Creating QFN-24 3.5mm x 3.5mm


#28

No i did not have to remove any solder mask. The part fit perfectly. Extra pad lengths could be a good thing i think. Short ones would make it harder to solder. But all in all, it seems Van’s pad length was a perfect match. im currently doing testing on my PCB now, and they all seem to be integrating with the chip just fine. Vanepp is a life saver.


#29

It’s all thanks to you really. :slight_smile: i would have been behind couple weeks if i had to redesign the PCB in EAGLE/ALTIUM or try to design the footprint from inkscape, as i would have to learn it from scratch. But i see now, that its definitely required to know some graphics editting tool such as inkscape for my work.

As for the soldering, i was quite surprised to as how to came to be decent. as the solder paste was a mess on the foot print, but i think perhaps the tining and flux played a part to it. heat gun manipulation can be tricky, if it gets too hot, then you could destroy the silk screen.

I do hope they continue to support fritzing more, one thing that is lacking is the routing feature to bend 45 degrees, this i do it manually, although it isnt too bad, but when doing it manually, sometimes it might not be as precise as we expect.


#30

Firtzing has lots of shortcuts, like ALT from a bendpoint starts a trace, and ALT+shift does 45º


#31

I had no idea you could snap on 45º. I just checked and on Linux it uses the “windows” button instead of ALT. Also just simply holding SHIFT while moving a trace tries to snap to a 45º but Fritzing is unsure which side of the trace to base the angle off of.


#32

Certainly for Frtizing Inkscape knowledge is necessary and there is a fair learning curve. For me (having been thought the learning curve at least somewhat :slight_smile: ) the footprint was easy when it was going to be anything but easy for you. It does get easier with experience but getting the experience is painful.

From what I’ve read, that seems to be how it is supposed to go (nice to see that it works in practice :slight_smile: ) the flux is supposed to wick the solder it contains on to the pad which it seems to have done, although I’m still impressed with how good that looks.

They in this case is likely us. Fritzing started out (and got to where it is I think) as a funded research project. As such it has a solid underlying design (if with a few bugs in the implementation), when the research grant ran out, it became an open source project and I think the original developers have mostly had to fight for new research funding and then work in those projects which is leaving little time to develop Fritzing further, so now it becomes up to us to grab the source and try and fix bugs and extend it if we can. To that end I have a development environment up and am poking at the bugs that have bit me, but this too is going to have a big learning curve I expect. I can see where and why the most annoying bug (Fritzing hangs when fed a bad part) occurs, but how to fix it is so far less obvious.

Peter


#33

If you want to learn Fritzing and Inkscape in one hit - it’s quicker - here is a video series. There video series might seem long, but it’s a lot quicker than fumbling around. It took me a year of fumbling around, because there was no one to help back then - you can see my non-replied posts -, so 3 hours is orders of magnitude faster than what I had to do.


#34

Ive read about this. Unfortunately for some reason for me, whenever i ALT+shift, it produces 90 degree bends instead of 45.


#35

With the limited time resource i have. I will see what i can do in order to pitch in and help the community. :slight_smile:

Thank you, although it was mostly a fluke really, i have not much experience soldering SMD type components of such size.

A noble endeavour indeed :slight_smile: i will try to stick around and help as much as i can. I can see the bug about feeding a bad part could hang. It has hung on me quite a while. Also, i notice that the the software would run painfully slow when the gridsize is more dense. I think this seems to be more of system limitations. but weirdly i dont notice the sluggish response when changing the grid size in altium. perhaps, it could use a bit of optimization there. just a thought. But seeing that fritzing is not funded anymore, it would be a definitely challenging to set time resources to work on it.


#36

Thanks. Subbed :slight_smile: . Will look into it


#37

yeah cool, i don’t know that alt forks a new trace from the bend point
but by starting from a bend point and connect to some pads, it moves the bend point a bit
and vise versa , a trave connecting to a bend point moves the bend point a bit


#38

Fritzing is like that.

