HOPI HP-9800 Power meter (recent) differences and safety modifications

[Gyro]

I recently purchased a UK variant HOPI HP-9800 Power meter. I've been envious of Big Clives for a while, I've tried a few cheap ones but have been disapointed with their performance, particularly on Power factor accuracy. The Hopi uses a dedicated CS5463 electricity meter IC for all the heavy lifting for measurement and calculation with good accuracy (energy data linearity of +/-0.1% of reading over a 1000:1 dynamic range as a headline figure) although this is of course dependent to some extent on the surrounding circuit implementation. Clive's review video is 4 years old now so I thought it might be useful to highlight the H/W changes that have occured since. Clive's review can be found here:

 


So what's unchanged:

- Still based on the CS5463 (good)
- Still has a Chinese universal exposed contact 'death socket' (bad)
- Still has sprung speaker terminals for bare wire connection (very bad for a 20A rated unit!)
- Still has the slide latch removable rear flap exposing all of the live connections (incredibly bad!)
- Still has a captive mains lead (in this case a with a fused UK plug but still 1mm² cable, at least it seems to be copper this time) but with the same tack soldering.

What's changed since Clive's review:

- The PCB is now a little shorter and the tiny 20A fuse has been removed. Probably good for UK buyers (fused mains plug) where it probably posed more of an explosion risk (see below) than protection, but maybe a concern for other countries.
- On first look, a normal 13A fuse in the plug, but on removing it, it becomes clear that HOPI are determined to give the customer their full 20A money's worth by soldering a 1mm² across the end caps to increase its rating to 2013A (probably higher!).
- The number of crystals has been reduced from two to one (on the CS5463). The microcontroller now appears to be internally clocked even though the CS5463  has a CPU clock output pin available, this just goes to a test pad. The datasheet indicates a master clock range of 2.5 - 20MHz, with the internal oscillator suporting crystals of 2.5 - 5MHz (external clock beyond).  Despite this, HOPI seem to have got away with fitting a 16MHz crystal.
- The 8051 based STC8H microcontroller has been changed from a 48(?)pin QFP to a TSSOP20 package. At the same time, the number of 74HC164 shift registers driving the displays has increased from 2 to 4. Probably good as an HC164 is easy to replace (compared to a dead GPIO).
- The FRAM chip has been deleted, presumably replaced by internal EEPROM on the micro.
- The little SMPS has changed a little but is performing the same function. Isolation slots seem similar.
- Mine came with the USB interface populated. From what I can see of Clive's unpopulated one, the layout is different.
- I have yet to see the "Frequence" (still) display deviate from 50.00Hz. Whether this is due to our very stable mains frequency (being smoothed?), or whether it defaults to either 50 / 60 Hz remains to be seen. It would have been nice to see a more useful readout in there, for instance, one of the several other power measurements the CS5463 calculates

I've attached some photos. On the photo with the rear of the case removed, I had already snipped and removed the wires to the speaker terminals and the terminals themselves.

In part 2, I will cover my mods...

[Gyro]

... and so to the modifications:

Mains outlets:

Firstly the speaker terminals had to go. I have a Cliff 'QuickTest' (recommended) for bare wire testing, much safer as the clips are isolated when the lid is lifted and the Earth is brought out too. The front panel holes are covered by a thick vinyl panel sticker for now.

Secondly the 'death socket'. Legrand produce a compact push-fit UK 13A socket (RS stock no. 815-9696) which fits reasonably in the front panel (they make these modular sockets for other country standards too, probably footprint compatible, I didn't check). The body width fits the cutout nicely, with just a few minor notches to be cut. The bottom of the cutout needed to be extended (by 10-12mm irrc) but is easily done. The retaining clips are intended for a thicker panel, this can be addressed by an 'overlay' frame cut  to size from something like FR4 and slipped over the rear of the socket inside the case. I opted for slow setting epoxy which is nice and secure. Luckily the depth of the socket is fine - the original death socket had to have its rear contact tabs bent flat to fit!

Mains Wiring:

The short captive mains lead was a liability, and being only 1mm², unsuitable for 13A, let alone 20A. I cut the bottom of the case to accept a chassis IEC plug. This of course limits the current rating to 6A or 10A, but is sufficient for my needs and the original lead was unfit for purpose anyway.

I replaced all of the internal mains wiring with decent 1.5mm² stranded copper (the original socket ones were solid core), again now overkill for the IEC connector but I only though of that afterwards! The Legrand socket has screw terminals so needs ferrules.

