PMP8740: PMP8740

Hi Anbarasan,

I am fine thanks: I hope you are doing well too! Vacation was great!

You don't need to change the compensation, if you add a Schottky diode in series to the load, because when the diode conducts, it's actually a short circuit regarding the power stage transfer function.

Also, for the second question, you are right: you can set a different reference for the current loop by changing the value of R50.

In my design, Iref is modified by the PWM, provided by the microcontroller, as you already know.

Best regards,

Roberto

Hi Anbarasan,

Thanks for your update: I am happy to see you are proceeding with testing.

If you have a working current sharing circuit, you can increase the steps also up to 10A.

Just, please, be sure during the load switching that there is no dangerous output current in each converter. For this purpose you should add a current probe on the output, and check during the transitions that there is no spike in both currents.

Best regards,

Roberto

Hi Anbarasan,

The connection between one primary full bridge and one secondary full bridge you showed looks correct.

From my point of view, this type of connection works well only if the two controllers are placed on the same board, or the distance between the controllers is also small (millimeters). If you have the controllers placed at tens of centimeters distance, I am afraid the signals can be corrupted by noise.

Here is a link to an example, of paralleling two UCC28950, but taking place on the same board, see PMP6712 schematic:

The whole documentation is available at the link:

Since in your case you have two separated boards, I suggest to run them separately, with two voltage loops, to be balanced by using the paralleling technique explained in the UCC29002 (load share controller) datasheet, or using the digital parallel I employed in my PMP8740.

Regarding the PFC boosts, there is no need to make any current or power sharing, since they are connected to mains, which has very low impedance.

Please let me know if you have further questions.

Best regards,

Roberto

Hi Anbarasan,

The boards look pretty close to each other, so it might work.

In order to minimize any noise, captured by the wires, please provide some filtering (from example from the COMP output of the main converter, to the input of the second one, by using a RC filter (keep the filter on the secondary controller as close as possible to the IC).

Regarding the oscillator and synchronization, please check that the signal doesn't corrupt while transferring from one board to the other.

Regards,

Roberto

Hi Anbarasan,

No need to sorry! :-)

I didn't know that you removed the two diodes.

I suggest to follow the schematic of the PMP6712 that I already sent the link. If you referrer to this schematic, the output of the error amplifier is available on R30 (net OPTO-E), then there is the internal error amplifier with gain = 1 (R31/R32), where the reference voltage is on EA+ of U5 (half of REF1).

The output of the primary (main) converter is now the net COMP1 (which is used by U5 to close the loop).

The information of COMP1 is now sent to the secondary full bridge controller (see U11), which enters in the pin EA+. As you can see here, there is already a series resistor R74, so you need only to put a capacitor between pin 2 of U11 and ground, let's say 470pF (which together with 10KOhm resistor R74 makes a filter at ~ 32KHz). After that, R75 connects EA- with COMP of the secondary converter (with a gain of 1).

This should be sufficient to have a primary/secondary architecture.

If you are still unsure how to connect the primary controller, please send me a bigger section of the schematic you are using.

Regards,

Roberto

Hi Anbarasan,

You are right! This last connection works. The point here is to avoid that the connection between pin EA+ of both controllers does not take noise.

For this purpose, it should be sufficient to connect a local capacitor (220pF...470pF) in each controller from EA+ and local ground. In order to decoupled these two capacitors, you might add a small resistor (100 Ohm) in series to each connection to the two diodes D1 and D3, which shouldn't be removed.

I created and attached a schematic only for the controller sections. The filter capacitors and resistors are Rf1, Rf2 and Cf1, 2Cf.

Best regards,

Roberto 

PMP8740 2KW Rev_E Full Bridge - parallel.pdf

Hi Anbarasan,

The connection regarding the parallel is also fine. I would recommend to remove the resistor R2 in the secondary controller, since it makes no function.

I am afraid you get noise into the SYNC pin of each controller, so it will be difficult to make it working without having both controllers on the same board.

Regarding the sync-FETs, the losses in the body diodes at mid to full load will be too high and can lead to failure. You may test the converters up to 10%...20% of nominal load, unless you replace the sync-FETs with Schottky diodes, then, with appropriate heatsink, you may reach full load.

Best regards,

Roberto

Hi Anbarasan,

It is possible to parallel converters by using C2000 series (like the F28335) without the need of serial communication (UART).

In order to do that, you have to sense and read the two output currents (best is to sense both inductor currents) in real time.

Since it concerns software, and I am actually an analog engineer, you can post this request into C2000 threads.

