Chuck D. Bones
Circuit Wizard
It's Way Huge!
Well, sorta. 5 knobs and only one of them does much of anything, unless you roll the guitar volume down. I'm baffled as to what Jorge Tripps had in mind, unless it's all an elaborate practical joke. Steve picked this one up on eBay, sent it to me, then I tested and traced it. Except for a few resistor values, the schematic matches the trace found on FSB. If you find the schematic below confusing, you're not alone. D2, D3 & D5 do nothing because there is never enough signal at that point to turn them on. C11 is shorted out by the HIGHS trimpot, which is more of a gain control than a treble control. C12 is too small to do anything audible. Ditto for R8 & C10. C8 is redundant since C6 & C13 provide DC blocking. The input impedance is 15K at mid and high freq, so it loads the guitar quite a bit. There is no input buffering on the I/O board. The 1st stage (Q2 & Q3) has 12dB gain and runs clean. Bass is rolled off below 100Hz. The 2nd stage (Q4) may look familiar, it's essentially a carbon copy of the DS-1 2nd stage. The gain from input to Q4-C is 40dB at 1KHz and is easily driven to saturation, just like in the DS-1. The 3rd stage (IC1) uses a CMOS opamp, similar to the CA3130. The two yellow LEDs (D4 & D6) in the feedback loop provide some clipping, but since the 2nd stage is overdriving the opamp, the 3rd stage is also easily driven to saturation, even with DISTORTION at zero. The tone network that follows has around 35dB attenuation, which leaves around 120mVp-p at the "clipping" diodes D2, D3 & D5. Not nearly enough to drive them to conduction. The signal is further attenuated by the TONE control and HIGHS trimpot before getting to the CMOS inverter 5th & 6th stages. IMHO, the one clever part of the circuit is the last two stages. Mr. Tripps wisely installed a current limiting resistor, R14 to drop the CD4049's power down to a reasonable voltage. 4000-series CMOS inverters perform better in analog applications, and use less power, when powered by 4 to 5 volts. Part of the clever thing he did was to use all six inverters in the package, paralleling them up for more output current and using 100Ω resistors to help balance the currents. Like the 2nd & 3rd stages, the last stage is easily driven to saturation, even with the DRIVE trimpot turned all the way down.
I have a modified version on a breadboard, but that's a topic for another discussion.
Well, sorta. 5 knobs and only one of them does much of anything, unless you roll the guitar volume down. I'm baffled as to what Jorge Tripps had in mind, unless it's all an elaborate practical joke. Steve picked this one up on eBay, sent it to me, then I tested and traced it. Except for a few resistor values, the schematic matches the trace found on FSB. If you find the schematic below confusing, you're not alone. D2, D3 & D5 do nothing because there is never enough signal at that point to turn them on. C11 is shorted out by the HIGHS trimpot, which is more of a gain control than a treble control. C12 is too small to do anything audible. Ditto for R8 & C10. C8 is redundant since C6 & C13 provide DC blocking. The input impedance is 15K at mid and high freq, so it loads the guitar quite a bit. There is no input buffering on the I/O board. The 1st stage (Q2 & Q3) has 12dB gain and runs clean. Bass is rolled off below 100Hz. The 2nd stage (Q4) may look familiar, it's essentially a carbon copy of the DS-1 2nd stage. The gain from input to Q4-C is 40dB at 1KHz and is easily driven to saturation, just like in the DS-1. The 3rd stage (IC1) uses a CMOS opamp, similar to the CA3130. The two yellow LEDs (D4 & D6) in the feedback loop provide some clipping, but since the 2nd stage is overdriving the opamp, the 3rd stage is also easily driven to saturation, even with DISTORTION at zero. The tone network that follows has around 35dB attenuation, which leaves around 120mVp-p at the "clipping" diodes D2, D3 & D5. Not nearly enough to drive them to conduction. The signal is further attenuated by the TONE control and HIGHS trimpot before getting to the CMOS inverter 5th & 6th stages. IMHO, the one clever part of the circuit is the last two stages. Mr. Tripps wisely installed a current limiting resistor, R14 to drop the CD4049's power down to a reasonable voltage. 4000-series CMOS inverters perform better in analog applications, and use less power, when powered by 4 to 5 volts. Part of the clever thing he did was to use all six inverters in the package, paralleling them up for more output current and using 100Ω resistors to help balance the currents. Like the 2nd & 3rd stages, the last stage is easily driven to saturation, even with the DRIVE trimpot turned all the way down.
I have a modified version on a breadboard, but that's a topic for another discussion.
Last edited: