Fullscreen HTML5Med Res (??MB)HTML4

Balanced Input & Attenuator for the USB Part 1 – by Phil Prosser This compact balanced input attenuator is designed to fit into the same instrument case as the USB SuperCodec. It provides four attenuation settings of 0dB, -10dB, -20dB and -40dB and has performance to match the superlative SuperCodec. Together, they form a potent recording and/or measurement system. T he SuperCodec USB Sound Card described over the ent devices (the measurement system and the device under test or DUT). last three issues has excellent recording and playback Another thing that the Audio Precision devices have but performance. So it can form the core of a high-perforthe SuperCodec lacks is input attenuators. The AP systems mance audio measurement system. One thing that it lacks compared to our Audio Precision can measure a wide range of signals from line level (well systems is a balanced input. Our AP System One and Sys- below 1V RMS) up to the output of multi-hundred-watt amplifiers (50V+ RMS). tem Two devices both have balanced and unbalanced inputs, As we mentioned previously, you can build our 2-Chanand you can select between them. There are times where you need those balanced inputs; nel Balanced Input Attenuator for Audio Analysers (May sometimes, you want to measure the performance of a bal- 2015) and hook it up to the SuperCodec inputs. That would solve anced audio device. both problems and But even with give you a test instruunbalanced devicment with flexibility es, it is common to approaching that of get better results usthe AP System Two ing balanced meas(and in some senses, urements. That’s exceeding it). because it helps to Fig.1: one channel of the Balanced Input Attenuator. There is an RF However, then you eliminate the com- filtering and DC-blocking stage before the relay-switched resistor-based would have two boxmon-mode noise attenuator. After the attenuators are the over-voltage protection stages, inherent in con- buffers and differential-to-single-ended converters before the signals are fed es or three boxes, two different power supnecting two differ- to the ADC inputs on the SuperCodec board. 44 Silicon Chip Australia’s electronics magazine siliconchip.com.au Features & specifications • Adds stereo balanced inputs (6.35mm TRS sockets) to the front panel of the USB SuperCodec • Balanced inputs replace the rear-panel unbalanced RCA inputs of the original design • Unbalanced outputs (RCA) remain on rear panel • Retains the 192kHz/24-bit recording & playback capabilities of the original SuperCodec • Fits into the SuperCodec case and uses the same power supply • 0dB, 10dB, 20dB and 40dB attenuation settings selected via front panel switch • CMRR: >60dB <at> 50-100Hz; >70dB <at> 1kHz; >50dB <at> 10kHz • SNR: 114dB <at> 0dB, 113dB <at> -10dB, 114dB <at> -20dB & -40dB • THD: 0.00010% (-120dB) <at> 0dB; 0.00014% (-117dB) <at> -10dB; 0.00028% (-111dB) <at> -20dB • Signal handling: 1V RMS <at> 0dB; 3.6V RMS <at> -10dB; 10V RMS <at> -20dB; 50V RMS <at> -40dB plies, cabling connecting them etc. That’s less convenient than having a single ‘all-in-one’ do-everything device. Also, the May 2015 project only has three attenuator settings (0dB, 20dB and 40dB) and we think that it doesn’t quite have the performance to match the SuperCodec, for reasons we’ll explain shortly. Hence, we came up with this project. It does a similar job to the May 2015 attenuator but with the addition of a -10dB attenuator setting and lower impedances for less noise. Importantly, it has been designed to integrate with the USB SuperCodec and fit in the same case, by keeping the PCB assembly compact and designing it to run off the same power supply rails. So with the addition of this balanced input board and some free or low-cost software, you can build an audio testing Here is the finished add-on board, with low-profile components to fit under the SuperCodec PCB. The inputs, RF filtering and AC-coupling components are at right, with the divider resistors in the middle. To their left are the attenuation selection relays, with the buffer op amps next to them, then the balanced-to-single-ended conversion circuitry at far left. siliconchip.com.au Australia’s electronics magazine November 2020  45 0 SuperCodec Balanced Input CMRR left channel, 0dB left channel, -10dB left channel, -20dB left channel, -40dB right channel, 0dB right channel, -10dB right channel, -20dB right channel, -40dB 10 Common Mode Rejection Ratio (dB) 23/07/20 10:59:20 20 30 40 Recording professional audio 50 60 70 80 90 100 20 50 100 200 500 1k Frequency (Hz) system that only a few years ago would have cost many thousands of dollars. 