Fullscreen HTML5Med Res (??MB)HTML4

Circuit Surgery Regular clinic by Ian Bell Measuring the frequency response of a circuit or device using a PC sound card, part 2 L ast month, we started discussing how to measure the frequency response of a circuit or device. This was motivated by the need to measure the response of the example digital filter from the recent DSP series, recognising that not all readers will have test lab equipment capable of this. The alternative is to use the “sound card” (audio interface) of a computer or laptop to generate the input signal and analyse the response using software. Such programs are mainly aimed at measuring audio systems/acoustics, but can be applied to circuits acting on signal frequencies in the audio range (20Hz-20kHz). In the previous article, we mainly focused on the basic principles of circuit frequency response of linear circuits such as amplifiers and filters, which we will recap briefly below. We also introduced use of PC sound cards to make measurements, concluding with a test set-up like that shown in Fig.1. Here, we see one stereo channel used for measurements and the other for reference, but other configurations are possible. Audio interfaces are limited compared to lab test gear in that they only generate and measure AC over a limited amplitude range. Therefore, the I/O signal levels, DC offsets and load drive capabilities for the device under test and sound card I/O may not be compatible. This may require signal conditioning circuits at either or both the input and output of the device, as shown in Fig.1. That figure includes an audio amplifier that could be used to listen to the test signal. It can also be used to listen to the output from the circuit under test (connected in parallel to the sound card line input). I have used a switch set-up to enable easy comparison of the two signals. If you do set up the capability to listen to test signals, be careful and keep volume levels low. Test signals are not pleasant to listen to and have more potential to damage your speakers or even your ears than standard audio recordings. 44 A recap of the key concepts signals and transfer functions can be represented with a Cartesian format using complex numbers, which facilitates mathematical circuit analysis. Still, this advanced maths is not essential knowledge for basic frequency response measurement. To measure a frequency response, we can apply a sinewave input, wait for the circuit to stabilise (steady state approximation), measure the input and output amplitude and phase difference, then repeat for as many frequencies as required. This is straightforward, can be implemented manually (with a signal generator and dual channel oscilloscope), but is relatively slow. Alternatively, other inputs, such as sinusoids that are swept continuously in frequency, or various types of wideband noise, can be used. The frequency response results are obtained by mathematical analysis of the measured For a linear circuit with input X and output Y, we write Y(f ) = H(f )X(f ), where H is the transfer function – the relationship between the input and output of a circuit. The signals and transfer function vary with the frequency, f. The term “frequency response” often refers to graphs showing how the transfer function H(f ) varies with frequency. Each point on the graph represents a single frequency, which implies a steady state (infinite duration) sinusoidal input, and linearity implies a sinusoidal output. At each frequency, the output waveform may be shifted in time and may have a different amplitude with respect to the input. The time shift is usually expressed as phase shift – the proportion of the cycle time (⅟f ) of the waveform at the frequency of interest (f ), typically in degrees. The fact that two aspects change from input to output (amplitude and phase) means that H at a given frequency cannot be represented by a single number. We can use the polar form of gain magnitude/phase angle, which is what is shown on the Bode plot. Alternatively, PC Sound card R Optional audio L amplifier In Line R out L Out Loopback connection Line L in R USB Input signal conditioning In Device under test Out USB In Output signal Out conditioning Fig.1: one possible setup for using a sound card to measure circuit behaviour. Possible USB connection – eg, for MCU dev board Practical Electronics | November | 2025 response; for example, using a fast Fourier transform (FFT), rather than simple individual measurements. In this case, the instrument or app does the maths for you to calculate and present the results. Still, some understanding of the theory will help when choosing settings/preferences and interpreting the results. An introduction to REW Prior to writing these articles, I had not made extensive use of sound card apps for circuit measurement, so this article covers my experience and experiments so far. I have tried a few programs, but have spent most time using one called Room EQ Wizard (REW), so it will be the focus of this article. I cannot claim to be a REW expert, but I have obtained some useful results. I may look at other features and other apps in the future. REW is aimed at room acoustic measurement, loudspeaker measurement and audio device measurement. For