EEZ Bench Box 3 (BB3)

[prasimix]

About CT insulation requirements: why don't you use a CT without a primary coil and use a suitably insulated wire (or wire + silicon sleeve) to reach the required insulation rating?

I know CT w/o primary are bigger, usually for larger currents etc... but you still can find something small enough I think.

Yes, the main reason why I don't want to use such packaging is their dimensions. So far I didn't find anything that can fit within e.g. 10 x 10 mm. They can by default offer reinforced insulation that is important if CT is on the secondary side as @jbb warned me, but if I succeed to manage measurement on the primary side then functional insulation should be enough. Also I already find few single primary CT in small package the can withstand 1000 or 1500 V_rms, if Murata 53100C with its 500 V_rms is not suitable.

[prasimix]

The problem with my first attempt to measure current on the primary side was that existing PCB wasn't designed to include CT and distance between C_bulk and hi-side MOSFET is very short. Without lots of thinking I was cut V_in trace before C_bulk instead of cutting that short trace between Cbulk and MOSFET! When I realized that I've managed in the second attempt to insert Murata 53100C (1:100) as shown below:



New schematic for current measurement is now looks like this:



Output from CT is measured directly on the CS pin of LM5041B, PCB looks now even more messy but still everything works just fine



Measurement for Vin=120 Vac (170 Vdc) and I_out=2,5 A:



... or I_out=5 A and compare it with previous measurement reported in #20:

[prasimix]

Now with current measurement in place it's good to check how stressful is start-up. LM5041B as many other controllers has SS (soft-start) pin for limiting duty-cycle on power up. Soft-start sequence is also initiated every time when fault condition (e.g. V_cc under-voltage or V_in under-voltage) is no longer present. First I've tested it with SS capacitor with 10 nF that is apparently too low. Start-up current with 110 W load connected (V_out=44 V, I_out=2.5 A) looks like this:



Much better result can be achieved with considerably higher value (e.g. 100 nF):




 

[prasimix]

A clarification/correction is needed about SR controller that is presented in #6. Infineon IR1168 min. operating voltage is set to 8.1 V (typ.) by ULVO protection therefore +7 V rail shown in that schematic does not work. Actually, if you try to power up the circuit, sync MOSFETs (Q6, Q7) will be disabled, but rectification will be in place using MOSFET's body diodes giving an impression that everything is working fine! But it is not a case and that can be easily proven by checking MOSFETs dissipation and of course gate drive signal that does not exist.

In my first prototype I had about +9 V for driving IR1168 but in latest prototype QR flyback offers just  about +6.5 V (that is step-down with LDO to +5 V). Therefore in the final PCB I'll use IR11688, a similar model which working voltage goes down to 4.5 V.

Anyway, that IR1168/IR11688 is a great controller that withstand up to 200 V (V_ds) without any extra circuits (like e.g. SRK2001). So far I didn't find anything even closer to it in the same price range. Also don't look for IR1163 as lower operating voltage IR1168 version announced for 2014. It seems that it is branded/labeled as mentioned IR11688.

[prasimix]

Finally I'll present aux/bias power supply section that is required to power on primary side LM5041B (and indirectly Si8233 A-side through LM5041B) and on secondary side NFB control/tracker, Sync. rectification (SR) controller and provide all required voltages for connecting post-regulator similar to one used in EEZ H24005 Power board.
Coming to the final solution was an expectedly long journey caused with lots of presumptions and lack of experience. On the first CF-DIC PCB prototype I tried to use PI's TNY266GN flyback controller with few off-the-shelf transformers that has enough secondaries (four in total, or 3 + 1 AUX if you like). I was especially eager to see how Feryster's 2399 transformer will behave but end result was unacceptable, I suspect because of its primary inductance that is too low.
Mentioned controller is well packed and integrated (e.g. does need external switch and protection circuitry) with reasonable price.
In the meantime I found few interesting Quasi-resonant (QR) flyback controllers also well integrated and for even better price. I've decide to give a try to the latest (5th generation) ICE5QR4770A controller. Once again I found that mentioned Feryster is not a solution, and I tried another off-the-shelf transformer this time from WE (750811914). This time I found another issue, ICE5 for unknown reason cannot start with been "kicking off" by applying shortly required bias supply. I tried to solve that with assistance of Infineon support but for unknown reason it didn't want to start, possibly because some of its protection mechanism is triggered.
Finally I found ST's VIPer family of QR controllers and decide to use VIPer35 in my latest prototype. In parallel I've arranged with people from Feryster to design a custom transformer in accordance with my specification.
That was a wining combination: nice flyback with broad input voltage range that works in QR mode with all needed output voltages. The latest schematic for that part of the CF-DIC is shown below. AC input section will be shared with CF-DIC power section.



