I used to use high quality BBDs (clocked as fast as they'd go to keep the quality up) as part of an analogue signal processor. This allowed me to monitor and process a signal in real time a few milliseconds before it got to the listener. Using an analogue discriminator and comparator, it was possible to recognise a scratch on a vinyl record just before it got to the output and replace it with something else less offensive to the human ear. Another comparator tracked the average background noise level from the vinyl, so the system became fully automatic with no need for any twiddly knobs to adjust anything. This was back in the late 80s, and I can still remember to this day the satisfaction of designing it all from scratch (pardon the pun). It took a few weeks to iron the bugs out, but the finished project still works very well to this day.
This is a higher quality explanation and demonstration than you would likely find in a university. Really nice work. What I love about the BBD is that it blurs the line between digital and analog; Two things that most people consider to be sort of mutually exclusive. You get a quantization of time, but the amplitude is fully analog.
agreed - plus i love how simple it is, especially compared to something like a PT2399. BBDs really feel like a super precise solution to one specific problem!
Your channel hits the perfect sweet spot of "technicality"! (At least for me - I studied EE/CS, but since college, I had zero experience with circuits and forgot all the annoying transistor calculations) Still engaging, skipping on some of the unnecessary details and calculations, but not "dumbed down" and just perfectly enough to appreciate the beauty and smartness of those designs, explaining exactly what was challenging and how it was solved. :) And while this might not be enough to build such a circuit entirely from scratch without your designs, it's again perfectly enough of a starting point if someone wants to dig deeper.
Exactly, I teach analog electronics and digital signal processing at university and I'm always dumbstruck by Moritz didactic quality. I do recommend his videos to students and colleagues!
Superb video - this is quality content that makes UA-cam worthwhile. I've no desire to build my own BBD based delay, but learning how they work is fascinating.
I’ve always wanted to design a BBD with a variable output filter tied appropriately to the clock frequency. This is a great point to jump off from and an amazing general learning resource as always. Thank you for your continued work on increasingly complex projects!
@borututuforte I know right?! Too distracting I need just hands and components or I stop learning. . Jk Moritz. . Happy to see your smiling face. This made my day!
I don't know if anybody mentioned this already, but I am pleasantly surprised by the little doodles and graphics present on the front plate of this new module =) I noticed something similar on the panel of Labor already, and here it is once again with this new eurorack module. Little arrows and squiggles identical to the ones we see in the animations of your videos, Moritz. I really like them, hehe, they give the panels a personal touch without being distracting. Quite elegant too, I have to say. I would like to see more modules in the future come out with front panels featuring similar graphical decorations. Thumbs up from me 👍
As expected, another superb video from Moritz. I got just as much out of your knowledge of the PLL as I did from the BBD circuit approach. Fantastic. Thanks Moritz.
This is an AMAZING teaching video. I have basic electronic knowledge and understood everything. Even if you’re not looking at building a delay/echo system, there’s many basic electronics lessons contained in this video so it’s a good teaching lesson.
SO many instructional videos will just say "We won't do *x* because it causes problems." and move on. The way this video makes the problems happen and demonstrates why they're an issue before fixing them makes it such a great educational resource.
These videos are masterpieces not only of engineering but visualization and narration / explanation / education. Thank you. I bought a couple of bucket brigade chips to experiment with building a chorus effect pedal and was puzzled why the chip data sheet recommended use of a specific related timing IC. This explained why a dual clock source is required.
i think low output impedance on the clock generators is also important (cause the mosfet gates do pull in current when switching on). so that’s also why they made those special companion chips.
Great video! Thanks for making it! I’ve built a couple PT2399 delays, but that chip contains much of the external components needed here. I had wondered why BBD delays were so much more complicated. For anyone interested,look up the data sheet for the PT2399, it shows a very basic delay that really just needs a voltage regulator,a couple caps, couple resistors, and a couple pots. It’s a fun and easy circuit to play with.
the PT2399 is notably not a BBD but in a way it's sort of the logical next step. turn the signal into a binary stream using delta-sigma and then you can use a really long shift register and cut down on how accurate/big the capacitors and transistors of each "stage" need to be on the die.
@@famitory yeah, I'm quite aware of that, but it's the only real frame of reference I have to compare it to. And while it is digital, it was designed to emulate the sound of the old bucket brigade type sounds. A lot of the control options are pretty similar too, such as being able to drive the clock speed with a modulated source, the raising and lowering of pitch as the speed is adjusted up and down, and the gradual darkening of the signal. It also does some wonky distorted sounds when you try to make the delay length more than it's designed for, similar to what he did in the video.
YES! I've been asking for a BBD video several times, and right as I wanted to try my own hand at it, a wonderful Moritz Klein video appears to help clearly explain everything about it! Thank you for these videos (and this one in particular)!
The obvious thing to do is to use more BBD chips for longer delay at higher clock rates. The clock rate should be more than double the max signal frequency. Then add proper higher order filters to get rid of all unwanted noise. So adding more BBD devices will increase delay at higher sample rates without distortion. Also one can tap signal between BBD chips for more interesting echo/delay effects. CCD chips used in digital cameras work the same as BBD devices, but initial charge stored in each capacitor in line is determined by light level - it discharges them (IIRC). So process of getting the image data from the chip starts by clocking it out across the sensor to the ADC at the end of the line. In case of CCD chips propagation delay is a bad thing, causing horizontal tearing when recording fast movement.
Back in the late 70s and early 80s Radio Shack sold a BBD chip for projects. I built the standard one: voice actuated cassette recorder. When audio turned the cassette recorder on, the BBD gave the audio enough delay for the recorder to start recording so the beginning wasn't cut off. But the audio quality of the BBD was barely audio cassette quality. I recall it more like AM radio quality and I assumed that was the cost of using a bunch of capacitors to store audio. I was surprised to hear later on that BBD circuits could actually produce high quality audio.
