Zener diode knee current to obtain best efficiency

Status
Not open for further replies.

gdylp2004

Member level 5
Joined
Dec 4, 2011
Messages
81
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
2,192
Hi guys,

I've attached a shunt linear regulator in attempt to provide a load regulation which consumes about 40mA in average. The regulated load voltage should be of 12V and thus a 12V Zener diode is chosen.

If we looked at the setup attached, there is a significant amount of power being wasted in R2 because of the high differential voltage of ~88V across V1 & cathode of Zener. So in order to reduce this wastage, the current through this resistor (R2) must be as low as possible, but not lower than the o/p current (which is 40mA for my case) + the knee current of the Zener.

And ensuring as low current through R2 also means that the I(z) must be also as low as possible. So my question is, is it correct to look for a Zener with the least test current since the knee current is never provided by the manufacturer and even if this information is valid, at I(z) = knee current, we would also expect that V(z) would most probably not maintained or regulated at 12V.

Also, is there anyway we could determine the di/dv characteristic from the datasheet like this one here: https://www.farnell.com/datasheets/69971.pdf

Thanks.
 

Attachments

  • edaboard_shuntreg.jpg
    375.7 KB · Views: 153

A zener diode does not break down abruptly at a single voltage point. The knee is always a bit rounded so that the operating current must not be too low. It should be well above the leakage current. The di/dv characteristics can be derived from the given dynamic resistance, but the value is not constant as the dynamic resistance decreases with increasing current.

Wastage of power by dropping from the source to the regulated voltage cannot be entirely avoided unless switching techniques are used. There are means to reduce it, though. One way is to use a transistor as a buffer instead of shunting the load directly with the zener diode. This allows using a lower current through the diode and the power saving is especially effective when the load current is not constant. Regulation will be slightly worse than with the diode alone due to variations of VBE with load current.

 

Yup, thanks for the heads up. The one you've recommended is a series linear regulator which purportedly to be of higher efficiency.

I did a load test and measured the I(avg) to be of 40mA. Although I've used a current probe or a 1ohm sense resistor in series in attempt to observe the waveform on how the load draws this 40mA, no current waveform could be registered on the o-scope. So I can't really determine if the load is usually ON or OFF.

By the way, the load is a gate driver which drives the nMOS of a SMPS, so using switched-mode as you suggested is out-of-the-question.

Are there any other better efficient linear regulator that you know of?

Thanks.

---------- Post added at 22:36 ---------- Previous post was at 21:13 ----------

Also, instead of using BJT, could I subsitute it with a MOSFET for the series design?
 

If we have to exclude switched-mode operation and use only linear regulation, the minimum power usage is limited to (source voltage)*(load current). In the present case, that's 100V*40mA = 4W at full load current. The useful power at the load is 12V*40mA = 0.48W. The loss of 88V*40mA or 3.52W is unavoidable and the theoretical maximum efficiency is 0.48/4 = 12%.

With the shunt zener technique, the full 4W is drawn from the source at all times, no matter what current is drawn by the load at any particular instant. An advantage of the series regulator is that current is drawn from the 100V source only when demanded by the load. If the load current is always at the full 40mA, then the series regulator has no advantage as far as efficiency is concerned.

The preceding description does not include extra losses caused by the need to pass a minimum current through the zener diode to maintain breakdown even when the load is drawing the full load current. This applies to both the shunt and the series regulator. The zener current can be only a very few mAs, so the overall efficiency can be not much worse than the theoretical maximum of 12%. Therefore, there's not much to gain by using a more efficient (and more complex) design.

A more efficient regulator could be made to reduce the standby current to a fraction of a mA and to improve the regulation. But the circuit will have to be more complex and the gain in efficiency will be small.

Also, instead of using BJT, could I subsitute it with a MOSFET for the series design?
A MOSFET can be used but regulation with changes in load current will be poorer. The zener diode provides the voltage reference VZ. With a BJT, the output voltage is VZ - VBE, and VZ - VGS with a MOSFET. Both VBE and VGS change with load current, but VGS changes much more than VBE for the same change of current. Therefore, regulation is poorer with a MOSFET.
 
I don't understand why 'a switching regulator is out of the question'?