Measurement on VIPer35 Drain pin without load (Vin=230 V_ac):



... or with moderate load applied:



It's clearly visible that VIPer35's internal MOSFET is switching on at the lowest point. That can be adjusted for the used transformer by changing the value of R_lim (R19).
Usual way of providing precise enough and cheap control loop with TL431/TL432 is not used since all critical outputs are additionally regulated by BJT or LDOs. Therefore just zener diode (ZD3) is used to set a main secondary output voltage.
Perhaps "double" rectification of AUX secondary shared between powering VIPer35 (D8, ZD2) and LM5041B (D1, Q1, ZD1) is not needed and can be reduced to single circuit. I have to test that further.

Your suggestions and comments are welcomed as usual.

[prasimix]

I've decide to make CF-DIC as a separate/stand-alone module that is within size of Mean Well LRS-150-48 module previously used in EEZ H24005 project. All files are now available on the GitHub under CF-DIC project. I'm going to order new PCB soon after some additional checking is performed since new layout is quite different from previous one but I believe that it should work nicely.

[jbb]

Would you like some review help with it?  I can see a couple of points that are quite close (and could still be fine, given the scale involved...)

[prasimix]

Of course, just go ahead! I can provide separate images of only top and bottom layer traces for better visibility if you cannot open Eagle file.

[prasimix]

In case that converter is housed in metal chassis I'd like to insert insulation sheet between bottom PCB layer components and metal plate.
I'd like to know what is a trade name for such insulation when ordering from suppliers. Here is how it looks like in Mean well module:

[prasimix]

I've added in latest revision an optional temperature alarm based on simple LM211 (IC10) comparator and NTC (see Sheet 3of3). It is intended to pull-down PWRGOOD signal when temperature near one of sync mosfet become very high (currently set about 85 oC). NTC is SMT and it's located near Q8. Maybe that is not the best way how to accomplish temperature monitoring neither the best component: we can expect that some of SiC MOSFETs will be much hotter during the operation but my logic is that sync mosfet heating will increase proportionally.
Also instead of controlling PWRGOOD signal that require external agency for turn-off input voltage or remove connected load, another possibility is to block NFB optocoupler (OK2) or even LM5041B by inhibiting ULVO input while error state is reached. Former should be simpler while later asks for another optocoupler or moving a whole temperature monitoring circuit to primary side and power it from +12 V that is just fine for the LM211.

[prasimix]

Thermal protection on the primary side of the converter could be something like this:



LM211 comparator is monitoring voltage on non-inverting input that comes from voltage divider with NTC (e.g. 10K/3570K). Reference voltage (that is stable enough) comes from LM5041B (REF pin) via R56/R59 divider. When temperature becomes critically high (and NTC value will decrease), LM211 will trigger Q5 that will pull to ground ULVO input until new power recycle. I've selected this method over comparator with hysteresis since it should be used only when something went terribly wrong.

[Giaime]

Are you sure you can guarantee the minimum holding current of the SCR, with such high value resistors?

[prasimix]

I've tested the circuit with a little bit larger SCR (MCR08) instead of specified "baby" NYC0102BLT1G (SOT-23 case), and triggering and holding is working fine repetitively.

[prasimix]

What I see as a possible issue here is unwanted triggering due to some glitch in AC line that is picked-up by LM211. I've tested it on the breadboard and can simulate such situation with powering on ancient Weller soldering station that is on the same power line. I believe that situation when everything is on the PCB will be better, or suggested comparator/SCR circuit has to be improved?

[prasimix]

Ok, it seems that quick remedy is to simply reverse a whole circuit and move up threshold level that will add extra immunity:

[Giaime]

Just to be sure, I would place some capacitors to ground on both LM211 inputs, to aid immunity. You don't need us triggering speeds for temperature protection. And maybe a cap in parallel to ZD5?

I've tested the circuit with a little bit larger SCR (MCR08) instead of specified "baby" NYC0102BLT1G (SOT-23 case), and triggering and holding is working fine repetitively.