example, to help optimise the acoustics of a listening room, studio or home theatre and find the best locations for speakers, subwoofers and listening positions. Having said that, its use for electronics is recognised by the developer. Some of its extensive measurement capabilities, including for frequency response and distortion, are useful for circuits via a setup such as the one shown in Fig.1. Others, such as reverberation time (RT60) are more specifically acoustics-related. REW is available as a free download from www.roomeqwizard.com for Windows, macOS and Linux. I have only used the Windows version. REW is supported by donations, but there is also pro upgrade available for a one-time fee that provides multiple-input measurement capabilities. At the time of writing, the current version is 5.31 (released in 2024). REW was created by John Mulcahy over twenty years ago, and he has continued to develop it since then. Originally, REW was mainly focused on setting up equalisers, but its measurement capabilities have been extended over the years. It is still under development, with various beta releases of version 5.40 published over recent months. If I understand it correctly, a key new feature in this future version is an application programming interface (API) which will, for example, allow advanced users to integrate their own code with REW, to make use of its measurement or calculation capabilities. REW is a complex piece of software with many features, settings and capabilities. As such, it may present a steep learning curve for new users. However, Practical Electronics | November | 2025 it has comprehensive documentation (in the program, via tooltips and online), which provides a good introduction to getting started and using the software. This seems to reflect a philosophy of making software suitable for both enthusiasts and professionals (indeed, it is used by both). The main preferences dialogs include help text in the window, which means you don’t have to search separately to find out what the settings do. There is also an active support forum on the AV Nirvana website (https://pemag. au/link/ac8k). It is recommended that new users read the first few sections of the help before jumping to more specific details. I will cover some key points here to provide an overview of REW’s relevant capabilities, and explain how I used REW for circuit measurements. Of course, I will not fully describe all aspects of the software. Different information may apply to different computers, operating systems and audio interfaces, so you may need some details which were not relevant to me. Units and levels Various units of measurement will be encountered in REW (and other software and instruments making similar measurements), so it is worth defining what these mean before proceeding with discussing its use. For testing circuits, we also need to be aware of the voltage levels of audio interface I/O, as this will determine compatibility with the device under test (DUT) and the design of any signal conditioning circuitry. Signal levels in audio and acoustics are commonly expressed using decibels (dB); for example, in dBV or dBu units. Decibels express the ratio of two numbers, which is straightforward for gain, where we have an input and output level; for example, gain (in dB) = 20log10(vout/vin). For signals in isolation, we need to use a reference level to express the value in decibel form. For example, for a voltage signal, vS, we need a reference level of v0. The decibel value for the signal is then 20log10(vS/v0). For power (p) measurements, the decibel value is 10log10(pS/ p0). If the signal value equals the reference, the level is 0dB. Consumer audio line levels are typically about 0.316V RMS (root-mean-square), which is 0.447V peak (0.316V × √2) or 0.894V peak-to-peak (0.447V × 2) for a pure sinewave signal. A signal of 0.316V RMS can be expressed as -10dBV – that is, minus ten decibels relative to a reference level of 1V RMS. For an RMS signal voltage of vS, the dBV value is 20log10(vS). In early audio systems, connections were made using matched 600Ω input and output impedances, which was based on the practice in early telephony systems. The reference level used here was based on the voltage required to deliver 1mW of power to a 600Ω load, which is 0.775V RMS (V2 ÷ R = 0.775V2 ÷ 600Ω = 0.001W). Voltage values referenced to 0.775V are expressed as dBu, where the u is for “unloaded”, as modern audio systems do not use 600Ω matched loads, but the reference level has persisted in common usage. dBV and dBu reference levels differ by a factor of 0.775 (multiply or divide as appropriate). Due to the log function, this translates to an addition or subtraction of decibel values, specifically of 2.21dB, as 20log10(0.775) = -2.21. So, we have dBu = dBV + 2.21dB and dBV = dBu – 2.21dB. Signal power can be expressed in dBm, that is, decibels relative to one milliwatt (1mW). If we are measuring voltages, we need to know (or assume) the load resistance the voltage is driving to calculate the power value. Professional audio systems typically use higher voltages than consumer ones, with a nominal level of +4dBu (1.228V RMS, 1.736V peak for a sinewave). In both cases, there can be variation in levels between