My try at Daves DC load? opinions

[ahnuts72]

Looking for general comments on layout routing and such.

[c4757p]

Well, it looks very... small. Can you upload something bigger?

[BravoV]

Q1 position in the middle, is that intentional ? Assuming you've thought about the Q1 heatsink's mount position.

[c4757p]

One thought - pick one point to call 'ground', and try connecting everything there. Basically a star ground. This way, currents being carried through ground traces won't cause a voltage offset.

Alternately, use a ground plane, but only after you've traced out all ground paths and made sure they are decent. You still should keep the heavy current-carrying "ground" trace out of the plane.

Example of what I mean is attached.

[ahnuts72]

BravoV  Q1 will be mounted separately on an external heatsink.

c4757p thanks for the picture I knew the ground wasn't exactly right I looked at it for ever trying to figure a way of
doing it different but I'm still trying to learn this layout stuff and not using the auto router in diptrace .

[c4757p]

Keep autorouters well away from analog circuits, they screw them up massively. This is a really simple circuit to lay out.

Any time you need to read a small voltage drop in a sense resistor (really, a small voltage, period), you should be using something like Kelvin sensing. Don't mix current and voltage - give them separate traces. Remember that "ground" can both be a return path for current and a reference point for voltage, and you should pay close attention to how they interact.

Sometimes you'll hear people advocate one grounding scheme for pretty much all purposes, but grounding is something that has to be done on a case-by-case basis most often. Note that I recommended something similar to a star ground, but didn't follow that completely in my example. If I connected the load input's ground back to the star instead of directly to the sense resistors, then the PCB traces would add to the sense resistance. With a 1 Ohm sense resistor, it would be pretty minimal (especially with those big-ass traces you're using everywhere), but still, it should be avoided.

One more thing: you have the pins of Q1 ordered source-gate-drain, as far as I can tell, but the transistor itself will be gate-drain-source. I'm sure you know this, but you should try to keep the pinout consistent to minimize the chance of accidentally wiring it wrong.

And another: The point where you read the top of the sense resistors has the same problem as your grounds. Don't mix current and voltage. Run the trace out of pin 6 of U1 straight down to the sense resistors.

And a third  : You ought to post the schematic along with the PCB layout next time. Makes it easier to read.

[ahnuts72]

Thanks for the reply.
Should the sense resistor traces be smaller?
I figured they where carrying the most current so they should be a bit bigger.

I was gonna put up the schematic but I haven't back annotated a couple of changes i made
while i was laying it out and I haven't added the fan control schematic to it ill try to get that done and add it.

[c4757p]

Should the sense resistor traces be smaller?
I figured they where carrying the most current so they should be a bit bigger.

Make them as big as possible. If you weren't paralleling resistors, it wouldn't really matter, as long as you separate current and sense like I said (and they're actually big enough to handle the current), but with ten resistors in parallel, the trace resistance will contribute a small amount. Best to make it as small as you can.

[ahnuts72]

Something like this. Any better?




Also how do you use a thumbnail then make it bigger when it is clicked on.

[BravoV]

Don't forget 4 mounting holes at the corners.

[c4757p]

Yes, that looks good to me!  Well, except that I do have one question: what's U2? It may or may not be acceptable that its ground merges with the sense resistor ground before the star. Sorry, I'm not really familiar with Dave's load, and I'm lazy as hell so I'm not going to look it up... 

[ahnuts72]

U2 is the transistor for the the fan.
Dave's load doesn't have a fan but as this is a learning experience
for me I thought I would try to add one.

Not sure why diptrace put U instead of Q.

Maybe like this or is it bad to run the ground all the way around?

[c4757p]

U2 is the transistor for the the fan.

Ah, I should have recognized that!

Looks fine to me.

Also how do you use a thumbnail then make it bigger when it is clicked on.

Stop using TinyPic and just upload it as an attachment.

[ahnuts72]

Thanks for the help its much appreciated.

[ahnuts72]

Here's the schematic.