Hopefully you never come across a "right by the limit outlier" component  Usually for extremely low currents, an SCR made up of standard bipolar transistors works better, it has very low holding current.

[prasimix]

And maybe a cap in parallel to ZD5?

Perhaps I can just move existing C42 to other side of R45, but I think that its current placement in combination with ZD5 that I must added to not exceed abs. max. ratings of ULVO (i.e. +7 V) for upper V_in give us a robust enough solution against glitches.

[Giaime]

That's ok as well. It's just that I don't like fast switches discharging capacitors with no current limitation. That as well can cause glitches. Putting C42 directly on IC pin limits its discharge current.

That's really nitpicking on my side, of course 

[Rerouter]

You mention on your website that the QR flyback is routed poorly, might I ask what parts in particular you feel need to be rerouted (its in the name :p), I figured out a nicer way of packing the LP2951, but I believe your referring to the viper35 side?

[ogden]

Have you considered putting a CT in series with the transformer primary? It should be voltage clamped by the secondary output voltage.

Alternatively, you could place a 0.1R (approx) current sense resistor in series with the return current from the push-pull FET source pins. This effectively measures the total inductor current for you. Losses would be quite small.

Neither of that looks promising to me, especially adding anything in series with the source pins. Maybe with ordinary MOSFETs but not with SiC. Therefore I'll proceed with testing what is possible to do with adding CT at the buck input as started in post #20. I believe that better results can be achieved.

Most likely I am out of my league here, but IMHO shunt and isolation amplifier, like Si8920 could be solution... perhaps. It's Up to 5000 Vrms for 1 minute after all.  Just two cents obviously

[prasimix]

You mention on your website that the QR flyback is routed poorly, might I ask what parts in particular you feel need to be rerouted (its in the name :p), I figured out a nicer way of packing the LP2951, but I believe your referring to the viper35 side?

You mean on the GitHub not website? The latest PCB that was an attempt to clean up a jungle that grew up over the time (and shown on #10 and #26) and to separate pre-regulation from post-regulation stage since that gives at the end much more flexibility.
CF-DIC (for pre-regulation) works on that PCB almost suspiciously good , but during the reorganization of the whole PCB layout I made a fatal mistake by putting LM5041B beneath power inductors . The result is a sort of "over-synchronization" when RT input pin (used also for ext. sync) obviously is picking up noise when load is connected (or a sort of positive feedback loop is created) and working frequency jump into the sky (more then 20x higher then set frequency defined with RT resistor!). That alone make a whole PCB completely unusable.
Another issue is connected with QR feedback that is not so problematic to rectify: I tried to supply both VIPer and LM5041B from the same rectified AUX voltage, but it seems that VIPer don't like that. Therefore two separate diode + cap is required. But, it's also possible that such weirdness is connected to the first issue and that LM5041B starts to sink too much power trying to drive MOSFETs (PP stage in the first place) with insane frequency. I'm working on new PCB, not r3B8 that is on the GitHub but currently r4B4 .

[prasimix]

Have you considered putting a CT in series with the transformer primary? It should be voltage clamped by the secondary output voltage.

Alternatively, you could place a 0.1R (approx) current sense resistor in series with the return current from the push-pull FET source pins. This effectively measures the total inductor current for you. Losses would be quite small.

Neither of that looks promising to me, especially adding anything in series with the source pins. Maybe with ordinary MOSFETs but not with SiC. Therefore I'll proceed with testing what is possible to do with adding CT at the buck input as started in post #20. I believe that better results can be achieved.

Most likely I am out of my league here, but IMHO shunt and isolation amplifier, like Si8920 could be solution... perhaps. It's Up to 5000 Vrms for 1 minute after all.  Just two cents obviously

Thanks for that, but I don't like to put anything in series with source pins. Solution with CT properly connected in series on the input side looks good, and cheap transformers with functional isolation exists (i.e. 500 V), and for one who wants better isolation it's possible to use with the same footprint with Hi-pot of 1 kV (possibly even 2 kV).

[prasimix]

... I figured out a nicer way of packing the LP2951

I forgot to mention that I'll probably move LDO's to the post-regulator board to have them closer to the load (i.e. control circuits).

[ogden]

shunt and isolation amplifier, like Si8920 could be solution

Thanks for that, but I don't like to put anything in series with source pins.

I said nothing about source pins, just mentioned shunt+Si8920. Put shunt where CT was originally planned, not on source pins

[prasimix]

Sorry, didn't got it at first.