different products. Sound card outputs are often designed to be used as headphone outputs, so they may differ from standard line outputs. Full-scale levels All signal processing systems have maximum input and output levels at which signal integrity is maintained. This is often slightly ‘soft’ – close to maximum, the signal quality will tend to decrease as distortion increases before a full limit (hard clipping) is reached. For ADCs and DACs, there is a very specific hard limit set by the maximum numerical value that can be represented in the conversion – the full-scale (FS) value. Ideally, the analog circuitry associated with such a converter will not be too close to its limits at the converter full-scale, so the system as a whole will have a simple hard limit at the converter’s full-scale value. This full-scale value can be used as a reference for a decibel full-scale unit, denoted dBFS. dBFS is useful because we can immediately see how much headroom we have available to increase signal levels (eg, for a better signal-to-noise ratio), or cope with the highest input levels (eg, due to resonant peaks) without causing clipping in the measurement system. Measurements in dBFS cannot be converted to other units using a generally applicable formula because the reference level depends on the specific hardware used and settings such as volume or gain controls. However, for a given setup 45 Fig.2: the Room EQ Wizard (REW) toolbar. Fig.4: the Room EQ Wizard (REW) oscilloscope. Fig.5: the Room EQ Wizard (REW) real-time spectrum analyser (RTA). 46 Practical Electronics | November | 2025 where the full-scale voltage is known, the conversion to or from other units is straightforward. REW makes use of dBFS values and has the means to calibrate the reference value (define the full-scale voltages). For frequency response plots, it can use the relative input to output dBFS levels to plot transfer functions (denoted as dBr), or the voltage levels (V/V in dB). Tools and measurements When REW opens, it presents a set of toolbar buttons (Fig.2) above a largely blank window containing a welcome message. The black area is used for response measurement graphs once these have been made or loaded from saved data. The Measure button opens a window that allows you to configure and run a frequency response measurement. However, before doing so, you need to set up the computer correctly and, using the preferences window (rightmost button), configure REW’s use of your audio interface(s) and then run a sound card calibration. We will look at this in more detail below. REW features various tools including a signal generator (Fig.3), oscilloscope (Fig.4), and real-time spectrum analyser (RTA) – see Fig.5. These are launched using the toolbar buttons and can be used independently of frequency response measurements. They are useful in their own right for work with circuits, subject to the limitations (discussed last month) that are imposed by using sound cards for measurements. The signal generator shown in Fig.3 is configured to produce a 1kHz sinewave 0.5dB below the full-scale output level. The tabs in the top part of the window indicate the range of other signal types that can be generated. The signal level can be set using dBu, dBV, volts (RMS) or dBFS units. Calibration is required to set the output voltage accurately (see below). The “Lock frequency to RTA FFT” box, which is checked in Fig.3, means the frequency is adjusted to the nearest FFT bin centre for the spectrum analyser (RTA), which improves spectral resolution, but restricts the frequencies to specific values based on the FFT. For example, the actual output with the setting shown in Fig.3 will be 999.76Hz, not 1kHz. This is handy when using the spectrum analyser. The “Add dither” option adds 2 LSB variations in the output signal, which helps to remove quantisation noise from distortion measurements. This is useful for analysing electronics, rather than acoustics, where other noise dominates. The lock frequency and dither controls just mentioned give some insight into the comprehensive and complex Practical Electronics | November | 2025 Fig.3: the Room EQ Wizard (REW) signal generator controls. nature of REW. The number of options and settings may seem daunting to new users, particularly if the terminology and related theory are not very familiar. However, REW’s documentation is very good, and clearly explains things such as when various options are appropriate to use. The two input channel Scope tool (Fig.4) provides the basic functionality and controls you would expect from an oscilloscope, including measurement cursors and some maths functions. The spectrum analyser will show the spectrum of any input signal, but is particularly useful in conjunction with the signal generator. The example in Fig.5 shows a 1kHz (actually 999.7Hz FFTlocked) sinewave. The harmonics (distortion components) are labelled 2, 3 etc. A distortion analysis is shown at upper left. A peak can be seen at the mains frequency (50Hz); the other parts of the spectrum are noise. Frequency response in REW REW makes frequency response measurements using a logarithmically swept sinusoidal signal which, as noted earlier, is faster than a stepped sine approach. The input sweep and response are