[2347] BBDs are not analog

Date: January 5th, 2018 | Comments : [10] | Categories: DIY.

A bucket-brigade-device (BBD) is not analog. It is a sampling circuit that takes small slices of the incoming voltage just like an analog-to-digital-converter (ADC), and the main difference between the two methods is in how the samples (slices) are stored. The BBD samples are stored as voltage levels on tiny capacitors in the BBD chip, while the digital system converts the voltage levels to a digital measurement and stores the binary word(s) in memory.

Both systems have common problems of Nyquist sampling rates, aliasing and bandwidth limits.

After a time delay, the binary words of a digital system can be retrieved from memory, then sent to a digital-to-analog-converter (DAC) and restored to an analog voltage that is exactly the same as when it was first sampled. However, the BBD generates delay by passing the sampled voltage through a series of capacitors and mosfet switches before it goes to the output. There are a thousand or more of these capacitor/transistor cells in series inside a BBD, and small losses are incurred while the voltage is being held, and as it is being transferred down the chain of capacitors.

If the capacitors and transistors of a BBD were perfectly uniform, the output voltage would be equal to the input level, exactly as with a digital capture. But they are are not perfect and the small voltage losses from the tiny capacitors result in distortion of the reconstructed output voltage in the BBD.

The MN3007 1024-stage BBD chip has 0.5% distortion from the chip alone. It is also limited to a signal-to-noise ratio (SNR) of less than 80db, which is about equivalent to 13 bits of digital conversion. Approximate SNR = 6.02*n where n is the number of bits. (Math!!!) The larger BBD MN3005 4096-stage chip has 1.0% distortion typically, though it could be as high as 2.5%, and its SNR is less than 75db (slightly over 12-bit equivalent). Real world measurements yield lower SNR for BBDs than the datasheets claim.

Conversely, undithered 16-bit digital captures will have quantization error distortion spikes, but they are lower than -110db and certainly not audible in guitar pedals. Furthermore, the software in the digital signal processor can dither the output signal and eliminate the distortion spikes.

The nearly infinite resolution of the analog sample capture by the BBD will make no difference. In comparison, a 16-bit capture of a 1 volt full-scale signal represents only 0.00001526 volt for each bit! No one can hear the difference that might occur between an analog signal at 0.50001234 volt and a rounded digital conversion at 0.50000000 volt, especially after the effect of the reconstruction filter.

The higher distortion and lower SNR of BBDs are the main differences in sound. It would seem that adding some mild distortion (0.5% to 2.5%) and reducing the bit depth to 12 or 13 bits would be the key to getting an equivalent BBD sound with a digital pedal. It is a start but not the whole story. There are clock feedthrough, nonlinearities, aliasing distortion and other effects to be considered too, including the compression/expansion that is used in many BBD-based delay circuits.

There is definitely a difference between the digital and BBD circuits, but neither is strictly analog since they are both working with voltage samples rather than the continuously varying analog voltage.

You can read my follow up post called How BBDs Sample Audio.


10 Responses to “BBDs are not analog”

[782140] xolotl studio Says: 1:14 pm, January 12th, 2018

according to common electronic definitions found in almost all textbooks,BBD are analog since they do not quantize amplitude, but they are not continuous-time device, so they are subjected to aliasing problems. some confusion may arise since normally we are used to deal with digital devices, which are both discrete-time and digital (that is: they quantize amplitude) and continuous time analog devices, like an analog synth.

[782142] admin Says: 3:52 pm, January 12th, 2018

Analog audio is a continuously varying voltage. When you start to slice it up into samples, it is no longer continuous. It is discrete time slices used as a sample of the original signal, and that is not analog.

The chunks are stored in memory for the time shifting but whether it is stored in silicon memory or in a capacitor is immaterial. It is still a voltage sample that it delayed in time and then reconstructed at the output.

Best regards, Jack

[783607] admin Says: 6:40 am, March 4th, 2018

Practical Modeling of Bucket Brigade Devices by Colin Raffel and Julius Smith:


[784188] Brad Says: 10:00 pm, March 17th, 2018

It’s analog. Delaying part of the analog signal in a BBD chip does not make it digital, if it’s not digital, what is it?

[784373] admin Says: 10:36 pm, March 23rd, 2018

Analog audio is a continuously varying voltage. When you start to slice it up into samples, it is no longer continuous and no longer analog. It doesn’t matter how you store those voltages slices, the point is that they are discrete voltages captured at a moment in time. If you were to measure those captured voltages one-by-one with a multimeter, you could not tell if they were from a BBD or a DAC.

[784405] Valvicus Says: 8:36 pm, March 24th, 2018

@Brad: “… if it’s not digital, what is it?” To paraphrase Otis Blackwell, it’s “all chopped up”! LOL

[784472] Adriano Says: 6:10 am, March 26th, 2018

“Digital” usually refers to quantised samples (discrete values), and its discrete-time characteristic is a precise choice that allows for consistency and repeatability together with value quantisation (digital electronics is basically “restricted” analogue electronics which makes it easier to do lots of things). A discrete-time continuous-value device is not digital, you may decide not to call it analogue as the output is not “analogous” to the input, but it’s not strictly digital.

[785481] pink Says: 10:46 am, May 29th, 2018

hahah now i know why everything i build with bbd’s always busts my stones as bad as digital stuff 😉
thanks jack!

i was thinking perhaps a better term than analog may be just silicon. its not digital, but its not analog either, for the reasons stated… but its definitely silicon! 😉
i gotta come by more often 😉 peace!

[785673] Mark Hammer Says: 8:11 pm, August 10th, 2018

BBDs are chips. I’ll leave it at that. 🙂

One of the things I’d like to see more discussion of, and especially scope pictures, is the impact of clock-pin input capacitance on clock-pulse waveform and resulting audio quality. The cell-to-cell/cap-to-cap “handoff” within a BBD has to be as smooth and seamless as an Olympic relay race. BBDs have varying amounts of input capacitance on their clock pins that act like lowpass filters on the clock pulses, when they exceed some maximum frequency. Buffering and some current drive can overcome that and permit faster clocking. But the unbuffered input capacitance introduces a degree of lag to the clock pulse that can make the “handoff” less seamless. That’s why the Matsushita datasheets indicate a maximum clock frequency of 100khz most of the time. I wanna see scope shots of what happens to the audio output when one approaches and exceeds that unbuffered maximum.

[785682] Steve Demedash Says: 4:19 pm, August 24th, 2018

Mark – if the input capacitance of the clock pins acts like a low pass filter, wouldn’t the upper limit simply be the point when the clock frequency is above the cutoff point of that filter? That is to say, when the clock pulses are coming in at a rapid enough pace, the pin would take a number of microseconds to charge up to the pulse’s voltage level, and if the pulse is over before the charge reaches the BBD’s V-on level, it would not register. This is what I’d think the maximum unbuffered clock frequency would be.

Or do you think there may be a range of frequencies between Matsushita’s 100kHz maximum and the ‘Maximum’ I describe, in which something else weird happens?


AMZ Home Page

Guitar FX PCBs