There are a number of issues with V2. A team of talented individuals have invested a lot of time investigating and updating the software and hardware to address these issues. The goal has been to keep hardware modifications to a minimum to be able to retrofit existing V2 boards. There are two versions of the boards - ElectricMusicStore and SAModular which have some differences which will be noted.
Numerous bugs have been addressed - see the Software section at the bottom for details. Two major usability issues addressed were CV "jitter" and the non-functional Expander.
Note that with default software, you need to pull Boot0, pin 60 on the processor, to 3.3V and set DIP switch 3 to off in order to reprogram it. Pin 64 is 3.3V so you can jumper pin 60 to pin 64 during programming.
These are the modifications I made to my 248 and Expander. I describe each as essential or improvement to performance or functionality. I believe them to be correct but offer no warranty or guarantee nor accept any liability for any issues that may arise from these modifications.
CV Output Jitter
The sliders are all read by a single ADC input on the processor. The signals are multiplexed by CD4051s and there is considerable noise in the system that causes a wide variation in converted values. Three main efforts have been utilized to minimize this variation.
The CD4051s used to multiplex the signals have been replaced with MAX4617s. These are pin compatible devices with a smaller over voltage spec (see Over/Under Voltage Protection section below). Replace IC24 on PCB1 and IC22, IC23, IC27, IC28 on PCB2. This modification improves performance but is not essential.
The software has improved filtering on the analog slider values.
A new processor daughter board moves critical components from PCB2 to as close as possible.
The software alone offers a great improvement in reducing jitter. The hardware modifications offer further improvements but are not essential. No plans have been made regarding this new processor daughter board.
SAModular V2.3s PCBs have been redesigned to eliminate the daughter board and associated long traces by incorporating the processor and critical components on the PCB.
Other areas including improved power and ground routing and analog filtering have been investigated but have not shown significant improvements.
CV Output Offset
There is no offset adjustment for the CV outputs so they have some positive or negative voltage with the sliders at minimum. This modification adds an offset trimmer for CV output. After implementing the modification let the module warm up, set the offset (mine set to +0.18 mV) and then check the scale trimmer. This modification improves performance but is not essential. Offset without the modification may be positive or negative in the tens of millivolts.
The CV outputs have adjustable gain (scale) and are non-inverting. This modification changes the ground reference to an adjustable reference. The trimmer needs +/-15V which is not easily accessible so I simply soldered the trimmers to the tops of the ICs. The modification consists of lifting a resistor, and adding two resistors and a trimmer for each CV output.
Some users report having to change the 100K to 49K9 in order to achieve zero offset.
IC18 is used for CV 1
IC20 is used for CV 2.
Over/Under Voltage Protection
There are 20 diodes on PCB1. Most are used as over-voltage protection by clamping various signals to 3.3V. The 1N457 diodes have a voltage drop around 0.55V which clamp the signals at 3.85V. The over voltage specification for the CD4051 is Vcc+0.5V. The over voltage specification for the MAX4617 and the processor is Vcc+0.3V. These parts contain internal clamping diodes and while with series current limiting resistors they will not be damaged, exceeding this specification can result in odd behavior.
An example is using the Time Multiplier CV with the control. When the voltage is clamped to 3.85V, the Stage Address control is affected and advances to the next stage. The solution is to use Schottky diodes such as the SD103B in place of all the 1N457 diodes. This limits the over voltage to within specifications. This modification improves functionality but is not essential. Over voltage can be avoided by keeping the CV input to 10V or less and not adding in more control to exceed a 10V summation.
Slider Minimum and Maximum Values
The processor and Expander sliders are powered by one LT1117-3.3 regulator and the main sliders are powered by a second LT1117-3.3 regulator. The specification over temperature for these regulators is 3.235 V to 3.365 V, a difference of 130 mV. A 10 mA load at 25°C reduces the tolerance to 66 mA. The main regulator load is over 100 mA so a typical difference at 25°C could be 80 mV.
If the slider regulator is less than the ADC regulator the sliders will not reach maximum value. With voltage drops across traces and connectors it is also possible sliders may not reach minimum value. This modification removes IC4 and powers the sliders from the same 3.3V supply as the processor. In addition, a small dead zone in the firmware at each end of the sliders improves achieving minimum and maximum values. Since the sliders only operate over 90% of their travel, this additional dead zone is insignificant.