When ever you make a connect it loses the snap so you have to grab it and snap it back. This is also a good way to check if it’s connected because when you move the bendpoint back you can check that all traces move with it.


#39

You have demonstrated your capabilities with the SMD QFN-24 part.
I think you are at the point, fieryfire, to move from THT resistors, to SMD resistors.
It turns out that smd “1206” pads are fairly close to size of THT pads, so really not difficult to hand solder.
And if you have a few SMD parts, and order a solder-paste mask, … it reduces the assembly time to have as many parts as possible SMD. Plus, SMD parts are considerably less expensive, of course.

I have found that it totally tightens the board up, from a design standpoint.

This month I moved to SMD 1206 parts. After a few boards, and a little practice, I will move to the next size down 0805 parts.

-allan


#40

Thank you allanschwartz, yes looks like that seems to be the case :slight_smile: i am planning to use the SMD components for my next design with the XR18910, i like excited to see how much more compact the design would go to :slight_smile: on the other hand, if you have any experience with bridge amplifiers such as XR18910 do let me know. I think i have power supply which causes voltage fluctuations so i play to add an LDO, and perhaps 50/60Hz IC filters before the amplificiation stage of the strain bridges. im not sure if this is the best idea.


#41

Not specific experience with strain bridges, but high precision op amps often use a linear regulator followed by an lc filter to get cleaner power to the op amps. In theory there is high power supply variation resistance in most op amps, but cleaner power is always better (especially if your input power source is a switching power supply which generally has high frequency noise in it). I’d be tempted to use a 78lxx non ldo regulator from a higher supply voltage (if you have one) to give the regulator more headroom, but if you only have 5V a ldo linear regulator should still give you a cleaner supply voltage.

Peter


#42

Hi Vanepp,

Is it necessary to have an IC filter following an LDO? i was under the impression that i could perhaps use a low noise 250mA LDO - 3V to power the bridges directly. And IC filters just before the IN+ and IN- of the op amps. Design wise i am not sure if this is a stupid thing to do in my case. As my XR18910 (8:1) amplifier doesnt seem to be linearising the output very well. And i suspect that it could be due to a bad voltage regulation stage. Since i am currently powering my bridges directly, from DF robot power supply that has a AMS1117 LDO. I would suspect that i would need to add another LDO after this stage, preferably the https://MAX8902Awww.maximintegrated.com/en/products/power/linear-regulators/MAX8902A.html
to power the bridges. As for the filters before IN+ and IN-, was thinking some type of 50/60Hz rejection filter?


#43

Looking at the XR18910 data sheet (which I didn’t before) I’m not sure the ldo is going to buy you much. There already is one in the chip that appears to be used to drive the bridge power so Exar seems to have already dealt with this. How long are the wires from the chip to the sensor? If you are having noise problems that’s where I would look and perhaps add extra bypass capacitors to the bridge power wires out at the sensor to see if that helps if the wires are long. Putting a scope on the bridge power wires at the sensor to see if there is ripple or noise would be a good bet too. Are you sure the linearity problem is in the Exar chip and not your A/D setup? Either one could be the cause which makes life exciting :slight_smile: . I’d expect the differential input from the bridge to be pretty much cancelling 60hz noise on the input as it should be pretty much equal on the two input wires (assuming they are running parallel to each other and perhaps twisted pair or shielded, so the induced noise is the same and gets cancelled), that’s why the inputs are usually differential (as in this case) from a bridge. If you have a high accuracy voltmeter and a lab supply you could replace the bridge with the lab supply and check the linearity of the XR18910 A/D combination to see if that is the problem or if the non linearity is coming from the sensor. If it is the Exar/ A/D that is non linear them move the power supply to the A/D input and see if it is the Exar of the A/D which has the problem (on a board this may not be easy or even possible to do though). This is why noise problems in an analog circuit are a nightmare because there are so many possible causes and measurements have to be made very carefully so as to not add new unexpected errors to mislead you.

