An analysis of the DR-110 Cymbal, Hi-Hat and Clap,This page:Theory; Metal/Noise, VCA, Hi-Hat, Cymbal
with suggested modifications and proposed clones
Page 2: Hand Clap, Cloning
Page 3: Clones continued
I'll begin by stating that, to me, the DR-110 is the benchmark for how synth cymbals should sound, with the TR-606 a very close runner-up. Similar methods are used in both, as well as in the '808, but the '110 circuits are more economical, and more suitable as a basis for D.I.Y. experimentation and replication. ("Cloning.")
At least, that's how I see it. So let's begin...
Waveforms...is a good place to start.
Check points 1-4 are the cmos oscillators' waves, given in period, which I have converted to frequency. The manual warns that these are critical and any deviations can cause dissonance -- as if that would be a bad thing.
Waveforms 5-14 are the various envelope voltages, with each purpose listed to the right of the box. Waveforms 5 and 6 are, respectively, the Hi-Hat open and close envelopes. Waveforms 7 and 8 create the Cymbal "ping" sound, and 9 produces the Cymbal "body" sound. Two separate VCAs that have different noise spectra inputs create the complete Cymbal sound. The Hand Clap envelopes 10-12 will be described elsewhere.
The hi-lighted components in the following schematics are modifications that can be found at these sites:
theninhotline site also has a zip file of recordings of all mods of all '110 voices. Unfortunately the filenames don't identify which modifications are being demonstrated.
Metal/Noise SourcesAlthough it's possible to "get by" using only white noise to synthesize cymbals, a clangorous metallic sound greatly enhances the simulation. (A little resonance helps, too.) The key to the DR-110's great sounding cymbals is its metal/noise mix.
IC1 and part of IC2 form four "metal" oscillators. (This can be greatly simplified with four inverters from a hex schmitt trigger.) These are mixed in equal amounts by resistors R2, 5, 9 and 10. Digital White Noise from a CD4006 shift register is mixed in by R1 at roughly one third the level of the oscillators. Changing R1 would affect the amount of "hiss" the mix contains.(TR-606 uses only mixed oscillators for its cymbals. The lack of hiss is where, IMO, its cymbals fall short.) The last oscillator output also goes to INTerrupt 0 of the CPU, where it's used for scanning the buttons and switches. I wouldn't mess with it.
The metal/noise mix then goes to two bridged-T resonant low-pass filters that separate the mix into two bands. (Coupling apacitors and high-pass filters in the Hi-Hat and Cymbal circuits reduce the low frequency content.) I call IC5a the High Filter and IC5b the Low Filter. The High noise is used by the Hi-Hat and the Cymbal "ping" circuits and the Low noise is for the Cymbal "body" sound. Replacing R16 with a potentiometer would allow tuning of the Cymbal "body". Doing the same for R12 would affect both the Cymbal and Hi-Hat "ping", which may or may not be useful. R11 and R23 affect the filters' Q or resonance.
Later I will show an alternative metal/noise source
For the metallic and noise instruments, the DR-110 uses brilliantly simple VCAs: each is a single transistor with the envelope supplying the transistors' collector voltage.
For a.c. coupled metal/noise, used in the Hi-Hat and Cymbal, the transistor gets a 1 megohm base bias resistor. (In the 606 and 808, where the pull up is to +15V, this resistor is usually 2M.) For d.c. coupled digital noise, as used by the Snare and Hand Clap voices, the transistor gets a 100k series resistor on the base. The transistor is then in switching mode -- it's basically a logic inverter with the output level controlled by the voltage on its collector. Capacitors C2 and C3 provide high-frequency bypassing if and as needed by a particular instrument sound.
The circuit will distort if the input signal is greater than ~ 1.5Vpp, which is approximately the base-to-emiter saturation voltage of small signal transistors.
A shortcoming of this, and most, single supply VCAs is that they tend to produce a thump when activated. In the '110 and other Roland drum units they're followed by a hi-pass or bandpass filter which mostly block the low frequencies of a thump. A nice feature of this simple circuit is that envelopes can be very simply combined to produce a more complex contour, as will be seen in the Cymbal description. Another advantage is the circuit can have more than one output, with each output having a different sound envelope and even different harmonic content. The DR-110 uses this capability to have one transistor serve as VCA for more than one "voice."
Hi-HatOne consequence of this economizing is that it can make the schematic(s) difficult to interpret. Therefore, in the schematics below I have grayed-out components not relevant to the voice being analyzed.
