PWM to +/-10V

[AshGork]

I found some very interesting discussions on GitHub about the use of a ColorLight with LinuxCNC.
I own a ColorLight 5A75E and I would like to retrofit an old 3axis CNC machine with AC Servo and drivers that I have to command with +/-10V (no possibility of step/dir command).

I would like to know if there is a solution not too complicated that uses LitexCNC, a ColorLight 5A75E and a breakout board (I can design and produce a PCB) in order to get a +/-10V output signal for each axis.
And then I will close the loop with the input board to connect the encoders to the 5A75E and allow LinuxCNC to do the PID calculations.

I saw on the LitexCNC webpage that they only offer step/dir or PWM signal but that's not what I'm looking for...

I hope I'm clear enough, and sorry for my bad English.

Thank you in advance

[AshGork]

Peter van Tol answered :

Your English is much better then my French. And also, English is my second language because I'm located in The Netherlands. Let our languages not stand in the way of retrofitting your machine.

I think it is possible to do it with the current modules in LitexCNC. The +/- 10 V signal, does that control speed or position?

In case it controls speed, you need an encoder to determine the position, feed it as feedback to the PID. The PID then calculates the required speed (input is the commanded position), which is fed to the PWM module.

In case it controls the position, you might be able to omit the PID completely for a first test.

The PWM module supports up/down mode. In that case the PWM has two pins on the FPGA: one for direction and one for the PWM output. For conversion you can either buy a board or create PCB yourself. The hardware is not yet available.

At this moment I'm designing a new line of PCBs to connect to the supported FPGAs. Each board has a connector with 4 GPIO, enable signal and power. Each board has a dimension of 71 mm long and 16 mm (or multiple of it) wide. The input and output is located on the short side. These dimensions have been chosen so multiple boards can be placed in a standard DIN housing (UM72).

Currently I've developed the following boards:

• GPIO out (supporting 500 mA 24 Volt outputs sinking or sourcing)
• GPIO in (supporting both NPN and PNP sensors, 12~24 V)
• shift out, when you need more outputs. This is an experimental board, because shift out is not yet supported by the released firmware.
• parallel in. Converts a 2x13 IDC box header to 4 extended PMOD connectors. This can be used to connect HUB75HAT to this new line.

Upcoming boards that are on my list at this moment are encoders (4 differential inputs)
and stegen (2 diffential driven stepgen on a board, with support for ENA signal and Alarm inputs).

I think your development fits really nice in this line!

[AshGork]

Ok, good news !

The +/-10V signal controls speed.
Indeed, there are encoders physically connected to the servo's shafts and send the signal to the drive. Then, there are output ports on the drive that allow me to re-use the encoder signal.

When you say "feed it as feedback to the PID", do you refer to the internal PID control of linuxCNC or to a PID control that I should implement into the electronics that will take place between the 5a-75E and the drive ?

The only boards I found online are able to convert PWM to 0-10V but not PWM(+dir) to +/-10V. Do you know what board I can buy to convert PWM(+dir) to +/-10V ?
Or any advice to build my own pcb to do that ?

The new PCBs you are designing look very interesting !

[AshGork]

Peter van Tol answered :

I'm referring to the PID controller in LinuxCNC. The only extra hardware you need are a couple of GPIO in or encoders and a PWM board for the output.

The encoder output from the drivers: is rhat single ended (GPIO is good enough) or differential (board needs a design)? Then I know where to focus on with my development.

Tonight I'll create a basic setup for a single axis for you. Can you supply me with the LED card type you have?

For tuning the PID I have also an idea, but will elaborate on that as soon as I am on a PC.

[AshGork]

Ok !

I'm pretty sure the encoders from the drivers are differential ones since I have PAO/PAO*, PBO/PBO*, PCO/PCO* signals.
On the LitexCNC webpage, they talk about a breakout board capable to support differential encoder, but I can't see any design/file (cf. attached picture).

Also, please find attached the model of my LED card.

[AshGork]

Peter van Tol answered :

Attached to this e-mail you find a configuration for your machine, it has:

• some example GPIO;
• PWM which supports direction (i.e. positive and negative output);
• encoders.

In the encoder section you can also find 3 encoders for tuning the PID. You can use the output of these encoders as input for your PID. Some cheap encoders from AliExpress will do to give you some control over tuning the PID in real-time.

When moved to the discussions, I will start creating a HAL-file. It's much easier to type in the code-editor over there, but also we can collaborate in creating the config for your machine.

cnc_configuration.json

[AshGork]

@Peter-van-Tol

Thank you for the .json configuration file !

