7 Costly DSP Richauto A11 Mistakes CNC Operators Must Avoid
The Richauto A11 runs a lot of entry-level routers. It also hides a few traps that turn a good part into scrap in one move. This page lists the seven mistakes we see most often, what each one costs, and when the controller itself is the limiting factor.
What this guide covers
These are operational errors, not brand complaints. Most of them are cheap to prevent and expensive to ignore.
Tool setting and zero offsets get skipped after a change
The controller keeps a work coordinate system in memory, and that system does not know a new cutter just went into the collet. Every tool change needs a fresh touch-off on the stock surface or a tool setter. Skip it and the Z axis drives down by the difference between the old and new tool length. The cutter usually loses that argument.
The damage is rarely limited to the tool. A plunge into a nearly finished pocket gouges the part, loads the spindle bearings, and can shift the fixture. During a run of thirty parts that single event can scrap the one that was almost done, and the scrap count climbs from there. A cutter worth $40 to $200 is gone, plus the downtime for re-calibration.
Overriding soft limits and rushing the homing sequence
Soft limits are the last electrical barrier between a moving gantry and a hard stop. They live in parameters, and parameters can be wiped. After a power cycle, a firmware reset, or an aborted homing cycle, those values may come back as zero or as a wide default that no longer protects anything. The machine still moves. It just moves without a fence.
Homing first, then verifying the limit values before you load a program, takes under a minute. Operators who skip that step and jog toward a fixture corner are gambling with the ballscrew. Over-travel can bend a screw, crack a linear guide block, and knock the frame out of alignment. Repair on a mid-size router runs past $2,000, and the machine sits idle for weeks while parts arrive.
If the machine cannot home reliably, stop. A homing failure is a symptom, not an inconvenience. Check the switch, the wiring, and the encoder before running production again.
Running G-code from the wrong post-processor
The A11 does not read every dialect of G-code the same way. A post written for Mach3, LinuxCNC, or a Fanuc-style control will usually include canned cycles, arc formats, or M-code behavior the DSP simply ignores or misreads. The file uploads fine. The first few moves look correct. Then an arc goes the wrong way.
Common symptoms are arcs that reverse direction, spindle speed commands that never take effect, and coolant or dust-collection M-codes that land on the wrong output. On a profile cut with many small arcs, a reversed arc can cut into the part instead of around it.
Pick one post-processor, verify it against a short test program, and freeze it. When you change CAM software, re-verify. Never assume two posts that look similar in the preview will behave the same at the machine.
Ignoring backlash and rigidity compensation
A router frame flexes. The A11 has no closed-loop feedback on most builds, so the control has no idea the axis did not quite reach the commanded position. Backlash appears as a small lost step each time the axis reverses. On a circular pocket it shows up as a flat spot on one side. On a square part it shows up as a size that differs depending on which way the cutter approached.
Measure it before you compensate it. Indicate a known surface, jog away, return, and read the difference. If backlash is under the tolerance your part actually needs, leave the compensation alone. Piling compensation on top of a worn nut hides the wear and makes the next setup unpredictable.
Rigidity is the other half. A long tool at high stick-out deflects under load, and no controller parameter fixes that. Shorten the tool, reduce the stepover, or add a support. If the part demands ±0.005 mm, a light router frame is the wrong machine for it, whatever the controller settings say.
Acceleration, deceleration, and jerk settings left at default
Default accel and decel values are tuned for a machine the factory imagined, not the one on your floor. Too aggressive and the gantry rocks on every corner change, leaving chatter marks and shocking the gearbox. Too soft and the cycle time doubles while the tool rubs instead of cutting.
Jerk, the rate of change of acceleration, matters more than most operators expect. A high jerk setting snaps the machine into motion and out of it. On a 3D relief with thousands of short moves, that constant snapping shows up as ripple marks on the surface and heat in the motors.
Tune with a scrap block and a dial indicator, not by feel. Change one value at a time. Raise acceleration until the surface finish starts to degrade, then back off by a comfortable margin and record the number. Those numbers belong in your setup sheet, not in someone's memory.
No parameter backup and firmware updates done blind
Every offset table, limit value, and motor tuning number on the A11 lives in internal memory. A failed board, a bad flash, or an accidental factory reset takes all of it. Rebuilding from scratch on a machine you have not touched in a year is a long afternoon.
Back up the parameters to a USB drive after every change, and label the file with the date and what changed. Keep the previous version. If an update goes wrong, you want the file that worked yesterday, not the one you just overwrote.
Firmware updates deserve the same caution. Read the release notes. Confirm the version matches your hardware revision. Run a test part before returning to production. A firmware change can alter how the controller interprets an existing program, and the program did not change with it.
Electrical noise and shop conditions treated as background
A stepper drive is a switching power stage. Poor grounding, a VFD sharing a conduit with signal wiring, or a spindle cable routed alongside an encoder line will inject noise into the step and direction signals. The result is a random lost step, a position that drifts over a long program, or a machine that faults only on humid days.
Separate high-voltage and signal paths. Ground the frame at one point, not several. Use shielded cable and land the shield at one end only. If a fault appears only when the dust collector starts, the fault is electrical, not mechanical.
Temperature and humidity matter too. A cold shop changes the machine geometry between the morning warm-up and the afternoon run. Let the machine idle through a warm-up cycle before the first cut, and check the part after the warm-up, not before.
Which failures come from the operator and which from the controller
Use this to decide whether to tighten procedure or change the machine.
| Symptom | Likely cause | Practical response |
|---|---|---|
| Z plunge after tool change | Offset not re-touched | Re-touch every change, use a setter |
| Axis runs into hard stop | Soft limits wiped or ignored | Homing check before each program |
| Arcs cut the wrong way | Wrong post-processor | Freeze one verified post, re-test after CAM change |
| Size differs by approach direction | Backlash or frame flex | Measure first, compensate only if needed |
| Ripple on 3D reliefs | Jerk and accel too high | Tune on scrap, record the values |
| Random lost steps | Electrical noise | Separate signal and power wiring |
Common questions
Is the DSP Richauto A11 good enough for precision work?
It depends on the tolerance. For wood, signage, plastic, and light aluminum work at Ra 1.6–3.2 μm and tolerances in the tenths of a millimeter, a well-tuned A11 router does the job.
For metal parts at ±0.005 mm with tight flatness, the open-loop control and frame stiffness become the limit, not the operator. That work belongs on a machining center with closed-loop feedback.
How often should I re-touch the tool offset?
After every tool change, every collet change, and every time the spindle is removed. Also after a crash, even a light one, because the tool may have shifted in the holder.
On a long run with one tool and no changes, a check at the start of each shift is enough. Write the check into the setup sheet so it is not left to memory.
What is the fastest way to check for backlash?
Mount a dial indicator against a flat surface on the table, jog the axis away in one direction, then command a return to the original position. Read the difference.
Do it in both directions on each axis. If the number is well below your part tolerance, leave compensation off and inspect the nut and thrust bearings for wear instead.
Can I run the same G-code on the A11 and a machining center?
Not reliably. The two controls interpret arc formats, canned cycles, and M-code assignments differently. A program that runs clean on a machining center may reverse an arc on the DSP.
Keep separate posts for each machine. Validate a new post with an air cut or a scrap block before it touches a real part.
Why does the machine lose position only on long programs?
Short programs finish before the error accumulates. On a long run, small noise-induced step losses add up into a visible drift.
Check grounding and cable routing first. Then check motor current and driver heat. If the drift follows the toolpath length rather than the load, suspect electrical noise.
When should a shop stop tuning and switch machines?
When the part tolerance is tighter than the frame can hold, or when the required surface finish needs closed-loop control. Tuning a light router past that point costs more in scrap than the upgrade.
A useful test: run the tightest feature on the drawing ten times and measure the spread. If the spread exceeds the tolerance, no parameter change will save the job.
Send us the part that keeps failing
Upload the drawing and the problem feature. We review the geometry, the tolerance, and the process, then come back with a quote and a DFM note within 12 hours.
12-hour quote100% inspection±0.005 mmNDA on request