processed using an FFT and compared to obtain the transfer function and impulse response, from which other measurements, such as distortion, can be derived. It can also use stepped sines for distortion measurements via the RTA, an approach which is slower but more accurate at lower distortion levels. REW uses one audio I/O channel for primary measurements (the default is the right stereo channel, as in Fig.1). The other input does not have to be used, but can be looped back (as in Fig.1), where REW can use it as a timing reference to compensate for delays in the sound card and operating system. The sound card itself may not have a perfectly flat frequency (gain) response, 47 Getting Started with the B-L475E-IOT01A1 We were contacted by a reader who asked for more detail on configuring the development board used in the article last month. We did not provide a detailed account of getting started using the SMT32 ecosystem in the articles because there are a number of online resources on this (video and text), one of which we referenced in the June 2025 issue. Typically, these describe ‘blinky’ projects that flash LEDs. However, STM32 may be more challenging to get started in than some other systems, like Arduino, so we will provide a few extra notes here. The configuration of the board should not be a problem because the STM32CubeIDE app does the work for you. You can start a new project for a specific board and all code to configure the MCU to use that board will be automatically written for you. You start with a working project that has no specific function. One potential problem is that the online examples refer to various boards, so you may have to use a different pin to control the LED, but much of the general content of these examples will be applicable. DigiKey provides a ‘blinky’ example at https://pemag.au/link/ac8l As is typical, this uses a different board, but the changes needed are minimal. During project creation, search for the board we used (B-L475E-IOT01A1) in the Target Selector, not the one in their example. The main code is in Core → Src → main.c (not just Src → Main) This difference is probably because the example uses an old software version. The B-L475E-IOT01A1 has two LEDs, so you have a choice of which one to flash. One option is to click the PA5 pin in the CubeMX view and select GPIO_ Output (the default is SPI clock) and save the .ioc file. Alternatively, use LED2, which is configured for use by default on the BL475E-IOT01A1. You don’t need to save the .ioc file (“Save” will be greyed out unless you change something, but it is not needed in here). Use “HAL_GPIO_ TogglePin(GPIOB, GPIO_PIN_14);” to toggle the LED instead of the code in the example. When looking at the autogenerated main code, you will see there are more peripheral initialisation calls, as the B-L475E-IOT01A1 board has more features. After that, create new projects in the same way (with the default configuration), then follow the configuration changes described in the articles. No changes to the board hardware are required. but REW can compensate for this using a measurement made with a direct connection between the input and output. This could also be used to compensate for the frequency response of any conditioning circuits, which should also ideally have a flat response. Setting it up REW obviously needs control of your sound card, so it is best not to have any other audio/media programs running, or have any media playing in a browser when you start REW or one of its tools. On the other hand, you can play media on the PC and observe the audio waveform on the REW scope via a loop-back connection or your device under test. To get the best results, the REW instructions recommend not simultaneously running any other apps or processes with high processor demand, disconnecting from the internet and switching off wireless networking (to reduce interference). I tried a few quick experiments while running a real-time spectral analysis and observed significant (temporary) disruption to the measurement when switching Bluetooth on and off and launching a demanding program. If this occurred during 48 a frequency response measurement, it would likely invalidate it. Therefore, it makes sense to keep the PC doing as little additional activity as possible when making important measurements. Sound card drivers (or the operating system) may provide audio processing with the aim of enhancing your music listening or gaming experience or applying processing, such as noise cancellation on inputs. These features need to be switched off while making measurements. For example, in Windows settings (System → Sound → Properties), for the selected output device (eg, Speakers/Line Out), there may be options such as “Audio Enhancements” or “Spatial Sound” under “Advanced Settings”, which need to be deselected. You probably also need to set the “Allow applications to take exclusive control of this device” option selected in the Windows audio properties for the device. The sound card manufacturer may also provide utilities/apps for interface control and/or sound processing as part of the driver install or as separately installed apps. You may also need to use these make sure that all processing is off and the I/O connector settings