JLH - 822 Voltage Switcher Theory of Operation The 822 voltage switch has only one function. It adds fixed amounts of DC voltage to the incoming control voltage to offset the output (it actually adds current not voltage -- more on that later). What makes this circuit unique is that it adds these offsets at fixed increments to provide voltage shift at exact musical semitones and octaves when used in conjunction with any 1-volt-per- octave voltage controller oscillator. And, the circuit is configured to be very accurate so that the keyboard 1-volt-per-octave scale is not compromised. Let's examine the circuit. The functions can be broken down into five sections for analysis: * Section # 1 is the semitone resistor divider network. This series string of resistors provides the individually adjustable offset voltages to our "step" selector switches. * Section # 2 is the octave resistor divider network. This series string of resistors provides the individually adjustable offset voltages to our "octave" selector switches. * Section # 3 is the summing amp that actually adds the incoming CV with the step and octave voltages. A by-product of this summing process is inversion of the voltage. * Section # 4 is a unity-gain inverting amp, which restores the CV to the correct polarity. * Section # 5 is the output driving amp. It is configured to offset the capacitance of cables and provide a low impedance output capable of driving multiple oscillators without error. Sections 3, 4, and 5 are duplicated in the "A" and "B" channels. For the purpose of discussion, we will examine the A channel, but the same logic applies to the B channel. First, let's examine section # 1. All of the shifting for the semitones is in an upward direction. So, we need to establish positive voltages to accomplish those shifts. What voltages do we need? Well, since 1 volt = 1 octave, and there are 12 semitones to an octave, we need 1/12 of a volt per semitone shift or "step." Step # 1 = 1/12 or 0.08333... volts Step # 2 = 2/12 or 0.16666... volts Step # 3 = 3/12 or 0.25 volts And, so on through Step # 11 = 11/12 or 0.91666... My prototype design featured just that exact voltage delivery to the summing amp for unity gain. However, after construction of my first 822, I found a problem with that approach. The load of the summing op amp on the semitone resistor divider network was causing the voltage to vary slightly through the network when the loading changed (by rotating the selector switch). Therefore, I decided to change the relationship between the total R of the network and the load imposed by the summing op amp. My calculations indicated that changing this by a factor of 4 or 5 would make the undesired voltage fluctuation so small it would never be noticed. I settled on resistor values that would produce 4 times the necessary voltage from the semitone resistor divider network. To calculate all but R12, let's assume 1mA of current (we will see later that I actually use 4mA). To get 0.0833 volts drop across R, I need 83 ohms. Remember Ohm's law says V=I*R. So, 83 ohms X .001 amps = 0.083 volts. Ideally, the multiples of 0.083 volts would occur around the center of the trim pots. Notice, all of the resistors except R1 and R12 are 33 ohm and the trim pots are 50 ohms. So, that works out to 83 ohms quite nicely. Why is R1 not 33 ohm? Because we want the 0.083 voltage point to be near the middle of VR1. Since VR1 is 50 ohm, we need to add about 1/2 of that to 33. So, 33+25 = 58 ohms. 59 ohms was the closest selection available. Now assume that 1mA of current flows in this circuit. The voltage at the point between R1 and VR1 is 0.059 volts, between VR1 and R2 is 0.109 volts (50+59=109 X .001 amps = 0.109 volts). You can calculate that through the entire string up to R12 and will find that the set points we are looking for (multiples of 0.083 volts are near the middle of each trimmer VR1 though VR11. Now, in reality, we need the voltage at these trimmer mid points to be four times what we calculated. So, we can accomplish that by leaving all the resistor values the same and causing 4mA of current to flow in this series resistor network. How do we do that? By the value selected for R12. Remember Ohm's law? V=I*R also works out to I=V/R or R=V/I. So, total R determines the current that will flow in this series string. I know my source voltage is 15 volts. Our calculation is R=15/.004. That equals 3750 ohms as the desired total R. You must subtract out the total resistance of R1 through R11 and VR1 through VR11. So, 3750 - 972 = 2778 ohms. Therefore, 2.8 ohms is just about perfect. That gives us 3772 ohms total R approximately. We know it will be slightly different because of resistor and trimmer tolerance. Anyway, it is close enough and that is why we have the trimmers. Next, let's examine section # 2. Shifting for the octaves will be in an upward and downward direction. So, we need to establish positive and negative voltages to accomplish those shifts. What voltages do we need? Well, since 1 volt = 1 octave, we do not have to think too hard on this one. So, this network is really composed of two identical series resistors strings. One is from +15 VDC to ground. The other is from - 15 VDC to ground. You will notice the resistor selection is the same in each. R13 = R20, R14 = R19, and so on. As in the semitone resistor divider network, I really want the voltage to be 4 times what we need. You will see why later. For now, let's calculate all resistors except R13 and R20 assuming 1mA and desired voltages of 1.0, 2.0, and 3.0. Our calculation is R = 1.0v/.001a = 1000 ohms. So, I want the midpoint of each trimmer to be about 1000, 2000, and 3000 ohms from ground. I selected 953 ohms for R16 and 17. The center of VR14 and 15 is therefore 1003 ohms from ground. The perfect value then for R14, R15, R18, and R19 would have been 899- 900 ohms. Well, they don't make common 1% resistors that size. So, I had to go with 909 for R15 and R18. That is not a problem, but it moves the correct trimmer setting away from mid point about 9 to 10 ohms. No problem, it is a 100 ohm trimmer. I went the next size lower (887) ohms for R14 and R19 to offset that value back the other direction so that VR12 and VR17 would be closer to mid point instead of pushing their setting further toward the edge. Now, R13 and R20 are selected to provide 4mA of current in this resistor dividing networks. We know from our last calculation in section # 1 that we need about 3750 ohms on each side. Looking at the either side we have 3750 - 3049 = 701. 