Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

DCM Full Bridge is best because no shoot through current (cf CCM Full Bridge)

Status
Not open for further replies.
T

treez

Guest
Hello,
We are doing a Current mode, DCM Full Bridge SMPS,
VIN=390VDC (PFC output) ;
VOUT=400VDC max;
Pout=100W max ;
Application=100W LED lamp
(V(LED) = 340-400VDC at 250mA);
F(sw) = 100KHz
Dimmable down to 10W.
Mains isolation

We have chosen to do the Full Bridge as a DCM Full Bridge.
The reason for DCM is because CCM Full bridge’s suffer the problem where the bottom/top FET in a ‘leg’ gets spuriously turned ON when the top/bottom FET in the same leg turns ON..resulting in momentary ‘shoot-through’ current.
This can be avoided in a DCM Full Bridge because the turn-on of a DCM full bridge can be heavily damped (slowed) so that this spurious effect does not happen. Heavily damping the turn-on in a DCM full bridge does not incur heavy switching losses because turn-on involves zero current.

Do you agree that this is the best plan?

LTspice Simulation of DCM , Current mode Full Bridge SMPS provided.
 

Attachments

  • FULL BRIDGE DCM_400vout.txt
    14.9 KB · Views: 53

Shoot through isn't a problem in adequate designs, and it isn't especially challenging in full bridge converters (otherwise LLC converters would be a nightmare). Using a full bridge in DCM means you'll never be utilizing the true power capacity of the converter, at 100W you might as well use a forward converter.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Using a full bridge in DCM means you'll never be utilizing the true power capacity of the converter, at 100W you might as well use a forward converter.
Thanks, though the Vout is 400V, and as you know, the diode off state voltage ends up being far more with a forward converter, whose Duty is limited to 0.45, than in a Full Bridge, where the Duty can go up to 0.8
(I am talking out Boundary conduction mode here...but even with DCM , you can still use far lower voltage rated diodes with the full bridge than the forward.)
Shoot through isn't a problem in adequate designs, and it isn't especially challenging in full bridge converters (otherwise LLC converters would be a nightmare).
The type of shoot through that I speak of in the top post certainly isn't a problem for the LLC. As you know, in the LLC, the FET turn-on happens when the FET Cds capacitances have charged such that the voltage across the fet that’s turning on is almost zero, so there’s none of this spurious turn-on with an LLC.
-The reason we didn’t do an LLC was because it can't manage the full Vf and dimming range without having too high switching frequency at lower load (very dimmed).
 

The type of shoot through that I speak of in the top post certainly isn't a problem for the LLC. As you know, in the LLC, the FET turn-on happens when the FET Cds capacitances have charged such that the voltage across the fet that’s turning on is almost zero, so there’s none of this spurious turn-on with an LLC.
Even when hard switching, shoot through should not be a problem. If it is then its a shortcoming of the drive circuitry, and maybe the FETs. Hard switched bridge converters are used successfully in the range of many kW.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Even when hard switching, shoot through should not be a problem. If it is then its a shortcoming of the drive circuitry
The attached pdf shows the shoot-through that I speak of (in a CCM Full Bridge). As you know, those current spikes that you see are due to the fact that when either the upper or lower fet of a leg turn on, the other FET gets temporarily turned on by the sudden flow of current through its junction capacitances, which result in a temporary Vgs voltage which momentarily turns on that fet..resulting in shoot-through which is visible here.

The LTspice file here of this CCM full bridge shows these current spikes due to shoot through. (The pdf attached here is actually a CCM full bridge but the primary current looks DCM due to the magnetising current being large in relation to the referred secondary current.)

Yes I agree that these current spikes can be reduced by making the fets turn-on more slowly, but that then incurs more switching loss.
…Unless, one uses DCM, and then the turn-on of the FET can be very slow because in DCM a slow turn on of the FET does not incur much turn-on switching loss.

This is the point that I am making in the top post.

Hard switched bridge converters are used successfully in the range of many kW.
..Well yes, though as we know, a very highly experienced engineer has recently been on this forum to warn about that. But as we know, the only reason those multi kw full bridges work is because the heatsinking is just increased to take care of it.
...............................................................................................................
The point is that in DCM you can really slow down the switch-on transient of the Full Bridge FETs and reduce those current spikes, without incurring lots of turn-on switching loss.
…whereas in CCM, if you really slow down the switch-on switching transient of the FETs then you incur lots of switching loss…this is why I am preferring DCM for this particular Full Bridge application.
 

Attachments

  • FET current.pdf
    33.4 KB · Views: 77
  • FULL BRIDGE CCM_400vout_.txt
    13.4 KB · Views: 36
Last edited by a moderator:

That's not shoot through, that's just the charging and discharging of Coss. The spikes have a fixed amount of charge in them, independent of CCM or DCM operation.
 
