EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: nemail2 on March 02, 2021, 10:27:02 am
-
Hi
for my PSU design (0-1A @ 0-24V output) I've got to a point where I'd need a current shunt resistor with the following specs:
- 0.8 Ohms
- 1W power dissipation capability
- 100ppm max, the lower the better
- 1% tolerance max, the lower the better
- stock at Mouser, LCSC or RS-Components
- cheap
Unfortunately I found something like this is a bit hard to find. 0.82 Ohms seems to be a more common value but still quite rare.
I know that one could put multiple resistors in parallel which is why I for now ended up using 7x 5.6 Ohms in parallel (MBB02070C5608FCT00) but variety isn't too big there as well and they get substantially hot (80 °C and still heating up) when fully loaded @ 1 Amp. They seem to heat up quite evenly though.
How could I solve this? Change the design so I can go with something like 0.1 Ohms which seems to be more common? Is it feasable/smart to put multiple resistors in series rather than parallel to get to the 0.8 Ohms?
I'm not planning this to be a precision device but it should still deliver decent performance. With the current setup I'm fine with the performance but as mentioned the shunt resistors get too hot, I'm afraid.
Also: do resistors usually fail open or fail short (I guess fail open)? Just in case they let their magic smoke escape...
Thanks!
-
8 is not preferred number in E series (https://en.wikipedia.org/wiki/E_series_of_preferred_numbers), so no wonder you do not find single resistor.
You can use two 1.6 in parallel, they are E24 and should be available as 1W or 2W resistors.
-
2x UB5C-1R6F8 digikey etc. will give you 0.8R at 10W with low drift. 0.8R is very high for a current sense resistor.
I am dreaming that maybe your limit circuit operates by generating a Vbe......mmmm . In which case resistor accuracy and drift are irrelevant.
If your circuit is not stuck in the 70's, why not boost the gain of your current sense amp and drop the shunt resistance and heating?
Or use a ready-made reistorless current sensor eg ACS70331 0.8V/A
-
8 is not preferred number in E series (https://en.wikipedia.org/wiki/E_series_of_preferred_numbers), so no wonder you do not find single resistor.
You can use two 1.6 in parallel, they are E24 and should be available as 1W or 2W resistors.
This. Alternatively use two 0R2 in series. The calculator linked below can help you find other values.
http://jansson.us/resistors.html (http://jansson.us/resistors.html)
I would redesign it to use a standard value such as 0R75 though.
-
For a 1 A range a 0.8 Ohms shunt is rather high. The power dissipation will cause quite some heating and this would change the resistance. For use as a precision shunt one should use only a small fraction of the rated power. Especially the ceramik type power resistors can be rated for rather high temperature - so more derating needed there.
So it would be more like looking for a 0.8 Ohms and maybe 5-20 W power rating. Some of the chassis mounted wire wound resistors may be suiteable and some are even reasonable low TC.
The alternative would be a lower resistance like 0.2 Ohms maybe even 0.1 Ohms.
One may than need a littel more care with the amplifier.
It is not ideal, but one can also have multiple resistors in series.
With multiple resistors combined and naturally close by, the power ratings may not simply add, espeically for SMD types that get most of the coolung from the PCB.
-
Change the design - sense amp gain, current feedback loop gain, digital calculation, whatever the design is I'm 99% sure there is another variable in the complete current sense equation than just Rs (or you can/should add a gain stage).
Single resistor allows easier Kelvin sensing.
Lower resistance allows you to reduce power dissipation. Still pick a resistor rated at higher power (i.e., derate it); this allows smaller drift given the same tempco. Lower resistance means you need a more accurate and lower-drift sense amplifier circuit, though, but that isn't too hard with today's good integrated opamps specifically designer for current sense applications.
Go for an easily available number. In addition to standard E12 numbers, shunt resistors are widely available in 2, 4, 5.... but apparently not 8.