[RobB]

At the risk of seeming to be a nazi but you really should convert your .bmp bitmap files to jpeg.
For example the picture for the schematic alone is over 1 megabytes. The equivalent jpeg is just about 30 kilobytes or less than 1/30 of the size without scaling or loss of detail.
Now think of how many time this posting is likely to be read and the pictures maybe downloaded you will see that the use of a .bmp has unnecessarily increased the data sent adding up to a larger bill for the forum hosting.
Again not trying to be a asshole. It's just commom practice and courtesy.

Cheers
Rob

[Dave]

Converting it to JPEG will ALWAYS result in loss of detail.
PNG is the way to go for images with sharp lines.

[IanJ]

Here's the schematic.

What are you doing with that unused op-amp U1,4.............best thing to do is to actually use it. Set it up as a unity gain buffer and tie the input to some part of your circuit or to a simple voltage divider.

Leaving the inputs unused and floating can lead to a saturated output and can increase current consumption to the device. The inputs can also pick up electrostatic fields or noise and oscillate the output....which can then cross-talk across to the other op-amps in the device..............etc etc etc.

If you tie the inputs together it doesn't actually help......both are still effectively floating.
Don't simply tie both inputs to 0vdc even with resistors, it can stress the op-amp, increase current consumption etc.

Ian.

[Alana]

Is Q1 pins in order Drain Gate Source? Check your datasheet because you may have an error.

[ahnuts72]

Thanks for the tip Ian I wasn't sure of it would hurt to leave it floating.

Q1 is going to an external heatsink so I wasn't worried about the pinout

Sent from my Milestone X using Tapatalk 2

[brabus]

Is Q1 pins in order Drain Gate Source? Check your datasheet because you may have an error.

I am also worried about that.

Also, provide a compensation network like the one you can see in this thread (R6-C5): https://www.eevblog.com/forum/projects/dummy-load-problem/msg203534/#msg203534

The compensation network lets you control the dynamic response of the load, and easily solve stability issues.

I remember having made a dummy load which seemed perfect to the multimeters measurement, but exhibited horrible instability if analysed with an oscilloscope! Be careful, and check this! 

[c4757p]

Is Q1 pins in order Drain Gate Source? Check your datasheet because you may have an error.

I am also worried about that.

Yeah, I addressed that earlier. (Though I suspect it might be too late now,) please don't depend on remembering to wire it correctly. Murphy thinks your reliance on memory is hilarious.

Also, provide a compensation network like the one you can see in this thread (R6-C5): https://www.eevblog.com/forum/projects/dummy-load-problem/msg203534/#msg203534

The compensation network lets you control the dynamic response of the load, and easily solve stability issues.

Good point. These do tend to be unstable.

What are you doing with that unused op-amp U1,4.............best thing to do is to actually use it. Set it up as a unity gain buffer and tie the input to some part of your circuit or to a simple voltage divider.

In the name of doing things "right", yes, this is important - but in my experience, LM324 really doesn't care all that much. It tends not to oscillate and the power consumption doesn't increase.

[ahnuts72]

Sorry I missed everyone's comments been away from comp.

Here is the board etched and almost done and the front
panel design for the enclosure.
I know the soldering isn't that great but been awhile been
using the breadboard for everything.

c4757p I can hear you yelling about the pinout already
Ill try to incorporate your suggestions into revision 1
as this was just me getting used to making my own boards again.

[Harvs]

Seriously you need to compensate the feedback network.  It doesn't take much, and even if you don't etch another board, just hack it in there.  How you've got it at the moment will oscillate under certain conditions, no two ways about it.

[ahnuts72]

OK from all the reading Ive done on compensation networks
these are the 2 Ive come up with,as I'm still learning all this stuff
which one is best for this application.
Or does it really matter?
Also any chance someone could recommend a good book on op-amps
preferably in print as I hate eBooks.
I'm reading Practical Electronics For Inventors right now but something
on just op-amps would be nice to have.