@@MoritzKlein0 A first-order all-pass behaves like a delay in the low frequency limit, specifically a delay that's twice the RC-time, or 1/(π*f₀), but at higher frequencies its group delay decreases towards zero in a non-linear way: if T is its low-frequency group delay then at frequency f its group delay is T/( 1 + (π*f*T)² ). This means the group delay is still 99% of T up to f=0.032/T, but it's down to 90% of T at f=0.1/T, and 50% at f=1/T. The phase error is only 1͏.1° at f=0.1/T but then rapidly increases to 10͏° at f=0.22/T and 60͏° at f=0.48/T. If we pick the f=0.1/T limit and say we want that accuracy in the entire audio band up to 20 kHz, that means a first-order all-pass is a pretty decent delay up to.... 5 microseconds.
19:15 Neither my cats nor I appreciated this +8dB LU increase in loudness. ^^" (Perhaps 10dB lower than equivalent LUFS would be ideal for such a shrill sound.) Great video though - I had never thought about the implementation details of BBDs before.
I am building guitar pedals, not eurorack, but still you are the most helpful resouce currently available and an endless source of inspiration. I have watched all your videos several times and each time come away with some new, deeper insight. So thanks for making this available. Thats what i am trying to say.
The minute the question was posed I immediately said capacitors, though in my head I imagined an array of capicators that could duplicate charges at different heights of the wave. I’m not an electrical engineer just a guy that watches electrical engineering videos.
That acid pattern is beautiful AF!!!! I either need a full version of it or just the sequence written so I can make it for myself lol. Beautiful video too big up!
Holy Smokes! I have an old 1980s Ross flanger, which had a beautiful sound, but was noisy as heck and then recently died. For some reason, none of the VST (computer-based) flanger effects seem to achieve the beautiful destructive interference which my old flanger gave me, and I've been looking for a good replacement. Your circuit seems to be less noisy than my old unit, and sounds incredible. I'm going to look into the kit you spoke of. If I have any success, I'll seriously consider putting this into a good enclosure for my guitar.
An *excellent* tutorial. Personally, I would have pegged Moritz to be noticeably older than he appears to be. So, kudos on being such an excellent teacher at such a young age! BBDs contain the sorts of stages nicely illustrated here, but the capacitors they use to store charge for transfer, are also pretty small and leaky. That means that clock rates lower than a certain rate will result in enough leakage for the charge that gets transferred to be less and less like the original, the more stages it is passed through, and the more slowly it is passed from bucket to bucket. Ultimately, EVERY bucket brigade chip has a lower clock-rate limit. It also means that they do not and can not "store" samples the way that digital memory can. But there are also *upper* limits to clock rate for every BBD. The input pins for the complementary clock pulses have their input capacitance. That capacitance tends to impact on the shape of the incoming clock pulses, when they exceed that upper limit, such that the "handoff" between complementary sections is not as instantaneous as intended, degrading audio quality. The Reticon BBDs had clock pin capacitances that were a fraction of those used in the Panasonic/Matsushita chips (and Coolaudio clones of them), making them more suitable for instances where one was deliberately aiming for very high clock rates. The V3205 used here is a 2nd or 3rd-generation BBD. Earlier 30xx BBDs were engineered differently, and were aimed at higher supply voltages. The 32xx series will work with +5V. Why the difference? Remember that, in the earliest days of BBD-based effects/processing, while some effects had onboard transformers and a power cord, external power supplies were very much a rarity, and many effects pedals assumed battery operation. Since the DC bias voltage used to make the input to a BBD appropriate was taken from the supply, as the battery wore down, the bias voltage would change. The 32xx series allow the BBD to operate with supplies as low as +5V, even though the supply to the overall circuit might be higher. Regulating the voltage provided to the BBD down to +5V, and deriving the bias voltage from *that* meant that any pedal operating from a +9V battery would be able to function properly until the on-board battery wore down to around +7V. And by that point, it would probably be too weak to power the audio path and any LFO and electronic switching circuitry. Put another way, the 32xx series strikes me as really a solution to problems inherent to battery operation, and not an "improvement" to audio quality, per se.
in "reconstruction sampling" you are essentially adding a fixed delay to the clock rising edge, to produce another rising edge for your sampler. this is not relative to the delay frequency, and would have to be set relatively conservative, in order to fit within the fastest delay pulses, and the slowest, and seems very clunky, to me. instead, why not generate a multiple of your clock frequency, use a clock divider to get your bucketing pulse, and a counter to derive your reconstruction sample trigger? this way you can ensure the reconstruction sampling happens at the most optimal point, at all delay times. thank you for this video, i have always wanted to know how "analog" delays work.
Yes, using a counter to get phase-offset square waves makes a lot of sense. With a bit of logic you could also ensure that your square waves have a dead-time between them, in case shoot-through is a problem.
that would have been my preferred approach, but in the end i decided to go with “good enough” to keep the circuit as lean as possible. good call though!
@@MarcoGualtieri ideally you would take the clock, apply a 90 degree phase shift (delay it by 1/4 of a wavecycle), and then turn that into a trigger. alternatively, you could also use more comparators/buffers/logic gates to add more propagation delay, but this will break at very fast clock speeds.
I still have few Reticon SAD 4096 chips in my bin from ~1985. I used them to add some "ambiance" to my music. I should put them back to work (if they're still good) because the circuitry is simple to build.
Only halfway through, but this is looking remarkably like a synchronous dual sample+hold circuit I designed a while ago, designed to be a high-speed peak detector. Also, with a 4046 as your clock source, you could make a comb filter out of it and use it as a VCF.
Thanks so much for sharing your knowledge. Your videos really helped me get a grip on audio electronics and being able to breadboard stuff and modding some of my synths.
