Author Topic: Voltage "divider" after PSU to highten the resolution? (Read 4470 times)

FriedMule · « **on:** February 16, 2019, 03:30:02 am »

I have a standard power supply with 10mV resolution and max 30V, what I need is a far better resolution.
My first thought was just to put two resistors on the output as a voltage divider, i.e. 1,000:1 but that would also reduce max voltage to 0.03V and that's fare less then the 10V I want and is wary inefficient!!

Is there an other way to increase the resolution by adding some circuit between the PSU and DUT?

beanflying · « **Reply #1 on:** February 16, 2019, 03:41:25 am »

First Question - What power supply because it matters a lot is it DC-DC or Linear what Brand and Model etc?
Second Question - What actual resolution and at what current and voltage?

As to a resistive divider for better resolution it then depends on the above factors and how 'stable' that supply is or can be made. They are also not good for regulation of a voltage due to what the unspecified load can and most likely will do to that ratio.

ejeffrey · « **Reply #2 on:** February 16, 2019, 03:56:26 am »

The easiest thing would be just to make a post-regulator on the PSU output like a LM317 or any number of more modern linear regulators. You can adjust the supply voltage so that you avoid dissipating too much power in the linear regulator.

What is actually best depends on what you actually need in terms of current, accuracy, and stability.

Keep in mind that your 10 mV resolution is already 0.1% of your desired 10 V range. This is not extraordinary, but your garden variety supplies and resistors won't be better than that. It isn't hard to get 0.1 mV resolution but you won't get that much accuracy or long term stability without some effort.

FriedMule · « **Reply #3 on:** February 16, 2019, 04:24:50 am »

Thanks for your reply, I have a Korad KA3305P power supply it is just a standard, not wild unit but normally it is my best unit.
About the change on the PSU I have just found this circuit that looks wary interesting but can it be modded to my needs?
https://www.electroschematics.com/945/01v-50v-power-supply/

My needs is that I would like to make as high a resolution I can without having to call NASA, in other wird a "converter" that not have to reach the power sources 100nV but still a lot better then 1mV. Abut stability and so on, this shall be a "standard" PSU I can have on my bench when I want to play with more precise values.

beanflying · « **Reply #4 on:** February 16, 2019, 04:40:19 am »

Very quick check of it's spec shows it as fairly horrid for compared to what you actually want. So you either use it as an initial supply for an external board (easiest) or try and 'fix' it's specs (buy another supply long before this option).

Tied up at present but use your Google FOO to look at accurate variable linear regulators as a stand alone external option to tame the spec.

FriedMule · « **Reply #5 on:** February 16, 2019, 05:23:26 am »

My idea is to use some sort of PSU that have alle the functions I need, then attach some sort of contraption to better the resolution but still have more voltage then a standard resistor divider would give.
So if the ripple is 100ma, it will also be divided by a voltage divider.

beanflying · « **Reply #6 on:** February 16, 2019, 10:06:24 am »

Just for some additional info. I was looking for some details on my Agilent 6632B today so I snipped it's specs for you.

Have a serious think about what is sensible and practical to make. Personally dump any idea you have of a voltage divider being the answer unless you only require maybe sub 50mA currents even then your voltage will vary a lot with load.

Somewhere between these specs and your Korad is where you can hope to finish up so think about what you really need and let us know.

FriedMule · « **Reply #7 on:** February 16, 2019, 06:19:06 pm »

Great reply!! What is wondering me is that you can get a PSU with 1mV for almost free but have to cut your arm off if you want something better.
Since that PSU shall be made / bought to my lab as a low voltage precision PSU to whatever I find interesting, it would be better with better specs but not for any price. :-)
I am thinking above 5V and maby best with 10V to simulate a 9V battery with maybe 1A.

Dave do even have PSU's that deliver 1pV and even 1attoV!! :-)

FriedMule · « **Reply #8 on:** February 16, 2019, 10:07:18 pm »

Would this be hard to buy / build?
From: 0-10V - 0-1A
Step: 100uV / 10uA

beanflying · « **Reply #9 on:** February 16, 2019, 11:45:30 pm »

So what you want is a power supply with better specs than the recently discontinued Agilent 6632B ($3000+ USD) for $100

There is no point having 0.1mV steps if the Ripple and or Accuracy is 5-30 times more than that. It is like calculating a number that has far more precision than you measured and it is fairly pointless. The 6632B is specified at 0.3mV ripple and at say 5V it's full 'Accuracy' would be 2.5mV + 10mV = 12.5mV so @ 5V +-12.5mV with 3mV ripple is more like reality.

