Thank you Kerry for this nice and very well narrated teardown. I am impressed by the build quality of this meter. Your knowledge and patient explanation is also very well appreciated. One thing that as a retired technician surprises me: Why does one maker not stick to a common box size? As a former industrial equipment designer myself, I would have built in a shielded toroid transformer and used a very quiet fan. I am still in close contact with scientists using high tech gear. We often scratch heads to as why very expensive gear sometimes has a rather short service life. The button cell found inside here is usually good for years. But its placement could be put up to discussion to what might happen if its voltage reaches a critical level causing operator frustration. I bet some makers will place an indicator onto the front advising if a fuse had blown.
The E-core transformer is probably OK and it is done as an economy measure compared to a shielded toroidal power transformer. The meter is usually set up to take an average reading across n number of power line cycles to help null out the effect of stray fields at power line frequency. Keysight does the same thing - E-core transformer and averaging in the 34460A multimeter which is what I presume this meter is aimed to compete against. Keysight also uses the LM399 voltage reference as do a lot of 6.5 digit multimeters. I don’t recall what ADC Keysight uses. It looks like there is a fair amount of quality per Dollar in the Uni-T meter. You give up very little if any performance when used as a bench meter, but save around $500 versus the equivalent Keysight. Uni-T’s warranty is longer too.
You have not paid attention to the current source for measuring resistances and voltage dividers, which allow you to get 2.5V for ADCs and other nodes.
The LM399 is around for maybe 30-40 years. It is very stable, yet not very accurate. I guess, the stability of the voltage reference is most important. Also the stability of voltage divider resistors The Caddock maybe is listed as 0.01%. Accuracy can be provided by digital calibration. I suspect there is a EEPROM somewhere. I suspect, that the 6.5 digit accuracy refers to the DC +/-10V range... the only range with GOhm input resistance. All other ranges have much lower input resistance, because it is impossible to manufacture GOhm resistors with any degree of stability and accuracy. RMS to DC chip is ancient as well and you are lucky if the AC measurement is 3.5 digits accurate. It is a relatively cheap instrument and 6.5 digit maybe a bit exaggerated. These days, 24b delta sigma converters are everywhere and with additional post filtering you may get a stable measurement, but is it accurate? Why do they have an old fashion (low noise) power transformer and then switching power supplies to get 3.3V ? It is a good DMM... anything better than that and you are spending real money.
@@KerryWongBlog Yes this ADC is very suitable for a high resolution ADC. The difficulty in high resolution DMMs is not the ADC but just about everything else like clock jitter, high frequency noise at the input which will be aliased into the pass-band. Note, it is a SINC filter, not a FIR. Even the quality of the capacitors right at the input is most important... better be NPO. Then there are thermocouple effects as well as RF filter at the input opamp. Common mode noise rejection is huge according to data sheet but only when the frequency of the noise is within the opamp's range. RF at the input is rectified and appears as offset error. Since the analog part of the DMM is galvanic isolated, the capacitors from both sides of the input to ground must be closely matched.
That switching 3.3V supply just goes directly off to the front panel. That'll be the digitial-side control MCU, front panel display, etc.. There's no measurement frontend up there so that doesn't matter about a bit of noise or ripple.
@@leonerduk no, the switching regulator generates noise into the MHz range, which is radiated all over the board. If this energy hits the front-end (whose common mode rejection is only good for low frequency) the RF will be rectified and cross-modulated with the clock of the ADC, just like in a superheterodyne receiver. The result is sub-harmonics (noise) which alters the offset of the ADC... measurement error. At 120dB dynamic range (ADC) attention to detail matters.
Mhmm ... Siglent SDM3065X or UT8806E ? :D The Uni-T does not have the (siglent patented) rust as Dave would say :D Guess the Siglent would be the "more trustworthy" (for the price)?
