An encapsulation of what I have been able to glean from the 57 prior pages. This information was invaluable in getting my 5-axis system to work.
With thanks to you all...

Essentials:
  1. Never plug a motor in when its powered up. The poor grounding (note4) may destroy the 6560 chips.
  2. Check that pin 25 for every driver IC is not shorting. Shorten it if it is too long.
  3. Check dip switches all work. Use a multi-meter on the back of the board to check for short when switch is on. Some have been found to be faulty.
  4. Ground the heatsink for each TB6560 through its mounting screws. Solder a wire to the large ground plane next to the small cap (there is a via there to use) and then wrap the other end around a screw on each TB6560, or use a solder tag. Can also use pin 10,15 but not 6 (5V earth) for grounds. Pins 10,15 also route to the stepper motor pins for convenient soldering.
    In the current design, there is a danger of blowing all chips if one goes bad. Adding the grounds helps protect them all.
  5. Add a protection diode across the spindle relay coil. (Diode bar on the +12V side). There is no protection here. A diode will help to prevent noise and protect the transistor.

  6. Disconnect the spurious current save feature. This generates extra pulses and is not useful. The simplest way is to remove the 10k resistors on the back of the board. These are located differently for each design but the 10k and transistor will be very close to each other and near the top of each driver chip.
  7. Pin 7 on driver chip has a cap that sets the internal frequency. The only value that is tested to work, and calibrated for the three decay modes, is 330pf. Replace the 1000pf which is installed. Forum members have tried several values (including not using one at all). All offered improvement over the stock 1000pf.
    However the data sheet clearly indicates that for speed decay and microstepping the only supported value is 330pf. (The supplied 1000pf value is too high and generates audio frequency noise).
  8. Cleaning up the pulse to Clock(Step) and Direction pins.
    The pulse from the optos is very slow on rise time. The chip triggers on the rising edge so this pulse needs to be better shaped. The negative edge is good but the chip ignores that.
    Two methods:
    • EITHER bypass the optos for step signals by removing them. Put a link between pin 2 and pin 4 of the step optos.
      This has the negative side effect of removing the tenuous isolation benefit of the opto but will not effect counting if using a display board.
      The on-board LEDs will display properly.
    • OR add an inverter after the opto to clean up the pulse shape and invert it for proper trigger on rising edge.
      This will maintain the opto isolation but you will need to set the step to active high if using mach3.
      Unfortunately as a result of the mach3 change, the on-board LEDs will now be on all the time. If you have the 5 channel board, with the extra 74HC14 driving the LEDS, then you can jumper this inverter to get the LEDs back.
      But the LEDs are a crude indication of whether an axis is working or not so you might simply choose to ignore them.
    • OR jumper the opto and resistors out completely. This will deliver a nice clean pulse but leave you without manual pendant and external display control (both of which use the DB15 port). You won't have LEDs flashing for each axis either. The optos will no longer be providing the tenuous isolation support.


    Changing any of these will have an effect on direction pulse also. We have changed the timings between signals. There will be a delay between direction and step pulses. This may cause a +/-1 problem if changing direction a lot.
    So if you jumper the opto for step - you should jumper it for direction also. Enable is less of an issue and does not need to be changed.

    You have more control over these timing issues if using LinuxCNC than you do if using Mach3. This pulse delay will only affect you if using very fast pulses. You can make a test by stepping and changing direction many times in a row. move10, change dir, move10, change etc. The end result is the stepper should not have moved after many pulses. Try 1000. If it has moved then you have a delay that needs fixing. See Checks section below.


Optional changes:
  1. Increase diameter of sense wire to better return current to the chip for current protection.
  2. Change sense resistors to proper values for your stepper motors. E.g. 0.47 ohm for 1 amp motor. Turn current dip switches to full.
    The current dip settings are not supposed to be the primary way to control motor current. Instead it supposed to be the sense resistors, with the dip switches being secondary and therefore matched to the motor.
    Proper decay made behaviour requires correct sense resistor selection. This sets maximum motor current. I = V/R. Where reference V=0.5V
    • so R=0.5 ohms = 1A, 0.25=2A, 0.15=3.3A
    • e.g. specific motor 57BYGH56-401A rated at 2.8A = 0.18 ohm sense resistor. (3-5W)
  3. The 12V regulator (7812) is close to being overloaded and runs very hot. Lighten the load by adding a second 7812 direct from the incoming DC to drive the fan.
  4. Startup order is bad. Do not use a switch on the power line coming from the power supply. Instead enable on/off by turning on the power supply itself. This will help a little as modern supplies have a soft start. But fundamentally you can expect breakage of 6560 chips when powering on or off. This is not widely occurring but its a design flaw.

Manual Pendant:
  • The display board may not reliably read pulses from the 25pin computer port unless the manual pendant board is removed.
    The board places too much load on the circuit and pulses are not read if it is attached.
  • The manual pendant board is useful for initial testing but once installed with EMC or LinuxCNC, the keyboard should be used for pendant operation instead.


Replacements:
The replacement design for this chip is the TB6564AHQ also called THB6064AH. It handles 40V instead of 30V and does not have startup sequence problems. It is not pin compatible with the 6560HQ.
Two good designs, reasonably priced, from:
ldt THB6064 Driver kit and:
THB6064 MassMind Stepper Motor Driver Kit

Note Geckodrive Step Motor Drives Gecko drives FTW (e.g. G540) but expensive.
(I am not associated with any of these.)

You're going to spend a bit of time getting the blue boards to work. Even using these tips. Your time may be better spent by instantly moving to a better chip/board design. Many users have found this to be true.


Checks:
  • Is your computer putting out pulses on the right pins ?
    You need to set your bios to ECP+EPP. Bidirectional is not the right setting.
    Here is a simple procedure for testing your pins using multimeter. (use jog control and check direction pins with trial Mach3 program).
    [ame=http://www.youtube.com/watch?v=uglCm_qsojk]CNC Electronics 5 -Testing the Parellel Port.wmv - YouTube[/ame]
    If the direction pins are working there is a strong likelihood that the other pins are working.
  • Here's a refresher on the DB25 Parallel port.
    Parallel port pins diagram. Inputs and outputs.

    Its worth noting that on some(most) blue cards Pin15 is not used.
  • Testing for loss of pulses
    Do 1000 +20,-20 steps. if there is a delay between change direction and step pulses you will see the stepper drift off from zero (in steps of 20) after a lot of pulses. This means the delay between changing direction and sending the step pulse is not large enough, or your signals are not well shaped or possibly inverted. If the offset is not in multiples of 20 steps then you may be losing single pulses, in which case your timings are probably too fast for your computer. There is more fine grained control over this in LinuxCNC than in Mach3.

I found this thread incredibly useful. Thanks to all who have contributed.