Keith
 

I don't understand why 'a switching regulator is out of the question'?

Keith

If I were to use SMPS to power the gate driver for the main SMPS, I would need another gate driver to drive the former SMPS and this driver would then need another auxiliary supply like the main SMPS do. And this goes on and on and never ending.

The main SPMS is a 100W system and hence 4W loss is only 4%. Afterall, the driver which I'm using is considerably a smaller load in the whole system and thus I've chosen the linear method instead. The driver I'm using now (IR2117) draws an avg of 40mA and the other driver (which is an optocoupler draws an effective load current of only 7.1mA which makes linear reg an even better choice.
 

Now that you've provided some more details about your intended application of the 12V supply, there's a way to further reduce wastage of power if the main SMPS has a 12V output. In the circuit below, the auxiliary supply we've been discussing puts out about 11.3V (zener voltage minus VBE). With this arrangement, the load (what you call the gate driver) will draw power from whichever voltage of the two sources is higher. Once the main SMPS is running, its 12V output is higher than the 11.3V from the auxiliary supply. The gate driver will be powered by the main SMPS and the auxiliary supply will remain in standby mode, drawing only the zener current from the 100V source.

 

Thanks Pjdd, the method you suggested is the start-up strategy which I've read this somewhere where the converter usually utilises a transformer configuration to provide isolation. And because a transformer is being used, they could tap any o/p voltage of any values they desired.

My converter is a fixed 100-28V non-isolated configuration. However on second thoughts, perhaps I could replicate a series reg where the i/p now is 28V and its o/p connected to the "+12V from main SMPS" terminal of your latest circuit?

Although the similar problem persist, the voltage drop, hence power wastage is reduced since p.d. is no longer 88V but 16V (Power waste:16X40mA=0.64W only!).

---------- Post added at 12:44 ---------- Previous post was at 12:34 ----------

Also, if you're interested the attached image is my HV DC setup (albeit the HV now is 100V instead of 30V). I did post a thread on it: https://www.edaboard.com/threads/233047/

By the way, could you share what software are you using to draw those little circuits? They are pretty nice to see.
 

Attachments

  • edaboard_IR2117.jpg
    1.4 MB · Views: 98
Last edited:

Yes, it will work as long as they can share a common ground. You could also use a 3-terminal 7812 since 28V is comfortably lower than the max input voltage rating of a 78xx regulator. Power dissipation in the regulator will be the same 0.64W you've already calculated. You could provide a small heatsink to keep it cool but that's not essential as it can easily deal with that level of dissipation without a heatsink.

Also, if you're interested the attached image is my HV DC setup (albeit the HV now is 100V instead of 30V). I did post a thread on it: https://www.edaboard.com/threads/233047/
I'll go through the thread later as my internet connection has been behaving erratically for the past 24 hrs.

By the way, could you share what software are you using to draw those little circuits? They are pretty nice to see.
For complex diagrams, I use CAD programs like CircuitMaker 2000, Protel and others, but for those small simple circuits, I use a very unsophisticated process - namely art progs like MS Paint, Ultimate Paint, etc. It's a legacy from the days when I used an Amiga as my main computer. Some Amiga art programs had provision to store small bitmap images as brushes, somewhat like the way CAD programs have symbol libraries.

Basic Windows programs don't have such a provision. I use either of two methods with MS Paint: 1) I made a standard bitmap image with various symbols at the top of the page. I Cut/Copy and Paste the symbols and arrange them as needed on the lower part of the page. 2) I save each symbol as a small individual bitmap image file and store them all in a common folder. Then I use the "Paste From...." tool in MS Paint to call them up as needed.

After the drawing is finished, I save the page with an appropriate name and crop it as necessary. Irfanview is very nice for cropping and other manipulations. The whole process is rather tedious and not really suitable for large complex diagrams. OTOH, I like the ease with which new symbols and elements can be created, and the way the schematics can be saved directly as bitmap images with a wide range of formats to choose from. PNG is perhaps the most efficient format for images with a small number of different colours. Many of the small schematics I've posted in this forum are less than 1KB in size.

