How can a chip (HX711) work without a ground pin connected?

[Tonn]

Hi,

I am working with a HX711 board. They are used with load cells and weight sensors. You can find them in many Arduino projects and they are probably sold by the thousands.



This is the schematic. It's copied all over the internet.



By comparing the board, the schematic and the HX711 datasheet I found several issues, marked in the schematic.
I's #4 that has me confused.

The HX711 chip has three power pins (VSUP, DVDD and AVDD), but only one ground pin (AGND).
The AGND pin should be connected to the supply ground but it isn't.

Power is supplied to JC9.
VCC (JC9, pin 1) is connected according to the datasheet.
Ground (GND - JC9, pin 4) goes to RATE (input pin, configures the chip for 10 samples/sec) and X1 (crystal pin, configures the chip for using the internal oscillator). It does not go to AGND.

The datasheet confirms that the power ground should go to AGND. No surprise there.

I checked five different boards from two suppliers (at least two different batches) and indeed AGND is not connected to GND.

Yet, the boards work well.
What am I missing?

Thanks,
Tonn

[theworldbuilder]

Are you sure they aren't connected on the PCB layout despite what the schematic says?
Like a connection under the IC or something along those lines?

[In Vacuo Veritas]

You mean like how pin 14 in connected to GND via the symbol at legend 1? And the trace on the PCB as well?

I don't follow the question. Granted, that's a piss-poor schematic.

[wraper]

I checked five different boards from two suppliers (at least two different batches) and indeed AGND is not connected to GND.

Did you measure actual resistance between AGND and GND?

[NiHaoMike]

The ESD diodes can provide a ground path, but I would expect charge injection to adversely affect the accuracy of a precision analog part. (Had that happen with a dsPIC in my senior design project since I only used series resistors to connect it to the 5V Arduino that handled the networking.)

[SiliconWizard]

I suggest looking at the HX711 datasheet.

They connect the AGND pin to the common GND.

I'm just guessing: in the board you're mentioning, AGND is made accessible through the J1 connector (E-). Probably comes from people that still think it could be a good idea to separate digital and analog grounds. Long story short: it usually isn't. (A more detailed point on this could take pages...)

E- on J1 is obviously meant to be connected to ground and not be left floating.
Doing so would not make any sense.

[Tonn]

Thanks for responding.

I measured the resistance between AGND and GND: 11 Mohm one way and OL with probes reversed.

So I did a diode test and got 580 mV, indicating an internal diode from AGND (anode) to GND (cathode).

I then powered the board and measured the voltage on AGND. It started at 470 mV and slowly rose to 840 mV (capacitor charging?).
840 mV indicates a diode drawing a current that is relatively large for that particular diode.

A clamping diode on the RATE input pin, or similar, would explain this.

Could that be it?
The chip being powered through an internal protection diode? And no one ever noticed this obvious design flaw?

[edavid]

AGND is not necessarily a power connection.  It could be a reference voltage that is only connected to some MOSFET gates, so no current is drawn.

Have you tried connecting it to ground to see what happens?

(Ooops, I also thought there was a DGND.)

[SiliconWizard]

AGND is not necessarily a power connection.  It could be a reference voltage that is only connected to some MOSFET gates, so no current is drawn.

So there would be absolutely no ground connection on this chip? 
As per the datasheet, it IS meant to be connected to ground.

[SiliconWizard]

A clamping diode on the RATE input pin, or similar, would explain this.

Except in very specific cases, IC pads usually have internal clamping diodes, so yes.

Could that be it?
The chip being powered through an internal protection diode? And no one ever noticed this obvious design flaw?

Unless people connect E- to GND externally, in which case it would work properly. But the schematic would let you believe that you shouldn't.
So yes, I think you're correct.

As said in my post above, if you weren't correct, a chip without ground (or otherwise return path for the power supply) connection? That would seem to defy electronics laws. Wouldn't it?

[wraper]

AGND is not necessarily a power connection.  It could be a reference voltage that is only connected to some MOSFET gates, so no current is drawn.

So there would be absolutely no ground connection on this chip? 
As per the datasheet, it IS meant to be connected to ground.

There is GND connection to IC, digital GND. The issue is that resistive bridge is connected to AGND but not actual GND, therefore is powered through AGND pin.

[SiliconWizard]

AGND is not necessarily a power connection.  It could be a reference voltage that is only connected to some MOSFET gates, so no current is drawn.

So there would be absolutely no ground connection on this chip? 
As per the datasheet, it IS meant to be connected to ground.

There is GND connection to IC, digital GND. The issue is that resistive bridge is connected to AGND but not actual GND, therefore is powered through AGND pin.

What is the digital GND pin on this IC?

[wraper]

What is the digital GND pin on this IC?

Oops, yep there is no digital GND pin, it's powered through protection diodes on X1 and Rate pins.

[Tonn]

The board works well with both E- to ground or not. It seems the HX711 is not much affected by the diode voltage drop.
Accuracy or noise may be affected, but I don't have the means to verify that.

It seems I can call this one solved. Very educational.
@NiHaoMike: nice catch; you made me look for a diode, not just connection or resistance.

Everyone thank you very much for your help.
Tonn

[Dave_PT]

Sorry to come dig up this topic.

But I have some problems with these chinese boards from a project point of view  .

Clearly they have design mistakes (GND, etc...). But can someone explain me the feedback resistances?
Why they use R1=20k and R2=8.2k?
To get a AVDD of 1.77V?  Why?


[EDIT]
Vbg=AVDD*R2/(R1+R2)
AVDD=Vbg*(R1+R2)/R2 -> the datasheet equation is wrong.



Another thing, has anyone used the oscilloscope to measure the AVDD? Can anyone explain me why the AVDD comes up as a "square wave" of about 50Hz?

Actually I can use the IC to take measurements ... but I can not figure out how the analog supply regulator works.


Thank you all

[free_electron]

A clamping diode on the RATE input pin, or similar, would explain this.

Except in very specific cases, IC pads usually have internal clamping diodes, so yes.

Could that be it?
The chip being powered through an internal protection diode? And no one ever noticed this obvious design flaw?

Unless people connect E- to GND externally, in which case it would work properly. But the schematic would let you believe that you shouldn't.
So yes, I think you're correct.

As said in my post above, if you weren't correct, a chip without ground (or otherwise return path for the power supply) connection? That would seem to defy electronics laws. Wouldn't it?

not necessarily.
'ground' is just a reference ...
just like 'power'

you could make a chip that refers everything to VCC ... in fact, that has been done. Many POTS (Plain Old Telephone System)  SLICs (Subscrier Line Interface Circuit ) work that way , simply because POTS uses negative voltages. The chip is powered with 'ground' and -48 volts. 'ground' is the positive terminal. All signal are referred to that 'ground'. to bring out the control signals you use level shifters using two transistors to a positive voltage 'above' ground. modern slics have those built in.