Fairly rough setup below in the photos but 6632B unloaded in the first two DCV and ACV (ripple) at the outputs
Next two 16 $\$\Omega\$$ resistor from the same 5V no remote sensing. Notice the drop in voltage even with a modest current across 'reasonable' test leads.
Last two with remote sensing connected and measured by the 34401A at the DUT, the supply has picked up some of the cable losses but still lower than the NL set point.

All this shows is that the Supply is better than spec (expected really) but it takes a lot of careful design and planning to get below 1mV in 'Resolution' let alone the 'Accuracy' at that resolution. If Agilent couldn't do it do you think you can.

So complete your specification before even trying to see if it is possible. Dave's Keithley's are specialized bits of kit and not general purpose Bench power supplies that said I have never read their specifications. I think there is an answer and it's name may not be Korad and some grasp of what is possible needs to be understood before you consider trying to make your own

FriedMule · « **Reply #10 on:** February 17, 2019, 02:27:11 am »

Thanks, you are completely correct:-)
My first choice is to buy a unit that I know is build correctly, if not possible, then try build.
What I can not understand is that you can buy a Siglent PSU 1mV/1mA for 377$ but have to go to fantasy-land to get one more zero?

beanflying · « **Reply #11 on:** February 17, 2019, 02:33:03 am »

In what I used to do for a real job (when I had one) there was a 'Rule of 4' going from Cast Iron to S/Steel as a build material. Each extra Zero with electronics I reckon is more like a 'Rule of 10'

FriedMule · « **Reply #12 on:** February 17, 2019, 02:49:53 am »

LOL yes, like from 7.5 to 8.5 DMM :-)
I have to say that this cheap Siglent or more expensive Rigol, sounds nice, but are they any good??

https://www.siglent.eu/siglent-spd3303x-power-supply.html

https://www.aliexpress.com/item/RIGOL-Programmable-linear-DC-power-supply-digital-programmable-power-supply-DP1116A/32479881210.html?spm=2114.search0104.3.1.28ef2fe63IwMkP&ws_ab_test=searchweb0_0,searchweb201602_10_10065_10068_319_317_10696_10084_453_10083_454_10618_10304_10307_10820_10821_537_10302_536_10902_10059_10884_10887_321_322_10103,searchweb201603_51,ppcSwitch_0&algo_expid=79c0ead5-6aea-4484-a36b-1033ad17f5a5-0&algo_pvid=79c0ead5-6aea-4484-a36b-1033ad17f5a5&transAbTest=ae803_3

Calvin · « **Reply #13 on:** February 17, 2019, 06:56:39 am »

Hi,

Quote

My needs is that I would like to make as high a resolution I can ..... when I want to play with more precise values.

That doesn't sound like a need or requirement to me, but just a wish.
Do You, resp. Your applications really require better than 1mV, or even 10mV of resolution??
Anyway, as suggested before a dedicated post regulator may get You closer to Your wish.
You may have a look at LinearTechs LT30xx series of regulators ... those look very promising for this kind of application .... search for "High Performance Portable DC Bench Power Supply: Save Money and Free Up Bench Real Estate by Building Your Own" and other Design notes.
See datasheets of LT3045, LT3081, LT3094 etc.

Regarding the Siglent supplys ... when You can live with programmability over PC than the SPD3303C (10mV/10mA) is as good as the SPD3303X-E.
It's a really very well behaved, precise, good power supply ... especially for the price ... and a considerable upgrade from a Korad and crapalikes.

regards Calvin

beanflying · « **Reply #14 on:** February 17, 2019, 07:02:08 am »

Or you spend a whole lot less on one good Power Supply a lot closer to home eBay auction: #292071246164 and keep your Korad for general less precise use on the bench.

Unless I needed automated control over a power supply I wouldn't spend my $ on unneeded 'features' over actual specifications of output. So in the short to medium term I will be sticking to my Agilent and a pair of Manson Linear supplies (0-15V @ 60A, and a Triple output with better specs than your Korad) for general use.

hli · « **Reply #15 on:** February 17, 2019, 08:20:21 am »

Just as a thought: you want your PSU to supply up to 1A. Assuming that you are not using separate sense wires (I think none of the Rigol / Siglent ones have them), an accuracy of 1mV means that your wires from the PSU to your powered device must have less than 1mOhm (including all connectors).
So even when then readout / voltage setting on the PSU has this accuracy, it does not mean that this what your device gets in reality. In addition to the ripple problem that why PSU with such specs are so expensive - sooner or later physics gets in your way.
Maybe you should first define _why_ you want to have such a accuracy? Whats the problem you are trying to solve?

ebastler · « **Reply #16 on:** February 17, 2019, 08:40:06 am »

@FriedMule -- I don't think you ever mentioned why you believe you need a power supply with that high resolution?!

