Silicon ChipAUDIO OUT - December 2021 SILICON CHIP
  1. Outer Front Cover
  2. Contents
  3. Subscriptions: PE Subscription
  4. Subscriptions: PicoLog Cloud
  5. Back Issues: PICOLOG
  6. Publisher's Letter
  7. Feature: The Fox Report by Barry Fox
  8. Feature: Techno Talk by Mark Nelson
  9. Feature: Net Work by Alan Winstanley
  10. Project: Easy-to-build Digital AM/FM/SW Receiver by Charles Koslna
  11. Project: Balanced Input and Attenuator for the USB by Phil Prosser
  12. Project: Mini Digital Volt/ Amp Panel Meters by Jim Rowe
  13. Feature: Max’s Cool Beans by Max the Magnifi cent
  14. Feature: KickStart by Mike Tooley
  15. Feature: Circuit Surgery by Ian Bell
  16. Feature: AUDIO OUT by Jake Rothman
  17. Feature: Electronic Building Blocks by Julian Edgar
  18. PCB Order Form
  19. Advertising Index

This is only a preview of the December 2021 issue of Practical Electronics.

You can view 0 of the 72 pages in the full issue.

Articles in this series:
  • (November 2020)
  • Techno Talk (December 2020)
  • Techno Talk (January 2021)
  • Techno Talk (February 2021)
  • Techno Talk (March 2021)
  • Techno Talk (April 2021)
  • Techno Talk (May 2021)
  • Techno Talk (June 2021)
  • Techno Talk (July 2021)
  • Techno Talk (August 2021)
  • Techno Talk (September 2021)
  • Techno Talk (October 2021)
  • Techno Talk (November 2021)
  • Techno Talk (December 2021)
  • Communing with nature (January 2022)
  • Should we be worried? (February 2022)
  • How resilient is your lifeline? (March 2022)
  • Go eco, get ethical! (April 2022)
  • From nano to bio (May 2022)
  • Positivity follows the gloom (June 2022)
  • Mixed menu (July 2022)
  • Time for a total rethink? (August 2022)
  • What’s in a name? (September 2022)
  • Forget leaves on the line! (October 2022)
  • Giant Boost for Batteries (December 2022)
  • Raudive Voices Revisited (January 2023)
  • A thousand words (February 2023)
  • It’s handover time (March 2023)
  • AI, Robots, Horticulture and Agriculture (April 2023)
  • Prophecy can be perplexing (May 2023)
  • Technology comes in different shapes and sizes (June 2023)
  • AI and robots – what could possibly go wrong? (July 2023)
  • How long until we’re all out of work? (August 2023)
  • We both have truths, are mine the same as yours? (September 2023)
  • Holy Spheres, Batman! (October 2023)
  • Where’s my pneumatic car? (November 2023)
  • Good grief! (December 2023)
  • Cheeky chiplets (January 2024)
  • Cheeky chiplets (February 2024)
  • The Wibbly-Wobbly World of Quantum (March 2024)
  • Techno Talk - Wait! What? Really? (April 2024)
  • Techno Talk - One step closer to a dystopian abyss? (May 2024)
  • Techno Talk - Program that! (June 2024)
  • Techno Talk (July 2024)
  • Techno Talk - That makes so much sense! (August 2024)
  • Techno Talk - I don’t want to be a Norbert... (September 2024)
  • Techno Talk - Sticking the landing (October 2024)
  • Techno Talk (November 2024)
  • Techno Talk (December 2024)
  • Techno Talk (January 2025)
  • Techno Talk (February 2025)
  • Techno Talk (March 2025)
  • Techno Talk (April 2025)
  • Techno Talk (May 2025)
  • Techno Talk (June 2025)