5k 10k 20k Fig.2: we tested the common-mode rejection ratio (CMRR) for both channels on our prototype, at four different frequencies and all four possible attenuation settings. The resulting plot is a bit messy but gives you an idea of the CMRR spread. A higher CMRR is better since it rejects proportionally more of the hum, buzz and EMI that may be picked up in cables etc. Another reason you might want to build the balanced input attenuator is to interface the USB SuperCodec with professional audio equipment. It gives you much greater recording flexibility, allowing you to use either balanced or unbalanced signals. And with the attenuator, it can handle much ‘hotter’ signals than the 1V RMS of the original Sound Card design. The 10dB attenuation setting puts professional +4dBu signals right in the sweet spot of the analog-to-digital converter (ADC), with good headroom. In this configuration, it can handle up to 3.6V RMS without clipping, or you can switch to the -20dB setting to handle signals up to 10V RMS, with relatively little degradation in performance at ‘normal’ signal levels. The design provides very well balanced inputs, with common mode rejection typically better than 60dB. The attenuation ranges of 0dB, -10dB, -20dB and -40dB allow full-scale inputs of 1V, 3.6V, 10V and 50V RMS which correspond to 1.4V, 5V, 14V and 71V peak or 2.8V, 10V, 28V and 142V peak-to-peak. This allows low-level signals, preamplifier outputs and power amplifier outputs to be used as signal sources (among other devices). Operating principles Fig.3: the noise floor of the combined Balanced Input Attenuator & SuperCodec ADC with the attenuator set to 0dB and the inputs shorted out. This shows that the new board adds minimal noise to the overall system. Fig.4: the same plot as Fig.3 but this type the attenuator has been switched to -10dB. As explained in the text, this is the setting where the Johnson (thermal) noise contribution of the attenuator resistors is highest. Despite this, the noise floor has only increased by around 1dB compared to Fig.3. 46 Silicon Chip Refer now to the block diagram, Fig.1. If you have a copy of the May 2015 issue, (or a download from siliconchip. com.au/Article/8560) you might also like to read back over the earlier Balanced Input Attenuator design, as this design has many similarities. The balanced input is via a 1/4-inch (6.35mm) standard tip-ring-sleeve (TRS) type connector (also often referred to as a “jack socket”). This was chosen over an XLR connector to save space, so that it will fit in the SuperCodec case. 6.35mm TRS is bog-standard, and often used for balanced signals, which makes this a versatile choice. We’re sticking with the standard TRS pinout of tip = “Hot” or positive, ring A view inside the "new" SuperCodec with the added PCB at bottom. It is designed to slot into the edge guides in the recommended Hammond 1455N2201BK aluminium case. Australia’s electronics magazine siliconchip.com.au = “Cold” or negative and sleeve for signal ground/screen. The balanced signals pass through an RF filter and DCblocking capacitors, then into the resistor and relay-based switched attenuator. After that, they pass through a clipping stage to provide over-voltage protection before going onto a set of buffer op amps. The buffered signals are then converted from balanced to single-ended signals, which are then fed to the inputs of the USB Sound Card already described. Performance We thoroughly tested the performance of the Balanced Input Attenuator to make sure it was up to SuperCodec standards. Fig.2 shows the measured common-mode rejection ratio (CMRR) value for both channels of the prototype, at all four attenuation settings and measured at four different frequencies. As you can see, the CMRR is between 71dB and 89dB at 1kHz for both inputs, and at all attenuation settings. Those are pretty good figures, and 1kHz is a typical test frequency. CMRR is slightly worse at lower and higher frequencies, but is better than 63dB at all tested frequencies below 1kHz, and better than 53dB at 10kHz. CMRR will be almost entirely a function of matching of the attenuator and balanced receiver resistors. So if you pay more attention when selecting those resistors, you could beat our prototype figures. With the 