Film resistor vs Carbon resistor.

[jpanhalt]

spec
I agree with your circuit, except for the source resistors. I don't feel they are necessary. 4QD and others have discussed how the positive tempco for RDS(on) of mosfets effectively provides balancing. Consider the IRF1010E as just one example. RDS(on) increases 2.5 fold over the temperature range of 20° to 175° at the junction. In addition, assuming the design of 30A per mosfet (another thread), power dissipation with high duty-cycle from each resistor would approach 56 W. I agree that lower value source resistors can and possibly should be used for shut down in the event of too high a current demand.

John

 

[gary350]

What happened to the list of companies on this forum that sell parts I can not find it to check prices and order parts?

 

[chemelec]

JUST GO HERE and type in your Part Number:

https://findchips.com

 

[jpanhalt]

What happened to the list of companies on this forum that sell parts I can not find it to check prices and order parts?

Parts for which design?
John

 

[gary350]

Parts for which design?
John

I need all the resistors for the PWM circuit. I have coffee cans of parts but I don't want to search for parts. I already searched for resistors and I don't have most of these values, to save time and make it easy I will buy packs of 25 or 50. Ebay has some good prices $1 packs free postage.

.1 ohm resistors, 1/2 watt . I need some parts to experiment with before I build anything big. 2 or 3 mosfets in parallel will be a good start.

I don't have anymore 555 timers I might just buy that ready made PWM circuit board.

I am going to play with both circuits a little just for FUN but I will build the PWM circuit.

It will be a month before parts all arrive.

 

[spec]

spec
I agree with your circuit, except for the source resistors. I don't feel they are necessary. 4QD and others have discussed how the positive tempco for RDS(on) of mosfets effectively provides balancing. Consider the IRF1010E as just one example. RDS(on) increases 2.5 fold over the temperature range of 20° to 175° at the junction. In addition, assuming the design of 30A per mosfet (another thread), power dissipation with high duty-cycle from each resistor would approach 56 W. I agree that lower value source resistors can and possibly should be used for shut down in the event of too high a current demand.

John

Hi John,

thks for your comments.

There is a misnomer about MOSFETs and current sharing in parallel which has arisen for historical reasons.

The early MOSFETs were lateral construction. which gave a high RDS, typically around 0R8, but it also gave good linearity (gm). They also have a positive temperature coiefficient, which means that they will current share in parrelel without source balancing resistors. Both, the high high RDS and the positive temperature coificient made the horizontal MOSFETs bullet -proof. In fact, there is a standard audio amp design dataing from the 1980s, which you just can not blow up, unlike other 100w amps. For that reason, that amp was/is very popular with musicians. The lateral types, having a more linear gm, are good for hifi audio amplifiers too.

When SWPS became widely used the manufacturers went over to a vertical construction with propritry name like VMOS, TRENCH FET and so on. These MOSFETS are basically switches and are not idal as amplifiers, especially because of their colossal and non linear parasitic capacitances. Due to their verticle construction, the temperature coiefficient is negative. That is why you now see SOA graphs on MOSFET data sheets. The net result is that if you parrallel vertical MOSFETs, which they all are these days, you must use some form of current equalisation, source ballast resistors being just one.

There are heaps of designs knocking around that don't comply with the rules. There are also heaps of dead VMOSFETS and also frustrated customers. Sorry to say this, but the circuit that gary posted made me cringe for two reasons; one was the absence of source resistors, but then there is no MOSFET type stated and neither is 'extreme current' quantified. The other is the gate drives, especially the negative excusion, which will not quickly empty the gates of the many parallel MOSFETS quick enough.

MOSFETs have feelings too- just like you and me

chuck

(not my post, by the way)

​

​

 

[chemelec]

As a Point of interest in the PWM Circuit I Posted, You will Notice that ALL Source Leads Go to a COMMON POINT GROUND.
THESE WIRES should ALL be the Same Size and Length.
Each one of these could also have a 0.1 Ohm resistor in it to help Offset Minor Differences in the Mosfets.

GARY350, Sometimes these CHEAP PARTS are NOT such a good deal.
You may just get what you pay for.
Cheap Price, Cheap Junk.

The 0.1 Ohm Resistors Should be a 2 or 5 WATT, Wire Wound Type.