AUDIO OUT AUDIO OUT L R By Jake Rothman Analogue Vocoder – Part 2 L ast month, in Part 1, we Pretty curves introduced our brand-new design for an Analogue Vocoder, which creates a unique and fascinating vocal effect – human speech is superimposed on a musical instrument. Before we look in more detail at the design, first a quick conclusion to the discussion of the ‘analysis’ (multiple channel) filters discussed at the end of last month’s article. A plot of the speech (or ‘analysis’) filters is shown in Fig.13. The output/synthesis filters should have exactly the same response, but are harder to measure (and hence produce plots for) because there are VCAs in-between both sections. The highpass and low-pass curves have a much wider bandwidth with 24dB/oct slopes. This is because nearly all speech information is in the midrange from around 200Hz to 5kHz. In general, fil20 Hz 50 100 200 500 1k 2k 5k 10k 20k ters should always 0dB use 1% tolerance –10 resistors, which fortunately are now –20 cheap. The prob–30 lem components –40 are the capacitors, Fig.13. Plot of the analysis filters of the prototype. A little uneven which are often only available in due to the use of 5% capacitors. For musical applications this 5% tolerance. Ocis of little consequence – we are not building precision test gear. c a s i o n a l l y, 2 % Often the upper and lower band-pass filters are made with a polystyrene and wider bandwidth than the ones in the middle, where the speech silvered-mica types information is most concentrated. so In p u t cke ts so O u tp u t cke ts 1 9 - i n ch r a ck B u s b o a rd e n cl o su r e , co Fig.14. The rough photocopied board layouts in J M Hawkes’ student project report were smoothed into a form suitable for etching using Photoshop. are available. These are the best capacitors for filters, but they tend to be larger and much pricier compared to the cheaper and easy-to-find polyester types. Board stiff Life is too short to make a 14-channel vocoder prototype on Veroboard! So, to make the prototype shown in Fig.5 (see last month’s Audio Out) I started off with IE C n ta in in g 1 4 b o a r d s M o l e x, r i g h t - a n g l e co cke t n n e ct o r s P h a n to m b o a rd ( o p tio n a l) 2 x d r i ve a m p lifie r b o a rd s 2 x d r i ve a m p lifie r b o a rd s T r a n sf o r m e r * T o o th e r b o a rd s M i cr o p h o n e p r e a m p lifie r b o a r d ( o p tio n a l) L P /H P B P 1 /2 B P 3 /4 B P 5 /6 B P 7 /8 B P 9 /1 0 B P 1 1 /1 2 P S U b o a rd * T o m i n i m i se h u m , p l a ce a s fa r a w a y a s p o s ib le f r o m a m p l i f i e r b o a r d s. F i l t e r b o a r d s x1 C o n tr o ls so 4 P o w e r s w i t ch Fig.15. The Analogue Vocoder consists of many interconnected boards. The microphone preamplifier, driver, summer/output and power supply boards are all useful audio building blocks in themselves. Putting all the circuits together on one big board is not a good idea since it inhibits experimentation and reduces versatility. (Commercial production would of course dictate much greater integration to reduce labour intensive wiring, but we don’t face that pressure.) 60 Practical Electronics | December | 2021 n microphone preamplifier, two driver boards, 12 band-pass channels on six boards, an optional low-pass/high-pass channel board and two summing/output amplifiers. These are shown in Fig.15. The vocoder channels are all plugged into a bus board using right-angle connectors, as shown in Fig.16. Control signals The output of the filters are fed into precision full-wave rectifiers (Fig.17) which are smoothed by second-order low-pass filters (Fig.18) whose cut off frequency is proportional to each frequency band. If the smoothing frequency is too high, the effect is ‘spittiness’ and distortion. If too slow, the intelligibility of the speech is impaired. The idea is to get a general trend of the energy fluctuations in each particular frequency band. The cut-off frequency should be around 5 to 20-times less than the lower frequency of the band-pass filter, with a Q of about 1. The Fig.16. Since all the channel connections are wired in parallel (the ‘bus’) a long piece of tolerances for these filters are uncritical stripboard