Peter


#44

Hello Peter,

Yes I did infact consult to use the LDO built-in the chip. but this only delivers 20mA maximum . And i want to currently use 4x120Ohm strain gauge excited by 3V which would consume 100mA. So this option is unfortunately ruled out. Other wise it would have been terrific to use the XR18910 to the fullest.

The wires are about 6 to 10 meters. That is the gauges are kept about 6-10 meter away from the bridge.

I would assume the linearity problem would be in the A/D setup. The ADC was not fit with a proper reference, hence this could be a reason where the non-linearity arises.

Your reasoning is extremely right behind the 60Hz noise being differentially inputted. Which is why i am confused that there is infact this noise. I am unsure as to where it arises from. Perhaps ill add a few by-Caps atop the bridge.

Currently here, goes my design. It looks a bit. Lame. I will add the by caps for the xr18910 as recommended by the data sheet.


#45

Ah, I hadn’t considered current (assuming the gauges were low current) the low resistance gauges are an advantage in this case as they reduce the induced noise (it has to be much stronger to affect the readings).

That’a a fair distance and noise will likely be an issue, however my guess is going to be that the problem is in the 3v power supply to the
bridge. Noise such as 60hz on there is likely to be more of an issue as it isn’t differential so bypass caps at the bridge may help (possibly a lot). The balance of the bridge should reduce that but perhaps not by enough.

That’s also a good bet, the more bypass caps the better in general.

While it is likely too late, had you considered the hx711 chip?

If you could put this module at the bridge somehow your 10 meters of wire would have digital signals not analog and a much higher res A/D (24 bit against 10 or 12 for most micro A/Ds). Note I’ve only seen these on Ebay not actually used one so I don’t know if they are good in reality, but Sparkfun has a breakout board and they are usually pretty careful. Another thing to try would be to move a sensor to within a foot or so of the board and see if the same 60hz noise appears without the 10 meters of cable. If it doesn’t then the cable length is the issue, if it does then its board/chip/ad which helps to isolate the problem.

Peter


#46

By induced noise do you mean that, since its low resistance, it draws more current and hence better signal? So which means inorder to have the same relative quality of signal for a higher resistance gauge like the 3000Ohm gauge, my excitation voltage must high enough to allow the same amount of current that would otherwise flow through a 120Ohm gauge?

Yes, it would indeed be an issue, in the future i would want to keep the signal conditioning as close as possible to the sensors, but for now it seems unlikely. Bypass caps between IN+ and IN- before the preamp stage could possibly reduce this noise. The frequency of the signal i would be measuring wouldnt exceed 50Hz. so an RC circuit would be ideal as a filter. I would assume that i must keep the R part as small as possible otherwise i would get even lower current to the op amp.

This was infact the dream IC to use. Hx711. But the biggest problem i have here is the low sampling rate. I can only get a sampling rate at a maximum of 80 SPS. My requirement here is at 1000SPS unfortunately. Otherwise i did consider the Hx711 being a very viable solution.


#47

With the warning that I haven’t actually done much analog after about 30 years ago :slight_smile: , yes that is the theory. You are basically in the area of signal to noise ratio, and the more power that the interference has to provide to change your signal level the better off you are.

A signal that close to the pass band of the filter is going to be a problem. I expect you may be better to look at an active 60hz notch filter (an rc filter is unlikely to be able to attenuate much that close to the corner frequency). The problem is (although technology after 30 years may have fixed this :slight_smile: ) getting an active filter that doesn’t add undesirable dc offsets and/or temperature effects. Analog has always been difficult and I don’t think it has gotten all that much easier over the years.

Yep sample rate could be a problem. I expect most of it is in the 24bitA/D, getting that much accuracy generally takes a fair bit of time. Although with a 50hz or less input signal change 1000 SPS seems sampling overkill. You won’t see much change between samples because the gauge won’t react that fast (at least I wouldn’t expect it to). That is why the 80 SPS limit on the Hx711 sounds about right to me (they look to be assuming about a 40hz max update rate or possibly even smaller)

Peter