 

[spec]

As a Point of interest in the PWM Circuit I Posted, You will Notice that ALL Source Leads Go to a COMMON POINT GROUND.
THESE WIRES should ALL be the Same Size and Length.
Each one of these could also have a 0.1 Ohm resistor in it to help Offset and Differences in the Mosfets.

Hi, chem,
A very good point which I had forgotton about; your approach is even better than you say: Because the temperature coificicent of copper is positive, it improves on the current sharing more than a pure resistor would.

GARY350, Sometimes these CHEAP PARTS are NOT such a good deal.
You may just get what you pay for.
Cheap Price, Cheap Junk.

Very true. Some components you can buy cheap, but for critical items always go to a reliable source. Resistors are so cheap anyway, that it is not worth going for 'bargains''.

https://www.cirris.com/learning-cen...-topics/177-temperature-coefficient-of-copper

 

[jpanhalt]

When SWPS became widely used the manufacturers went over to a vertical construction with propritry name like VMOS, TRENCH FET and so on. These MOSFETS are basically switches and are not good as amplifiers, especially because of their colossal and non linear parasitic capacitances. Because of the verticle construction the temperature coiefficient is negative. That is why you now see SOA graphs on MOSFET data sheets. The net result is that if you parrallel vertical MOSFETs, which they all are these days ,you must use ssome form of current equalisation, source ballast resistors being just one.

​

​

Hi Chuck,

Thank you for the additional information. It is still unclear to me why a vertical construction would give a negative tempco. Of course, if it is negative, I have no argument with your design whatsoever.

None of the mosfets I have mentioned were intended for anything but switching. The IRF1010E has RDS(on) = 0R012, and as mentioned, the resistance doubles over a range of 25° to 150°. I looked up a couple of Fairchild "Power Trench" mosfets. They also had a positive tempco (e.g, FDBL0260N100 -- chosen only because it was second in the list). In fact, its normalized RDS(on) vs. temp was about the same as the IRF1010E. Same went for graph in an app note from IXYS of its V-DMOS (loc. cit.):

I relied a lot on IR's and 4QD's application notes when designing those motor controls. I didn't ignore Fairchild or the others, IR's devices were cheaper at my supplier. Can you point me to a good discussion of vertical power mosfets with a negative tempco?

Regards, John

 

[spec]

Hi John,

I'm back on this thread now.

I haven't found the references I am after yet, but here is a starter which mentions the negative temp co of MOSFETS.

http://www.onsemi.com/pub_link/Collateral/AND8199-D.PDF

This quote supports what you say about MOSFET positive temp co, but there is a caveat:

​

http://www.microsemi.com/document-portal/doc_view/14692-mosfet-tutorial

I havent read through these references on MOSFETS yet. They are just a random collection:
**broken link removed**

**broken link removed**

http://coefs.uncc.edu/dlsharer/files/2012/04/J8a.pdf

http://www.ixys.com/Documents/AppNotes/IXAN0061.pdf

http://www.fairchildsemi.com/application-notes/AN/AN-9010.pdf

https://www.electro-tech-online.com...vs-carbon-resistor.146607/page-3#post-1242706

http://documentation.renesas.com/doc/products/transistor/apn/rej05g0001_pmf.pdf

http://epc-co.com/epc/Portals/0/epc/documents/articles/bp_2012_09_SafeOperatingArea.pdf

http://www.exicon.info/why-use.php

**broken link removed**

http://educypedia.karadimov.info/library/BDE0033-03_catalog.pdf

I'm trying to decode this statement:

Parallelisation of MOSFETS
MOSFETS have a negative temperature coefficient on the drain-source contact which means that the body resistance increases with current. As a result when MOSFETs are paralled to increase current capability, current sharing occurs naturally as current rise results in a rise of the channel temperature and consequently the excess current is redirected to parallel chips.

**broken link removed**

 

[spec]

Gary,

These book looks right up your street:

(1)) https://books.google.co.uk/books?id=vDQHzeEmSfUC&pg=PA63&lpg=PA63&dq=mosfet+negative+temperature+coefficient&source=bl&ots=SMjNWbUb77&sig=eDITz4_Z9cYrakgFC4nUwv7ya20&hl=en&sa=X&ved=0ahUKEwjPgumPxeLJAhVJthoKHTPvA4o4HhDoAQgvMAA#v=onepage&q=mosfet negative temperature coefficient&f=false

**broken link removed**

 

[kubeek]

I just looked into a random datasheet from IPB107N20N3G and the curve of Rdson vs temperature is definitely rising, so it shoud be able to be paralleled.