was originally used to connect them all together on the prototype. However, and 10% tolerance capacitors are fine. in the design we will be building over the next few months, a proper PCB bus board will A linear voltage-to-current converter be used. It has plated-through holes for extra strength. using an op amp and transistor (shown in Fig.19) is used to feed the control pin of the VCA’s an edited version of JM Hawkes’ PCBs. Unfortunately, all I had were transconductance amplifiers. This ensures better tracking scratchy photocopies of taped artwork. These were soon cleaned at low levels because the (PNP) transistor provides a much up in Photoshop (shown in Fig.14) and printed out for etching. higher drive impedance than just a resistor. Once I had made my own tweaks and produced a working system I got my PCB guru, Mike Grindle, to design a professional set of PCBs. The vocoder comprises 14 boards, including an optional Chips and VCAs The transconductance VCAs are the biggest source of noise in the whole system. However, this is masked by the muC * C * d e p e n d s o n f r e q u e n cy b a n d 3 9 n F I C 1 * * u se T L 0 8 2 o r si m i l a r sical sounds generated. Interestingly, it does not detract C a n ce l l a t i o n o f h a l f - w a ve r e ct i f i e d from the overall effect, the slight roughness introduced si g n a l w i t h t h e p o si t i ve h a l f o f t h e 1 % i n p u t si g n a l o cu rs h e re . makes it sound more ‘vocal’ and ‘breathy’, which is a positive effect for this system. My prototype shown in Fig.16 used the classic CA3080 1 % 1 % transconductance amplifier, which ceased production in – IC 2 b * * 2005. Later vocoders and this design use the LM13600/700 1 N 4 1 4 8 C la m p + O u tp u t d io d e which is effectively two 3080s with added Darlington buffers. Linearising diodes are also added for distortion In p u t 1 N 4 1 4 8 1 % 0 V – 0 V + – V t F u l l - w a ve - r e ct i f i e d o u t p u t ( i n ve r t e d ) + V IC 2 a * * + V I W h e n i n p u t vo l t a g e si t s a t 0 V th e o p a m p o u tp u t is a t – V B E . t 0 V – V 0 V 2 2 n F Fig.17. A precision full-wave rectifier circuit is used to generate DC from the output of each speech analysis filter. The circuit depends on accurate subtraction of a half-wave rectified signal from the rectifier input. Note that 1% resistors are needed. t + 2 2 n F O u tp u t to V - t o - I co n ve rto r – t – V – V S m o o t h e d o u t p u t vo lta g e Fig.18. The output from the rectifier is smoothed by a low-pass filter. These filters need different values for each frequency band. Practical Electronics | December | 2021 – I n p u t n e g a t i ve g o i n g vo l t a g e 0 V IC 3 b + T R 1 B C 5 5 9 C t – V 1 0 0 n F In p u t fro m r e ct i f i e r 1 N 4 1 4 8 N e g a t i ve cu r r e n t t o t r a n sco n d u ct a n ce a m p l i f e r co n t r o l p i n I H i g h i m p e d a n ce O u tp u t Fig.19. The voltage-to-current converter controls the transconductance op amp to form a VCA. The VBE of the transistor is cancelled with negative feedback. This circuit is a design from Jacob Moskowitz’s article, Current-compensated op amp improves OTA linearity, Jan 1977. I can’t trace the original publication, but it is available on ResearchGate.net. I’ve made minor improvements consisting of a diode to protect the base-emitter junction from reverse bias and a capacitor to prevent HF instability has been added. 