So 0.5 ohms, 0.47 ohms, 0.4 ohms are all good options, for example. Or why not even lower. Give it a shot.
-
To answer the question: why not go lower? On the original poster's behalf. My guess is it's for a lab power supply, with a variable current limit. Low drop-out and efficiency are often compromised, for increased accuracy. 0R8 is high for a current limit, but it starts to look very small, for lower current limits. The intention is probably to set an accurate current limit from around 1mA, to 1A and a value of 0R8 was chosen as a reasonable compromise. If going down to 0R75 isn't feasible, it might be reasonable to go up to 1R and adjust the component values elsewhere to compensate.
-
8 is not preferred number in E series (https://en.wikipedia.org/wiki/E_series_of_preferred_numbers), so no wonder you do not find single resistor.
You can use two 1.6 in parallel, they are E24 and should be available as 1W or 2W resistors.
This. Alternatively use two 0R2 in series. The calculator linked below can help you find other values.
Erm… wouldn't two 0R2 in series total 0R4, not 0R8? ;)
-
I can double that 0.8R is too much for 1ADC max rated current (resistor will be too hot as for me). It is much better to do a small redesign to it was 0.47R or 0.22R or even 0.1R.
If TS don't want to redesign it's power supply then he may use 0.75R 2W (from E24) resistor instead of 0.8R.
-
The 0.8 Ohms range may still be OK, if one needs good performance also at low currents. One than needs to look for a high power resistor - like 20-50 W chasis mount or the like. There are also ISA-plan PBV sereis 4 wire shunts, though not that cheap at some 5-10 EUR.
For measurment use the shunts should be used well below there rated power, so they don't get that hot.
-
100ppm/C doesn't do you much good if you are running it at 100C over ambient at full load. Larger resistors have to dissipate the exact same amount of heat as smaller ones, but they may be easier to keep cool. I would suggest a resistor or combination of resistors designed to dissipate at least 10 watts, then make sure they are heat-sinked or raised off the board with fan cooling if you have it.
-
If the difference between 0.80 and 0.82 ohms is important to your design, then you must deal with both the manufacturer's tolerance and treat the parasitic lead resistance (or PCB trace resistance) carefully. The nominal difference is only 2.5%. With seven 5.6 ohm resistors in parallel, the sense-point connection may be important, as well.
-
8 is not preferred number in E series (https://en.wikipedia.org/wiki/E_series_of_preferred_numbers), so no wonder you do not find single resistor.
You can use two 1.6 in parallel, they are E24 and should be available as 1W or 2W resistors.
This. Alternatively use two 0R2 in series. The calculator linked below can help you find other values.
Erm… wouldn't two 0R2 in series total 0R4, not 0R8? ;)
You're right. I don't know what I was thinking. It can't be done with two standard E24 values in series. 0R47 + 0R33 would work.
-
I have used bare nichrome wire for shunts in a DC load and power supply.
You can create any custom low or sub-ohm value you need.
For example, I needed 0R25 at <5A this was my procedure:
a. cut 8in(203mm) of 24AWG (1.67R/ft) nichrome wire, bend it in half,
b. twist the open end a couple of tight turns to close the loop,
c. place a small bolt on each end inside the loop (holding one bolt in a vice), pull the bolts appart to make the wires parallel,
d. measured between the bolts; should read close to 0R25 (since we have two 0R5 resistors in parallel),
e. if it measures too large, twist the once open end a little more to bring the bolts closer, do opposite if too low, and
f. complete by twisting the wire pair between the bolts snuggly.
Tada! Custom shunt with eyes for bolts on either end.
You could also crimp spade connectors instead of bolt holes. Then coil/bend it to make it occupy less space.
According to https://www.allaboutcircuits.com/textbook/direct-current/chpt-12/temperature-coefficient-resistance/ (https://www.allaboutcircuits.com/textbook/direct-current/chpt-12/temperature-coefficient-resistance/) nichrome has a 170ppm temp. co. However, maybe you can compensate, in software, by tracking air temp. and power through the shunt.