I'm currently trying to design the board that will convert pwm (+dir) to +/-10V.
If we assume that the PID loop of LinuxCNC has a frequency of 1 kHz, then my response time of the board I'm designing must be lower than 1ms, isn't it ?
With the current values of capacitors and resistors, my response time is larger than 30ms...

[Peter-van-Tol]

I bought a couple of 74LVC245 to replace the actual buffers in order to use some pins of the 5A-75E as inputs. Please find attached the last version of my circuit. I don't understand where should I place the optocoupler to separate the grounds ? And if I understand well, I should convert my 24V to 5V and then use the b0505s-1w to feed the optocoupler ?

The optocouplers should be placed between V3 and R7 and between V4 and U6. When using the TLP2161, these require a 5 Volt signal on the driver side. This power can be supplied with the b0505s-1w. The optocouplers should be fed with +-10 volt if they are rail-to-rail. You can also feed them with a little bit more like 12 Volt. Then you can create the 12 volt and -12 volt using 7812 and 7912 voltage regulator (assuming you have 24V input voltage).

As an example, you can look at the design of the GPIO out board from the zip-file in the previous post. However: the pinout of the TLP2161 is incorrect. Pins 1 and 2 are reversed.

[AshGork]

Ok, I'm currently designing the electrical schematic of my board with all your suggestions. I will show you once it will be done.
Thank you

[besriworld]

It is very nice that the project has support for multiplexed inputs. This works very well in my board. It takes 6 pins from the fpga and gets 24 inputs. Good deal :d

[Peter-van-Tol]

For multiplexed inputs I did not design or ordered the board yet. That will be in the next batch I'm going to prototype. The shift_in will be based on the 74HC165 parallel to serial shift register. With that chipset it will be possible to chain many inputs together.

Thinking I will limit the design to 256 inputs, or 32 chips (!), per line (4 GPIO of the FPGA). Reason to limit these inputs is the clock cycle of LinuxCNC. Assuming a base period of 1 ms (or 1 kHz), linking 256 inputs would require a frequency of over 250 kHz to transport all the data in time before the next cycle of LinuxCNC starts.... Those frequencies are bringing possible interference problems into the cabinet. And I sincerely think 256 inputs should be enough.

It should be noted that these inputs are NOT realtime. For homing sensors I would strongly recommend to directly attach them to the FPGA.

[Peter-van-Tol]

@besriworld : I've received and build my shift-out boards using a 74HCT595.

In #127 I've added support for this board. After an evening of debugging, it is confirmed that it works. Coming week (I hope) I receive the shift in boards (using a 74HCT165) and will finish the driver....

More info on configuration can be found in #86

[AshGork]

breakout_board_axis.pdf

Could anyone check my design ?
I'm pretty confident with the PWM to +/-10V conversion but I'm less confident with management of the signal from 5A75E board to my conversion module.

[AshGork]

Good idea.
According to PWM_DAC.pdf, you modified the input type of signal ? Now you use step/dir instead of PWM (+dir) ?

[besriworld]

This is because I am using my big board. I have differential outputs for step direction. But in the firmware of the mesa this output is PWM . I'll just make a board that will connect to my other board, instead of creating a whole new one when everything else is working fine. But it can easily be made to fit with this project.I'm not a professional at this, but if everything works well this time I'll be able to make a board suitable for this project.

[Jaibadon]

Hey! I am in the exact same situation of needing a +-10V differential signal from the 5A75E . @AshGork, did you ever finalize your board design and if so, would I be able to have a look? Many thanks.

[horcsab]

Hi,
I am attaching the schematic diagram., this is working!
+-10V_pwm_colorlight.pdf

[EhsanRasouli1371]

Hi, I am attaching the schematic diagram., this is working! +-10V_pwm_colorlight.pdf

Can you please share the Gerber file for this electronic circuit with us?

[horcsab]

Hi, All gerber and schematic include this: https://github.com/horcsab/Colorlight-CNC-step-dir-to--10V- Regards, Csaba 2025. 11. 07. 16:04 keltezéssel, EhsanRasouli1371 írta:

…

Hi, I am attaching the schematic diagram., this is working! +-10V_pwm_colorlight.pdf <https://github.com/user-attachments/files/21104622/%2B-10V_pwm_colorlight.pdf> Can you please share the Gerber file for this electronic circuit with us? — Reply to this email directly, view it on GitHub <#133 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AM7TNDDAMKF6BJBCOK65SFD33SYG5AVCNFSM6AAAAABZEBHZK2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTIOJQGM2TONQ>. You are receiving this because you commented.Message ID: ***@***.***>