USB:

Lastly, the USB port, this requires a little description. It actually wouldn't be hard to retrofit to non-populated units. On first glance, the isolation looks hideously dangerous but on closer examination is only moderately dangerous as it stands. The interface to the microcontroller is via serial data, the Rx and Tx signals are galvanically isolated by a pair of EL817 opto-isolators. Note that everything on the PCB other than the USB section is mains live referenced! The USB interfacing is performed by a CH340N (the variant with internal oscillator rather than crystal). The USB area is also footprinted for what looks like another microcontroller and accompanying crystal. I'm not sure what this would be for, the unit speaks Modbus via a micro-USB connector. Power for the USB section is of course derived from the USB connector rather than the unit itself.

In terms of creepage and clearance, the most obvious weak point is next to the unpopulated microcontroller footprint (and USB +5V on the underside). The PCB is slotted here but track clearance is still only around 3mm. I inserted a strip of 1mm insulating plastic into the slot with generous overlap to address both. For the serial data and opto-isolators, a little care has actually been taken in the layout. The mains referenced signals are routed to the area between the two isolators, with the USB side signals taken from the outer ends, routed away from the area and isolated by being on the bottom side of the 1.6mm thickness of the PCB. The Isolators look like through hole types but are the surface mount lead option (not chopped!), with the mains referenced connections on the top side of the PCB. The closest creepage distances are 4mm. This is short of the 8mm reinforced insulation distance that would be required for Class II (which the USB interface is as it stands) but ok for basic insulation (Class I). To address this, I have taken an earth lead directly to USB Gnd / Connector shell. This probably isn't completely foolproof in terms of failt excursions on the USB data signals, but is as close as it's possible to get.


I think the result of the mods makes the HOPI as safe as practically possible, what you see when you open the rear flap looks positively safe compared to the original - it isn't of course, the shunt and PCB are still touchable. The final step that I haven't carried out yet will be to insert a small screw to lock the slider, and possibly glue the seams, there's no reason at all to open it now even to replace a non-existant fuse.

I'm not sure how much use the USB interface will get, it responds correctly when fed the right Modbus command using Terminal (9600 baud), but I haven't got either of the two Github interface S/Ws working yet. At the moment I'm working on the basis of 'if it's present, it ought to be safe'.

[elecdonia]

For many years I’ve used the “Kill-A-Watt” power meter. More recently I purchased a “Poniee PN2000” from Amazon:

https://a.co/d/59Bn52g

Compared to the Kill-A-Watt, the PN2000 meter has 2 additional features which I like:

1) It stores elapsed kWH while unpowered. It has internal EEPROM for this purpose.
2) Also it is capable of accurately measuring Watts down to slightly less than 1 Watt and current down to about 5 mA. This is useful for testing small power supply bricks and for testing the standby power consumption of larger devices.

I’m interested in hearing more about other power measuring devices which are compatible with USA 120V mains and plugs/receptacles.

I would also like to find a device which can measure V, A, Watts, and power factor on USA 240V circuits for large devices such as central air conditioners. A maximum range of 20A would also be desirable.

[DaneLaw]

Got quite a few of these AC power meters with LCD screens.
Some of them have Bluetooth, some of them have Wifi, and some of them his also temperature based, while some of them have apps for different mobiles & tablets & PCs, some of them work with these voice assistants, "Internet of smart things" so you can control it with your voice-assistant, Alexa, Google, Siri etc.
Some of them also seem to have pollution metrics.
Comes with years-long logging, kwh-expenses, multiple advance timers over a year, power settings, countdowns, & power-consumption with graphing over the week, months, year, and energycost
A lot of features, to an extent that I can't keep check of it all. (maybe too much, or not)

It's these... the blue ones from Zhurui were a tad over 40 US quite an old model from more than a handful of years back,  I liked the first one a lot from around 7 years back, so I got another some years later, and it's the same, only the screen seems to have been updated as its a tad sharper. (very nice product IMO with tons of features and control and the build quality matches the price and the accuracy seems good)

The white ones were around 17 US each at sale (incl. 25% VAT & shipping) and existed in 3 different models at release, purchased one of each to try them out, it's those that got wifi & Bluetooth, and temperature control with dedicated probe.

 

White ones..

Different pictures of the different interface templates & settings etc, amongst these models  https://tinyurl.com/2ub7ypwx

- as mentioned some of them are only wifi that fx support Alexa or Google's own smarthome-apps for monitoring & control, while the other are Bluetooth-based and rely on their own dedicated apps.. and the last one is centered around "temperature" adjustment/control and rely on "Wifi so compatible with third-party "internet of things"- though also got that full voltage overview & control - its the unit with the cobber or bronze probe-sensor that you can plug in as needed.
The app for the Bluetooth version' seemed a lot like those you get with the more advanced USB meters from China, so quite rudimentary.

HKJ Review of the Zhurui version.
https://lygte-info.dk/review/Power%20Zhurui%20Power%20Recorder%20PR10%20UK.html

With an accuracy that is quite decent for a household AC socket meter. (vs. Metrahit Energy)
https://lygte-info.dk/pic/Power/Zhurui%20Power%20Recorder%20PR10/Test.png