The thing is, whether its shoot through or charging and discharging of the Coss of the upper and lower fets of a leg, we still want to minimise the di/dt of it and the magnitude of it. As you know, this can be achieved by slowing up the fet turn-on transition.
…and this is my point of the top post, because slowing up the fet turn-on transition is much better done in dcm than in ccm. –Because in DCM, there is zero current in the FET at turn-on, so slowing up the fet turn-on transistion in dcm results in less switching loss than in CCM.
 

The thing is, whether its shoot through or charging and discharging of the Coss of the upper and lower fets of a leg, we still want to minimise the di/dt of it and the magnitude of it.
Why? Changing di/dt or dv/dt won't change the amount of energy dissipated associated with Coss. That energy is only a function of the total charge and voltage on Coss, which is independent of load current/duty cycle. So in order to maximize efficiency, you should try to maximize the energy delivered to the load per cycle.
 

thanks, but we wish to reduce di/dt to reduce EMC problems. We wish to do this by slowing the turn-on of the full bridge fets, and reducing the turn on of full bridge fets is better done in dcm, and its less dissipative to reduce turn-on loss in that case.
As you know, the input and output current of this particular full bridge is so low that its ok to be in dcm.
 

Hard switching of Coss should not be a substantial contributor to EMC at 100W. Do you actually have some EMC test data or is this just based on simulation?

If Coss is causing an EMC problem then your FETs have far too large of a Coss and you should choose a lower power device. Picking oversized FETs and then kneecapping the entire converter by running it in DCM with sluggish gate drive isn't going to give good results.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Hard switching of Coss should not be a substantial contributor to EMC at 100W.

…Yes that’s right…but I feel the point is being missed..the point is that if this Full Bridge had instead been done in CCM, then it wouldn’t only be the Coss based losses that would be in there….as you know, if it had been done in CCM, there would be those extra turn-on switching losses involving “overlap” of voltage and current in the FET at turn-on.
The whole point of doing this in DCM is that those “overlap” type switching losses don’t exist at turn-on of the FETs.

Admittedly, when in DCM, there are “overlap” type switching losses in the FETs at turn-off, but the turn-off switching transient can be made very fast, without incurring a bad EMC penalty.

So this is why we are doing it in DCM instead of in CCM..And yes we will choose to slow up the FET turn-on switching transient..might as well, it comes with little extra switching loss penalty.

- - - Updated - - -

kneecapping the entire converter by running it in DCM with sluggish gate drive isn't going to give good results.
As discussed, its only the fet turn-on transient that we will slug, the turn-off transient will be very fast.
 

As discussed, its only the fet turn-on transient that we will slug, the turn-off transient will be very fast.
And due to the higher peak current in DCM, the turn-off losses will be increased vs a CCM design...

I think if you were to do a full side by side comparison of two designs, CCM would perform at least as well. At least I see no convincing reason to upend conventional wisdom based on a single LTspice simulation.
 

Yes, true, but the peak primary current is only 2002mA with the 1mH primary inductance and in dcm, so not much turn-off losses anyway, and with the fast turn-off, its even less of a problem. In fact most of the primary current peak is magnetising current which will be pretty much the same whether dcm or ccm, so not much difference in primary peak current in this application between dcm and ccm.

I actually believe that due to DCM’s absence of “overlap” turn-on switching losses, and the fact that turn-on in CCM (which does involve “overlap”) cannot be fast due to noise problems, that overall, DCM will give less switching loss, surely you agree?

In summary, it’s the fact that turn-on switching losses are relatively higher in CCM, and the fact that DCM doesn’t have turn-on switching losses, which means that DCM will give less switching losses in this case.

Another point about this full bridge converter is that it has to be dimmed down from 100W to about 10W….and DCM gives a greater range of peak primary currents over this dimming range, so it will be less noisy for the wide range dimming than CCM.
 

This is very related to the above thread, but is a new rendition of it....

Do you agree that this Current Mode Full bridge (vin=390vdc, vout=400vdc, Pout=100w, fsw=100khz) is indeed best done in DCM and not CCM?

Attached here is the CCM simulation, and the DCM simulation in LTspice.
(Also the pdf basic schematics)

Even in DCM, the peak primary current Is only 940mA at max load. –so the turn-off switching losses will not be too bad in DCM. Also, in a Full Bridge SMPS, the turn-off switching transient can be made extremely fast without incurring an EMC penalty. –so the turn-off switching loss will be made even less.
In a Full Bridge SMPS in CCM mode , the FET turn-on transient is particularly noisy, because there is a sudden “step” drawing of current from the input capacitor. This is due to the fact that in CCM mode, the FET gets switched on ‘into’current. The current loop involving the input capacitor is usually unfortunately quite a large-ish current loop, so the noise problem is made worse. Therefore the FET turn-on transient in CCM is very noisy and needs to be slowed up by damping, -this incurs high switching loss in CCM because of the voltage and current ‘overlap’.

