PCB Layout Help

[akahrim]

Hello All,

I am working on a Constant current load (adjustable from 250mA-2A) for a project. The load will be batteries around .6V-1.5V I bread boarded the circuit and it worked great. I transferred it over to PCB and not so great. It was a disaster. This is my first mixed signal pcb. Needless to say I didn't pay attention to where signals were routed and I overlapped digital and analog signals. I've since redid the layout. I'm hoping this will keep my analog and digital signals clean not talking to one another. If you all could provide any feedback and pcb layout conflicts that would be greatly appreciated. How could I improve this layout? Would ground plane cuts help with this design? I have attached the pcb file and schematic.

Amir

 

[dougy83]

Can you provide PDFs of the schematic and PCB layers? Not everyone will have PCB Artist installed.

 

[akahrim]

Hello Dougy,

Sorry for the delay. I wasn't able to get a PDF generator so I have attached jpegs of the layers. It is a 6 layer board and the layers are as follows
1 - Top Signal Layer
2- GND Plane
3- Power Plane (+5)
4- GND2 Plane
5- Unused signal layer
6- Bottom signal layer

Planes 4 and 5 are not necessary so I was going to convert this to a 4 layer board. Please let me know if the images are adequate. Thanks for the help.

 

[alec_t]

I am working on a Constant current load (adjustable from 250mA-2A) for a project. The load will be batteries around .6V-1.5V

If you're charging batteries be aware that different cell chemistries require different charging profiles for safe charging. Which battery type are you using?

 

[kubeek]

Also a schematic would be of great help.
Either way, if it worked on a breadboard, but doesn´t work on a six! layer board with two ground planes then something is amiss, like a bad solder joint or a component in a wrong place. Or something in the schematic that didn´t show up on the breadboard, like the added stray capacitance of the ground plane.
The board looks so simple that you should be able to comfortably do it in just two layers without any issues, I have no idea why you went for six on the first try, it must have been hugely expensive.

 

[akahrim]

Hello Alec_T,

I won't be charging batteries with this circuit. Only discharging from 250mA-2A.

Amir

 

[CapeCAD]

Looking at your board from a power viewpoint, I would suggest the following:

Your thermal reliefs are very thin. I would flood over all vias and either flood over or increase width of power connections at a minimum.

Use multiple vias for connecting higher current to ground/power. Unless you are using heavy copper (>2 Oz) your vias will be plated with less than 1 mil and will get hot.

Power traces are thin for 2 Amps. Judging by component pad widths, this would be rated for 1.5 Amps max with 1 Oz copper. Fill in the outer layers, the less copper you remove the better. Flood with ground where there are no connections, this will help with heat dissipation and shielding.

ICs U1, U2, U3, and U7 look like they would route better (shorter) if they are rotated 90 degrees counter-clockwise. U8 looks like 90 degrees clockwise would be better.

I would also add bypass capacitors to each IC and add some bulk capacitors to the board.

 

[akahrim]

Hello CapeCAD,

Thank you very much for your advice. I just had a couple questions. What do you mean by "fill in the outer layers, the less copper you remove the better"? The term "flood" is new to me. What do you mean by flood the ground? Sorry for my lack of knowledge, I'm still a novice at PCB design. Thank you for your help.

Amir

 

[akahrim]

Also a schematic would be of great help.
Either way, if it worked on a breadboard, but doesn´t work on a six! layer board with two ground planes then something is amiss, like a bad solder joint or a component in a wrong place. Or something in the schematic that didn´t show up on the breadboard, like the added stray capacitance of the ground plane.
The board looks so simple that you should be able to comfortably do it in just two layers without any issues, I have no idea why you went for six on the first try, it must have been hugely expensive.

Hello Kubeek,

Sorry for the delay. I have attached a hand drawn copy of the circuit. I hope its legible enough.

Amir

 

[kubeek]

First of all I don´t see any power supply decoupling caps anywhere, which might be a hint as to why it worked on a breadboard but not on the pcb. Just a thing, how did you manage to connect those smd chips to a breadboard?
Also how exactly was the build a disaster? Did the output oscillate?

Second thing that hit me is that 150uF cap right at the output of the opamp, this is a rather strange configuration, and the next thing is that si2312 transistor, what is its purpose? PWM regulation?
If the only problem was instability of the loop, you could try some small caps between output and negative input of the opamps.

 

[akahrim]

Hi Kubeek,

The breadboard design reaches the constant current as soon as the circuit is powered up. The load current on the PCB spikes and then settles down on the constant current value. ts a steady decay from the peak .

