Help with designing an electronic load

[OM222O]

Hello, I'm designing a simple electronic load and here are my specs:
0 to 5 amps continuous (accuracy of 1mA for at least 0 to 4A range, 10mA is acceptable for the 4 to 5 A range)
0 to 30 volts (accuracy of 5mV for 0 to 5 volts range, 10mV for 5 to 30 volt is acceptable)
Maximum power is gonna be capped to 100 watt as I will be using a CPU cooler which are usually rated for 95 watts but can handle 100 watts easily.

I am at a predicament here: on one hand, I found this bad boy which is a 1% initial accuracy (can be calibrated in software to basically get close to 0 error), 30 watt 1\$\Omega\$ resistor:
https://www.mouser.co.uk/ProductDetail/Bourns/PWR221T-30-1R00F?qs=sGAEpiMZZMtlubZbdhIBIPNKsuXHpjQqm7P7fPg44Qo%3D

This would do just fine and if I use 12 bit DACs and ADCs with a 4.096V voltage reference, I can get the desired accuracy and use a x2 gain amplifier for the 4 to 5 A range and a similar method to divide by 10 to do the voltage monitoring (this would enable me to do upto 40 volts as well).
There's also the added benefit of the resistor carrying the majority of the load and the mosfet can run a lot cooler. but this means that I need to add another mosfet as I plan to have the ability to "short" the input as well. The resistor would also get really toasty at about 105 degrees at 5A load due to it's thermal resistance. I could use 2 , 2\$\Omega\$ resistors in parallel to dissipate the heat more evenly but it would increase the cost.12 bit DACs and ADCs are also quite expensive and I plan to sell this unit.

on the other hand I can drop the value of the resistor to 10m\$\Omega\$ and use divide by 100 op amps or even more (to use the 8 bit adc of the microcontroller itself),but then I need to use precision op amps and a split supply to get the full desired range with the desired accuracy which is gonna drive up the cost again and I'm also not really sure about the gain errors of a /100 or larger op amp ... it can turn out to be pretty horrific and inaccurate. The mosfet also needs to be beefed up considerably which can ruin all the gains in other areas.I also probably need to use a negative rail for op amps anyways as I'm not sure about their minimum voltage (even on precision op amps) and it's gonna be bad if the load can't go below lets say 50mA due to that offset.

It's really frustrating and I'm really lost as what approach to take to keep reasonable margins for selling it, as well as keeping it at the 40 to 50 GBP selling price. As this is the first time for me trying to make a commercial product, I don't know how much interest there is for a such device with the mentioned ratings. It would be fairly accurate and reasonably high power, but it would be a mid range product which is probably going to be avoided by the hobbyist and professionals at the same time. If you have specific part recommendations for the ADC, DAC and Vref, I'm open to suggestions. I would also love to gauge the interest in the device so feel free to comment about it. (It will have constant power, constant voltage, constant resistance, transient mode, internal resistance testing, data logging, automatic battery type detection (alkaline,lipo,NiMH, etc) and possibly even wifi or usb functionality to be used on a PC (I'm making a decent user interface using java. it won't be a basic useless app as I've studied computing science and know how frustrating a poorly designed interface is). It could also be used as a fairly accurate ammeter in the "short" input mode and a decent voltmeter as well). All in all, I think it has a really nice set of features which are mostly achieved in the software so they won't add any cost to the basic electronic load. I'm thinking about having different models with optional parts that might cost more, so people can choose if they want that high of accuracy, data logging capability, wifi / usb interface to pc which are obviously gonna cost more.

Thanks for your time and looking forward to your replies 

[IDEngineer]

There are already such loads available on eBay, Alibaba, etc. for around $50 shipped. You might consider buying one for study and experimentation as you develop your own. Here's an example of a China-built, US-sourced unit:

https://www.ebay.com/itm/200V-20A-150W-Adjustable-constant-current-USB-electronic-load-DIY/183195778904?epid=20018516513&hash=item2aa751bf58:g:wrUAAOSwO0Ra4C1U

It's nobody's idea of a precision lab instrument, but it's a good starting point for figuring out what you want to do.

