Hioki RM3545 mini teardown

[gamalot]

Just a few pictures 

It's very easy to open the case, much easier than all my other instruments.

The main processor is TI's LM3S5C31 (cortex-M3)

[gamalot]

Analog board, the shield is also very easy to be removed 

It has a temperature measurement unit near the rear panel.

[gamalot]

The main ADC is ADS1251, voltage ref is LT1027, and there is an AE resistor 

[lukier]

Some DC-DC converters in the top right corner. Maybe the AC-DC supply module provides only one voltage, earth mains referenced that goes to the digital board via the white cable and then the same output goes via the green cable to the DC-DC converters that provide isolation and analog voltages (bipolar supply etc). Then five TO-220 flapping in the breeze as Dave would say are perhaps LDOs to make the DC-DC outputs a bit more quiet. Interesting, I've always expected a transformer + linear supply, at least for the analog section, in precision instruments. Keithley's DMM7510 is a notable exception with its crazy coax transformer.

Nice precision resistors around the analog board, with some AE 10K000 being the most prominent. ADS1251, hmm 19 effective bits, not really 6.5 digit, but almost.

Something is crazy there. There seem to be 37 single channel optoisolators on the digital board alone (probably floating the signals on the big DB connector), and then 25 optoisolators on the analog board - crazy! Why not put MCU/CPLD on the floating side and just transfer some SPI over isolation barrier.

Thanks again for the teardown, this is very interesting and unconventional design.

[gamalot]

Main power supply is switch mode, and there are five 3-terminal regulator and filters after it 

[lukier]

Aaa so these chunky SMD inductors are just filters? (T506 etc) I thought these are some isolated DC-DC converters, but I don't see controller chip and feedback optocoupler.

[PA0PBZ]

Something is crazy there. There seem to be 37 single channel optoisolators on the digital board alone (probably floating the signals on the big DB connector), and then 25 optoisolators on the analog board - crazy! Why not put MCU/CPLD on the floating side and just transfer some SPI over isolation barrier.

Yeah, I was wondering the same, this is just insane. Is that where optocouplers breed? 

[gamalot]

Current output stage, The 4-pin black square block in the up-left corner, I can read it's part number clearly, I guess it's Panasonic's optical relay

That's all, thanks 

[gamalot]

Aaa so these chunky SMD inductors are just filters? (T506 etc) I thought these are some isolated DC-DC converters, but I don't see controller chip and feedback optocoupler.

I think maybe it is just a common mode filter. 

[gamalot]

Some DC-DC converters in the top right corner. Maybe the AC-DC supply module provides only one voltage, earth mains referenced that goes to the digital board via the white cable and then the same output goes via the green cable to the DC-DC converters that provide isolation and analog voltages (bipolar supply etc). Then five TO-220 flapping in the breeze as Dave would say are perhaps LDOs to make the DC-DC outputs a bit more quiet. Interesting, I've always expected a transformer + linear supply, at least for the analog section, in precision instruments. Keithley's DMM7510 is a notable exception with its crazy coax transformer.

Nice precision resistors around the analog board, with some AE 10K000 being the most prominent. ADS1251, hmm 19 effective bits, not really 6.5 digit, but almost.

Something is crazy there. There seem to be 37 single channel optoisolators on the digital board alone (probably floating the signals on the big DB connector), and then 25 optoisolators on the analog board - crazy! Why not put MCU/CPLD on the floating side and just transfer some SPI over isolation barrier.

Thanks again for the teardown, this is very interesting and unconventional design.

The basic accuracy is 60ppm, so 19 effective bits is good enough. 

[gamalot]

Something is crazy there. There seem to be 37 single channel optoisolators on the digital board alone (probably floating the signals on the big DB connector), and then 25 optoisolators on the analog board - crazy! Why not put MCU/CPLD on the floating side and just transfer some SPI over isolation barrier.

Yeah, I was wondering the same, this is just insane. Is that where optocouplers breed? 

I was shocked by those optocouples too! 

[Kleinstein]

The big blocks (T506) and similar seem to be transformers. They should provide isolation from the digital side to the analog side. There are 2 small SOT23 transistors for each transformer - so maybe a kind of royer converter or similar (though I am missing the series inductors for a classical royer). There is no need to have feedback  - there are lots of 78xx (TO220) behind it to do the fine regulation.

The effective number of bits is not that critical. For slower readings averaging can give a higher effective number of bits, even more than 24. The limiting factor is more the INL and maybe gain / offset drift. If the INL is well behaved it might be even possible to use some compensation / corrections.

[gamalot]

Aaa so these chunky SMD inductors are just filters? (T506 etc) I thought these are some isolated DC-DC converters, but I don't see controller chip and feedback optocoupler.

I think maybe it is just a common mode filter. 

I think I was stupid! They must be DC-DC converters, cuz there is no way to have negative voltage without them!

I need to re-open it tomorrow and see if there is anything on the bottom side of analog board, it's 5 in the morning, I wanna sleep ...... 

[gamalot]

The big blocks (T506) and similar seem to be transformers. They should provide isolation from the digital side to the analog side. There are 2 small SOT23 transistors for each transformer - so maybe a kind of royer converter or similar (though I am missing the series inductors for a classical royer). There is no need to have feedback  - there are lots of 78xx (TO220) behind it to do the fine regulation.

The effective number of bits is not that critical. For slower readings averaging can give a higher effective number of bits, even more than 24. The limiting factor is more the INL and maybe gain / offset drift. If the INL is well behaved it might be even possible to use some compensation / corrections.

YES! I think may be  maybe you are right, the royer convertor has lower noise, it is perfect to be used here, I will try to figure it out tomorrow. 

[gamalot]

The big blocks (T506) and similar seem to be transformers. They should provide isolation from the digital side to the analog side. There are 2 small SOT23 transistors for each transformer - so maybe a kind of royer converter or similar (though I am missing the series inductors for a classical royer). There is no need to have feedback  - there are lots of 78xx (TO220) behind it to do the fine regulation.

The effective number of bits is not that critical. For slower readings averaging can give a higher effective number of bits, even more than 24. The limiting factor is more the INL and maybe gain / offset drift. If the INL is well behaved it might be even possible to use some compensation / corrections.

YES! I think may be  maybe you are right, the royer convertor has lower noise, it is perfect to be used here, I will try to figure it out tomorrow. 

They are all switching power supplies without negative feedback. 

[gamalot]

Another interesting thing is how they mounted LEDs (red and green) on PCB behind the power button 

[gamalot]

Since a forum member PM me asking for the firmware for the MCU LM3S5C31 on the RM3545, I put it here in case someone else needs it.