-
Tada! Custom shunt with eyes for bolts on either end.
Pictures or it didn't happen.
-
Tada! Custom shunt with eyes for bolts on either end.
Pictures or it didn't happen.
See attached.
-
Beautiful. The only problem I have with nichrome wires as shunts/sense resistors is that they can't be soldered. Is the connection with bolts reliable?
-
Beautiful. The only problem I have with nichrome wires as shunts/sense resistors is that they can't be soldered. Is the connection with bolts reliable?
I don't see why not.
Transistors are bolted to heat sinks without issue.
Otherwise, you can use lock washers, Loctite, double-nut even.
Or, crimp these on: https://media.digikey.com/Photos/Panduit%20Corp%20Photos/PN10-14R-D.JPG (https://media.digikey.com/Photos/Panduit%20Corp%20Photos/PN10-14R-D.JPG)
-
Dug this out of toolbox: This is a 0.2 ohm resistor, it uses a slice of 'tuna can metal, 1/4 inch strip (or was it cat food ???). I had built it to get good accuracy, and the 10 watt resistor puts a 1 ohm trimming.
I'm not familiar with (what) temperature coefficient is in place for a coated tin strip, but both the strip and trimmer have decent heat shedding surfaces.
I used the accurate resistor value, while checking out a scooter motor, at about 2 to 4 amps range.
-
According to https://www.engineeringtoolbox.com/resistivity-conductivity-d_418.html (https://www.engineeringtoolbox.com/resistivity-conductivity-d_418.html) the temperature coefficient of resistivity for mild steel is about 50% worse than that of copper (its actually fractionally higher but the tin-plate will bring it down a little), and as any non-newb knows we don't use copper wire or tracks as current shunts if there is any alternative due to its objectionable temperature coefficient. I suspect you only got away without temperature compensation because the dissipation was low and the enviroment temperature stable . . .
-
Beautiful. The only problem I have with nichrome wires as shunts/sense resistors is that they can't be soldered. Is the connection with bolts reliable?
I don't see why not.
Transistors are bolted to heat sinks without issue.
Otherwise, you can use lock washers, Loctite, double-nut even.
Or, crimp these on: https://media.digikey.com/Photos/Panduit%20Corp%20Photos/PN10-14R-D.JPG (https://media.digikey.com/Photos/Panduit%20Corp%20Photos/PN10-14R-D.JPG)
Crimps are a good idea. Another option is PCB mount screw terminals.
(https://www.mouser.co.uk/images/keystoneelectronics/lrg/7780_SPL.jpg)
https://www.mouser.co.uk/ProductDetail/Keystone-Electronics/7788?qs=zBlGwAMnvtfxCRT2qpr1FA== (https://www.mouser.co.uk/ProductDetail/Keystone-Electronics/7788?qs=zBlGwAMnvtfxCRT2qpr1FA==)
-
The better material for a DIY shunt is Manganin, like they use in most shunts too. It is reasonable well solderable. The conductivity is higher than with NiCr, but this is more like a good thing for low resistance values.
Stainless steel has a reasonable low TC and may thus be a makeshift alternative, if really needed. Still many types and not all are good. However soldering is not much easier than with Nichrome.
The ceramic case wire wound resistors are not that bad and not that expensive. Just de-rate them a lot, as they get very hot at the rated power. As a shunt, stay below 1/10 the rated power.
-
Yes, 0.8 Ohm is difficult, because it is sort of in the no man's land category. Below 0.1Ohm, you have the metal shunts, that exhibit good thermal properties, high dissipation. Above 10 Ohm, you have the thin and thick film parts, and bulk metal foil. If you really need this value, your best bet is to custom order it.