61 the control current and the Darlington input to obtain both a high impedance and a high maximum current output over the full range of control currents. The LM13700 is a later version where better processing gave higher current gain in the Darlingtons, allowing this link to be dispensed with. The LM13700 is currently the only transconductance amplifier IC in production, and I use the surface-mount version in my Dubreq Theremin. The CA3280 is the best choice, but is hard to find. It also has linearising diodes, but no buffer stages. Using dual devices, such as the LM13600 means that two vocoder filter channels have to be placed on one board. There is also the NE571 telephone compander chip that includes a couple of full-wave rectifiers, in addition to two VCAs. The Paia 6710 eight-band vocoder kit uses these. Another odd ball is the CA3060, which has three 3080s on one chip. All these VCA chips are shown in the IC rogues’ gallery of Fig.20. Transconductance amplifiers are voltage in / current out devices. However, here we are using a gain-controlled transconductance amplifier with an additional linear current-driven gain control pin. There are even more complicated VCAs available, such as the 2180 from That Corp. These have logarithmic control characteristic, which complicates the circuits excessively. Also, their Hi-Fi specification is not necessary, possibly making the vocoder too clean sounding! I suspect the soft clipping produced by non-linearised transconductance Fig.20. Suitable transconductance VCA chips for vocoders. Here we see, the classic CA3080, the LM13600/700, the oddballs NE571 and CA3060 – plus the most recently developed and best, the CA3280. Only the LM13700 and ‘That Blackmer Corp’ topology 2180 chips are still in production, although the DIL version of the LM13700 will soon be discontinued by Texas Instruments (TI). All are available from the AO shop – see page 59. cancellation, allowing a higher drive level, and thereby giving lower noise. This is activated by installing R25 (see Fig.21). There is no difference in practice between the LM13600 and LM13700. Originally, the LM13600 provided a link between T P 1 * C 2 3 3 n F T o o th e r ch a n n e l 1 . 2 5 kH z In p u t sp e e ch 8 6 0 H z D 1 1 N 4 1 4 8 R 9 – IC 1 a T L 0 8 2 R 2 C 5 1 0 n F – + V – C o n tro l o u tp u t IC 1 b T L 0 8 2 R 8 R 1 2 R 1 1 – – + R 5 D 2 1 N 4 1 4 8 IC 2 b T L 0 8 2 IC 2 a T L 0 8 2 + + R 1 0 0 V 0 V C 7 1 0 0 n F R 1 5 R 1 6 – IC 3 a T L 0 8 2 C 8 2 2 n F C 9 2 2 n F + T R 1 B C 5 5 9 C C 1 1 3 3 n F In p u t syn th z 0 V 4 (1 3 ) – + R 3 6 R 2 4 1 1 2 (1 5 ) 1 (1 6 ) 3 (1 4 ) C 1 2 3 3 n F R 2 8 R 3 5 6 R 2 6 7 (1 0 ) R 3 1 M i x b u se – R 3 3 IC 5 b 5 5 3 2 s O u tp u t r ig h t T o o th e r ch a n n e l R 3 0 8 (9 ) O u tp u t le ft R 3 2 + R 2 9 IC 4 a /(b )* * L M 1 3 7 0 0 5 (1 2 ) T P 3 8 6 0 H z – + R a il b u se s V – 1 5 V * * I C sh a r e d w i t h o t h e r ch a n n e l R 2 3 R 2 2 R 2 7 0 V 0 V F i l t e r va C 1 6 + 1 0 µF 2 5 V C 1 3 3 3 n F R 2 5 R 2 1 IC 5 a 5 5 3 2 R 2 0 0 V V + ( o p tio n a l) C 1 0 3 3 n F R 1 9 C 1 4 4 7 0 n F V + 1 5 V R 3 4 R 1 8 0 V 1 . 2 5 kH T o IC p o w e r p in s 0 V + C 1 5 + 1 0 µF 2 5 V R 1 7 D 3 1 N 4 1 4 8 IC 3 b T L 0 8 2 – T o o th e r ch a n n e l R 1 4 C 6 3 9 n F R 1 3 R 6 C 3 3 3 n F R 4 T P 2 * V C A b ia s V R 1 * R 7 C 4 3 3 n F R 3 C 1 3 3 n F R 1 V + * M o u n t o n to p e d g e o f b o a r d f o r a ce s F ilte r o u tp u t l u e s sh o w n a r e f o r 8 6 0 H z a n d 1 . 2 5 kH z si n g l e ch a n n e l o n ly . V – Fig.21. Assembling all the building blocks to form a complete vocoder channel. We will build two channels per PCB – see Fig.15. 62 Practical Electronics | December | 2021 D r i ve b u s e s S p e e ch S yn M i xi n g b u se th L (o d d ) R s ( e ve n ) 20 Hz 50 100 200 500 1k 2k 5k 10k 20k 20 Hz 50 100 200 500 1k 2k 5k 10k 20k 20 Hz 50 100 200 500 1k 2k 5k 10k 20k 0dB –10 L P 1 3 0 H z H P 8 kH 1 1 5 0 H z –20 –30 z –40 0dB 2 –10 2 1 0 H z –20 3 3 0 0 H z 4 4 5 0 H z 5 6 5 0 H z 6 9 0 0 H z –30 –40 0dB –10 –20 –30 –40 1 . 