are configured correctly (for example, the jack socket is set to be a line input rather than side speaker outputs) and to control settings like sampling rates. Some audio interfaces will attempt to automatically change their configuration when devices are plugged in (eg, in response to headphone impedances on output connections), so it is a good idea to first plug in the leads you will use for measurement and then check the all the sound settings are correct. If possible, switch off auto-detection before adjusting settings for measurement work. Sound card preferences REW can use all the audio I/O interfaces that it detects on the computer, and you need to select the appropriate ones and which channels to use (eg, stereo left or right) for input and loop-back. This is done via the REW Soundcard Preferences window (Edit → Preferences → Soundcard). You also need to select the driver to use (Java or ASIO) and the sampling rate, which must match the setting in the operating system. There are various other settings in the Soundcard Preferences, and you should read the REW help page on this, which has detailed instructions and advice on how to configure things. REW supports a range of sampling frequencies depending on the sound card. As discussed in depth in the recent DSP series, the maximum frequency that can be measured is half the sample rate (the Nyquist frequency). The sound card should have an anti-aliasing filter, but is still better to avoid using signals with content above this value during measurements. The instructions recommend using 44.1kHz or 48kHz for acoustic measurements, assuming a 20kHz upper frequency requirement, as higher rates do not improve the results. Higher sampling rates are more demanding (memory and processing), but are required for results at higher frequencies. Signal levels and calibration As mentioned above, the sound card will have a maximum input voltage it can handle without clipping – this is the full-scale (FS) value of its ADC. Full-scale would be expected to occur at an equal magnitude for positive and negative excursions of the AC signal. The specific full-scale voltage will vary for different sound cards, but will typically be in the order of a volt for line inputs. As discussed last month, clipping must be avoided when making frequency response measurements. It can occur in the external circuitry or the sound card. Clipping in the sound card will be detected by REW, and notification can be Practical Electronics | November | 2025 From soundcard line output Buffer Device under test Out Gain 0 to 3 Buffer Audio line driver GainOut 0 to 0.5 Out 2 Gain To soundcard line input Fig.6: a block diagram of my signal conditioning circuitry for testing the DSP filter. provided to the user. Measurements can be automatically stopped if clipping occurs. REW will not detect clipping in the device under test or signal conditioning circuits if the sound card input does not reach full-scale, so care must be taken to avoid this. The signal generator and scope tools can be used before a frequency response measurement to check for clipping or other problems in the circuitry. The output of the sound card will also have a full-scale output voltage, which will depend on the output volume control setting and possibly other settings. We can calibrate the signal generator and input so that the numerical values of voltages set and measured by REW are accurate. After calibration, REW will display the full-scale output voltage in the signal generator window (Fig.3) and full-scale input voltage in the RTA window (Fig.4). The full-scale input voltage is also shown in the scope channel setup dialog, via clicking the leftmost of the three cog icons, near the “CH1-CH2” button – see Fig.4. To calibrate REW, you need a DMM, audio millivoltmeter or oscilloscope capable of measuring AC RMS voltages at around the 1V level. A basic meter that’s accurate at mains frequencies is sufficient. I/O voltage calibration After dealing with the operating system settings and REW preferences, proceed as follows. 1) Connect the sound card line input directly to the line output (loop-back in both channels) with a connection that allows you to measure the signal voltage between the ground and signal lines of the output. I used two audio cables with one connector cut off and the signal and ground (shield) wires soldered to male jumper wires which could easily be plugged into a solderless breadboard or pin headers, but you could also use a splitter cable and clip the meter to a jack plug or other connector. 2) Set the volume control to an output to a low level using the operating system control. Set the stereo balance to centre. The input (recording) volume in the operating system can probably be set at maximum, but it may depend on the specific setup. If the interface input has a physical gain control (more likely with an external USB unit), this may also need adjusting as part of the process. Practical Electronics | November | 2025 4) Run the signal generator, scope and RTA tools in REW, with the signal generator producing a sinewave on both channels (L+R) at a frequency the meter can accurately measure. Fig.3 shows the signal generator configured for a 1kHz sinewave, which was fine for the true RMS meter I was using, but other frequencies can be used. 5) Set up the scope to show both channels (Right and Left) with the voltage scale set so that the full-scale lines are visible. The scope display in Fig.4 shows the input (blue) and output (yellow) waveforms and the full-scale levels (dotted horizontal lines) with signal levels just below full-scale. 