But I think you are better, if you modify your circuit. Typical shunt is designed with 100mV drop, so you would use 0.1 Ohm. Dissipation would be 0.1W, so an 1W shunt would be about 20 degrees above ambient, resulting 1000PPM-2000PPM error. Not a big deal, since it is a power supply design.
-
You're giving contradictory information.
"Better" then 1% but cheap".
100ppm but "not planning to be a precision device".
80 degree celcius is a normal operating temperature for resistors near their rated power handling.
If you want a lower temperature, then use a higher power rating, such as for example 5Watt resistors, or spread it over more resistors.
Or, as already suggested, lower the value, so it does not dissipate as much in the first place.
It's easy to add a bit of amplification with an opamp.
And why a "precision" resistor? For small quantities you can easily calibrate amplification with a potentiometer, for higher production numbers you can calibrate in software.
As the big Davey Jones has said many times. Accuracy of such components is not relevant. Stability is. (Lower temperatures may help with long term stability).
But would re-checking your own power supply once a year be a problem?
-
Remember that in series connections, the connecting track and solder joints (copper tempco 4000 ppm/degC, similar for solder) become part of the circuit. They seriously vary from unit-to-unit, this can be calibrated out but the tempco (practically) can't.
In parallel configuration, where you are connecting the sense wires? To one of the resistors? Even if the resistors match exactly, they do not share the current due to the different track and mounting resistances.
1% accuracy is already in a territory where I'd suggest following proper current measurement practices. The most important IMHO; this also makes your work easier:
Use a single resistor. Use a resistor designed for current sensing. Use Kelvin connections.
I'd go for a 2512 packaged SMD current sense resistor; they are rated at 2W or 3W. I'd choose R so that maximum dissipation is 1W (with a 4-layer PCB so that ground plane right under the resistor cools it; with 2-layer and reduced cooling area, maybe just 0.5W!). And again, Kelvin sense it. Use a low-offset shunt amplifier. Chopper auto zero types are cheap and good when you don't need very high BW (up to some 50-100kHz). And even if the gain=200 model seems good on paper, don't go overboard; gain=50 is fine if you can accept the heat in Rs.
So, to sense 1A, I'd maybe choose a 0.5ohm shunt, for 0.5V voltage drop and 0.5W dissipation, in a 2512 SMD shunt, then go for a gain of 5..6 to interface 3.3V ADC, or whatever.
Or if low dissipation is more important, I'd maybe choose a 0.1ohm shunt, for 0.1V voltage drop and 0.1W dissipation, maybe in a 1210 SMD package if I need a small solution, then use a gain=20..30 current sense amplifier.
-
Hi all,
wow I completely lost track on this topic, sorry for not answering earlier and thanks for all your replies, answers and suggestions, really appreciated.
The issue why I even went with something like 0.8 Ohms was because using 0.1 Ohm and the MAX4080T (20x Amplification) instead of the MAX4080F (5x Amplification) which I'm using with the 0.8 Ohm Shunt was because weird stuff happens with the control loop/the output voltage when using the 20x Amplification and the 0.1 Ohm Shunt (shown later in this post).
I have attached the schematic of the analog part/control loop. VREF is a MCP1541 and as you may already have noticed, this design is derived from Daves µSupply publications from years ago. Almost forgot: D2 and R4 are not populated and R5 is 20k.
This design/my implementation is not particularly fast in terms of constant current mode but I'm ok with the performance and I wouldn't want to completely change the control loop and start from scratch, if possible.
So, like I said, when using the MAX4080F (5x Amp), no matter what current shunt I use (0.1R or 0.8R), all works fine in CC and CV mode. But when I put in a MAX4080T (20x Amp) + a 0.1R Shunt, as soon as the PSU switches to CC mode, I seems to have some kind of weird 50 Hz oscillation/noise which isn't there when using the MAX4080F.
My guess was, that maybe because of the higher amplification, the slopes of the MAX4080T output voltage are some kind of steeper, which then in turn somehow causes this issue.