Me to. Back in the 70s when Bucket Brigade chips were invented, I saw an article in a DIY electronics magazine which used a BB chip to create a click/scratch cancellation circuit for vinyl records. It worked like this. The source from the phono cartridge preamp was fed to the input of the BB and the BB output went through a mute circuit and from there to the power amplifier. The source signal also went directly to a mute circuit and from there to the amplifier. Finally, the source signal went to a filter which could detect the leading edge of very fast loud scratches. Because a scratch has a very short transient time, if you have a delay line which is longer than the scratch, at the moment the scratch should emerge from the delay line, you simply switch off the delayed output and switch directly to the input source for a few milliseconds. A bit like using a DAW to cut and paste a tiny short blemish in a recording. If you simply muted the BB output, the user would hear the silence, but by switching to the source which because of its temporal proximity to the scratch, the output level and sound will be virtually identical to the tiny muted segment. The listener never notices. I wish I could find the circuit again as I would love to build it. Maybe you would take it on as a challenge.
Vibrating wires in large coils was also used as a delay. The bucket brigade and the vibrating wires are also used as computer memories. I worked with both many years ago, the wire in a military testing lab, the BBD when I worked in a manufacturer's R&D dept.
Cool ! I had a bbd it had mic and guitar inputs with volume knobs and the out ouput had a switch for different dB levels so you could use it like a pre amplifier. It had some mental sounds if you did too much feedback, it was quite hissy. I also had flanger and chrorus pedals before the days of digital delays. I had the first digital delay pedal as well when it came out.
This is great! Is this, using a S&H to eliminate clock noise, ever done before? I never seen it and experimented a lot with BBD’s. One obvious improvement would be using a MN3005 instead of the lower voltage type you’re using, which has better headroom. I’ll definitely are going to experiment with this concept!
erica synths did something similar for the Black BBD module, but afaik it was a lot more complex. and yeah agreed, a chip with more headroom would improve the signal-to-noise ratio for sure!
In 1980 or 1981 my uncle brought back from Japan one of those combo radio/cassette players. It had a switch with 3 positions : mono, stereo and "expand". If you listened to music with some headphones the "expand" switch really made the sound sound "expanded" as if the sources were more "separated" : I can't explain it better than this. It was quite pleasant and I've never seen it again in any other device. Could have just been a slight additional delay between left and right channel ?
Usually when sampling it is a good idea to add a low pass filter to the input to prevent aliasing. It probably isn’t that necessary in this circuit since wildly varying the sample rate is part of the fun and the aliasing adds to the glitchy sound.
@@andrewpensinger3586 i agree, i actually like the artifacts. but yeah, if you’re going for fidelity, you need to filter, compress and then filter and expand at the output :)
Your videos are masterpieces! But i think you got one little thing wrong: you animate the electrons for example comming from gnd in the cap and traveling "through" the cap. Capacitors interrupt the circuit, so they can't flow through the caps (in DC).
26:10 min: “I won’t pretend I understand how this works in detail.” This applies to most information you provide but I am at least understanding the main principles. Charging a row of capacitors with the previous ones charge, eh? No wonder they call it a bucket brigade. But the biggest surprise is that BBD’s work like analog samplers by “Digitising” waveforms. That will scare a lot of analog effect fans. 🤣 One is never to old to learn. Thanks! 👍👍👍
Wow, man, I just discovered that website on the root of that CircuitJS simulator, and It's pure gold !! tons and tons of simulations of the barebones maths formulas describing the most fundamental concepts with acoustics sound waves, adding up to some complex simulations of sound modeling and synthesis!! down the rabbit hole we go :D By the way is also yours, that website?
Just started to watch the video when the flu and cvd shots kicked in and forced to bed, but this is absolutely intriguing. The PT2399, mentioned in the comments is a cool thing and small boards can be bought from China for a few bucks. There's one PT2399 in the MIDI Ultimate from Soundtronic and I will try to squeeze one into Ray Wilson's microsynth, the Noise Toaster for my grandson. Else I have a pair of TDA1022 and am looking for a suitable project for these. Guess this video will shine some light when I am back on my feet.
Ancient tech. Personally I prefer to play with ICs like PT2399, which record and playback a signal from a 44K digital ring buffer, giving a delay from 30ms to a few seconds, without too much distortion.
The concept of a delaying a signal by a bucket brigade arrangement of capacitors predates 1955. I have a copy of the book "A Palimpsest on the Electronic Analog Art" which includes this. It is true that audio applications - and chips - came much later.
@@MoritzKlein0 It is a very strange though beautiful delay sound, the degradation is lovely. I have made delay selectable between two BBD chips - however At first I had no audible clock, now I have audible clock at certain delay frequencies - Still de-bugging this rather annoying occurrence. Something has changed in the filtering it seems, it's difficult to track down : )
@@MoritzKlein0 ha ha! Thank you. I will implement your S/H technique as I have another BBD to play with. I believe you have revolutionised the way of working with, and results from, these chips. Well done! Next I will experiment with the companding technique, this improves the S/N ratio a bit. It's described by electric druid and electrosmash
Moritz, so excited for this video! Can I use a BS170 for the switching FETs if I want to build the discrete BBD? I bought a bunch for exactly this type of project, based on other advice, but you mentioned that discrete MOSFETs can't handle the reverse current.
afaik, any mosfet with three terminals will allow current to flow from source to drain regardless of gate voltage. (you can’t turn it off in that direction.) i’m afraid you’ll have to invest in some JFETs 🥲
Awesome video, and a very interesting reconstruction method! I understood at 28:40 why traditional low pass capacitor was removed (to preserve high freqencies), but why dual tap from the second output was removed? Did it provide any downsides by itself?