All of the IC's that were used on the breadboard were the DIP equivalents. The resistors were all through-hole components. For the PCB, I just purchased their respective SMD package. The Si2312 is controlled by the ATMEGA. It turns on and off depending on the duty cycle that I would like (ex. 750uS on 250uS off). I threw in the 150uF cap because the circuit wasn't stabilizing. With the 150uF cap, the current is rock steady.

 

[akahrim]

Hi Kubeek,

The breadboard design reaches the constant current as soon as the circuit is powered up. The load current on the PCB spikes and then settles down on the constant current value. ts a steady decay from the peak .

All of the IC's that were used on the breadboard were the DIP equivalents. The resistors were all through-hole components. For the PCB, I just purchased their respective SMD package. The Si2312 is controlled by the ATMEGA. It turns on and off depending on the duty cycle that I would like (ex. 750uS on 250uS off). I threw in the 150uF cap because the circuit wasn't stabilizing. With the 150uF cap, the current is rock steady.

I am redoing the design to include the bypass caps mentioned by CAPECAD and the decoupling caps as you suggested.

 

[kubeek]

I am a little confused here, is this supposed to be a pwm controlled current sink, or a linear one? Or just with an off option?
Because the way you have the it connected from the top op amp to the main mosfet is really weird. The source o fthe si2312 is basically floating, and with 5V supply you will have trouble getting the mosfet to fully open.
The 150uF is also in a wrong place, as the RC constant relies on the internal resistance of the opamp output, and you should not need it anyway.

As for the second board, I suggest you start with addidng the decoupling caps to the board you have now, usually thin magnet wire and glue is the way to go. Get that board working correctly, and then go on designing another one.

 

[akahrim]

Hello Kubeek,

It is a pwm controlled current sink. I went and removed the 150uF cap and the circuit isn't stable. The output of that op amp (the amp connected to the 150uF cap) oscillates.

 

[kubeek]

The opamps make linear feedback, while the pwm is banging the transistor on and off. I don´t think this will ever work the way you want it to, at least the way I think you want it to work.
The way I see it now, the max6006 and the pot set the maximum current, and the atmega turns it rapidly on and off. This however completely upsets the feedback loop coming from the sense resistor.
Instead of having the 150uF right at the opamp output, use a resistor in series with the bottom left opamp and then a smallish cap like 100nF or something.
Another bad thing is that 200ohm resistor, the lt1001 cannot source that much current, so use 1k instead. And the si2312 is incorrectly biased so I wonder it works.

So from the beginning, do you want it to be linear or pwm? You cannot really have both (you could have a pwm with current limit, but the circuit would have to be a bit different).

 

[CapeCAD]

Hello CapeCAD,

Thank you very much for your advice. I just had a couple questions. What do you mean by "fill in the outer layers, the less copper you remove the better"? The term "flood" is new to me. What do you mean by flood the ground? Sorry for my lack of knowledge, I'm still a novice at PCB design. Thank you for your help.

Amir

Flood and pour are commands used in PCB design by many CAD software to instruct the software to automatically fill in the open areas.

In the attached image ground was flooded, after routing, to fill in and around the existing traces and pads.

 

[dougy83]

I agree with kubeek. The 150uF cap should not be there. Putting caps on the output of opamps can actually lead to oscillation (when there otherwise wouldn't be). This would normally happen with a lower ESR cap though.

Also, the si2312 is in a bad position, and you could move it to between the 0.6V reference and GND, to reduce the reference to 0V and maintain the regulation loop. You won't need the 200 ohm resistor. There is still the chance of oscillation caused by driving the IPB0... FET gate capacitance, so connect a small series resistor between the opamp output and the gate, perhaps 47 ohm.

You don't need the lower two inverting opamps; just a single opamp will perform the regulation task fine.

You should also put a small capacitor (10-100nF) across the MAX6006. Also the decoupling caps near the ATMEGA as mentioned above.

You should be able to do all this on your existing PCB.

 

[alec_t]

The circuit looks overly-complicated for a PWM-controlled current sink to discharge batteries.
Here's a simpler version:

 

[ronv]

Battery load

Here is one with all the suggestions. It has the compensation Dougy talked about so maybe it won't oscillate. Also a different op amp. Are you really using the IPB011N04l?

 

[akahrim]

Flood and pour are commands used in PCB design by many CAD software to instruct the software to automatically fill in the open areas.

In the attached image ground was flooded, after routing, to fill in and around the existing traces and pads.

View attachment 74498

Hello CapeCAD,

I've seen this before on boards before but I never knew what the purpose was. I still don't actually. The feature is available in the PCB layout software I use (PCB Artist). Would you be able to explain how cooper fill areas affect the performance or help the pcb? Is it standard practice to use copper fill areas on the signal planes?

Amir