0.1% resistors specified, the attenuation error is less than ±0.1dB across all tested frequencies. Fig.5: we measured the total harmonic distortion (THD) with a -7.66dBV sinewave fed into the balanced inputs and a 0dB attenuator setting. The result shows very little difference from the same test without the Balanced Input Attenuator add-on. So it appears that the added circuitry is not introducing any extra distortion to the signals. Fig.6: the same test as Fig.5 but with the attenuator set to -10dB. Other than the signal level falling by the expected amount, there isn’t much difference. The increase in THD reading is mainly due to the change in signal level; increasing the input signal level by 10dB would likely give the same result as in Fig.5. And here's a view from the opposite end, with the lid removed, showing how the new PCB fits "upside down" above the existing SuperCodec board. siliconchip.com.au Fig.7: and the same test again with an attenuator setting of -20dB. The same comments as for Fig.6 apply. Note how the signal level drops by very close to 10dB and 20dB in these two tests, showing off the excellent attenuation accuracy. Australia’s electronics magazine November 2020  47 Benefits of balanced signals Professional audio equipment uses balanced signals carried on three conductors: the positive “Hot”, negative “Cold” and a screen. Electromagnetic interference picked up in the cable (usually heard as hum or buzz) affects both the Hot and Cold signals similarly. The balanced receiver subtracts the Cold signal from the Hot, resulting in twice the signal with severely attenuated noise. In other words, if the Hot signal is signal x 1 + noise and the Cold signal is signal x -1 + noise, Hot – Cold gives you (signal x 1 + noise) - (signal x -1 + noise) = signal x 2 + noise x 0 This is a great way to reject noise and hum from things like ground loops, especially on long cable runs. Besides added complexity in the circuitry, the main disadvantage of this approach is that converting a balanced signal into an unbalanced signal generally introduces a bit of white noise; so while hum and buzz are rejected, the signal-to-noise ratio (SNR) can suffer a bit. When testing audio equipment, we often need to analyse the signal between two particular points in the device under test (DUT). We certainly want to avoid measuring any voltages within the ground system of the DUT or our test equipment itself. By using a balanced input in this situation, we can connect the Cold conductor to an appropriate ground reference point in the DUT. The Hot connection is then used to measure the signal of interest. Any noise between the USB Sound Card ground and the DUT ground is subtracted out of this measurement. When measuring low voltages and exceptionally low distortion levels on signals at moderate voltages, this is extremely important, as sometimes we are looking for microvolt or even nanovolt level distortion signals. As good as balanced interfaces are, Earthing remains essential. To achieve good results below -100dB, you will need to work on the test Earthing and layout. You might be surprised how much things like the orientation of the equipment being tested and its proximity to computer equipment and even the operator can affect the results! 48 Silicon Chip The noise and distortion performance is not significantly worse than the straight USB Sound Card with a 10kΩ input impedance (the input impedance options are described below). There is a small increase in THD on the -10dB range for the 100kΩ input impedance version. Fig.3 shows the output spectrum with the attenuator on the 0dB setting and the inputs shorted to ground. If you compare it to Fig.5 on page 27 of the August 2020 issue, showing the same measurement for the SuperCodec alone, you will see that there isn’t all that much extra noise being introduced by the Balanced Attenuator. Fig.4 shows the same measurement but with the attenuator on the -10dB setting, which is the worst case (as explained below). Overall, the noise has only crept up by about 1dB compared to the 0dB attenuator setting, so that’s a good result. Fig.5 shows the THD+N measurement for a test signal of around 300mV RMS being fed into the Balanced Input Attenuator with the attenuation setting at 0dB. This is virtually unchanged from the measurements we made previously without the Balanced Input Attenuator board. You can compare this to Fig.4 on page 27 of the August 2020 issue, but note that