IC4 must be removed and the POT7 pad nearest to PLD1 can be wired to the bottom of R168. This modification only improves reaching maximum values if the PCB1 regulator is significantly lower than the PCB2 regulator.
The regulators on my 248 differed by 25 mV with a couple of mV of ground difference. I chose to remove IC4 and hard wire the sliders to PCB2 to eliminate any voltage drop between the PCBs. I was able to achieve equal 3.3V and a 2.2 mV difference in ground between the processor, sliders, and the Expander sliders. You can easily burn out a slider by miswiring power. I added a dot of red and black paint to mark the correct pins in case I need to make repairs. This photo shows my direct slider wiring and you can also see my heavier power runs from that previous investigation.
CV Out Filtering
There is a lot of high frequency noise components on the CV outputs. This noise is fast enough it doesn't really impact operation. There is a 220R series resistor so I simply added 0.1 µF capacitor from the banana jacks to a nearby ground. This rolls off the high frequency components at 7 KHz with a time constant of 22 µS. This modification improves performance but is not essential. In this photo you can also see my offset trimmer modification.
Start / Stop Impedance and Levels
The software has been enhanced to enable external clocking by simultaneous driving both the Start and Stop input jacks. The input impedance of each is 6.9K through a diode. The Start and Stop inputs are attenuated digital inputs to the processor with a Vil is <0.95V and Vih is >1.79V. Ignoring resistor tolerance this requires a low signal at the banana jack to be < 3.4V and a high signal needs to be > 6.0V. Between 3.4V and 6.0V is not allowed/undefined. An issue arises with combination trigger/gate pulses. A 15/7.5V trigger/gate is seen as a single high input but a 10/5V pulse is seen as a high with the gate in the not allowed region.
The driving module with a 10V pulse to both jacks needs to have an output impedance less than 2K2. If it is driving all four inputs then it needs to be less than 1K. Remember patching to other modules will reduce this even further.
The Start and Stop attenuation resistors are low in value. I changed them all to 49K9 which raises the input impedance to 100K and attenuates the input voltage by half after a diode drop. A 5V gate is now a valid high signal. With my Schottky diodes, voltages over 7.5V will clamp to the 3.3V supply and at 15V only result in 222 µA of clamp current. I changed R61, R69, R101 and R109 from 4K7 to 49K9 and R63, R72, R103, and R112 from 2K2 to 49K9. This modification improves performance but is not essential.
Some modules, like the 281, can have the pulse outputs paralleled. Adjacent stages have a continuous gate with multiple trigger pulses. These resistor changes now allow a 5V pulse as a valid high so you should be able to capacitively couple to the Start input. This will block the gate DC voltage but allow the trigger pulse to clock. There is no path to discharge the capacitor with the input diode so it needs to be paralleled with a high resistor value to bleed down the capacitor. I would suggest trying a 47 nF capacitor paralleled with a 4M7 resistor. None of my modules will generate multiple trigger pulses with a continuous gate so I cannot test this. Let me know if you can verify externally clocking the 248 in this manner.
The panel on the 248 is floating and can pick up and radiate noise. While a properly grounded boat will ground the panel, I wanted it connected when on the bench so I added a wire from the ground on a capacitor to a standoff pad. You have to use a standoff which screws to the panel. You can also see my heavier power runs in this photo. This modification is not essential.
ABCD Input Impedance
The ABCD inputs are attenuated with a 47K/22K network with an input impedance of 69K. When you parallel all four to do calibration, the input impedance is 17K25. If you are generating the 10V reference with a module, the output impedance of that module will reduce this voltage resulting in an inaccurate calibration. For example if the source module provides 10V with a 1K output impedance the voltage on the ABCD inputs is only 9.45V. The 248 is best calibrated using a 10V power supply.
Time Multiplier Controls
The BOM specifies 10K - 100K for these controls. This isn't quite right. These potentiometers are voltage attenuators driving a 100K circuit. As such, the 100K input impedance affects the attenuation. If these controls are 10K then the front panel legends reasonably are accurate. If these controls are 100K, then the taper is more log and 1 is closer to 1.4 and 2 is closer to 2.3. The linearity of the Stage Address is also affected similarly. You should use 10K for these four controls.
DIP switch 3 was used to not save / save slider information. The software saves slider information but moving a slider overrides the saved position so this function is no longer needed. DIP switch 3 should be set to OFF to enable programming.
The V/Oct DIP switch positions in the V2 manual are wrong. This illustration shows the correct DIP switch settings for V2.5 software.