Whether in CCM or DCM, the FET turn-off transient is not as noisy because it doesn’t involve such a high di/dt change in current from the input capacitor. This is evidenced because in SMPS schematics, one often sees a fast turn-off being enabled by a diode being placed across the series FET gate damping resistor. (anode to FET gate).

Therefore, DCM was chosen for this Full Bridge converter, since in DCM there is zero ‘overlap’ switching loss at the turn-on transient.
Also, in DCM, the output inductor can be made smaller than in CCM.

So this is the fact:
If you are doing a Full Bridge SMPS and Peak primary curent would be quite low in DCM, then definetely use DCM (not CCM), because you have zero turn-on “overlap” switching losses in DCM, whereas in CCM, you have especially high ‘overlap’ switching losses at FET turn-on, because the CCM FET turn-on transient also has to be relatively highly damped to mitigate noise issues.
 

Attachments

  • FULL BRIDGE CCM_400vout_.txt
    13.3 KB · Views: 43
  • FULL BRIDGE DCM_400vout_.txt
    13.2 KB · Views: 39
  • FULL BRIDGE CCM_400vout_.pdf
    22.8 KB · Views: 43
  • FULL BRIDGE DCM_400vout_.pdf
    22.5 KB · Views: 63

The reason we didn’t do an LLC was because it can't manage the full Vf and dimming range without having too high switching frequency at lower load (very dimmed).

If your load current goes way down but the voltage doesn't, and if the input voltage is relatively fixed (i.e. PFC pre-regulated), an LLC will work great. You just have to use a largish leakage inductance (say 1/2-1/3rd of magnetizing L) and set the turns ratio so operating frequency at full load and nominal voltage is just below resonance. And then for very light load, use a controller that goes into burst mode (very common).
 

Thanks, but burst mode is not wanted as it may make the leds flicker.
Also, as you know, the LED voltage does actually go down when the led current goes down, especially at the lower led current levels.
 

Burst mode frequencies are typically tunable to go from 100's to 1000's of Hz. Eye can't see that. Lots of LED systems do dimming by PWM'ing at a few 100 Hz.

Also, LED's don't go down very much in voltage as current is reduced, unlike a resistive load.
Lotsa folks use LLCs with PFC for LED's with 10:1 dimming range or grater.
 
Last edited:
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Thanks, but we have 100 leds in series, and each led could be from a high vf batch, or a low vf batch, so the LED string voltage, even without dimming, can be as variable as 320vdc to 410vdc over temperature.
As you know, When you dim leds down to the "knee", they get more variable in vf......even though burst frequencies can be 100hz or so, the irregularity can still be seen by the eye....paricualrly because sometimes the burst frequency can be less than 100hz, you don't tend to have much control over it.
The LLC design template shows that our vin and vout are just too variable for LLC....after all, PFC output voltage is varying between 382 and 398v too.

- - - Updated - - -

I appreciate LLC can be used though, but its expensive….eg if you also vary your pfc output voltage as the LLC output voltage varies, then you can squeeze more out of the LLC…also, if you drive the LLC with a micro, and you can adjust the dead time as the laod varies than you can also squeeze more out of the llc….So I don’t debate what you say particularly, but with our cheap situation, we cannot do that with a straightforward LLC and using leds from any vf batch that comes along.
 

Attachments

  • LLC design template_1.zip
    400.8 KB · Views: 35

he reason for DCM is because CCM Full bridge’s suffer the problem where the bottom/top FET in a ‘leg’ gets spuriously turned ON when the top/bottom FET in the same leg turns ON..resulting in momentary ‘shoot-through’ current.

Using DCM and slow turn on, you have ch0sen a very curious way to "overcome" your power electronics problems.

Certainly DCM has much more dynamic response ability if you are using the converter to drive an LED string in 100Hz bursts...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Thanks, though the thread has become a little convoluted before your input, DCM was chosen over CCM in the end because….
1…..Smaller output inductor
2……No turn-on “overlap” switching losses in FETs for DCM.
3…..Even in DCM, The peak currents are very low, so turn-off switching losses are low in DCM.
4. The turn-off current is higher in DCM, but as you know, turn-off is less noisy than turn-on for the Full Bridge SMPS. This is because the turn off current is flowing in an inductor and so cant suddenly changeto be a high spike, wheras at turn on, there can be a sudden relatively high spike of current coming strainght from the input capacitor, and through the cds’s of the two fets of a leg.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top