Best regards,

Roberto

Hi Anbarasan,

Here are my answers to your points:

1) At low output current, it's normal that there is some unbalance: this is due to the difference of the peak ripple inductor current of each DC/DC stage, plus small difference in the internal blocks of the UCC28951.

2) No, the filter is not impacting the loop compensation as long as the pole, introduced by the filter, is far away from the crossover frequency (for example the pole frequency is 5...10 times the Fco).

3) In the main converter the soft start pin should be connected according to the PMP8740 design, while for the secondary converter it can be connected to a small capacitor to ground (1...10nF) and left it alone, because the secondary converter is controlled by the main one also during soft start.

4) Yes, this is also correct. BTW, you can leave a small 100pF capacitor here, to avoid some noise.

Best regards,

Roberto

Hi Anbarasan,

I think the performance of the current sharing is good, because it's within +/-2% (here the average output current is 18.1A).

The output voltage of each PFC should not impact the regulation of the respective DC/DC stage, so it sounds strange to me that you have oscillation on the PFC stages.

Of course you are not connecting the outputs of the PFC stages together, correct? If so, it could lead to oscillation and even fail of this stage.

The only reason I can imagine the PFC is varying, is that the DC/DC is also supplying oscillating output current; in this regard, please double check it by analyzing the output current of the PFC with a current probe, attached to the scope.

Regards,

Roberto

Hi Anbarasan,

I think I understood why the parallel of your two DC/DC converters is having some issues. The origin can be the connection between the two grounds.

I prepared a sketch, showing the secondary side connection of two DC/DC converters (named "1" as the primary or main, and "2" as secondary), see attached file:

PMP8740 PSFB - Ground connections to avoid.pdf

As you can see, you are connecting together GND1 with GND2 with the wire "A". But when you conenect two DC/DC converter in parallel on the output, you do that by tying both pin 1 and 3 of J8, with the wires "B" and "C". 

As you can see, the wire "B" and the wire "A" are shorting both current sense resistors R91-1 and R92-2. 

In theory you should have no signal on these resistors and collect no output on Iout-1 nor Iout-2. Probably you get still these outputs because the resistance of the connections "A" and "B" is comparable or higher than the value of each resistor R92-1 and R92-2.

That means, you should avoid the connection "A", but, of course, this is not possible, because you need both controllers UCC28950 sharing the COMP signal, and therefore having a single ground (otherwise the CMOP signal would be not usable).

In my architecture, when I connect together several modules of PMP8740, I use the connections "B" and "C" but I do not connect the grounds GND-x together.

The PMP8740 is designed to have all GND-x connected by means of resistors (see R123 and C98 on page 3 of the schematic). This way, regarding the signal on the current sense resistor, there is no short circuit between modules.

Let's go back to your case: here, in order to be able to connect also all GND-x (wire "A"), I suggest to move the current sense resistor R92 on the positive output (pin 1 of J8). If you do that, you may want to replace the operational amplifier connected to R92 with a high-side current-sense precision amplifier, from the INAxxx series. Please let me know if you agree.

Best regards,

Roberto

Hi Anbarasan,

I suppose you are employing R92 = 0.2 milliohm and the gain on U11 (OPA378) = 165K / 1K = 165.

With these values, at 35A load, you should get (35A * 0.2 milliohm) * 165 = 1.155V.

I don't understand why you get different values. If it is confirmed, it might depend to the what I explained in my last explanation, where you shorted both grounds of the DC./DC converters twice.

Please check also if the parallel (the hissing sound can be a sign that the converters are not stable) is getting better by using the INA amplifier on the positive output terminal, rather to the ground.

Best regards,

Roberto

Hi Anbarasan,

No problem for the gain change, I understand your point.

Please check whether during hissing noise you see jitter in the duty cycle, or if the synchronization between the two DC/DC is lost.

It might be that SYN pin is getting some noise, so you may want to protect the connection against it (please check whether a shielded or twisted cable helps).

Best regards,

Roberto

Hi Anbarasan,

I am doing good, thanks....I hope you are fine too!

I am happy to see that you solved the issue with noise. The architecture you built is very complex and I see it is working fine, well done!

I would keep the back to back FETs in case your load is a battery, also for reverse polarity protection, but it looks like your load is an inverter, so you can remove all FETs.

Regarding this type of load, even if during braking there is some energy flowing from the motor back to the output of all DC/DC converters, you can consider that the UCC28950/51 has the automatic DCM detection, which avoids reverse current flow, from the output to the input of the DC/DC converter.

Anyway, if this energy is high enough, you may want to dissipate it by means of a braking circuit (maybe relay + resistor?) to avoid overvoltage on  the output of all converters.

Best regards,

Roberto