![[UA] NAVI vs MOUZ | BO5 | IEM Rio 2024](http://i.ytimg.com/vi/CMBYk0cbGoA/mqdefault.jpg)
Thank you Kerry for this nice and very well narrated teardown. I am impressed by the build quality of this meter. Your knowledge and patient explanation is also very well appreciated. One thing that as a retired technician surprises me: Why does one maker not stick to a common box size? As a former industrial equipment designer myself, I would have built in a shielded toroid transformer and used a very quiet fan. I am still in close contact with scientists using high tech gear. We often scratch heads to as why very expensive gear sometimes has a rather short service life. The button cell found inside here is usually good for years. But its placement could be put up to discussion to what might happen if its voltage reaches a critical level causing operator frustration. I bet some makers will place an indicator onto the front advising if a fuse had blown.
The E-core transformer is probably OK and it is done as an economy measure compared to a shielded toroidal power transformer. The meter is usually set up to take an average reading across n number of power line cycles to help null out the effect of stray fields at power line frequency. Keysight does the same thing - E-core transformer and averaging in the 34460A multimeter which is what I presume this meter is aimed to compete against. Keysight also uses the LM399 voltage reference as do a lot of 6.5 digit multimeters. I don’t recall what ADC Keysight uses.
It looks like there is a fair amount of quality per Dollar in the Uni-T meter. You give up very little if any performance when used as a bench meter, but save around $500 versus the equivalent Keysight. Uni-T’s warranty is longer too.
I'm definitely going to buy one
Good look affording it. If you need to ask the price you can't afford it. Will cost several hundred $/£/€ and the rest.
رائع جدا استاذ
Those are CS5173 buck regulators.
You have not paid attention to the current source for measuring resistances and voltage dividers, which allow you to get 2.5V for ADCs and other nodes.
Thanks for pointing that out!
Cheers !
at 9:45 just above the ADC there is another botch, looks like they put extra a cap.
Yep, I also took a picture and posted on my website.
The LM399 is around for maybe 30-40 years. It is very stable, yet not very accurate.
I guess, the stability of the voltage reference is most important. Also the stability of voltage divider resistors The Caddock maybe is listed as 0.01%. Accuracy can be provided by digital calibration. I suspect there is a EEPROM somewhere.
I suspect, that the 6.5 digit accuracy refers to the DC +/-10V range... the only range with GOhm input resistance. All other ranges have much lower input resistance, because it is impossible to manufacture GOhm resistors with any degree of stability and accuracy. RMS to DC chip is ancient as well and you are lucky if the AC measurement is 3.5 digits accurate.
It is a relatively cheap instrument and 6.5 digit maybe a bit exaggerated. These days, 24b delta sigma converters are everywhere and with additional post filtering you may get a stable measurement, but is it accurate?
Why do they have an old fashion (low noise) power transformer and then switching power supplies to get 3.3V ?
It is a good DMM... anything better than that and you are spending real money.
Have you looked at AD7175-2's datasheet? The ADC is quite capable and the offset error is in the uV range.
@@KerryWongBlog Yes this ADC is very suitable for a high resolution ADC. The difficulty in high resolution DMMs is not the ADC but just about everything else like clock jitter, high frequency noise at the input which will be aliased into the pass-band. Note, it is a SINC filter, not a FIR. Even the quality of the capacitors right at the input is most important... better be NPO. Then there are thermocouple effects as well as RF filter at the input opamp. Common mode noise rejection is huge according to data sheet but only when the frequency of the noise is within the opamp's range. RF at the input is rectified and appears as offset error. Since the analog part of the DMM is galvanic isolated, the capacitors from both sides of the input to ground must be closely matched.
That switching 3.3V supply just goes directly off to the front panel. That'll be the digitial-side control MCU, front panel display, etc.. There's no measurement frontend up there so that doesn't matter about a bit of noise or ripple.
@@leonerduk no, the switching regulator generates noise into the MHz range, which is radiated all over the board. If this energy hits the front-end (whose common mode rejection is only good for low frequency) the RF will be rectified and cross-modulated with the clock of the ADC, just like in a superheterodyne receiver. The result is sub-harmonics (noise) which alters the offset of the ADC... measurement error.
At 120dB dynamic range (ADC) attention to detail matters.
Mhmm ... Siglent SDM3065X or UT8806E ? :D The Uni-T does not have the (siglent patented) rust as Dave would say :D Guess the Siglent would be the "more trustworthy" (for the price)?
Siglent rust is a thing from the past. Not happening since at least 5 years.
@@ofgjf Good to know, thanks!