3 kH z 8 1 . 8 kH z 9 2 . 6 kH z 1 0 3 . 8 kH z 1 1 5 kH Fig.23. The total frequency response of the stereo output of the vocoder is produced using the left channel (top) and right channel (middle) – here plotted on top of each other (bottom). The shaded area in the bottom plot highlights the difference between left and right channels. This produces a unique musical effect. 7 z amp. This amplifier will also drive output transformers. Summing up The outputs of the channels can either 1 2 6 . 8 kH z all be mixed together into mono or alterM i x r e si st o r s p o si t i o n natively a stereo mix se t o n b o a r d can be created. In Tim Orr’s vocoder Fig.22. I use a special mixing trick in all my vocoders to give designs, the phase of a stereo effect. The outputs of the channels are panned each alternate filter alternately to the left and right outputs apart from the lower was inverted bebands, which are mono to ensure an even power distribution fore being summed between left and right. to mono to give a smoother response. This is because there amplifiers is actually musically benefiis a phase cancellation where the final cial. I suggest leaving R25 out in your 12dB/oct filter slopes crossover. In the first build. design we are building I don’t do this, All the circuit elements described in the which gives a more spacious stereo at vocoder channel block diagram shown the expense of slightly duller sound in in Fig.4 (see last month) are assembled mono. The low-pass filter, the two lowin the complete circuit diagram for a sinest band-pass filters and the high-pass gle vocoder band-pass module in Fig.21. filter are fed into both output channels to maintain essential mono compatibility Driving in the bass. The remaining bandpass filThe total input impedance of all 14 filters are then fed alternately into the left ters is of the order of 130Ω, which is too and right channels. A block diagram is low for a single op amp to drive (one shown in Fig.22. Thus, the frequency section of a 5532 can drive 600Ω). This responses of the left and right channels means a small power amplifier or mulare opposite, where one goes down the tiple op amps are required. This can be other goes up, as shown in Fig.23. This achieved by adding a current-boosting gives a lovely stereo spread between the push-pull emitter follower stage to an op Practical Electronics | December | 2021 two speakers. This special mixing is accomplished in the driver PCBs. Next month Now that we have described the system, we’ll start constructing th driver amplifier PCB next month. This will be a useful module in its own right for feeding headphones, spring-lines, transformers and enabling amplifier bridging. The other boards will follow later. Acknowledgements An excellent reference on filter designs is The Filter Handbook by Stefan Niewiadomski (a previous contributor to Practical Electronics). Also, The Active Filter Cookbook by Don Lancaster. Vocoder principles were taught by senior lecturer Tim Orr at the London College of Furniture in 1985. Yes, I did study electronics at a furniture college! (now the London Metropolitan University). He wrote many articles in Electronics Today International magazine on electronics for music. These culminated in the ETI vocoder of September and October 1980, marketed as a kit for £175 + VAT by Powertran Electronics. He later expanded on the filter design in Bandpass and beyond (ETI, December 1980). This design was also based on a student HND project report from around 1982 by JM Hawkes. Where is he now, does anyone know? 63