If you actually want to supply power to some circuit (i.e. a supply current of several mA and up), the voltage should be far less critical than requiring mV resolution. If it is, I would argue that your circuit is not designed properly.

If, on the other hand, you need a precise reference voltage, e.g. to set a comparator threshold or such, you will not need a significant current. Such reference voltages are usually not provided via a big adjustable power supply, but a separate "voltage reference" circuit. You can find many integrated voltage references, with fixed or adjustable output voltages, with nice built-in temperature compensation etc.. Simple 2 or 3-terminal devices which are connected to a not-so-precise supply voltage and provide the required precise and stable reference voltage at their output.

Doctorandus_P · « **Reply #17 on:** February 17, 2019, 02:49:35 pm »

I'm a bit surprised "SMU" has not been mentioned here yet:
But it probably also does not matter much because of budget constraints

https://www.ebay.com/sch/i.html?_nkw=keithley%20smu

Quote from: FriedMule on February 16, 2019, 10:07:18 pm

Would this be hard to buy / build?
From: 0-10V - 0-1A
Step: 100uV / 10uA

Have you thought about how realistic those values are?
What sort of cables do you want to need if you want to keep a 5V supply within 100uV over a 0 to 1A current differential?

That's a voltage resolution of 1:100000
Have you thought about how you want to adjust such a thing? That is definately not going to work with a 10 turn potentiometer.

Things like these are very expensive, but most of the price does (probably) not go into the components, but in the engineering such things for a small market. But also because they are only made by companies who do not care about price, but only about qualtity and robustness.

As a hobbyist it is fairly doable to slap a few DAC's onto a microcontroller, add a display, some rotary encoders and some opamps for the output section.
Opamps don't deliver much current. One easy option is to add a buffer such as LME49600 (can these be parralelled?) Or on a low budget maybe even some integrated audio amplifier.

For a DAC you could for example take an MCP4922. It's only 12 bit, but in the datasheet the've shown a simple trick to increase it's resolution.
You can get more (absolute) accuracy by carefull calibration, or by using a good adc and fine adjustment in a software loop.
Fairly good ADC's have become pretty affordable. For example take a look at ADS1220, or other DAC's in that series.

Do you only want to measure current, or do you also want a current limit to work reliably at your resolution?
It might be a better idea to build and adjustable current source.

For measuring current, you can buy expensive high accuracy & stability resistors, But you can also use a cheap power resistor and callibrate it yourself. You can also for example glue a thermometer to it and after characterisation of the resistance deviation, you can partly compensate for it in software.

How good you can build something like that and what accuracy / resolution / reliability you get will mostly be dependant on the experience you have with electronics and the amount of time you put into it.
But with a reasonable amount of effort you will probably:
1). Get much better accuracy / resolution than your KORAD.
2). Have built designed something to be proud of, and have lots of fun in the process.
3). Learned a lot about electronics.
4). Have a usefull device for doing experiments.

alex-sh · « **Reply #18 on:** February 18, 2019, 09:06:40 am »

Quote from: FriedMule on February 16, 2019, 10:07:18 pm

Would this be hard to buy / build?
From: 0-10V - 0-1A
Step: 100uV / 10uA

Try it for yourself and see why it is bloody hard to build it or why Keithley is charging so much for their voltage/current source.
I wonder why or where you would need such resolution?

FriedMule · « **Reply #19 on:** February 19, 2019, 03:43:31 am »

First sorry for my late answer, I hade some family business to attend to.
I'll try to answer all your questions.

Calvin Siglent has a model without the X that goes down to 1ms on the interface with remote sense, I am glad you say that they are fine, cheaper for me, I am thinking of either buy or build but buy use to be a fare better quallety:-)
I think it's a combination on wich and Ehm sort of need, I have my Korad 10mS so one more cero does not do much difference. I am hoping on learning about components at wary low power, like Dave did with his nanometer and LED's, I want to play with BJT, Mosfet, LED and other components, to se how they behave, also to build circuit that can run with as low a power as possible. And on top of that, it would be a fantastic tool to have on the bench! :-)

beanflying my demands are not high in output voltage but would like to cover different types of standard batteries like 9V and 1A would be really nice. I have the Korad for just supplying power when I need it, think of it like my replacement for a car battery.