6) Set the generator signal level to 0.00dBFS and adjust the system output/ main volume until the signal amplitude on the scope is just below clipping. That is, if you increase the volume just above this value, you will see the clipped waveforms on the scope. The volume control may not provide very fine steps, so it may not be possible to set the output exactly equal to the full-scale input – use the highest value with no clipping. Measure the signal level with the meter and enter this as a calibration level in the signal generator and RTA via their “Calibrate level” buttons (see Figs.3 & 5). The signal does not have to be at full-scale when doing this – REW will calculate the full-scale values from the current output level setting and input level of the sound card’s ADC. If you change the output level in the signal generator, the meter reading should change to match the new setting, and the RTA and scope should show this accurately. You can switch the generator output level units and RTA y-axis to volts to make it easier to check this. Once the full-scale calibration is complete, do not change any of the computer or audio interface volume or gain settings. Conditioning circuit design Fig.6 shows a block diagram of the circuitry used for the DSP filter frequency Input Protection and DC shift Out response measurement. Of course, this can also be used directly for other circuits under test or adjusted to match different voltage range requirements. The circuit was designed to provide unity gain from input to output in the audio range, with the possibility to make gain adjustments to fine-tune the signal levels. It was constructed on a solderless breadboard to facilitate experimentation and development. There isn’t enough space to cover all the details of the circuit in this article, so we will just provide an overview here. The first buffer AC-couples the output from the sound card, provides a variable gain of up to three times (adjusted with a trimmer potentiometer), and has highfrequency roll-off (at around 250kHz) to help remove any RF signals that might be picked up at the input. With the sound card I was using, the maximum (just below full-scale) output from the sound card was 0.431V RMS, which is 1.22V peak-to-peak for a sinewave. The same value applies to a unity-gain circuit. For testing the DSP filter, the signal needs to be amplified to around 3.2V peak-to-peak (just under the microcontroller’s ADC’s 3.3V input range), which is 1.13V RMS. A gain of 2.6 times (3.2 ÷ 1.22) is required, hence the buffer’s maximum gain of 3 times. As described in the DSP series (August & September 2025 issues), the MCU digital filter implementation includes a protection circuit, AC-coupling and DC shift to bias the input to 1.65V (half the 3.3V supply). This is included in the DUT block in Fig.6, along with a reconstruction filter at the MCU DAC’s output (Fig.7). No anti-aliasing filter was used at the MCU’s ADC input because the frequency response test would not be applying signals with content beyond the Nyquist limit. The reconstruction filter used was a fourth-order low-pass op amp based circuit with a 20kHz cutoff, not the simple RC filter from the previous basic test SMT32 MCU Dev ADC in Board implementing DAC out DSP filter USB PC USB for power, programming and debug 4th-order 20kHz lowpass filter Out Output Fig.7: the connections for testing the DSP filter. 49 Fig.8: REW’s calibration measurement for my sound card (16 bits, 192kHz sampling rate). Fig.9: REW’s measurement progress display. further details on the “Getting set up for measuring” help page. It is suggested that REW output volume and sweep level settings (in Preferences) of 0.5 and -12dB to -6dB are used, respectively, but this is aimed at acoustic measurement. I tried various options, checking the frequency response and distortion graphs, and got good results with settings of 1.0 and -1.0dB, respectively. An example sound card calibration measurement is shown in Fig.8. A key aspect is that it is flat over the 20Hz to 20kHz range. After calibration, you need to save the results to a calibration (.cal) file via the “Make a cal file…” button in the preferences window (you are reminded to do this). You can have as many of these calibration files as you want, selecting the relevant one when you make a measurement. The calibration also appears like a regular measurement, with various graphs available to view. The frequency response (as in Fig.8) is referred to as “SPL and phase” in the set of buttons for selecting which graphs to view, and is displayed by default. SPL is sound pressure level and reflects REW’s primary use. REW also uses microphone calibration files, which relate to SPL meters, so are not needed for electronic circuit measurements. Any measurements, including those used for calibration, can be saved to a file and opened again in REW later. This is a different file type (.mdat) for the calibration file. Measurements are saved separately, via the list of measurements in the main REW window. Making measurements Fig.10: a plot of an RC low-pass filter response measured using REW. (September 2025). This means the reconstruction filter gain is flat throughout most of the stop-band up to the near DSP filter’s Nyquist limit of 24kHz. The RC filter was fine for quickly checking with an oscilloscope that the cutoff of the DSP filter looked about right, but would have had a significant effect on the overall frequency response if used in this measurement. The DSP filter and reconstruction filter have unity gain in the passband, so the output of 3.2V peak-to-peak needs attenuating back to 1.22V peak-to-peak for overall unity gain and to suit the sound card’s maximum input level. The output from the device under test goes to a second variable gain buffer and then to an audio line driver. The audio 50 line driver is more capable of driving capacitive loads and long cables than a typical op amp, and can drive both single-­ ended and balanced inputs. The line driver has a gain of about two times, which needs to be taken into account in the attenuation provided by the buffer. A gain of about 0.19 is required (1.22 ÷ 3.2 ÷ 2). Sound card calibration REW can measure the frequency response of a sound card and use this to compensate for any variations in sound card gain when making measurements. This is done via the “Calibrate sound card” button in the Preferences window. There are step-by-step instructions provided after you click the button and Measurements are made via the “Make a measurement” window after clicking the Measure toolbar button. There are a lot of settings and options, but the documentation explains these in detail, so we just give a brief outline here. The measurement type needs to be SPL for a frequency response. You can give the measurement a name and type in notes if you want. Key things to set are the frequency range and the output level. A level of 0dBFS should be OK with a unity-gain circuit without resonant peaks, but a slightly lower level may give better distortion performance. There is an option to abort the measurement if clipping occurs, which should usually be left on. There is a setting for sweep length (the number of sample points). Longer sweeps provide better signal-to-noise ratios in measurements and are suited to use with electronics (the help advises against their use with loudspeakers). The longest option (4M points) was Practical Electronics | November | 2025 www.poscope.com/epe Fig.11: the frequency response of the example digital filter as measured by REW. - USB - Ethernet - Web server - Modbus - CNC (Mach3/4) - IO - PWM - Encoders - LCD - Analog inputs - Compact PLC Fig.12: an LTspice simulation plotting the expected response of the digital filter. used for the measurements here. You can also run multiple sweeps and average the results. You can set some sound card options, but these should already be as required as they are based on the Preferences settings. You can select the calibration file to use, with the default being the last one created. Once you are happy with the settings, click the Start button to make the measurement. REW will show the measurement progress via a small rolling graph of the measurement and information on the remaining number of sweeps and time (Fig.9). It also shows the headroom in decibels below full-scale for the largest reading so far. After the measurement, REW may give a warning if the levels were excessively low, if clipping was detected, or high levels of distortion were present. Measurement results Before attempting to measure the DSP filter, I tried a simple RC low-pass circuit as the device under test (as in Fig.6). This used a 330nF capacitor and a 390Ω resistor. The cutoff (-3dB) frequency for this circuit is 1.24kHz; 1 ÷ 2πRC = 1 ÷ Practical Electronics | November | 2025 (2 ⸳ π ⸳ 390Ω ⸳ 330F ⸳ 10-9). Using measured values of the components (347nF and 386Ω), a cutoff of 1.19kHz would be expected. The results are shown in Fig.10. Using the cursors available on the results graph in REW, the cutoff was 1.20kHz, which is within 1% of the expected value. I moved on to measuring the digital filter using the circuits shown in Figs.6 & 7. The filter configuration was a 1kHz low-pass windowed sinc, with 37 coefficients, a 48kHz sampling rate, and a rectangular window running on the BL475E-IOT01A1 STM32 development board, as described in recent articles. The results are shown in Fig.11. Fig.12 shows an LTspice simulation of the same filter using the approach described in detail earlier in the DSP series. Comparison of the two graphs shows a good match, and confirms that the filter implementation is working correctly, and that REW is a useful tool for making such measurements. Coming up Next month, we will continue to explore this topic, focusing on the signal PE conditioning circuitry. - up to 256 - up to 32 microsteps microsteps - 50 V / 6 A - 30 V / 2.5 A - USB configuration - Isolated PoScope Mega1+ PoScope Mega50 - up to 50MS/s - resolution up to 12bit - Lowest power consumption - Smallest and lightest - 7 in 1: Oscilloscope, FFT, X/Y, Recorder, Logic Analyzer, Protocol decoder, Signal generator 51