'series' combination of capacitance, inductance and resistance, with capacitance being dominant at the lower frequencies, i.e. below self resonance. Inductors similarly equate to a 'parallel' combination of capacitance and inductance both being in 'series' with resistance, with inductance being dominant at frequencies below self resonance. For both, the resistive part comprises a fixed resistance value in series with a frequency dependant variable resistance. For inductors this fixed resistance is the dc resistance of the winding (DCR), the variable part results from the various frequency dependant losses which cause the 'Q' value to change with frequency. For capacitors the fixed part is the irreducible resistance measured at high frequency (TSR), largely comprising metallic connection resistances. The variable resistance derives from AC dielectric losses, described as tan d, being frequency dependant. At lower frequencies this can amount to many ohms, dependant on capacitance value. Any simulation based on 'ideal' components will differ significantly from practical measured results of working Filters. The maximum insertion loss from simulation will far exceed measured results. At frequencies below Filter resonance the insertion loss will be overstated while above resonance it will be understated. These differences can accrue to many dB's difference. The EMCFiltr software package is a node by node solution by frequency of Kirchoff, but considerably enhanced by using true three component models for each part with the frequency dependency variations required to simulate real world parts. This frequency dependency is derived from manufacturers published characteristics and stored as data in the software. Hence assuring realistic simulation results. The 'standard' version software is targeted to Filter users and provides a choice of base capacitor and inductor materials commonly used in the production of Filters. The 'professional' edition is targetted towards those intending to manufacture Filters, and while based on the standard version's simulator engine, this version provides the facility to 'tailor' the frequency dependant material models to your choice of materials. The high dielectric constants used with some ceramic capacitors and their physical size, can result in narrow band resonance modes, these also can be modelled given the precise physical descriptions of the parts. This is incompatible with the objectives of simplest possible no-learning curve approach of this 'standard' version hence is available only in the 'enterprise' edition software currently under development. The full version of the software, also any technical support you might need, is available from:- C.Bateman Engineering. 'Nimrod' New Road. ACLE. Norfolk. NR13 3BD. England. Tel. 01493 750114 'E-Mail' cyrilb@ibm.net 76251.2535@compuserve.com Also see Web Page at:- http://ourworld.compuserve.com/homepages/cyrilb About the Author. ----------------- He first trained as an RF Electronics Engineer in the mid '50's and subsequently became an Electronics applications and design engineer for one of the then largest capacitor manufacturers. As an application engineer he was deeply involved with both the birth of the Hi-Fi and Colour Television industries in England, and personally designed the first Aluminium Electrolytic Capacitors optimised for Hi-Fi sound. These were used in the first volume produced quality amplifier, the original Goodmans 90 and 100 series. Around the time the Apollo program got underway in America, he also inherited the EMC Filter Applications tasks. Subsequently as the use of EMC Filters increased, principally in Military equipments, he founded 'Mercator', a trading house tasked to import Components and EMC Filters not available from UK Manufacture. Subsequently the volume of EMC Filters increased such that he was then tasked to setup a UK design / development program to manufacture these Filters, and gain full European Quality approvals. For this operation he wrote all the design and measurement software needed to gain the approvals and for their production which was 100% Insertion Loss tested and Certified to 1 GHz. While with Mercator as a result of pressures from customers engineers, in 1983 wrote his first computer simulation program designed specifically to select the most appropriate Filters for differing source and load impedances. As a result he has had much practical experience of Filter simulation both with his own programs and commercial simulators, ranging from various Spice based programs to the most professional Microwave Design systems, and has been responsible for characterising RF capacitors using 'S Parameters' up to 3 GHz. He is now an independent design consultant working principally with RF topics, EMC Filters and Capacitors.