the test signal level is slightly different. Fig.6 shows that the distortion performance on the -10dB setting, with the same signal applied as for the 0dB setting, is barely any worse. So the attenuator does not appear to be introducing any signal distortion. Similarly, Fig.7 shows the result with the attenuator on the -20dB setting. The THD measurement has risen to 0.0003% / -111dB. However, note that if the applied signal amplitude were increased to a level that you would need the 20dB of attenuation to measure, the THD level would probably drop quite close to the 0.0001% / -120dB shown in Fig.5. Circuit details Refer now to the full circuit diagram, Fig.8, and compare it to the block diagram, Fig.1. Let’s consider the left channel signal path, starting at CON1; the right channel is the same. The input signal goes via a ferrite bead with a 22pF bypass capacitor to filter off the worst of any RF signals on the input. The USB Sound Card is Australia’s electronics magazine AC-coupled, so a DC blocking capacitor is included between the input RF filter and the attenuator. We want a lower cutoff frequency (-3dB point) an order of magnitude below 20Hz, so we have chosen 1.5Hz. This means that any non-linearities in the DC-blocking capacitors will not introduce any distortion, so long as they are not gross non-linearities (as is found in high-K ceramic capacitors, for example). For a 100kΩ input impedance, as used in the May 2015 Attenuator design, this demands the DC blocking capacitor be 1µF. But the Johnson noise in a 100kΩ resistance is enough to affect the performance of the USB SuperCodec, so we really need a lower input impedance, say 10kΩ. This demands a 10µF DC-blocking capacitor for the same 1.5Hz -3dB point. The current through these capacitors is extremely low, and pretty much any film capacitor will work well. You could use an electrolytic, but many people don’t like the idea of electrolytics in the signal path (even though they work OK for signal coupling). Also, they tend not to last as long as film capacitors. And as mentioned above, ceramic is a poor choice, so plastic film it is. The switched attenuator The input attenuator reduces the input signal level by 0, 10, 20 or 40dB. That means division ratios of 3.16:1, 10:1 and 100:1. We chose these values as 0dB (ie, straight through) gives the best sensitivity and a useful 1V RMS input level. -10dB is well suited to professional audio signal levels. It is also low enough to be useable with consumer equipment like CD, DVD and Blu-ray players which tend to produce an output signal of around 2.2V RMS. The -20dB and -40dB settings are handy for power amplifier testing. The attenuator is a simple resistive divider. The total series resistance sets the input impedance of the balanced interface, and as mentioned above, this has an impact on the noise performance and the size of the DC blocking capacitor required. Thermal noise The noise impact will depend on the attenuation setting. At 0dB, the divider is effectively bypassed and so the insiliconchip.com.au put impedance has no real effect on the performance. At the other three settings, the input impedance ‘seen’ by the SuperCodec is the upper and lower halves of the divider, bisected by the selected tap, in parallel. The worst case is the -10dB setting, at 21.6% of the overall input resistance (ie, 21.6kΩ for the 100kΩ option and 2.16kΩ for the 10kΩ option). For the -20dB setting, it is 9% of the input resistance and for the -40dB setting, it is 1% of the input resistance. Thermal noise in a resistance is calculated as √(4 x K x T x B x R) where K = 1.38 x 10-23, T is the temperature in Kelvin, B is the bandwidth in Hz and R is the resistance in ohms. At room temperature (around 300K), for a bandwidth of 20kHz and a resistance of 21.6kΩ, this works out to 2.67µV RMS, which is -111.5dBV. That is a higher level than the inherent noise in the SuperCodec ADC, so it would definitely degrade performance. A source impedance of 21.6kΩ to the buffer op amps would also increase their distortion contribution slightly. For 1/10th the resistance, that noise level drops by a factor of √10 = 3.16, to 845nV RMS or -121.5dBV. This is usefully below the noise floor of the SuperCodec, so it will have little impact on performance at -10dB, and even less on the -20dB and -40dB settings. In fact, the biggest impact on performance is likely to be EMI pickup due to the higher input impedance in this case. Consider errors