43,000 count voltmeter from 10 bit ADC - how?

[michaeliv]

Hi,

EDIT: There's actually an 18 bit ADC on the board.

I got a 5 digit pannel voltmeter and noticed that it only has a STM8S003F3 microcontroller(10 bit ADC), a few passives + what looks like an op-amp(couldn't find what the CAH4 SOT23-6 part is).
Question is: how do they achieve this ?
The front side doesn't have any components except for the LED display.

Specs are:
- 0-4.3000v, 0.0001v (0.1mV) resolution
- 4.3-33.000v, 0.001v (1mV) resolution
- ~300ms Update Rate

Item link: http://www.aliexpress.com/item/Green-LED-display-Color-DC-0-4-3000-33-000V-0-36-Digital-Voltmeter-three-Wires/1320678280.html

Thanks!

[bktemp]

U2 is a MCP3421 18bit ADC.

[michaeliv]

U2 is a MCP3421 18bit ADC.

Thanks, how did you manage to find the part from the code ?

[bktemp]

I have used this great little chip before.
After seeing these modules and the SOT23-6 device on its back, the MCP3421 was my first guess and I was right.

[Brutte]

Besides, this ADC is much faster than 3sps so you could get arbitrary resolution result from 10-bit ADC by using dithering.

[michaeliv]

Besides, this ADC is much faster than 3sps so you could get arbitrary resolution result from 10-bit ADC by using dithering.

Yup, that's what I thought was happening, they were doing funky stuff with the 10 bit ADC to get more resolution but less samples.
Do you have any links that describe how this is achieved ?

[retrolefty]

Besides, this ADC is much faster than 3sps so you could get arbitrary resolution result from 10-bit ADC by using dithering.

Yup, that's what I thought was happening, they were doing funky stuff with the 10 bit ADC to get more resolution but less samples.
Do you have any links that describe how this is achieved ?

 Did you not understand that the MCP3421 is performing the A/D conversion? The 10 bit A/D on the micro is probably not even being used.

http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en520011

[michaeliv]

Did you not understand that the MCP3421 is performing the A/D conversion? The 10 bit A/D on the micro is probably not even being used.
http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en520011

Yup I did understand that that is what is going on in this particular circuit - The MCP3421 has 3.75 SPS.
I was responding to Brutte who (I think) was saying that the on-board ADC is faster than 3SPS and could theoretically be used to get higher than 10 bit resolution with decreased SPS. I was asking if this is possible even though this is not what is going on in the voltmeter above.

[bktemp]

Atmel has some application notes on oversampling:
http://www.atmel.com/images/doc8003.pdf
http://www.atmel.com/Images/doc8498.pdf

It can be used to increase the resolution by 1 or 2 bits. More makes no sense in most cases, because of the nonlinearities in the ADC.

[michaeliv]

Thanks a lot, very interesting.
Another question: Seller says that in order to get the full 0.1uV resolution you need to use a separate, isolated power supply ( from the voltage source you are measuring ). Why would this make a difference ( assuming that both have equal noise ). I can't find any mention of this requirement in MCP3421's datasheet - though it does have differential input.

[edavid]

Another question: Seller says that in order to get the full 0.1uV resolution you need to use a separate, isolated power supply ( from the voltage source you are measuring ). Why would this make a difference ( assuming that both have equal noise ). I can't find any mention of this requirement in MCP3421's datasheet - though it does have differential input.

You don't need a separate power supply, just a separate power supply ground (Kelvin connection).

You might want to read the previous thread: https://www.eevblog.com/forum/projects/$5-voltmeter-with-5-digit-%280-1mv%29-resolution/

[Kleinstein]

The MCP3421 not only has the 18 Bit ADC, but also a resonably accurate reference voltage (typical specs are 5ppm/K and +-0.05% absolute).

Oversampling the µC internal ADC does not help that much on a DC reading. However it's quite powerfull if it comes to RMS readings.

[bktemp]

The MCP3421 not only has the 18 Bit ADC, but also a resonably accurate reference voltage (typical specs are 5ppm/K and +-0.05% absolute).

That was the spec from the old datasheet. In the more recent version it is 15ppm/K and no maximum value is specified, so use at your own risk if you need a certain accuracy...
They are great little chips, but without any specifications for the maximum values, they are useless for any decent design without measuring the drift of each unit.