Doctorandus_P about if I have thought about the realistic in my project, yes and no. You can buy a 0-60V 0-10A 1mV/0.1mA for an okay price and other with less power almost every ware, and you can buy SMU with insane specifications and price, and Dave have some "normal" power units down to 1nV, so I tought one more Zero would not be wild to go fore. I really like your advices about components and will read a lot more about them. :-)

alex-sh I know that a zero more does thing harder, but as I wrote abowe, did not think we was talking about a Mars mission. Yes you can get a source meter, but I thought that the price was due to the insane specs, the 6.5 digit DMM, brand name and so on. I do not need a DMM 6.5 or these great features on top of my PSU. "All" I wanted, was a PSU, one digit better then the cheap Hong Low unit to 377$:-)

hli · « **Reply #20 on:** February 19, 2019, 08:47:16 am »

Remember: a power supply is not a measurement device. Its intended to _supply _power_ to your device, that it also measures voltage and current is mainly so you can adjust its settings, and get a sense about overloading. When you want to do precision measurement, use a DMM (or e.g. a µCurrent).
Yes, there are PSUs out there that can measure very precisely too, but as you did already discover this combination has its price.

FriedMule · « **Reply #21 on:** February 19, 2019, 10:30:53 pm »

Yes you are right and I thought, if I could get / buy a PSU, I could always use a meter to adjust it by, no need for advanced DMM, "just" a great resolution.

ArthurDent · « **Reply #22 on:** February 20, 2019, 03:09:49 am »

As others have pointed out, finding a power supply with the specs you mention is not an easy (or cheap) task.

Resistors are the worst way to try to get this resolution because they only give you a ratio (at no load) and the ratio error will increase with load. If you want 10 volts at 10x more resolution and you use a 10:1 divider you get 10x more resolution at 10x less voltage so you have to increase the voltage into the divider by 10x (or 100 volts) to get your required 10 volts out and resolution wise you are right back where you started except you no longer have a regulated output – it is load dependent.

If your supply only has local sensing and you set the power supply output to 10.00 VDC and use two 3 foot number 18AWG (1.2mm) wires with a 1A load, the load will only see about 9.95 VDC. If your load varies from zero to 1A then the load will vary 50 millivolts caused by the lead drop, that on top of whatever load regulation the supply has.

Trying to build your own supply will be a huge problem unless you have unlimited funds to devote to research and development. Besides needing a supply with remote sensing capabilities as has been pointed out, the supply has to be temperature stable and able to maintain its stability and accuracy over some extended time period. Drift can be a big problem with low end power supplies. That is why the price you have to pay for precision goes up exponentially. Resolution can be meaningless if the other supply errors are large enough to mask the resolution problem.

Below is a photo of my HP e3632A bench supply similar to the HP 6632A supply beanflying has. Supplies like this can sell used for up to $1000 (I paid FAR less) but they are one way to get what you need. This video about the HP e3632A gives some good information to describe what you’re talking about.

The second photo is of one of my stable references that has been on for several months and has an output of 10.00000 VDC as read on my HP3457A meter. That precision comes at a cost and a lot of time for burn-in.

You have to ask yourself if you really need what others have said you are wishing for. One common way to eliminate lead drop would be to build the necessary regulator right at the circuit you’re building so the lead length is an inch or so and if you need 10 VDC then feed the input to the board regulator with the higher necessary voltage.

FriedMule · « **Reply #23 on:** February 21, 2019, 05:49:50 am »

Thanks. great info and video.

I have found this Maynuo M8851 model but can't seem to find a datasheet.
http://www.maynuo.com/english/pro.asp?tid=35

I know that it only goes to 6V but better then nothing, it do on the other hand, goes sky high in Amp: 60!!

Calvin · « **Reply #24 on:** February 21, 2019, 11:46:17 am »

Hi,

its under the folder Download at the top of the page.
Wouldn´t be the 8811 (30V/5A) or the 8831 (30V/1A) be closer to Your needs?
Welectron offers the Manyuo electronic loads ... maybe they can also supply for the power supplies.

regards
Calvin


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Voltage "divider" after PSU to highten the resolution? (Read 4470 times)

Share me