Unfortunately control loop compensation and such is complete black magic for me, I wouldn't even know where to start. All control loop compensation in this circuit has been done by very smart people from this forum or from mikrocontroller.net (german forum).
I have attached some scope shots, this is how the output voltage looks like in CC and as well in CV mode (completely identical) when using the MAX4080F + 0.8R shunt:
[attach=2]
This is how CV mode looks when using the MAX4080T + 0.1R shunt (obviously nothing special to see here, as we are in CV mode):
[attach=3]
Well, and this is how CC mode looks when using the MAX4080T + 0.1R shunt:
[attach=4]
This happens when I for example connect a 12R power resistor to the output, set the output voltage to 6V and the current to 1A and THEN lower the current limit to 400mA. As soon as CC mode kicks in, the output voltage looks like in the last scope shot.
Other than that, I'm really happy with the design and everything and if I can't get this to work with 0.1R and MAX4080T i will just stick with 5.6R + MAX4080F and maybe either put higher rated power restors in it or just lower the max output current spec a bit and everything will be fine.
However, it would be cool to get this working with 0.1R + MAX4080T, so any suggestions or hints where this weird crap on the output could be coming from would be very appreciated.
If you want to see the rest of the schematics, just let me know. Basically it is just an Atmega1284p, a 2004 LCD, some buttons, two mains transformers for 8VAC and 24VAC, a 7805 and a LM317T which makes fixed 27V. Oh and the MCP1541, of course.
If anyone wonders, this is how it looks like:
[attach=5]
Thanks!
another edit: LOL i just realized that the commercial product DP832 has 1000µF of capacitance right at the output so in terms of that, my PSU should be an order of magnitude better :-DD |O
-
The AC signal seen with the 0.1 ohms shunt looks like mains hum. With a smaller shunt wiring gets more critical and the wiring may be just a bit different from the case with the 0.8 Ohms shung.
The relatively large capacitor at the output of the supply is often there to limit the over-shoot for the transition from CC to CV mode. You know how much capacitance is needed after testing the CC to CV mode cross over. The way the compensation for the voltage mode is done, this part may be relatively good, especially as the current mode regulation is slow.
The relatively slow regulation for the current mode could be a weak point however. This kind of acts a littel like a simulated capacitance, though worse at low voltage, while the real capacitance is more a problem starting at high voltage. The critical test would be something like looking at the current when apllying a near short (just a small resistor to measrure the current) from a high or low set voltage. For the start maybe add some extra resistance to prevent the current from reaching damaging levels.
Ideally R26 should be used to add an additional fast current limit (a simple transistror to give a fixed limit of some 5-7 A).
-
anyway your shunt is tooo big, redesing for lower value like 0r1
and if you stick to your 0.8, just put whatever you want, parallel or whatever, you will need to calibrate at some moment the crt measurement, if you obtain more than 1 A max scale it won't hurt, cut it in sw/hw
-
The AC signal seen with the 0.1 ohms shunt looks like mains hum. With a smaller shunt wiring gets more critical and the wiring may be just a bit different from the case with the 0.8 Ohms shung.
OK that doesn't sound too bad, thanks. The wiring should be 100% identical as I have just made two PCBs (completely identical), only the shunt and the MAX4080F/T was different and I had both PCBs in the very same case with same cabling and everything (just swapped them back and forth few times).
So can this hum be of any harm or can it be accepted and just leave it be? I mean it is not even 20mV peak to peak and that is under load, including potential ripple...
The relatively large capacitor at the output of the supply is often there to limit the over-shoot for the transition from CC to CV mode. You know how much capacitance is needed after testing the CC to CV mode cross over. The way the compensation for the voltage mode is done, this part may be relatively good, especially as the current mode regulation is slow.