Great explanation,! You mention how discrete MOSFET's aren't suitable (and I surmise that's because source and substrate are shorted together). Now I'm starting to sketch out whether your demonstrator can be replicated with the 4007, where at least one of the CMOS pairs has isolated sources. (If not, a 4053 would surely work for three stages of the delay.) I didn't realize that the V3205 had cascodes between the stages, but it makes sense: isolate the voltages between the two stages while passing the current. The intermediate MOSFET functions as a current follower - analogous to the way a source follower duplicates its input voltage at a much lower impedance, a common-gate amplifier duplicates the input current at a much higher impedance. (One's Thévenin, the other is Norton.) I also once saw an ancient design that used a long digital shift register - the sort that used to be used to circuiate the characters on an old non-graphical display - together with a delta/sigma modulator: a digital delay with a 1-bit A-D/D-A. I'll have to see if I can dig up the ancient reference. In the end, I want to make a digital one - probably using an STM32 and an I2S chip - so that I can get a very wide range of delays. (I want to experiment with plucked-string and bell effects using a re-entrant delay line as the oscillator.) That's a longer-term project, I don't have anything to share just yet. The analog folks will hate me for it.
If I would have to desing a "delay" circuit I would try to do it like this: Digitize the analog signal with an ADC that has parallel outputs (so one output pin for each bit). Connect each ADC output pin to a lot of cascaded lowpasses (each one has a voltage follower after it). Then put that "delayed" digital signal into an DAC and you should get a delayed signal.
@@MoritzKlein0 Ugh, that would be difficult. Maybe "precharge" the capacitors so they don't take that lonkg to reach a "lgic high" (subtract the voltage at the end because otherwise it would never reach logic 0)
I used Boss analog delay in the 80s which used sad512 chip if I remember correctly. and there was audible high pitch. Your design seems to be better than that.
I used to use high quality BBDs (clocked as fast as they'd go to keep the quality up) as part of an analogue signal processor. This allowed me to monitor and process a signal in real time a few milliseconds before it got to the listener. Using an analogue discriminator and comparator, it was possible to recognise a scratch on a vinyl record just before it got to the output and replace it with something else less offensive to the human ear. Another comparator tracked the average background noise level from the vinyl, so the system became fully automatic with no need for any twiddly knobs to adjust anything.
This was back in the late 80s, and I can still remember to this day the satisfaction of designing it all from scratch (pardon the pun). It took a few weeks to iron the bugs out, but the finished project still works very well to this day.
Some BBD IC? Just curious
@@MrSlipstreem sounds like a really fun problem to solve. what did you replace the unwanted sounds with?
@@MoritzKlein0He probably used anything other than the tone in the video at #19:20 😉
Probably the best explanation of how a BBD works.
This is a higher quality explanation and demonstration than you would likely find in a university. Really nice work.
What I love about the BBD is that it blurs the line between digital and analog; Two things that most people consider to be sort of mutually exclusive. You get a quantization of time, but the amplitude is fully analog.
agreed - plus i love how simple it is, especially compared to something like a PT2399. BBDs really feel like a super precise solution to one specific problem!
Yeah. First it is analog, then discretized, and lastly digitized. This chip/circuit just omits the last step.
Lost me at dry/wet mixer 😮
Wait this isn't a fully analog circuit? Why?
I feel like this video should count for credit towards an electrical engineering course... great work!!
Your channel hits the perfect sweet spot of "technicality"! (At least for me - I studied EE/CS, but since college, I had zero experience with circuits and forgot all the annoying transistor calculations)
Still engaging, skipping on some of the unnecessary details and calculations, but not "dumbed down" and just perfectly enough to appreciate the beauty and smartness of those designs, explaining exactly what was challenging and how it was solved. :)
And while this might not be enough to build such a circuit entirely from scratch without your designs, it's again perfectly enough of a starting point if someone wants to dig deeper.
that’s exactly the balance i’m trying to hit - glad to hear it works for you!
Exactly, I teach analog electronics and digital signal processing at university and I'm always dumbstruck by Moritz didactic quality. I do recommend his videos to students and colleagues!
Superb video - this is quality content that makes UA-cam worthwhile. I've no desire to build my own BBD based delay, but learning how they work is fascinating.
@@chriswareham glad to hear :)
dammit I'm the other side of nerd, now there's a youtube video let's 'ave it, try and build one as well!! :P
@@MouldySoul that’s the spirit!
I’ve always wanted to design a BBD with a variable output filter tied appropriately to the clock frequency. This is a great point to jump off from and an amazing general learning resource as always. Thank you for your continued work on increasingly complex projects!
good luck with that project, sounds like a fun one!
Sampling the signal again at the BBD's output is genius! I also love the creative front panel design. Amazing video and amazing kit as always :)
@@taidi4038 glad to hear you like the front panel design - thought it’s time the modules get some visual spice :)
Moritz Klein face reveal!??
No he has shown his face multiple times in live streams
hottie
I almost spat out my coffee
(even though I've seen his face before on his ig)
@borututuforte I know right?! Too distracting I need just hands and components or I stop learning. .
Jk Moritz. . Happy to see your smiling face. This made my day!
I don't know if anybody mentioned this already, but I am pleasantly surprised by the little doodles and graphics present on the front plate of this new module =)
I noticed something similar on the panel of Labor already, and here it is once again with this new eurorack module. Little arrows and squiggles identical to the ones we see in the animations of your videos, Moritz. I really like them, hehe, they give the panels a personal touch without being distracting. Quite elegant too, I have to say.
I would like to see more modules in the future come out with front panels featuring similar graphical decorations. Thumbs up from me 👍
@@dr.getter7118 glad to hear, that was exactly the intention! and i do want to keep adding these to upcoming modules :)
As expected, another superb video from Moritz. I got just as much out of your knowledge of the PLL as I did from the BBD circuit approach. Fantastic. Thanks Moritz.
that PLL chip is seriously feature packed :)
This is an AMAZING teaching video. I have basic electronic knowledge and understood everything. Even if you’re not looking at building a delay/echo system, there’s many basic electronics lessons contained in this video so it’s a good teaching lesson.