And that reminds me of something else. If I may make a suggestion regarding your images: Take this latest one as an example. It's 1.40MB which is quite large, especially for those with slow connections, and it's not as clear as it could be. It can be much improved after you've downloaded it from your Blackberry. I did a few things to it and re-uploaded it here.The one on the left is 40.3KB and the smaller one on the right is 12.5KB. It took me a couple of minutes with Irfanview.

 


Thank you Pjdd for your suggestion. I would have to use the "another series regulator method" instead of a 7812 IC mainly because my objective is to build an "as simple as possible" circuit, hence using discrete components would be my first choice. And if you're interested to know, I would build a 2nd converter with an improved topology where I changed the catch diode with another mosfet as well as this "double" series reg with an All-in-one IC chip package like this: https://www.ti.com/lit/ds/symlink/tl783.pdf. But still, I appreciate your suggestion and I especially like your idea of the "change-over supply" method to reduce power wastage.



I think I would just stick to LTspice when posting a schematic. Everyone has their own style which suits them best, and most probably you've already mastered your way for some time. :-D


You definitely could suggest, because when you suggest, I learned. I am a little surprise as you knew I was using a Blackberry to capture these photos, though it is actually very simple to find out by copying the image URL into this site: **broken link removed**.

My intention for such a high resolution is because just in case someone would be interested to view a detailed information from the picture as sometimes I do post the actual connection on breadboard, etc. The image I attached earlier was abit hard to see despite large file size was most probably due to my shaky hands, my apology.

I think I'll start down sizing my uploaded images from now on, as I did forget that not all users have fast enough connection speed as I do. Thanks for the reminder!



---------- Post added at 00:49 ---------- Previous post was at 00:35 ----------

Here is some updates from my hardware progression.

Attached is the breadboard connection of a shunt linear regulator. The power resistor has a 10W rating and it is rather hot to touch after some time although only dissipating of only about 3.52W! And if you look at the long piece of red wire connecting to the right, it is actually being connected to the Vcc terminal of the IR2117 gate driver. The whole system works.

The edaboard_oscope_Vz.jpg image shows V(z) in CH1, a 11.86V value very close to the desired 12V.

But what I could not understand is why V(z) has a 4V when I had off the main 100V supply. This 4V only disappears when I switched off the Function gen. By the way, the function gen. is providing a 100kHz, 50% duty cycle, 10Vpk-pk voltage PWM. edaboard_oscope_Vz(off).jpg depicts this.
 

Attachments

  • edaboard_breadboard_shuntreg.jpg
    1.9 MB · Views: 119
  • edaboard_oscope_Vz.jpg
    72 KB · Views: 124
  • edaboard_oscope_Vz(off).jpg
    70.4 KB · Views: 122
Last edited:

I think I would just stick to LTspice when posting a schematic. Everyone has their own style which suits them best, and most probably you've already mastered your way for some time. :-D
Of course we should all use whatever we're comfortable with. I use LTSpice too, but mainly for simulations as I find the schematic art rather unattractive.

There's no big secret behind my knowing what you used to capture the image. Irfanview comes with an EXIF viewer that displays all the metadata in an image.

Here is some updates from my hardware progression.

Attached is the breadboard connection of a shunt linear regulator. The power resistor has a 10W rating and it is rather hot to touch after some time although only dissipating of only about 3.52W!
It's normal for power resistors to get quite hot when they dissipate appreciable power even if it's well below their max rating. This is why they usually have some form of ceramic casing and sometimes, as in your case, a metal casing on top of the ceramic one. I recall the specs of some stating that they reach something like 300°C or more at full power.

And if you look at the long piece of red wire connecting to the right, it is actually being connected to the Vcc terminal of the IR2117 gate driver. The whole system works.
Nice to know that. It's a good idea to put another filter capacitor close to the gate driver IC. A ceramic or plastic cap of 0.1µF should do the job.

That's probably because, when the 12V supply is off, the 100kHz signal passes through the IR2117 IC to its Vcc pin and the pulses are integrated by the 10µF filter capacitor which is in parallel with the zener diode. Basically, integrating 10V pulses at 50% duty should result in 5V DC, but there must be some losses in the IR2117's internal circuitry as the Vcc pin wouldn't be directly connected to the input pin. And then there's the finite output resistance of the function generator which may be 50 or 75 ohms.
 