The relatively slow regulation for the current mode could be a weak point however. This kind of acts a littel like a simulated capacitance, though worse at low voltage, while the real capacitance is more a problem starting at high voltage. The critical test would be something like looking at the current when apllying a near short (just a small resistor to measrure the current) from a high or low set voltage. For the start maybe add some extra resistance to prevent the current from reaching damaging levels.
Ideally R26 should be used to add an additional fast current limit (a simple transistror to give a fixed limit of some 5-7 A).
I did try smaller capacitances at the output but below a significant amount under 100µF, the thing started to oscillate like crazy when transitioning from CV to CC mode, if I recall correctly.
Shorting the output doesn't really seem to hurt anything in the PSU, maybe because of the ESR of the capacitors and because the current spike is relatively short until the slow CC kicks in? I don't know...
I should be able to measure the current spike using a low side shunt at the output + a scope across the shunt, right?
Nonetheless, core question for me: is there anything obvious I should or could do about that (suspected) mains hum in CC mode when using a .1R shunt + MAX4080T or should I just leave it be?
anyway your shunt is tooo big, redesing for lower value like 0r1
and if you stick to your 0.8, just put whatever you want, parallel or whatever, you will need to calibrate at some moment the crt measurement, if you obtain more than 1 A max scale it won't hurt, cut it in sw/hw
Thanks, playing around with 0.1R again, having that mentioned 50Hz noise issue...
-
As far as one can see from the pictuture the layout around the shunt and flarge fitler cap looks Ok.
The extra wire links for the supply (linkely lower voltage) are a bit unusual, so may something wrong with the ground link between the 2 transformer parts.
One could see if the supply to the LM324 is clean or has similar looking ripple.
There may be shunt resistors in between some 0,8 Ohms and 0.1 Ohms. The 0.1 Ohms are a bit on the low side for a 1 A power supply and would compromise on fine regulation.
-
The extra wire links for the supply (linkely lower voltage) are a bit unusual, so may something wrong with the ground link between the 2 transformer parts.
the connection should be good, the PCB which has the 50Hz hum doesn't have JST connectors but soldered wires, so there shouldn't be any contact issues (I was too lazy to crimp the JSTs on that one).
One could see if the supply to the LM324 is clean or has similar looking ripple.
SDS00006.png shows the LM324 supply voltage during CC mode, you can see a little bit of hum there, I guess. SDS00007.png shows the LM324 supply when the PSU output is turned off, so no load.
There may be shunt resistors in between some 0,8 Ohms and 0.1 Ohms. The 0.1 Ohms are a bit on the low side for a 1 A power supply and would compromise on fine regulation.
I wouldn't want to replace the MAX4080 with some other current sense amplifier because I'm afraid that I may have to redo (or let someone redo) the compensation of the control loop opamps. So I'm stuck with 5x Gain (MAX4080F), 20x Gain (MAX4080T or 60x Gain (MAX4080S). Because I have 4mV resolution on the ADC (VREF = 4.096V, 10 Bit ADC, 1 Bit = 4mV), I'm left with 0.8R @ 5x Gain or 0.2R @ 20x Gain. I actually just realized that using 0.1R @ 20x Gain I can't even measure 1mA (not taking into account noise, inaccuracies and stuff like that which adds error).
So without having to change a whole lot of stuff, I could (and will) try 0.2R by just connecting two 0.1R shunts in series.
Higher values would require some additional gain stage at the DAC output which controls the CC max current (IC3B, Pin 5) because I can only output 5V absolute maximum using the MCP4922 (it already has an internal 2x gain stage and can go above the 4.096V reference voltage). That additional gain stage again would add error. Alternatively I guess I could add some feedback resistors to IC3B pin 6 (current measure feedback path) to lower the voltage needed from the current limit setting DAC output (IC3B, Pin5)... But that definitely would require a new PCB if I don't want ugly bodges on the board..
I have also attached the layout for your reference.
edit: wow, using 0.2R it realy does seem to look better, see SDS00008.png (0.2R) in comparison with SDS00004.png (0.1R)!