Table 1 — Delay times available from 512-, 1,024-, 2,048-, and 4,096-stage Panasonic BBD chips at various clock frequencies and sampling rates.
Table 1 shows how much delay time can be expected from the various Panasonic/Matsushita chips at different sampling frequencies. A pair of 4,096-stage chips in series can produce around 500 ms with respectable bandwidth. A number of commercial delays have gone this route, often with a switch to select between the outputs of the first or second BBD. (It should be noted that, though there are two chips in these designs, both chips are driven by a single clock.)
All the Trimmings
All BBD chips require what is referred to as a bias voltage. If the audio signal does not come riding into the chip sitting on top of a particular DC voltage, it won’t make it to the other side. As a consequence, nearly all pedals that use BBDs, regardless of effect type, will have a small bias trimpot (separate from the one used for balancing output signals). Some pedals with cascaded BBDs may have a single trimpot providing the same correct bias for all BBDs used, but the effectiveness of such a setup would be due more to a lucky guess than anything. Optimal audio quality occurs when every BBD in the circuit has its own trimpot for custom-tailoring its bias. It’s more work and increases production costs, but it makes for a better transfer between chips and is worth the additional labor, parts, and space.
The trimpot for adjusting a chip’s DC bias voltage is usually located somewhere near the delay chip’s input pin. If that bias voltage is too high or too low, you won’t hear any signal from the BBD. If it is just a little high or low, you’ll hear output, but it will be distorted. So if you have an old BBD device that doesn’t seem to be working (or sounding quite right), it’s worth opening it up and fiddling with this trimpot to see if that fixes things. There’s no risk of damaging the BBD or overall circuit if you simply let your ear guide you as you adjust it to restore audio quality. But when there’s more than one BBD in series and you can't hear the impact of each separate adjustment, it’s handy to have an oscilloscope so you can measure each chip’s output for minimum distortion as you adjust its bias trimmer.
Typical BBD-based effects have anywhere from one to four trimpots on the circuit board. As previously noted, some designs also have a trimpot for balancing output between two BBD outputs—and this is also found very close to the chip. Further, BBD effects in which feedback is used (like flangers and echoes), often have another trimpot for adjusting maximum feedback before howling or oscillation set in. This is usually found close to where the feedback is returned. Finally, there may also be a trimpot for adjusting the maximum delay time. This is useful for when, between like-valued BBD chips, there’s variation in where the clock-noise filtering begins to roll off top end and, consequently, the clock speed has gone low enough to be audible.
As you’ve probably gathered by now, when there’s more than one BBD chip in an effect housing, these trimpots can appear in duplicate or triplicate. For example, as noted earlier, the much-coveted Maxon AD-999 delay eschews high-capacity chips in favor of eight 1,024-stage chips, resulting in 25 total trimpots! Meanwhile, Electro-Harmonix’s most recent reissue of the Deluxe Memory Man has a bias trimmer for each of its two 4,096-stage BBDs, and an output balance on the second one, but it also includes a small gain stage for each BBD to compensate for potential signal loss (all BBDs lose a tiny bit of signal in comparison to the dry path), which means there’s a trimpot for each gain stage, too. Needless to say, when it comes to repairing or recalibrating a BBD-based effect, it pays to know where all the specific trimmers are and what they do! To aid in the process, track down a schematic and—if possible—a service manual for your pedal’s circuit.
Even simple effects with BBD chips typically have one to four internal trimpots for adjusting things like bias voltage, output balance, maximum delay time, and feedback. As you can see from the guts of this Maxon AD-999 analog delay, each BBD chip can exponentially increase the number of trimpots. Photo by Mark Hammer
Whoa, Slow Down!
As mentioned earlier (and as revealed in Table 1), shorter and longer delays can be produced by clocking a chip faster or slower. But there are limits. You can’t transfer the samples inside the chip too slowly (which is why Table 1 doesn’t go below 10 kHz), but you also can’t clock them too fast either—otherwise the sample-transfer process gets corrupted.
A BBD chip’s clock input pins have what is referred to as input capacitance. For our purposes, we’ll simply say that this capacitance can “dull” the clock pulse and turn it from square-ish to more triangular. When the clock frequency is low enough, the capacitance has no effect and the clock pulses are able to instantly switch the transistors in the chip that feed the samples to the next. But as the clock frequency increases, that input capacitance begins to add some lag and to turn what started out as a pretty square (i.e., either on or off) clock pulse into something more triangular. The near-instant switching action that normally yields seamless taking and passing of samples now begins to have a gap, as it takes longer for the transistors to be switched on. The gap itself is not directly audible since it is on the order of microseconds. But imagine a movie with a blank frame inserted between each successive image. It simply won’t look the same—the seamlessness of the flow of samples matters. Consequently, data sheets for many BBDs indicate a maximum clock frequency, beyond which performance will suffer. Table 1 shows that many common Matsushita chips can’t normally be clocked faster than 100 kHz.
Second-generation Matsushita chips had about 700 pF of capacitance for each 1,024 stages on the chip. For example, a 4,096-stage MN3005 or MN3205 would have an input capacitance of 2800 pF. In contrast, Reticon chips only had 110 pF for the same pins. Why does this matter? For long delay, it doesn’t. For flanging, where the most dramatic effects are produced by shortening the time delay to something almost negligible, you want a chip that is tolerant of very high clock frequencies (remember, faster clocking results in shorter delay time). The Matsushita chips weren't comfortable being clocked much faster than 100 kHz, while the Reticons were good up to 1.5 MHz. If you ever wondered why that A/DA Flanger or Boss BF-1 sounded so much better than your Boss BF-2, it’s because your Matsushita MN3207-based BF-2 couldn’t delay any shorter than 1 ms, while Reticon-based pedals could zip well below 1 ms with ease.