 

[spec]

Hello again John,

Paralleling MOSFETs, MOSFET Temperature Coificient and Switching

After reading many of the links in post #50, I am confused, but it seems reasonable to draw the following conclusions:

(1) You cannot parallel most MOSFETS without some kind of current balancing mechanism, usually source resistors. Some lateral MOSFETS can be paralelled but, by definition, their high Rds, typically an order higher than vertical types, prevents them from handling high currents.

(2) In the main, MOSFETs, as you say, have a positive temperature coificient, but on the other hand this cannot be used to ensure safe current sharing. But, under certain conditions MOSFETS do have a negative temperature coieficient.

(3) Great care must be used when pysically connecting up high power MOSFETs, because at high currents the resistance of wire and solder joints can cause inbalance in a current sharing circuit.

(4) When MOSFETS are switching high currents, especially into reactive loads, it is vital to drive the gate with a fast edges from a source with plenty of current to charge up and down the parasitic and virtual capacitances at their gates. From this, it follows that when turning off an N type MOSFET, for example, it is not good enough to just take the gate to OV; it has to be taken negative.

Thanks you John; I have learned quite a bit about what is going on inside a MOSFET due to you.

Chuck

 

[jpanhalt]

Hi John,

I'm back on this thread now.

I haven't found the references I am after yet, but here is a starter which mentions the negative temp co of MOSFETS.

https://www.onsemi.com/pub_link/Collateral/AND8199-D.PDF

This quote supports what you say about MOSFET positive temp co, but there is a caveat:

View attachment 96159

​

https://www.microsemi.com/document-portal/doc_view/14692-mosfet-tutorial

That is one of the application notes I used too. The rest of the caveat goes on to say you need good thermal design too. No one will disagree with that either, I hope. Just consider for a moment that the actual construction of a power mosfet has many tiny junctions on the die that operate in parallel. They are thermally connected. None of those junctions has a load balancing resistor attached. The positive temperature coefficient derives from the physics involved, not how it is packaged..

As for the comment by kpap, I am not sure what expertise the Kilimanjaro Porters Assistance Project has with mosfets (**broken link removed**) . Maybe that comment is a translation error or typo.

As for the TS, his last post seemed to indicate that he was proceeding with his original design (variable resistor on the gate) and just scaling up to 20 or more mosfets in parallel. Hence he was headed out to buy some "parts." Hence, I have lost interest in continuing discussion of a design that will be ignored.

John

 

[jpanhalt]

@post #53
I maintain that in most cases, you do not need balancing resistors when the design incorporates the principles mentioned. It simply doesn't make sense to me to use modern mosfets with an RDS(on) in the range of a few milliohms to low teens milliohms and then add a whopping 62 milliohm source resistor. Not only does that resistor make a small space heater from a motor drive, by effectively reducing Vgs, it changes the mosfet from being fully turned on to being somewhat less "on," increases RDS(on), and increases heat dissipation in the mosfet. Of course, that is the mechanism by which such resistors work, but they are not necessary with good design.

John

 

[spec]

That is one of the application notes I used too. The rest of the caveat goes on to say you need good thermal design too. No one will disagree with that either, I hope. Just consider for a moment that the actual construction of a power mosfet has many tiny junctions on the die that operate in parallel. They are thermally connected. None of those junctions has a load balancing resistor attached. The positive temperature coefficient derives from the physics involved, not how it is packaged..

As for the comment by kpap, I am not sure what expertise the Kilimanjaro Porters Assistance Project has with mosfets (**broken link removed**) . Maybe that comment is a translation error or typo.

As for the TS, his last post seemed to indicate that he was proceeding with his original design (variable resistor on the gate) and just scaling up to 20 or more mosfets in parallel. Hence he was headed out to buy some "parts." Hence, I have lost interest in continuing discussion of a design that will be ignored.

John

last bit

About the many tiny junctions. This is effectively true for all semiconductors, at the xtal structure level. But there is a difference between transistors on different crystals and substrates working in parallel. The common substrate transistors have identical geometries identical doping levels and are also pretty close in temperature. I think there are other reasons why common virtual transistors share well too. This is not the case with physically separate transistors.