Gosh, I am a little upset as I think I've damaged my 2N5682 10W BJT. URL: MULTICOMP|2N5682|TRANSISTOR | element14 Singapore

Everything went fine when I interfaced an about 600ohms as dummy load and the collector current was measured to be 60mA. However the transistor was so hot that it burned and melted an insulator of a neighboring wire (see attached image).

While believing this is normal, I've proceeded to connect the IR2117 driver as my load and has observed a lower collector current (~40mA) but the transistor is still being very very hot.

And out of a sudden, the collector current drops to 0.4mA range and the gate is no longer powered by my series reg. That was the indication when I think my transistor is spoilt.

Upon double checking the datasheet of my 2 dollars BJT, I did see another power rating of 1W, a 10x reduction of power ratings as specified.

Could any kind souls tell me:

1) If the BJT can handle 88V across CE, I bet it could.
2) If the power rating is 10W as what I've interpreted? If yes, what is the 1W rating about?
3) If the answers are yes to the above 2 questions, why did my BJT failed only after about 1 minute of operation?

Thanks, please.
 

Attachments

  • edaboard_wirebunt.jpg
    213.7 KB · Views: 121
  • edaboard_seriesreg.jpg
    152.1 KB · Views: 107

With 17.5C/W junction to case it can only handle 10W with an infinite heat sink - the junction would be at 200C.

Without a heatsink it is 175C/W so can only handle 1W, but would have a junction temperature of 200C at an ambient of 25C.

Keith
 
With 17.5C/W junction to case it can only handle 10W with an infinite heat sink - the junction would be at 200C.

Without a heatsink it is 175C/W so can only handle 1W, but would have a junction temperature of 200C at an ambient of 25C.

Keith

Thanks Keith! But how do we heat sink from this type of package? It would be wiser to choose another BJT with better power dissipation capability w/o any external heat sink?
 

I think you should be able to find clip on heatsinks for that package, but you will need to do some calculations to see if it will be suitable. In general I find that even 1W can require heatsinking. A TO220 package has around 60C/W so at 70C ambient would be at 130C without a heatsink when dissipating 1W. For more than 1W you would need a heatsink. For reliability I wouldn't run transistors at 200C. The package you are using has a very high C/W from junction to case so even an infinite heatsink will struggle to keep it cool.

I would suggest looking at transistors in TO220- they are better designed for adding heatsinks.

Keith
 


Thank you for allowing me to understand the difference between power dissipation to case and to ambient.

However, could I know for datasheets w/o specifying the junction-to-ambient temperature coefficients (ie: **broken link removed**).

Is it wise to estimate the Thermal Resistance, Junction To Ambient by looking at other products with the same packaging but with stated Junction-to-ambient rating. Just like this: https://www.onsemi.com/pub_link/Collateral/NJW3281-D.PDF

Great thanks.

---------- Post added at 13:02 ---------- Previous post was at 12:55 ----------


And if I adopt Pjdd suggestion, could I use back the TO-39 package because it looks like the transistor is going to operate for a very very short while before the operation is being taken over by the another linear reg - +12V from main SMPS (but with a lower power rating since Vce has been reduced to only 16V!)?
 
Last edited:

Yes, it is fine to use data from another manufacturer for junction-ambient thermal resistance - the percentage error doing this will be small. This is not true for junction-case - you could be out by a factor of 2 or more from one manufacturer to another.

Keith
 

Pjdd!

Got a question. I've connected all the my buck from breadboard and transferred onto stripboard. Ironically, the connection (with all the various parasitic COULD work but once I've transferred onto stripboard, it doesnt work!

Did a check on the base of the NPN 2N3439 and realised the voltage can't climd to the desired 12V++ as you can see from the waveform. That waveform is when I've set V+ to 30V, and theoritcally its voltage should be sufficient to turn on the series linear reg. since 30V > 12V, but my experiment shows otherwise.

Do you know why? Please help me.

My main thread with schematic, layout drawings and other useful images or information: https://www.edaboard.com/threads/233047/
 

Attachments

  • VB_30.jpg
    109.6 KB · Views: 75
Last edited:

Status
Not open for further replies.

Similar threads

Cookies are required to use this site. You must accept them to continue using the site. Learn more…