First, we will subject our very famous Korad KA3005D power supply to a dynamic load test. It is a very popular price-performance product. Since no one has done this type of test so far, I guess its secret has not been revealed.
A good electronic load is required for these types of tests, but I do not have one. I used a resistor as a load and set up the mechanism in the diagram to switch the current. The switching pulses will come from the signal generator. I used TLP350 so that nothing bad would happen to the signal generator. External 13V is required to feed the TLP350. Trimpots are for adjusting the rising and falling edge slope of the current. After setting up the circuit, I set it to approximately 0.5A/us. We will see the voltage falling on the 0.05 Ohm shunt as 50mV/A on the oscilloscope. I have four 2.5 Ohm aluminum cased resistors that I can use in different combinations as load resistance.
DYNAMIC LOAD TESTS:The turquoise beam shows the current pulses, the purple beam shows the load transient responses of the korad. The first images are before the modification, the second ones are after "MOD 1".
The power supply is set to 20V / 5A. 100mA is drawn at its output with a 200 Ohm fixed resistor. In this case, when the 7.5 Ohm load resistor is switched with a 1ms pulse, the 100mA/2.76A load transition response is as follows.
When I increased the pulse frequency so that a new pulse would come before the current was cut off and the voltage could recover, things got a little uglier. No problem after the modification.
When I tried 400Hz with a 150us pulse, an oscillation close to 2V occurred. There is no problem after the modification.
If we look closely at the 100mA/2.76A switching response after "MODE 1", as follows. In general, there is no more deviation in the load transient response.
I limited the current to 2A and ensured that the voltage dropped to 15V during switching (again with a 7.5 Ohm resistor). There is a considerable overshoot when released. After applying "MODE 1" there is no more overshoot.
RIPPLE / NOISE TESTS:There is nothing in the room that produces significant common mode noise. Except one, which I saved for last.
To keep the external noise to a minimum, I twisted the wire connected to the terminals and connected it to the probe as shown. This is very important.
First, with the power supply turned off and unplugged, the background noise is as follows:
Below when 15V is unloaded.
As will be seen in subsequent tests, there is a low frequency oscillation on the output associated with the 50Hz network, no matter what. If this were not the case, the output would actually be flat as a rope. I did not consider this a problem and did not bother with it.
15V 2.5A CV mode below. The 20ms period peaks seen correspond to the moments when the main capacitor draws current. In the image on the next, it is seen that these peaks are no longer present thanks to "MOD 5". There is also a cleaner image in terms of noise. This must be thanks to "MOD 3".
Since the response is very fast, a ripple of the type associated with switching frequency in SMPS power supplies does not occur in linear power supplies. The main oscillation seen below is something different.
15V 2.5A CC mode below. We encountered a 50Hz oscillation. This is because the cheap shunt resistor in the form of a winding is affected by the transformer. After applying "MODE 2", the CC mode is clean as on the next.
That's if the crappy 12V SMPS powering my sound system is plugged in.
MODIFICATIONS:MOD 1:This is the simplest modification to implement and definitely the one that needs to be done. Because the dynamic response of the power supply is really unacceptably awful. Only two capacitors (C28 and C35) will be checked and if they are 10nf, they will be removed and replaced with 100pf.
MOD 2:The 0.1 Ohm shunt resistor in the form of a wire wrap, due to its proximity to the transformer, imposes a 50Hz distorting signal on the output voltage when the device enters current limiting (CC).
I canceled the original shunt resistor and used four DALE 0.025 Ohm resistors in series. I placed the resistors on the back of the card. If they are in front, they are still affected by the transformer, although they are much less than the original resistance. The wire bridges in the image have nothing to do with this mod. They are for another modification.
The change should be made considering that the shunt resistor will convert 2.5W of power into heat when 5A current is drawn.
MOD 3:With this modification, we are correcting the part of the power supply that reads the output voltage. The first image is the original version of the part that reads the output voltage, and the second is the modified version that we applied.
This change did not have much effect on the result, but it is an example of how not to design a power supply. Because here is a design that was thought correctly and implemented incorrectly.
The output voltage was taken from the banana jacks and passed through the voltage divider resistors and entered into the differential opamp with 1 gain. However, the application was done incorrectly. There is no need to divide the voltage before the opamp, and when you look at the structure on the pcb, the opamp does not actually work as expected from a differential amp structure. The GND of the control circuit, the + end of the power stage is carried with the sense line and combined next to the opamp. This is wrong. The sense lines must be carried completely independently from the source to the target and drawn as parallel lines on the pcb.
With the modification, we provide a differential voltage reading as it should be. We also removed the structure that divides the voltage before the opamp. Our new differential amplifier does not make a gain anyway, it works as an attenuator.
The voltage division ratio of the original version is 1/11.04, but after the modification it becomes 1/10.42. After the process, we correct the error in the voltage reading by calibrating the device. There is a YouTube video on how to calibrate current and voltage.
The tolerances of the resistors used are very important. I took the 10k resistors from the existing circuit. I measured the 39k and 4.7k ones between the 805 sheath resistors I had and used the ones that gave the exact same value.
On the back of the card, we cut the original paths of the VSense lines coming from the banana jacks. There is a via connected to GND in the square, we separate it by scraping around it. With the red cable, we connect the GND of the control circuit to the + end of the power stage from the right point. With the wrapped dual cable, we take the Vsense lines from the sense jack and carry them to the diffreential amp input on the front of the card.
I forgot to take a picture of the front of the card without processing. Note that there are missing elements compared to the original. This is how it looks after processing.
MOD 4:There is normally a 100 Ohm resistance in the marked section. I installed 1k instead of 100 Ohm here and added 100nf parallel to it. In this way, there was a slight improvement in the load transition response.
MOD 5:In Ripple / Noise tests, peaks were seen on the output voltage, corresponding to exactly 50Hz peaks and bottoms. These peaks are the moments when the main capacitor draws current. I minimized the effect by adding a 0.68uf capacitor that I removed from a scrap power supply to the bridge diode AC input.
MOD 6:The GND of the relays that switch the transformer windings on the power card comes to the motherboard. However, it is incorrectly connected to the analog GND. We disconnect the connection from the bottom of the socket on the front of the card and connect it to where it should be with the red cable seen on the back.
Home
Help
Search
Login
Register