caused by loading the DUT with 10kΩ. A preamp might have a 100Ω resistor in series with its output. If we measure this preamp with a 10kΩ input impedance balanced line test set, we will introduce a 1% scaling error. That probably does not matter in most cases, but it does need to be considered. We certainly would not want errors greater than this. So 10kΩ is the lower practical limit, especially when you consider that film capacitors with values above 10µF are expensive and bulky, and would not fit in the space available. We also need to consider power dissipation in the divider. With 50V RMS fed into the divider, the power dissipation is 0.25W for a 10kΩ divider. This is spread out through several resistors, siliconchip.com.au Parts list – Balanced Input & Attenuator 1 assembled USB SuperCodec without 2x12-pin headers attached or front/rear panels drilled but with loose MCHStreamer module (described in Aug – Oct 2020 issues) 1 assembled Balanced Input Attenuator board (see below) 1 set of Test Leads (optional; see below) 2 6x2-pin header sockets, 2mm pitch with pigtails (supplied with MiniDSP MCHStreamer) 1 180mm length of heavy-duty figure-8 shielded audio cable [Altronics W2995, Jaycar WB1502] 1 1m length of red medium-duty hookup wire 1 1m length of black medium-duty hookup wire 1 1m length of green medium-duty hookup wire 1 30cm length of 5mm diameter black or clear heatshrink tubing 1 30cm length of 2.4-2.5mm diameter black or clear heatshrink tubing Balanced Input Attenuator board 1 double-sided PCB coded 01106202, 99.5 x 141.5mm 2 6.35mm DPDT switched stereo jack sockets (CON1,CON2) [Altronics P0076, Jaycar PS0180, element14 1267402] 1 right-angle 3-pin polarised header (CON3) [Altronics P5513, Jaycar HM3423] 1 right-angle 4-pin polarised header (CON4) [Altronics P5514, Jaycar HM3424] 4 4-5mm ferrite suppression beads (FB1-FB4) [Altronics L5250A, Jaycar LF1250] 8 2A DPDT 5V DC coil telecom relays (RLY1-RLY8) [Altronics S4128B/S4128C, Mouser 551-EA2-5NU] 1 DP4T right-angle PCB-mount switch (S1) [Altonics S3008] Semiconductors 6 NE5532AP or NE5532P dual low-noise op amps, DIP-8 (IC1-IC6) 2 12V 1W zener diodes (ZD1,ZD2) 2 3.9V 1W zener diodes (ZD3,ZD4) 8 1N4148 small signal diodes (D1-D8) Capacitors 1 100µF 16V electrolytic 4 10µF 100V polyester film*, 15mm lead pitch [Mouser 871-B32562J1106K] 6 10µF 35V electrolytic 6 100nF 63V MKT 8 100pF 50V C0G/NP0 ceramic 4 22pF 250V C0G/NP0 ceramic Resistors (all 0.25W ±1% metal film unless otherwise specified) 4 1MW 2 3.3kW 1 82W 4 68W 4 39W* 4 33W 6 10W 4 6.81kW* ±0.1% [Mouser 71-CMF556K8100BEEK] 4 2.15kW* ±0.1% [Mouser 71-RN55C-B-2.15K] 16 1kW ±0.1% [Mouser 71-PTF561K0000BXR6] 4 900W* ±0.1% [Mouser 71-CMF55900R00BHEB] 4 100W* ±0.1% [Mouser 71-CMF55100R00BEEB] * for 100k input impedance, substitute these instead: 4 1µF 250V polypropylene film, 7.5mm lead pitch [Mouser 667-ECW-F2105HAB] 4 68.1kW ±0.1% [Mouser 279-H868K1BYA] 4 21.5kW ±0.1% [Mouser 279-YR1B21K5CC] 4 9kW ±0.1% [Mouser 71-PTF569K0000BYEK] 4 1kW ±0.1% [Mouser 71-PTF561K0000BXR6] 4 390W ±1% Test Lead parts 2 90° 6.35mm TRS line plugs [Altronics P0048 or P0049] 2 1.2m lengths of microphone cable (or length to suit) [Altronics W3024/W3029, Jaycar WB1534] 2 small red alligator clips [Altronics P0110, Jaycar HM3020] 2 small black alligator clips [Altronics P0111, Jaycar HM3020] 2 small green alligator clips [Altronics P0102] 1 30cm length of 6mm diameter black or clear heatshrink tubing 1 30cm length of 3mm diameter black or clear heatshrink tubing 1 30cm length of 2.4-2.5mm diameter black or clear heatshrink tubing Australia’s electronics magazine November 2020  49 50 Silicon Chip Australia’s electronics magazine siliconchip.com.au Fig.8: the circuit of the Balanced Input Attenuator add-on board. CON1 and CON2 are the new 6.35mm TRS jack socket inputs connectors, while CON3 and CON4 connect to the ±9V supplies and CON4 input header on the USB SuperCodec Sound Card board respectively. The attenuator resistor taps are selected via relays RLY1-RLY8, and the signals then pass to op amp buffers IC1-IC4 and the differential-to-single-ended converter stages based on dual op amps IC5 & IC6 before going to the ADC. siliconchip.com.au Australia’s electronics magazine November 2020  51 but heating in those resistors could lead to some inaccuracies. The ratings of the divider resistors would allow up to 80V RMS to be fed in, but besides this being possibly unsafe, we prefer not to run them at their limits. So