The limit on speed will always be the energy available to charge and discharge the gate. To go fast, it makes more sense to store charge in a DC cap referenced to the get source, and use an active pullup/pulldown switch to drive the gate from that. Passive pulldown has to be a low-value resistor to get decent fall times, and that fights the pullup drive. All of which takes continuous power, when the gate only really needs a zot of charge now and then.
A power mosfet will need a lot of gate drive, in both directions, to switch in 100 ns.
True, but: In most power applications (think switch gear closet) you don't need to switch in 100nsec. But you do need tens of usec. A SSR that yawns for a millisecond inside the linear range is fairly useless to me because it won't be fault-tolerant. If a control signal rattles for a while and it switches back and forth you'll hear a loud bang and a splat.
It would (usually) be no problem to pipe the control voltage also to the cabinet. But SSRs don't have a 5th pin for that. Beats me why not.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
On a sunny day (Sun, 30 Aug 2009 19:19:47 -0700) it happened John Larkin wrote in :
Well, any sort of generic remark like that is of course wrong in some circumstances. Transformers are from the era when the US could still have men walk on the moon. Not all super high level integration or not all silicon world can replace them.
100 uS.
Have you done the math? I did, and it is not really that much power (in case of constant long 'on' times). Take gate capacitance, frequency, rise and fall times, gives you R. Better (much) for logic level MOSFETS as U^2 / R.
If you were doing fast PWM, and the range of the PWM is limited (that is not
100% on or off), then the transformer is even more interesting as it can drive the MOSFET directly, or perhaps via a DC restorer circuit.
Perhaps less power then the whole of your secundary power suply plus its [in]efficiency.
Do the math. Q= C.U = i.t For a 10nF gate capacitance (normal), and 5 V gate (Logic level MOSFET) and 100 nS, R.C = 100 ns, R = 10^-7 / 10^-8 = 10 Ohm.
5 V in 10 Ohm = 2.5 W. That is about what your driver will have to do if constant on.
If it was a PV, tend of milliwatts. A small cheap commercial dc/dc converter, 200 mW maybe. If Joerg can steal a little of the load supply, r-diode-cap-zener maybe, he wouldn't need many milliwatts to charge a cap enough to switch a medium-size fet at 100 Hz.
On a sunny day (Mon, 31 Aug 2009 15:05:43 -0700) it happened John Larkin wrote in :
100 nS,
Yes, but Joerg had a problem with milli second switching. If he is happy with 1000x faster, say 1 uS, then R is 100 Ohm, and P 1 W for a 10 V gate drive MOSFET, 250 mW for 5 V gate drive.
Anyways you cheat if you use a DC/DC converter, those also have a transformer in it, something you claimed should be avoided.
Using a C reactance so as to not lose power to make a secondary voltage, could work if the application allows. The system I propose will not need anything special supply related on the MOSFET side.
Too complicated. Saturation. DCR. Various capacitances. Leakage inductance. Core loss. Cost. Size. Lack of second sources.
We buy a bunch of custom transformers, and the specification/purchasing/prototype-testing procedure is a real pain. I go for a "silicon solution" whenever possible.
Here's a transformer being loaded by one of our pick-and-place machines:
Cost? LAN transformers are dirt cheap and you can get a four-pack for a song. Even with common mode chokes in there. The rest, well, that's what engineers are for :-)
:-)
Hey, you still use a dot matrix printer? I thought they were almost extinct.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
We use ISDN transformers as power isolators in our isolated-channel ADC and DAC boards. But they don't move a lot of power, as you'd need to drive a mosfet in a simple gate rectifier situation. We recently had a custom transformer developed for us by Minntronics to replace the ISDNs in several applications. The LAN transformers are toroids with the primary and secondary all interwound, and have about 60 pF of pri-sec capacitance, enough to inject a heap of inverter noise into our floating channel commons. The new parts are EI cores with separated windings, more like 10 pF. Interestingly, a lot of tape between the windings works better than a shield.
$4 a pop now.
See what I mean?
Almost. I really want an 11x14 fanfold laser printer, but that's not in this year's budget. These old Panasonics seem to run forever. I have one at home, too, in the sump-pump room.
The best way to debug software is to print it and *read* it before it's ever run. You'll never find all the bugs by testing. I go through about a box of paper and two ribbons per embedded product design.
although it doesn't use nearly so many transistors.
formatting link
Too bad it's been blowing transistors. Something about temp compensation. Once it passes about 40C it's downhill, even with a small fan. Weird thing is, it's got a Vbe multiplier in there, it should be perfect. Do 2N3904s have less tempco than darlingtons or something?
Tim
-- Deep Friar: a very philosophical monk. Website:
The darls will be 4 Vbe's, not counting any other drivers. Maybe use bigger emitter resistors?
I just hate blowing power fets, especially 32 at a time. I have found fets to be more rugged than bipolars, and easier to bias.
My box has a fast ADC that digitizes everything 2000 times a second and runs a realtime software simulation of junction temperatures, and shuts down if they hit 150C. Works, mostly.
Tim Williams wrote: "John Larkin" wrote in message news: snipped-for-privacy@4ax.com... >> Here's a transformer being loaded by one of our pick-and-place >> machines: >>
Well, there's a factor of two from the Darlington having two Vbe drops in series, and also the driver stage runs at a lot lower current than the output stage. If the current densities are very different, the tempco of Vbe will be different too.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
Ok, yeah, at your typical quantities you can't be picky about pricing. At least not for custom. But if you do high volume and have places like Xfmrs in China wind the cores, different thing. The topper was a single-winding toroid for a switcher I designed. Found a COTS version for 60c or so. Client inquired anyhow and lo and behold they lined up a place in Taiwan that made them for 35c. Tested, bagged, shipped, no complaints in years.
Your sump pump has its own room? Maybe it even has a gold-plated discharge pipe :-)
I am amazed you can still get ribbons for them. We find it increasingly hard to buy cassette tapes for our church.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Quite typical. (Note: supply is actually +/-25V, that was a typo.)
Now, the darlingtons are 4Vbe + Re's (at 0.1 ohm, just a few tens of milivolts for a regular class AB bias). And the Vbe multiplier is the same. With about 50mA bias in the volt amp and current sink, the 3904 should be running fairly heavy I'd think. 'Course, as the outputs sink more current, they draw a few miliamps, reducing the Vbe mult's current density slightly, while dramatically increasing their own.
Is it just the case that I'm biasing this thing too low, so it will always have a positive current tempco? Which in turn means I actually need to bias it at a few amperes quiescent, which will need a much larger heatsink.
Hell, I might as well make a floating bias servo to hang on the Re's.
Tim
-- Deep Friar: a very philosophical monk. Website:
On a sunny day (Wed, 2 Sep 2009 01:49:51 -0700 (PDT)) it happened Tim Williams wrote in :
I am not sure the constant current sources that have the 1N914 and
220 Ohm resistor as voltage reference will not drift, and increase driver current, so also in the Vbe multiplier, with temperature. That way the current also depends on the supply voltage. A zener or LED or some other reference would perhaps be better here.
Why the "bias" pot? Its wide adjustment range is scary. Why not decide what you want and set it with a couple fixed resistors? Fixed resistors can't be misadjusted; pots can.
The 2n3904 should be thermally tied directly to the output transistors. Is it?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.