The Expander requires updated software to function. However, a modification might be required on the 248 PCB2 to supply 3.3V to the Expander ribbon cable connector. Pin 9 on the "To Sliders" IDE connector should be 3.3V. The square pad is pin 1 and that row are the odd numbers, so pin 9 is the 5th pad. 3.3V can be found on pin 16 of IC23 or IC31 (the photo shows IC31). You do not want to power the Expander sliders from the Expander 3.3V regulator as that adds another regulator with tolerance. This modification is required to operate the Expander with ElectricMusicStore boards and SAModular boards prior to V2.3s. SAModular V2.3s and higher include this missing trace.
V1 Expanders use a LM1117-5.0 regulator. V2 Expanders should operate on 3.3V so IC9 should be a LM1117-3.3. Expanders are likely to work without this modification but are driving 5V logic with 3.3V signals. Similar to the 248, replace the CD4051s on the Expander with MAX4617s. Replace IC1, IC2, IC6, and IC7. The BOM calls out 270R resistors. I chose to build mine with 330R although either will work fine.
This is the updated BOM with changes that I used for my V2 Expander.
248 V2 Exp BOM
Version V2.5b3 (beta) is available on the Muffwiggler thread MARF 248r source code available and also at SAModular under the Firmware tab.
External input for Stage Address 2 now calibrates.
1st stage after Enabled stage causing a stop now sends a Programmed Pulse.
1st stage after Sustained stage causing a stop now sends a Programmed Pulse.
Programmed Pulse timing on stages selected by External Stage Address input corrected.
Interval Time slider settings are now recalled with preset.
Updated slider value filter to reduce amount of CV jitter.
After loading a preset, moving a slider will activate that slider with correct values.
DIP switch 3 is disabled to allow the debugger to connect and program without having to pull Boot0 pin 60 high.
The Expander does not function with this version.
DIP switch 3 was used to not save / save slider information. The software saves slider information but moving a slider overrides the saved position so this function is no longer needed. Leave in the off position.
Version V2.5 is available on the Muffwiggler thread MARF 248r source code available
Expander functionality - elimination of slider data corruption, correct slider 17 operation, improvements to slider, switch, and LED serial timing, and improvements for more consistent minimum and maximum slider values.
Sliders (when calibrated properly with offset and scale) should reach full voltage range (all 32 of my sliders read 0.00V to 10.00V).
Reduce skew for Stage Address control and external CV.
Improve Advance switch operation.
Fixed AFG1 Cont/Strobe operation.
Enhancement: Output pulse LED swap option.
Enhancement: External clock.
Enhancement: Swing mode with external clock.
DIP switch 3 is unused. Leave in the off position.
V2.5 Enhancements This two page document describes the new enhancements and can be appended to the V2.0 manual.
Output Pulse LED Enhancement
The Output Pulse LEDs are swapped. The hardware fix is a simple modification to cross resistors R219 and R222.
The software addresses this issue but to be compatible with boards built with and
without these resistors crossed, a power-up sequence saves the correct
configuration in eeprom.
Power on the 248 while holding both Output Pulse 1 and 2 switches in the off/down position.
Output Pulse LED 1 will start to blink but this could be either LED based on whether the above modification has been done.
If Output Pulse LED 1 is blinking (resistors crossed) then press Output Pulse switch 1 to the on/up position to save this configuration.
If instead Output Pulse LED 2 is blinking (resistors not crossed) then press Output Pulse switch 2 to the on/up position. This will save the configuration to reverse these LEDs in software.
External Clocking Enhancement
External clocking is now enabled by driving both the Start and Stop jacks together. Advance will clock to the next stage. This photo shows the 248 driven by an external 300 Hz clock.
External Clock Swing Mode Enhancement
When driving the 248 with an external clock, the Time Range switches, Time Multiplier, and Interval Time sliders have no function. The Cont / Strobe switch is repurposed as a Swing Mode enable/disable switch. Pressing it down momentarily while externally clocking enables Swing Mode and pressing it up momentarily disables Swing Mode. Note this is a global, not a per stage, mode.
When enabled, it will cause a second pulse to be output on the All Pulses jack. The time between pulses is controlled by the Time Range switches, Time Multiplier, and Interval Time sliders. If you want to "disable" swing mode on a particular stage, set the Time Range switch to 0.002 and the Interval Time slider to minimum for this stage. There are still two pulses but they are now very close together.