SO many instructional videos will just say "We won't do *x* because it causes problems." and move on. The way this video makes the problems happen and demonstrates why they're an issue before fixing them makes it such a great educational resource.
glad to hear, that’s exactly what i’m hoping for :)
These videos are masterpieces not only of engineering but visualization and narration / explanation / education. Thank you. I bought a couple of bucket brigade chips to experiment with building a chorus effect pedal and was puzzled why the chip data sheet recommended use of a specific related timing IC. This explained why a dual clock source is required.
i think low output impedance on the clock generators is also important (cause the mosfet gates do pull in current when switching on). so that’s also why they made those special companion chips.
As a gear nerd who realised in the mid 80’s that not all BBD delay pedals were created equal, this video is fascinating.
Great video! Thanks for making it! I’ve built a couple PT2399 delays, but that chip contains much of the external components needed here. I had wondered why BBD delays were so much more complicated. For anyone interested,look up the data sheet for the PT2399, it shows a very basic delay that really just needs a voltage regulator,a couple caps, couple resistors, and a couple pots. It’s a fun and easy circuit to play with.
the pt2399 is extremely complex compared to a BBD chip - that’s why the driving circuit can be so much simpler :)
the PT2399 is notably not a BBD but in a way it's sort of the logical next step. turn the signal into a binary stream using delta-sigma and then you can use a really long shift register and cut down on how accurate/big the capacitors and transistors of each "stage" need to be on the die.
@@famitory yeah, I'm quite aware of that, but it's the only real frame of reference I have to compare it to. And while it is digital, it was designed to emulate the sound of the old bucket brigade type sounds. A lot of the control options are pretty similar too, such as being able to drive the clock speed with a modulated source, the raising and lowering of pitch as the speed is adjusted up and down, and the gradual darkening of the signal. It also does some wonky distorted sounds when you try to make the delay length more than it's designed for, similar to what he did in the video.
@@MoritzKlein0 exactly! That's why I had looked at BBD circuits in the past and had no idea what was going on 😁
Awesome video! Perfect amount of in-depth explanation yet keeping things easy to understand and consume. Great job.
One the best channels on YT.
Your videos are such a huge inspo for me as a synth DIY geek. Amazing as always!
Love this! Great explanation of how everything works, why everything works and in a lot of cases why something DOESN'T work.
Great Job Moritz. And thank you for your videos that manage to inspire even the most experienced sdiy nerds, like me. big up for your work🎉
Best BBD explanation I ever seen! Thank you!
This is just amazing. The video came out super clear and ultra interesting. By FAR the best one you've uploaded, keep doing this please! Thank you!!!
@@lucanotreally314 really glad to hear :)
I built a PAIA “Phlanger” in the late 70s with a bbd. It was quite effective.
YES! I've been asking for a BBD video several times, and right as I wanted to try my own hand at it, a wonderful Moritz Klein video appears to help clearly explain everything about it! Thank you for these videos (and this one in particular)!
perfect timing - hope the video will help!
The obvious thing to do is to use more BBD chips for longer delay at higher clock rates. The clock rate should be more than double the max signal frequency. Then add proper higher order filters to get rid of all unwanted noise. So adding more BBD devices will increase delay at higher sample rates without distortion. Also one can tap signal between BBD chips for more interesting echo/delay effects.
CCD chips used in digital cameras work the same as BBD devices, but initial charge stored in each capacitor in line is determined by light level - it discharges them (IIRC). So process of getting the image data from the chip starts by clocking it out across the sensor to the ADC at the end of the line. In case of CCD chips propagation delay is a bad thing, causing horizontal tearing when recording fast movement.
Ah, then instead of modulating the clock rate, you can just manufacture more BBDs … no, wait, that's software thinking.
Pitch...swing ... CRACK and OUT OF THE PARK!!! Another brilliant video and teaching session!!! Well done.
thank you :)
absolutely amazing, i always wondered about BBD and this explains it in the best way i can understand. thank you for making this!
glad the video was helpful :)
Back in the late 70s and early 80s Radio Shack sold a BBD chip for projects. I built the standard one: voice actuated cassette recorder. When audio turned the cassette recorder on, the BBD gave the audio enough delay for the recorder to start recording so the beginning wasn't cut off. But the audio quality of the BBD was barely audio cassette quality. I recall it more like AM radio quality and I assumed that was the cost of using a bunch of capacitors to store audio. I was surprised to hear later on that BBD circuits could actually produce high quality audio.
that's a really interesting use case. maybe they applied really heavy filtering to combat the sampling artifacts and clock noise?
There’s also always the all-pass filter for delays. It’s a phase shift, but you can build up longer delays by adding them together.
but afaik it affects different frequencies differently, right? which makes it more of a specialized tool
@@MoritzKlein0 A first-order all-pass behaves like a delay in the low frequency limit, specifically a delay that's twice the RC-time, or 1/(π*f₀), but at higher frequencies its group delay decreases towards zero in a non-linear way: if T is its low-frequency group delay then at frequency f its group delay is T/( 1 + (π*f*T)² ). This means the group delay is still 99% of T up to f=0.032/T, but it's down to 90% of T at f=0.1/T, and 50% at f=1/T. The phase error is only 1͏.1° at f=0.1/T but then rapidly increases to 10͏° at f=0.22/T and 60͏° at f=0.48/T.
If we pick the f=0.1/T limit and say we want that accuracy in the entire audio band up to 20 kHz, that means a first-order all-pass is a pretty decent delay up to.... 5 microseconds.
19:15 Neither my cats nor I appreciated this +8dB LU increase in loudness. ^^" (Perhaps 10dB lower than equivalent LUFS would be ideal for such a shrill sound.)