There is still a problem with current sharing on a common substrate though. This is evidenced by the safe operating area (SOA) graphs appearing on MOSFET data sheets, whereas at one time they were absent. BJT power trans have always had them. In their case, the many transistors in parallel model, applies especially.

A consequence of all this, is that, as the voltage withstanding capability of a MOSFET, and especially a BJT, goes up, its conductivity goes down. I don't know for sure now, but I seem to remember that the bulk resistance, or similar aspect of the silicon, is increased by the manufacturer to reduce secondary breakdown.

chuck

 

[spec]

I just looked into a random datasheet from IPB107N20N3G and the curve of Rdson vs temperature is definitely rising, so it shoud be able to be paralleled.

Hi kub,

Afraid that a positive temp co does not guarantee good current sharing. One of the main parameters is the different geometries between the two trans. See the quote in post 50. Sure you can put MOSFETSs in parallel and they will probably work OK, but you would never know if they are being abused. Normally, the only sign you would get is gross abuse where, as I have said before, the channel melts and forms a short between the source and drain.

Much of the data in this area is vague at best and contradictory at worst. If you have some data that would clarify the situation, I for one, would be very interested. I thought I had it all pretty wells sussed until we started discussing this area and I read a few references.

 

[spec]

@post #53
I maintain that in most cases, you do not need balancing resistors when the design incorporates the principles mentioned.

What principles do you refer to? Is it the good gate drive?

It simply doesn't make sense to me to use modern MOSFETs with an RDS(on) in the range of a few milliohms to low teens milliohms and then add a whopping 62 milliohm source resistor.

It does seem a waste, but it is universal that you lose something in order to get a reliable and characterised design. I assume that you are referring to the 62 m Ohm source resistors in my circuit. If so, I wouldn't pay to much heed to that. The circuit is only a concept and has not been detailed. In practice, the source resistors may be lower (and I would make the base gate driver more elegant too!). But even the 0R062 Ohm resistors would have no significant impact on the overall performance of gary's application. They would just give the MOSFETs an easier time and ensure that his system is reliable.

Not only does that resistor make a small space heater from a motor drive, by effectively reducing Vgs

Did you mean Vds. If so, yes that is true and undesirable, but I would say it is a penalty worth paying. In that case of the 62m Ohm the loss would be 620m V, which would not be significant in a 12V lead acid battery application where the battery voltage itself would be changing by much more than that. 2.38 W is hardly a space heater. The MOSFETs, motor, and battery would be generating orders more heat

it changes the MOSFET from being fully turned on to being somewhat less "on," increases RDS(on), and increases heat dissipation in the MOSFET.

. Don't like to contradict but that is not significant. You would be over driving the gate with around 12V so the odd 620mv, here or there is not significant.

Of course, that is the mechanism by which such resistors work, but they are not necessary with good design.

back to my initial question.

In summary, I take the opposite view to you. The drain resistors are a very good thing, and an essential part of a good deign, but it all needs to be well executed.

 

[jpanhalt]

I meant Vgs . Remember that when N-channel mosfets source current as in your design, the effective Vgs is reduced by the voltage drop across that load (i.e., R11). Now,look at your circuit:

R28 and R27 for a voltage divider the gate gets 82% of the voltage from your driver (ignoring any transient effect of C20). That driver has a voltage drop of approximately 0.6 volts across the top 2N3704 (Vce = 0.6 volts saturated). So, your top rail is 12V, which gives approximately 9.3 V at the mosfet gate. Gary350 was talking about 30 A or so per mosfet. That is 1.86V across R11, or a net 7.48 Vgs max. For many N-channel mosfets, including the one cited by kubeek above, that much change in Vgs is not insignificant. Also, 30 A across 0.062 Ω = 55.8W, which is wasted heat. It may be insignificant compared to an 11 kW drive, but it is not insignificant in terms of finding a 100 W resistor that is not terribly bulky.

Several posts ago, you asked me to validate what experience I had. You never countered with describing your own experience and have particularly avoided documenting your claim that modern versions of the mosfets under discussion have a negative tempco. Can you do that?

This thread, while "challenging," is not really going to help the TS at this point.

John

 

[KeepItSimpleStupid]

last bit


A consequence of all this, is that, as the voltage withstanding capability of a MOSFET, and especially a BJT, goes up, its conductivity goes down.

chuck

1/R?

https://dictionary.reference.com/browse/conductivity