698 ohms was the closest resistor available and almost perfect. OK, now it is time to examine the heart of the circuit - section # 3, the summing op-amp: U1A provides the summing of voltages. This is a very common configuration that will sum any voltages (currents actually, but that is a lesson on op amps) that are attached to its inverting (-) input. This configuration does invert the voltage. Why use an inverting amp? You can configure op amps so they don't invert. However, it is not an easy task to make a summing amp that is non- inverting. And, since we are VERY interested in accuracy in this circuit, the best method is the well-traveled inverting summing amp. The amp has 3 input resistors, they are R21, R24, and R25. So, we are summing 3 voltages - the incoming CV, the voltage from the step switch, and the voltage from the octave switch. We also have a feedback resistor (R22) which is used to set the gain of the amp. Let's look first at R21 and R22. Why are they 100K? I could have used many different values. However, 100K is a good value for R21 to provide high input impedance to the amp so that the incoming CV is not loaded down. The gain of the amp = R22/ R21. We want the gain of the CV to be exactly 1. We do not want that critical scaling to change. That is why R21 and R22 are hand matched to be exactly the same. The 1-volt-per-octave CV passes through this amp at the exact value it comes in (except inverted). What about R24 and R25? Why are they 402K? Well, remember, I selected resistors so that the voltages in my resistor divider networks were 4 times what I actually needed. This is where we make the op amp use that value but divided by 4. The gain of the step voltage = R22/R24. The gain of the octave voltage = R22/R25. 100K / 402K = 0.2488. So, all voltages applied to R24 and R25 are divided by 4 in the output, putting them right where we need them. Therefore, the output of U1A, is the sum of the incoming CV X 1, the step voltage X 0.2488 and the octave voltage X 0.2488, but inverted to the negative side of zero. Let's look at the other two resistors in this stage. R23 does nothing but establish a reference to ground for the incoming CV. It should be high enough not to place a load burden on the proceeding stage. The value is not critical. I usually use about 100K. But, since we have the A input normalled down to the B input, there are two in parallel. So, I selected a higher value. It could have been anything from 180K on up to whatever. What about R26? I selected R26 when this circuit was configured for R24 and R25 to be 100K. Since the change to 402K, I really should have changed R26. The value is non-critical though, and some designers just connect this to ground. So, I left it alone as "close enough" But, to be perfect, R26 should be 39K. Should you change it? Nope. R26 should equal the parallel equivalent of all input resistors plus the feedback resistor at amp U1A. So, if you calculate that for four 100K resistors you get 25K. That is why I originally selected 24K. For two 100K and two 402K resistors, that value calculates to around 40K. Hence, my suggestion that a "more correct" value here is 39K. But, 24 K is close enough. Section # 4 will not take much examination. U3B is an inverting amp. R33 and R34 are matched so that gain is exactly one. The value can be almost anything. 10K was simply my choice. All this amp does is restore the output of U1A back to the correct polarity. R35 is calculated the same as R26. Two 10Ks parallel equals 5K ohm. So, 5K1 was the easy choice. R39 references the output voltage of U3B to ground for the input of U3A (section # 5). Maybe it could be omitted. But, I like this approach. So, there it is. Section # 5 is a non-inverting unity gain op amp. I shamelessly stole this circuit directly from the MOTM-820 LAG processor. What makes this circuit slightly different is that it is designed to drive large capacitive loads (like long cables). So, you can drive multiple oscillators from the single output of the JLH-822 using a multiple. That would be especially useful if you were using the 822 for on-the-fly transpose. I could not provide ganged outputs because of space limitations on the 1U panel. It is completely full. The circuit descriptions for sections 3, 4, and 5, always referenced channel A. However, you will notice that channel B is identical with the exception of the trimmers VR18 and VR19, R30 and R31. Only one (actually one "step" & "octave") resistor divider network is provided for both channels. We want it to have the same effect on each channel. I could have hand matched the resistors to the exact same size. But, that would have meant that R21 and R22 not only would have to be matched to each other, but also to R27 and R28. Now, that would have been a HUGE pain. But, you will notice that the matched resistors I supplied with your PCBs were indeed very close in value between channels. To assure uniform effect on each channel, I inserted trimmers in the B channel. That is why when calibrating the octave and step trimmers you ALWAYS use the A channel. Then, trimmers VR18 and VR19 are used to match the step and octave in channel B to channel A. Lastly, I will give you a brief explanation of why I increased the current in the resistor divider networks to 4mA from 1mA. This of course provided four times the voltage I needed, and then I turned right back around and multiplied that by 1/4 in the summing amp. Why you ask? That is a very good question. First, accept that op amps really sum currents, not voltages. We talk in terms of summing voltages because it is easier for discussion purpose. Since the voltage at the wiper of our trim pots determines our shift points, we want that voltage to remain as constant as possible. What establishes that voltage? The current through the resistor network of course. So, any current added to this network from the load of the summing op amp will change our voltage. We don't want that to happen. How do we limit that current to an amount that will not be significant enough to effect our divider network? By increasing the value of R24, 25, 30, and 31 so that this current is limited. Why did I select 402K? Why not 1 meg? Well remember that the effect of this shifting voltage on the output of U1A and U1B is the ratio of R22/R25 (looking at the octave switch in channel A. That equals about 1/4. I know I need +3 and -3 at the output of this amp. So, the voltage needed to achieve that output is 4 times that (or -12 to +12). Since my divider network cannot provide voltages in excess of +/- 15 VDC, four was simply the largest multiplier I could use. Otherwise, we could not achieve the +/- 3 octave selections. Please feel free to e-mail me if you have additional questions. Larry Hendry