Great video though - I had never thought about the implementation details of BBDs before.
i am very sorry 🥲
Much respect to the guy who grows a moustache just to prove to his buddies he can grow a moustache.
Great tutorial
Feed the UA-cam algorithm with a comment. Great video. Thank you!
NEW MORITZ I CANT BELIEVE IT
OH HAPPY DAY!!!!!!!!!
loved it, thanks!!!!
What an epic overview! Well done!
Keep up the great content, Moritz Klein!👍
@@sjay4673 will do 🙏
great video my guy
@@iLightSoundGeometry thank you my guy
I am building guitar pedals, not eurorack, but still you are the most helpful resouce currently available and an endless source of inspiration. I have watched all your videos several times and each time come away with some new, deeper insight.
So thanks for making this available. Thats what i am trying to say.
Pretty cool how the BBD chip is just an analog memory chip.
@@poptartmcjelly7054 really bad memory because of capacitor leakage - but yeah :)
@@MoritzKlein0DRAM is even worse, which is why it needs constant refresh cycles.
The minute the question was posed I immediately said capacitors, though in my head I imagined an array of capicators that could duplicate charges at different heights of the wave. I’m not an electrical engineer just a guy that watches electrical engineering videos.
That is an excellent explanaition of the CD4046 PLL chip. truely excellent!
Hey, that's some really cool prototyping setup :)
Nice material on BBDs, I learned a few things. Thanks!
i am extremely happy to not be working on a normal breadboard anymore 🥲
For some reason I somehow knew how you look yet I've never seen you before. I imagined you very much like this. Good to see you. And great video too
That acid pattern is beautiful AF!!!! I either need a full version of it or just the sequence written so I can make it for myself lol. Beautiful video too big up!
Holy Smokes! I have an old 1980s Ross flanger, which had a beautiful sound, but was noisy as heck and then recently died. For some reason, none of the VST (computer-based) flanger effects seem to achieve the beautiful destructive interference which my old flanger gave me, and I've been looking for a good replacement.
Your circuit seems to be less noisy than my old unit, and sounds incredible.
I'm going to look into the kit you spoke of. If I have any success, I'll seriously consider putting this into a good enclosure for my guitar.
RIP to your flanger 🥲
Awesome work again Man.. Your way of explaining things is on point. !!
These BBDs were a huge breakthrough. I have a book that tells you how to make most effects with BBDs, not just delay. Flanger, chorus, etc.
what the name of the book ?
@@darmstard Wouldn't help you. It's long out of print and not in English.
@@CristiNeagu thank you anyway
just out of curiosity , whats the name of the book ?😅
@@CristiNeagu ok. what's the name of the book tho?
Great video, thank you. Now i feel like i want to build one (once i make some room in my projects backlog)
An *excellent* tutorial. Personally, I would have pegged Moritz to be noticeably older than he appears to be. So, kudos on being such an excellent teacher at such a young age!
BBDs contain the sorts of stages nicely illustrated here, but the capacitors they use to store charge for transfer, are also pretty small and leaky. That means that clock rates lower than a certain rate will result in enough leakage for the charge that gets transferred to be less and less like the original, the more stages it is passed through, and the more slowly it is passed from bucket to bucket. Ultimately, EVERY bucket brigade chip has a lower clock-rate limit. It also means that they do not and can not "store" samples the way that digital memory can.
But there are also *upper* limits to clock rate for every BBD. The input pins for the complementary clock pulses have their input capacitance. That capacitance tends to impact on the shape of the incoming clock pulses, when they exceed that upper limit, such that the "handoff" between complementary sections is not as instantaneous as intended, degrading audio quality. The Reticon BBDs had clock pin capacitances that were a fraction of those used in the Panasonic/Matsushita chips (and Coolaudio clones of them), making them more suitable for instances where one was deliberately aiming for very high clock rates.
The V3205 used here is a 2nd or 3rd-generation BBD. Earlier 30xx BBDs were engineered differently, and were aimed at higher supply voltages. The 32xx series will work with +5V. Why the difference? Remember that, in the earliest days of BBD-based effects/processing, while some effects had onboard transformers and a power cord, external power supplies were very much a rarity, and many effects pedals assumed battery operation. Since the DC bias voltage used to make the input to a BBD appropriate was taken from the supply, as the battery wore down, the bias voltage would change. The 32xx series allow the BBD to operate with supplies as low as +5V, even though the supply to the overall circuit might be higher. Regulating the voltage provided to the BBD down to +5V, and deriving the bias voltage from *that* meant that any pedal operating from a +9V battery would be able to function properly until the on-board battery wore down to around +7V. And by that point, it would probably be too weak to power the audio path and any LFO and electronic switching circuitry. Put another way, the 32xx series strikes me as really a solution to problems inherent to battery operation, and not an "improvement" to audio quality, per se.
in "reconstruction sampling" you are essentially adding a fixed delay to the clock rising edge, to produce another rising edge for your sampler. this is not relative to the delay frequency, and would have to be set relatively conservative, in order to fit within the fastest delay pulses, and the slowest, and seems very clunky, to me. instead, why not generate a multiple of your clock frequency, use a clock divider to get your bucketing pulse, and a counter to derive your reconstruction sample trigger? this way you can ensure the reconstruction sampling happens at the most optimal point, at all delay times. thank you for this video, i have always wanted to know how "analog" delays work.
Yes, using a counter to get phase-offset square waves makes a lot of sense. With a bit of logic you could also ensure that your square waves have a dead-time between them, in case shoot-through is a problem.
that would have been my preferred approach, but in the end i decided to go with “good enough” to keep the circuit as lean as possible. good call though!
@@MoritzKlein0is there a way of sampling it just before the end of the exponential decay, rather than just after it begins?