[coppice]

The MCP3421 not only has the 18 Bit ADC, but also a resonably accurate reference voltage (typical specs are 5ppm/K and +-0.05% absolute).

The gain stability is good, but not quite that good. Gain error with temperature is 15ppm/K typical, with no worst case specified. The absolute gain error is 0.05% typical, and 0.35% worst case. There is an additional error of 0.1% typical when the PGA is used. No worst case is specified for this. There is also a DC offset of +-40uV +-50nV/K to allow for at x1 gain. They don't seem to specify what happens to that at higher gains.

[JoeN]

The MCP3421 not only has the 18 Bit ADC, but also a resonably accurate reference voltage (typical specs are 5ppm/K and +-0.05% absolute).

That was the spec from the old datasheet. In the more recent version it is 15ppm/K and no maximum value is specified, so use at your own risk if you need a certain accuracy...
They are great little chips, but without any specifications for the maximum values, they are useless for any decent design without measuring the drift of each unit.

The old Microchip switcharoo...   

[dom0]

The MCP3421 not only has the 18 Bit ADC, but also a resonably accurate reference voltage (typical specs are 5ppm/K and +-0.05% absolute).

That was the spec from the old datasheet. In the more recent version it is 15ppm/K and no maximum value is specified, so use at your own risk if you need a certain accuracy...
They are great little chips, but without any specifications for the maximum values, they are useless for any decent design without measuring the drift of each unit.

The old Microchip switcharoo...   

Here, we have this great (datasheet rev 2) very good (ds rev 3) good part with absolute (rev 2) specs.

[Kalvin]

The tracking speed of the panel meter is quite poor because of the 3 Hz update rate. However, if one can get access to the source code of the panel meter, one could possibly create a small hardware and software hack so that the on-chip 10-bit A/D could be used too. When the signal amplitude changes "very much", the on-chip 10-bit A/D could be used as an approximate A/D for the measurement value. And when the amplitude is stable the 18-bit A/D will be used. In that way one can increase panel's responsiveness.

[Kleinstein]

If one could change the SW, it would be allready a speedup, if the 15 SPS mode(still 16 Bit)  of the AD is used. If 3 of these readings are avaraged even a noticable 50 Hz suppression is possible. 15 SPS is way fast enough for manual readings. Usually close to 4 readings per seconds is not that bad - thats abut the speed of normal handhald DMMs.   

[Kalvin]

If one could change the SW, it would be allready a speedup, if the 15 SPS mode(still 16 Bit)  of the AD is used. If 3 of these readings are avaraged even a noticable 50 Hz suppression is possible. 15 SPS is way fast enough for manual readings. Usually close to 4 readings per seconds is not that bad - thats abut the speed of normal handhald DMMs.

Your suggestion even better as no hardware modification is needed. 

[michaeliv]

Looking at the sister ammeter ( 0.0000-3.0000A, 30mR ) : 'CA2P'  SOT23-6, I'm assuming it's the MCP3421 again with a different code.
I've looked around AliExpress and the MCP3421 being sold have several different codes for the same part number : CACP / CAVV/ CA20 for MCP3421A0T-E/CH
Why do manufacturers do this - give different codes to the same part number ?

[bktemp]

Looking at the sister ammeter ( 0.0000-3.0000A, 30mR ) : 'CA2P'  SOT23-6, I'm assuming it's the MCP3421 again with a different code.
I've looked around AliExpress and the MCP3421 being sold have several different codes for the same part number : CACP / CAVV/ CA20 for MCP3421A0T-E/CH
Why do manufacturers do this - give different codes to the same part number ?

See datasheet page 35: The last two digits are "Alphanumeric traceability code", probably a datecode/production lot marking.

[michaeliv]

See datasheet page 35: The last two digits are "Alphanumeric traceability code", probably a datecode/production lot marking.

Thanks. Also, slightly unrelated - does the ammeter shunt in the above ammeter have a special name ( The wire-type shunt I mean ) ? What is the material called ? If I wanted to buy wire-shunt what would I search for ?

[Kleinstein]

There are also versions of the MCP3421 with different I2C adress.

There is resistor wire available. This can be Konstantan or the better Manganin type. However there is one difficulty: these materials are not easy to solder. On may need extra flux.