The triggers from the CPU Board are low-going. The pnp transistors Q1and Q2 invert them into high-going to produce envelopes of different durations by R-C combinations. Starting with the Open Hi-Hat (OH,) C8 and R20 produce an envelope voltage with a 700mS decay. (Check Point 5.) This voltage is buffered by Q3 and goes to Q7's collector through resistor R52 and diode D5. High Noise that's applied to Q7's base "stands up" at the collector, passes through C25 and gets high-pass filtered by Q8 and components. The Closed Hi-Hat (CH,) operates pretty much the same way, but with a shorter decay (Check Point 6,) due to C7 being smaller than C8. (Also because the voltage isn't buffered.)
But what about Q4?
By itself, CH produces a short duration envelope, but if OH and CH arrive at the same time, the short, Closed envelope will make Q4 conduct. Resistor R19 will cause C8 to discharge faster than it does when only an Open trigger is received, causing the Open decay to be shortened. When combined with the short decay produced by C7, the result is a decay somewhere between open and closed. If a Closed trigger arrives immediately after OH, while the Open decay is in progress, Q4 again conducts, this time "snubbing" the Open decay, an effect similar to hitting the Hi-Hat while it's open, then closing it before the sound dies away. As to what happens if OH arrives during the brief time a Closed decay is in progress... I don't know.
Click to hear closed-closed-open-closed
Modifications can be made at the components hi-lighted in yellow. Making R20 variable, say by changing it to a 22k or 47k resistor in series with a 1Meg ohm pot, would be a way to control the duration of the Open decay. Something similar could be done for the Closed decay, but since "Closed" is by definition of short duration I don't see much point in that, except to maybe swap the durations of Open and Closed. A resistor and pot for R19 could control how fast Open envelope gets snubbed by a Closed trigger, but I don't think that would be particularly useful, either.
Dan Lavin's modified Hi Hat
Finally, a resistor and pot for R50 could change the cutoff frequency of the hi-pass filter, altering the tone. However, this filter is shared with the Cymbal's "ping" sound, so that would be altered as well.
Cymbal...is a bit tricky to explain, because the sound generated consists of two frequency ranges of metal/noise and more complex enveloping: three envelope voltages of different durations and amplitudes controlling two VCAs with different signal inputs and different output filtering.
Low-going CY produces a high-going pulse at Q5 collector. Capacitor C17 produces an envelope voltage with a 60mS decay. (Check Point 7.) The pulse also produces a longer envelope at the base of Q6. (Check Point 8.) Buffered, it appears at the emitter and is mixed with the shorter envelope at roughly one-tenth the amount at (7) because R42 is 10X R41. Together, the two envelopes produce the Cymbal
"ping" CV for Q7 (doing triple-duty!) through the Diode That Has No Identifier. The resulting audio has an initially fast decay that drops to a lower level with a longer decay. The signal then goes to the same circuit that hi-pass filters the Hi-Hat signals, but with a larger value coupling cap (C22), passing more of the metal/noise low frequencies so the ping won't sound the same as the Hi-Hat.
Meanwhile, the full amount of that longer envelope from Q6 is further lengthened by C21 and becomes the CV for Q9, (Check Point 9,) VCA-ing the Low Noise, which becomes the Cymbal "body" sound. Note that, according to the chart, the envelope at (9) reaches a lower peak voltage than the others. There are, I think, two reasons for this; first, the voltage from Q6 is decreasing (decaying,) while it's charging C21; second, R45 slows the charging of C21, giving the resulting envelope an Attack time, which also prevents C21 from fully charging up to the Q6 emitter voltage before it decays to nothing. The audio produced by Q9 gets passively hi-pass filtered by C30, R74 and C42. Q9 is the only single-purpose VCA in the '110.
Click to hear cymbal X4
If R42 is made variable, for example with 22k in series with a 100k pot, then the ping trailing away longer decay voltage can reach Q9 at a higher level, making the ping sound more prominent. Similarly, if R45 is made variable, then the body Attack phase can be made shorter and C21 will have time to charge to a higher voltage, making the body sound more prominent. In that case, Decay will be longer, because C21 will have to discharge from that higher voltage. Varying R44 (labeled as DECAY in Magnuson's mods,) will afftect the length of the trailing away of ping as well as the body sound, and can produce, in Magnuson's words, "...some really strange effects." R47 (not hi-lighted,) also affects body decay.
Dan Lavin's modified Cymbal
As pointed out in the Hi-Hat section, changing the cutoff freq. of hi-pass filter Q8 (R50,) will affect the tone of both the HH and Cymbal ping. The body tone can independently be changed by varying R74.
Hand Clap analysis and proposed as well as actual clones continues on the Next Page.
|To Projects or Publications or use to go to where you came from||contact: rockingsofthome.net|