@@MarcoGualtieri ideally you would take the clock, apply a 90 degree phase shift (delay it by 1/4 of a wavecycle), and then turn that into a trigger. alternatively, you could also use more comparators/buffers/logic gates to add more propagation delay, but this will break at very fast clock speeds.
Once again, thanks a lot.
I still have few Reticon SAD 4096 chips in my bin from ~1985. I used them to add some "ambiance" to my music. I should put them back to work (if they're still good) because the circuitry is simple to build.
that does not sound SAD at all!
@@MoritzKlein0 Well, I probably missed something but, that's the marking printed on the chips. :)
@@Bob-1802 it was just a bad joke 🥲
Only halfway through, but this is looking remarkably like a synchronous dual sample+hold circuit I designed a while ago, designed to be a high-speed peak detector.
Also, with a 4046 as your clock source, you could make a comb filter out of it and use it as a VCF.
Dude this is so fascinating and explained so clearly!! Very very cool!!
thank you :)
Amber, nice album
These animations are gorgeous. I’d love a quick explainer video showing how you built them
good idea! might do a YT short about this
Thanks so much for sharing your knowledge. Your videos really helped me get a grip on audio electronics and being able to breadboard stuff and modding some of my synths.
BBD devices! FINALLY! Thank you so much!
Me to. Back in the 70s when Bucket Brigade chips were invented, I saw an article in a DIY electronics magazine which used a BB chip to create a click/scratch cancellation circuit for vinyl records. It worked like this. The source from the phono cartridge preamp was fed to the input of the BB and the BB output went through a mute circuit and from there to the power amplifier. The source signal also went directly to a mute circuit and from there to the amplifier. Finally, the source signal went to a filter which could detect the leading edge of very fast loud scratches. Because a scratch has a very short transient time, if you have a delay line which is longer than the scratch, at the moment the scratch should emerge from the delay line, you simply switch off the delayed output and switch directly to the input source for a few milliseconds. A bit like using a DAW to cut and paste a tiny short blemish in a recording. If you simply muted the BB output, the user would hear the silence, but by switching to the source which because of its temporal proximity to the scratch, the output level and sound will be virtually identical to the tiny muted segment. The listener never notices. I wish I could find the circuit again as I would love to build it. Maybe you would take it on as a challenge.
Как же это круто! Жаль нет столько времени чтобы разобраться во всех крутых вещах. И хорошо что есть те кто разобрался и делится знаниями. Спасибо.
Vibrating wires in large coils was also used as a delay. The bucket brigade and the vibrating wires are also used as computer memories. I worked with both many years ago, the wire in a military testing lab, the BBD when I worked in a manufacturer's R&D dept.
Fantastic explanation thank you for sharing.
Amazing!!
Cool ! I had a bbd it had mic and guitar inputs with volume knobs and the out ouput had a switch for different dB levels so you could use it like a pre amplifier. It had some mental sounds if you did too much feedback, it was quite hissy. I also had flanger and chrorus pedals before the days of digital delays. I had the first digital delay pedal as well when it came out.
This is great! Is this, using a S&H to eliminate clock noise, ever done before? I never seen it and experimented a lot with BBD’s. One obvious improvement would be using a MN3005 instead of the lower voltage type you’re using, which has better headroom.
I’ll definitely are going to experiment with this concept!
erica synths did something similar for the Black BBD module, but afaik it was a lot more complex. and yeah agreed, a chip with more headroom would improve the signal-to-noise ratio for sure!
I need such video for each IO instead of (in addition to) datasheet. It is so easy to get the idea of what each pin does.
sure, will take a couple years though 🥲
Just outstanding ❤
❤
thanks so much for these videos !
I just recently ordered Amber on vinyl, nice choice to have at the front of the stack :)
Great. May be you can try to do something interesting with AD538 ACU chip?
never heard of that chip. looks interesting. will check it out!
Dude, you are an excellent teacher 👌
You're a treasure! Thanks for all the lessons
This was great! Time to order some bbds!
This is gold.
Thank you thank you thank you for this video
Brilliantly explained. Thank you for such a great video. 👍😀
In 1980 or 1981 my uncle brought back from Japan one of those combo radio/cassette players. It had a switch with 3 positions : mono, stereo and "expand". If you listened to music with some headphones the "expand" switch really made the sound sound "expanded" as if the sources were more "separated" : I can't explain it better than this. It was quite pleasant and I've never seen it again in any other device.
Could have just been a slight additional delay between left and right channel ?
could be! reminds me of the “spatial audio” that you have on the airpods now - it also makes the sound more wide.
It was called "stereo wide" in the UK. Some boom boxes had it 😊
Usually when sampling it is a good idea to add a low pass filter to the input to prevent aliasing. It probably isn’t that necessary in this circuit since wildly varying the sample rate is part of the fun and the aliasing adds to the glitchy sound.
@@andrewpensinger3586 i agree, i actually like the artifacts. but yeah, if you’re going for fidelity, you need to filter, compress and then filter and expand at the output :)
Your videos are masterpieces! But i think you got one little thing wrong: you animate the electrons for example comming from gnd in the cap and traveling "through" the cap. Capacitors interrupt the circuit, so they can't flow through the caps (in DC).
that’s meant to convey that charge is flowing into one side (plate) of the capacitor and accumulating there!
26:10 min: “I won’t pretend I understand how this works in detail.” This applies to most information you provide but I am at least understanding the main principles. Charging a row of capacitors with the previous ones charge, eh? No wonder they call it a bucket brigade. But the biggest surprise is that BBD’s work like analog samplers by “Digitising” waveforms. That will scare a lot of analog effect fans. 🤣 One is never to old to learn. Thanks! 👍👍👍
@@marcbrasse747 yeah, it’s basically a stopgap between the worlds of analog and digital :)
Wow, man, I just discovered that website on the root of that CircuitJS simulator, and It's pure gold !! tons and tons of simulations of the barebones maths formulas describing the most fundamental concepts with acoustics sound waves, adding up to some complex simulations of sound modeling and synthesis!! down the rabbit hole we go :D
By the way is also yours, that website?
nah, i wish! it's by a guy called Paul Falstad afaik.
Very Informative! ✨
Nice Amber showcase ❤ Æ
Can we now replace all the broken BBD Chips in our old delay units? :-)
a bunch of them a being produced as clones now (and i think panasonic might be making them again as well?) - so maybe!
Great Video in any aspect 👍
heyy, i like your shirt! portrayal of guilt is great. also, thanks for publishing these
devil music is 🔥
Should not have taken acid before watching this...Amazing!
@@dirtypauwz2109 yeah that seems like a lot
😅
Just started to watch the video when the flu and cvd shots kicked in and forced to bed, but this is absolutely intriguing. The PT2399, mentioned in the comments is a cool thing and small boards can be bought from China for a few bucks. There's one PT2399 in the MIDI Ultimate from Soundtronic and I will try to squeeze one into Ray Wilson's microsynth, the Noise Toaster for my grandson.
Else I have a pair of TDA1022 and am looking for a suitable project for these. Guess this video will shine some light when I am back on my feet.
hope you feel better soon!
Ancient tech. Personally I prefer to play with ICs like PT2399, which record and playback a signal from a 44K digital ring buffer, giving a delay from 30ms to a few seconds, without too much distortion.
Answer me this, if I have a nice mustache, will I, too, be cool enough to make pedal effects?
@@XDjUanZInHO it is highly likely
Seems unlikely
The Vgg thing forms a cascade amplifier, which improves overall transconductance and reduces capacitive coupling of clock pulses.
i read it’s to combat the miller effect - might be the same thing!
@@MoritzKlein0 Yup, miller effect for high freq response, cascading and also helps stabilize operating point for a better gm.
The concept of a delaying a signal by a bucket brigade arrangement of capacitors predates 1955. I have a copy of the book "A Palimpsest on the Electronic Analog Art" which includes this. It is true that audio applications - and chips - came much later.
Thanks, I've been messing about modding a BBD for a while now, it's nice to read exactly how it works, and why two (one inverse) clocks are needed!
@@robinsutcliffe-video_art that mystified me too when i started looking into it. glad to hear the video is helpful!
@@MoritzKlein0 It is a very strange though beautiful delay sound, the degradation is lovely. I have made delay selectable between two BBD chips - however
At first I had no audible clock, now I have audible clock at certain delay frequencies -
Still de-bugging this rather annoying occurrence.
Something has changed in the filtering it seems, it's difficult to track down : )
@@robinsutcliffe-video_art godspeed :)
@@MoritzKlein0 ha ha! Thank you. I will implement your S/H technique as I have another BBD to play with.
I believe you have revolutionised the way of working with, and results from, these chips. Well done!
Next I will experiment with the companding technique, this improves the S/N ratio a bit. It's described by electric druid and electrosmash
Moritz, so excited for this video! Can I use a BS170 for the switching FETs if I want to build the discrete BBD? I bought a bunch for exactly this type of project, based on other advice, but you mentioned that discrete MOSFETs can't handle the reverse current.
afaik, any mosfet with three terminals will allow current to flow from source to drain regardless of gate voltage. (you can’t turn it off in that direction.) i’m afraid you’ll have to invest in some JFETs 🥲
Epic work! Thank you!
Awesome video, and a very interesting reconstruction method! I understood at 28:40 why traditional low pass capacitor was removed (to preserve high freqencies), but why dual tap from the second output was removed? Did it provide any downsides by itself?
simply because it’s not necessary - it doesn’t add much, since the sample-able region is big enough as is!
Latency as a feature ^^
i guess you could just use a livestream as an alternative ⚡️
Great explanation,!
You mention how discrete MOSFET's aren't suitable (and I surmise that's because source and substrate are shorted together). Now I'm starting to sketch out whether your demonstrator can be replicated with the 4007, where at least one of the CMOS pairs has isolated sources. (If not, a 4053 would surely work for three stages of the delay.)
I didn't realize that the V3205 had cascodes between the stages, but it makes sense: isolate the voltages between the two stages while passing the current. The intermediate MOSFET functions as a current follower - analogous to the way a source follower duplicates its input voltage at a much lower impedance, a common-gate amplifier duplicates the input current at a much higher impedance. (One's Thévenin, the other is Norton.)
I also once saw an ancient design that used a long digital shift register - the sort that used to be used to circuiate the characters on an old non-graphical display - together with a delta/sigma modulator: a digital delay with a 1-bit A-D/D-A. I'll have to see if I can dig up the ancient reference.
In the end, I want to make a digital one - probably using an STM32 and an I2S chip - so that I can get a very wide range of delays. (I want to experiment with plucked-string and bell effects using a re-entrant delay line as the oscillator.) That's a longer-term project, I don't have anything to share just yet. The analog folks will hate me for it.
Great video, very well explained, thank you!
If I would have to desing a "delay" circuit I would try to do it like this:
Digitize the analog signal with an ADC that has parallel outputs (so one output pin for each bit).
Connect each ADC output pin to a lot of cascaded lowpasses (each one has a voltage follower after it).
Then put that "delayed" digital signal into an DAC and you should get a delayed signal.
cool idea. how would you modulate the delay time?
@@MoritzKlein0 Ugh, that would be difficult.
Maybe "precharge" the capacitors so they don't take that lonkg to reach a "lgic high" (subtract the voltage at the end because otherwise it would never reach logic 0)
I used Boss analog delay in the 80s which used sad512 chip if I remember correctly. and there was audible high pitch. Your design seems to be better than that.
yeah many BBD designs have clock whine in them - it seems to have been widely accepted as a trade-off