Author Topic: (Not) another Dummy-Load...  (Read 373 times)

0 Members and 1 Guest are viewing this topic.

Offline zli

  • Newbie
  • Posts: 2
  • Country: ch
(Not) another Dummy-Load...
« on: September 19, 2019, 03:32:31 pm »
Hello together

Another dummy load thread... No, this time something else:) I thought to myself why not try to control a dummy load purely by software? Everybody always says "doesn't work", "too slow" etc... So I searched my pool and soldered the following components to a board:
- Teensy 3.5
- 2 x external ADC ADS7818 (12 Bit SPI)
- OpAmp (only as ADC buffer/amplifier) LT1490
- Darlington MJH6284+ various controller/Shunt/Fan-Control/etc

Everything was programmed with Arduino. Every 25 us both ADC (voltage/current) are read in by interrupt, the PI controller is calculated and a little bit filtered by software. Thus I reach a Risetime (according to definition 10-90%) at 10 A of 32 us. Noise is 12.6 mV at 10 A (see pictures, measured at the terminals).
DUT is a 2S-Lipo 7.2Ah, 90C
The biggest problem that still needs to be solved is the ADC noise. But I think with a SMD, 4-layer PCB that would be possible.
Once the noise is eliminated, the next step is a shorter Risetime. Currently I lose about 10 us by the SW filter. With faster ADC, improvements are certainly possible. The final target is a tasktime of 10-15 us. I can make the PI controller faster, but then I have the problem of the overshoot, that needs to be improved.

The load can be 13.8 A at maximum 110 W (all monitored by software), fan control is controlled via Teensy-DAC. OLED shows actual current, input voltage, power, temperature and the setpoint of the current. Control is via encoder with built-in switch, 1 x short press output on, 1 x long press output off.

Greetings
Franz
« Last Edit: September 19, 2019, 03:37:20 pm by zli »
 
The following users thanked this post: pigrew

Offline Jay_Diddy_B

  • Super Contributor
  • ***
  • Posts: 1793
  • Country: ca
Re: (Not) another Dummy-Load...
« Reply #1 on: September 21, 2019, 11:39:22 am »
Franz,

Welcome to the forum.

Hello together

Another dummy load thread... No, this time something else:) I thought to myself why not try to control a dummy load purely by software? Everybody always says "doesn't work", "too slow" etc...

Snip

Greetings
Franz

I have shared one of the most complete studies of a dummy load in this thread:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/msg288313/#msg288313

The thread contains detailed analysis of the control loop in a dummy load controlled by an op-amp.

In this type of control loop the circuit action is continuous.

When an ADC is included the control loop the system is considered to be 'sampled' or 'discontinuous'.

So what is the impact of sampling?

Nyquist tells us that the bandwidth is limited to half the sampling frequency. Try and increase the bandwidth beyond this and you get aliasing.
In practice the bandwidth that can be achieved is probably between 1/5 and 1/10 of the switching frequency.

Loop analysis

Because the control loop is discontinuous the Bode plots have to be approximated.

A sampled data has the same properties as a delay. It does not change the amplitude, it changes only the phase. Strictly speaking should not  really talk about phase because it is discontinuous.

The delay can be modelled like this:







The switch and the capacitor form a sample and hold circuit. The statements on the right implement a gain phase meter. I have used a 1us sample period, 1 Msps.

The result is




By inspection you can see that average delay is half of the sampling period.

Looking at this in the frequency domain, the 1us sampling period results in 18 degrees of phase shift at 100kHz (1/10th of the sampling frequency)



At the Nyquist frequency, in this case 500kHz, the phase shift from the sampling, is 90 degrees.

Laplace

The sampling delay can be modelled in the frequency domain using Laplace:




The delay has been set to 1/2 the sampling period.

The results are:




The results are 18 degrees of phase shift at sampling frequency/10 and flat gain.

System modeling

This block can be included in the analysis explained in the Dynamic Load thread to show the impact of using an ADC or sampling.


Regards,

Jay_Diddy_B
« Last Edit: September 21, 2019, 11:48:16 am by Jay_Diddy_B »
 

Offline zli

  • Newbie
  • Posts: 2
  • Country: ch
Re: (Not) another Dummy-Load...
« Reply #2 on: September 22, 2019, 10:07:33 am »
Hi Jay_Diddy_B,

Thank you for your analysis.
I know your thread ;)

But just one question, what is your actual statement?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf