Boost converter problems

[trobbins]

Sorry - didn't mean to make work for you - but unfortunately once design questions are raised then that sort of needs a modicum of design details to be collated. 

Regarding the FET, it is often about listing the available options you can obtain that generally meet the protection limits (eg. Vds in this case), and then hone in on package type that suits the application, and then look for trade-offs between Qg and Coss and Rds-on - and even then if you're keen the outcome would be to purchase the likely best 2-3 models, and try each of them under standard conditions.  It's not easy to compare final efficiency, as the frequency and duty cycle and drive capability and parasitic loop inductance and capacitances may all have some impact.

Is 150mA the peak current at turn-off ?

Do you get any transient overshoot in Vds?  You may need a good oscilloscope and probe and probing technique (ie. not just standard ground wire and alligator clip).

 

[JohnRoberts]

You can probably tell where your efficiency is lost with a temperature probe, or very carefully probing around with you fingers (perhaps in a rubber glove.

JR

 

[gemini86]

To be clear, i'm very grateful for all input in this matter, trobbins. I don't mean to sound as if I just want the answers, I really do want to learn all I can about this.

I think at this point i'm very limited in my ability to really fine tune this circuit. I don't have an oscilloscope (at least nothing that's capable of over 20khz).

John, I'll give that a try. I know nothing is getting "hot" and the efficiency numbers suggest that nothing dramatic is happening.

 

[JohnRoberts]

I suspect the lost efficiency is going to heat not RF... Let your fingers do the walking, carefully.

JR

 

[gemini86]

I suspect the same, since my current fet has Rds(on) around 1 ohm.

 

[trobbins]

Can I suggest working out how many watts is lost - this sort of does require accurate voltage and current measurements, but if you use the same meter and technique then that eases the inaccuracy.  Then as suggested, attempt to guesstimate which are the main parts and what is their relative surface temperature - my expectation is the fet, inductor and diode is where > 90% of heat ends up (and yes all the watts lost from efficiency will go to heat - there is no 'dark matter'/emi loss of energy here).  Then you can sort of estimate the watts lost in each part, and then do a simple thermal calculation of eg. Tj for the FET and diode (but that does also mean having a guess at Rc-a [case to ambient] for each part), and also look at their spec sheets as to whether you are getting in to derating territory.

Before you know it, you've started to compile a batch of relevant design details

 

[gemini86]

I did the measurements with the same meter. I used 1 ohm 1% resistors to measure current draw for both low and high sides. Although, I should probably go higher tolerance to get more accuracy. I have a spreadsheet drawn up where I entered in all my measurements, calculate power in and power out, then calculate efficiency.

 

[trobbins]

You're on your way

Spreadsheets are also good for adding part types and values and embedding all the basic design stuff.  There are heatsink tutorials for working through Tj-c, Tc-hs, Ths-amb.  The inductor doesn't just have DC resistance loss - their will be core loss plus possibly added loss from higher AC resistance due to skin effect - ahh, the joys of smps design just keep on keepin on!

 

[gemini86]

Yes yes. fortunately, I designed the converter as a plug in daughter board so I can continue working on improving the design in the future.

 

[Niketouille]

Hi,
I've been into some trouble with a dcdc converter design. It is based around a 34063.  My expecation was to get very little output noise. The two main issues were ground path... and the feedback network that needed to be decoupled to ground with a small cap.
That's all, I just wanna share experience.
Cheers
NKT

 

[gemini86]

Sorry for not replying for so long. My wife and I just got a new puppy and we've been tending to her day and night. Very little time for lab work (that's what I like to call my glorified tinker time).

Thanks for that NKT, So you had to bypass the FB pin to ground with what size cap?

Just an update on this;
Although I'm figuring out the efficiency problem and it is basically functioning, I am now encountering some odd problems when under load.
I've uploaded a video to youtube to show what's going on.

Sorry for the poor video quality. I have the scope set to roll mode, moving at .5 seconds/division. The upper trace is the output ripple, which is about 14V at ~20Hz. The lower trace shows what the fet gate is doing. The video starts with tube heaters off, and slowly warming up and tube gradually conducting from zero to full load. The controller seems to be diving hard at about 20Hz or so once the full load is applied. When I was first looking at the gate and output, I was looking for the 220KHz+ noise that I was expecting, so this low frequency noise went unnoticed until I started preamp listening tests, and obviously there was a horrible motorboat noise at 20Hz on the output. I'm struggling to figure out what is going on here. FB voltage reference unstable? Wrong Rsense value?

I've ordered a new fet, one with much better specs for this purpose than what I'm currently using. Once I have the new part I'll see if things improve, but I'm thinking there's still something wrong here.

Any help is greatly appreciated!

 

[Niketouille]

Thanks for that NKT, So you had to bypass the FB pin to ground with what size cap?

10 to 22nF was working ok for me.

 

[gemini86]

I ended up using 10nF from hv to fb.