there is no perfect answer. Hence, we are providing resistor values for the input attenuator that give either a 10kΩ or 100kΩ input impedance. Remember to choose the right value capacitor to go with them. Our inclination is to go with 10kΩ, but we fully understand why others might choose 100kΩ. We have used relays to switch between the four possible attenuation settings. This is a little bit expensive, as these are a few dollars each, but it makes the design nice and clean in terms of layout and avoids the possibility of noisy, unreliable wafer switches failing. The relays give a satisfying “clunk” as you switch across ranges, suiting such a high-performance device. Buffers The voltage divider output impedance varies depending on the range selected. This does not suit the balanced-tosingle-ended converter, so buffers are needed. We use two paralleled op amps to do this, driving two balanced-to-singleended converters. These are combined at the output to get a 3dB improvement in signal-to-noise ratio compared to using fewer op amps. The differential-to-single-ended converters subtract the Cold input signal from the Hot input signal. The matching of resistors in these is important, at least within each arm of each converter. So we have specified 0.1% tolerance 1kΩ resistors here. This tolerance is required to deliver the specified performance. We have chosen 1kΩ resistors as they have a low enough resistance to add negligible thermal noise to the convertor without loading the op amp outputs too much. And as many constructors will likely have plenty of 1kΩ 1% resistors, they could select well-matched pairs using just about any DMM and avoid the cost of 0.1% types. The output of the differential-tosingle-ended convertors is combined through 10Ω resistors (necessary to allow for the op amps having different offset voltages), which then feed into the USB Sound Card. We have included input protection comprising diodes clipping to a 3.9V rail. We have tested that this does not impact distortion performance. Note though that if you connect this to a high-voltage source on the 0dB range, you will damage this part of the circuit! There is additional protection on the power supply rails provided by 12V zeners, which again should only operate under extreme fault modes. Next month Unfortunately, we don’t have room for the construction details this month. That will have to wait for the next issue. As well as describing the construction, and what you have to do to get the Balanced Input Attenuator to fit into the same case as the USB SuperCodec, the second and final article in this series will also cover the testing procedure, and how to make some handy balanced SC test leads. Subscribe to SILICON CHIP and you’ll not only $AVE $AV AVE MONEY AV but we GUARANTEE you’ll get your copy! When you subscribe to SILICON CHIP (printed edition) in Australia we GUARANTEE you’ll never miss an issue! Subscription copies are despatched well before the on-sale date (the magazine is due on sale around the end of the previous month). It is unusual for copies to go astray in the post – but when we’re mailing out many thousands of copies, it does happen occasionally. So we make this promise to you: if you haven’t received your SILICON CHIP anywhere in Australia by the end of the month of issue, contact us and we'll give you online access to that issue and also post you a replacement copy in our next mailing. You can contact us via e-mail at silicon<at>siliconchip,com.au Remember: it’s cheaper to subscribe anyway - do the maths and see the savings! Remember: we pick up the postage charges – so you $ave even more! Remember: your newsagent may have sold out when you get there - and you’d miss out! Remember: you don’t have to remember to go out and buy your copy - it’s delivered right to your letter box! Remember: you can also subscribe to the online version, or get a discounted, combined print+online subscription Convinced? We hope so! And we make it particularly easy to take out a subscription – for a 6 month trial, a standard 12 month or a giant 24-month with even bigger savings! To subscribe, simply go to our website – www.siliconchip.com.au – click on the “subscriptions” tab and follow the prompts. We accept payment by Paypal, Visa, Mastercard, EFT/Direct Deposit or Cheque/Money Order (sorry, no Amex or Diners). We’re waiting to welcome you to the SILICON CHIP family! 52 Silicon Chip Australia’s electronics magazine siliconchip.com.au