[OM222O]

That exact unit is pretty useless as you can see here:


it advertises the features that it doesn't have, even the basic constant current. it's an analog circuit which is manually set using 2 pots which is where they cutting a lot of their cost. I want my unit to be actually useful and not a piece of junk. yes, I know such units exist and non of them are upto the task as they lack something important or they are falsely advertised.

[Mechatrommer]

That exact unit is pretty useless as you can see here:

it advertises the features that it doesn't have, even the basic constant current.

that is unfair judgement, not what the VoltLog summed it in the video.

it's an analog circuit which is manually set using 2 pots which is where they cutting a lot of their cost.

wrong, 2 pots are more expensive compared to 2 small push buttons usually made for up down adjustment in FW UI. granted pot lacks resolution and drift overtime but it is very quick adjustment.

I want my unit to be actually useful and not a piece of junk.

nobody designed their product to become a junk. there are specifications, there are target users. accurate, fancy and "useful" don't go along with cheap. good luck designing your accurate and useful product.

[OM222O]

A digitally controlled circuit would be way better and more accurate while being cheaper. Rotarty encoders instead of pots, DACs instead of a voltage divider, ability to correct for any drifts while the analog equivelant would just drift. Data logging is also pretty important, even some basic mulitmeters do it. If I drop the accuracy by using 8 or 10 bit DACs and ADCs, I can make the device for about 15 GBP! My only issue is not knowing what approach to take to achieve the desired accuracy while keeping reasonable cost. I'm also unsure about the target audiance as it's neither dirt cheap for the hobbyist, nor is it super accurate for the proffesionals. I would really apperciate if you can help me with the accuracy problem rather than recommending off the shelf units with similar power ratings or suggest somethings that I might be doing wrong. Soon I'll post schematics and the BOM after finializing the design, that might help a bit.

[David Hess]

Besides the resolution of the ADCs and DACs, consider other error sources and especially INL (integral non-linearity).  There are a lot of 12-bit converters, including low cost discrete ones, which only have 10-bits of accuracy.

Current accuracy will be limited by the temperature coefficient of resistance of the current shunt so I would design within its limits.  20 and 50 ppm/C parts are inexpensively available but power dissipation in a current shunt can result in considerable temperature variation creating a major error source.  A 10C change at 50ppm/C is 1 part in 2000 which is more than twice as large as your accuracy requirement.

Minimize this error by lowering the resistance of the current shunt to lower power dissipation and increasing the power derating of the current shunt to lower temperature rise.

[OM222O]

That's a really good point but 10c increase would be 0.0005 which is 0.5mA (half as much as the required accuracy  and when using the 2 resistors in prallel, each would have a maximum load of 6.25 watts which considering the 4.2 C/W thermal resistance means a maximum of 26.25C increase). I will be measuring the casing tempreature of each part using an NTC thermistor placed in the middle hole (screw mount) and intended to adjust for that in software. Although not perfect and pretty janky, it would result in the desired accuracy if calibrated per unit correctly! Anyways, I agree with you and I was considering using one of those thick wire shunts, but couldn't find any manufacturer who makes them. It would be easy to keep them cool just using a fan and they would be ideal for this application. Linear misfets with sufficient SOA are pretty expensive as well ... I was considering using multiple ones (6 using a 6 channel rail to rail op amp) and having the 12 bit ADC across a common shunt resistor with low value and low temperature drift, and have the 12bit DAC drive one of the mosfets and again have a low value ballast resistor across that mosfet too. The rest can be driven using a low accuracy (8 bit DAC) with large ballast resistors.  This way I can achieve the desired accuracy that way and even have the short functionality as well as spreading the heat significantly which means I can save money on using a cpu cooler as well (they cost about 10-15 GBP for a decent 100 watt one). I'll make a list of prices and design the board using the new approach and will post it here ASAP. Thanks for your feedback 

[In Vacuo Veritas]

I love the smell of naive optimism in the morning!