How fast can a MOSFET switch?
Bill yg
I would like to know the highest speed that can a MOSFET switch, from
the data in the datasheet.
In the datasheet, I have a few parameters, namely the turn-on delay time
(td_on), rise time (tr), turn-off delay time (td_off), fall time (tf) and
reverse recovery time (trr). Which one of them dominates the limit of the
switching speed of a MOSFET? Or, should I add them all up?
--
Thanks.
--
Bill y.g.
Winfield Hill
Add them up if you want, but the correct answer is much faster
than the numbers you'll get - for high-voltage FETs it's between
5 to 10ns for those of us who are ordinary mortal experts, but
rumored to be ~ 2ns for a few extraordinary folks. The subject
has been discussed carefully and in detail in these s.e.d. pages
during the last few months or years; a google search will turn up
entire threads, with much more than we can quickly say here now.
Try this URL, http://groups.google.com/advanced_group_search
Thanks,
- Win
Winfield Hill
Rowland Institute for Science
100 Edwin Land Blvd
Cambridge, MA 02142
John Larkin
wrote:
Win,
the guys who make intensified CCD cameras use mosfet shutter drivers
that switch voltages around 300 with shutter times down to about 2 ns.
But they won't tell me how they do it.
Here's one:
click on ICCDs. (My little digital delay generator board is just under
the pagoda top of this camera.)
John
Winfield Hill
>
> the guys who make intensified CCD cameras use mosfet shutter drivers
> that switch voltages around 300 with shutter times down to about 2 ns.
> But they won't tell me how they do it.
>
> Here's one:
>
> http://www.andor-tech.com/
>
> click on ICCDs. (My little digital delay generator board is just
> under the pagoda top of this camera.)
Wow! I'm impressed, but I'll bet they're very expensive, right?
Actually I'm looking for an affordable CCD camera with 1us gate
time. Not necessarily with an MCP - I've got *lots* of light.
Bill yg
"Winfield Hill" <hi...@rowland.org> 撰寫於郵件
news:3B913C85...@rowland.org...
John Larkin
wrote:
>John Larkin wrote:
Win,
I don't know pricing, but I'll try to find out.
What are you photographing?
John
Winfield Hill
>
> sorry.. so the answer for my question is??
It appears that there's agreement, the correct answer to
your question, "How fast can one switch an FET?" is 2ns.
If your question was, How fast can *I* switch an FET?
or, How do I learn how to do this? I answered that too:
use Google to read from numerous completed s.e.d. threads.
You can add the names Win Hill and Winfield Hill to your
search if you want to read some of what I've written on
the subject, and the responses to it. :-)
Also read these IR app notes: AN-937, -944 and -957
http://www.irf.com/technical-info/an937/an-937.htm
http://www.irf.com/technical-info/an944/an-944.htm
http://www.irf.com/technical-info/an957/an-957.htm
Boris Mohar
>Bill yg wrote:
>>
>> sorry.. so the answer for my question is??
>
> It appears that there's agreement, the correct answer to
> your question, "How fast can one switch an FET?" is 2ns.
The speed killer is the package lead inductance, primarily the source lead
inductance created by the bonding wire from source bond pad to the source
lead. This inductance of few nano henries develops a substantial voltage
drop caused by driving the gate very hard which opposes the gate drive.
Regards,
Boris Mohar
Viatrack Printed Circuit Designs
Genome
"Bill yg" <ygdr...@yahoo.com> wrote in message
news:9mr8ni$s5q$1...@news.ctimail.com...
> Dear All,
> I would like to know the highest speed that can a MOSFET switch,
from
> the data in the datasheet
> Bill y.g.
It FUCKING says so in the FUCKING data sheet
Hmmm.... I see you haven't bought my book yet.
Why not use a real e-mail addy on your own account...... or are you just
asking boring twat questions...... next thing you will make statements like
'Thanks for constructive comments'...
YAWN
If you have the data sheet, why not read it a bit. You might like to look at
the column labelled 'Test Conditions' and then look at the circuit it refers
to.......
Then you might like to look at the 'Typical Capacitance Vs Drain to Source
Voltage' and the 'Typical Gate Charge Vs Gate to Source Voltage.... (and the
test circuits) With a bit of head scratching you'll be able to figure out
why the various times have specific values for the test circuit shown......
or at least how they relate to each other........ go on try it.
Well Done
Now you know that the highest speed a Mosfet can switch depends on the
circuit you are using it in.......... you might also realise how to
improve those times.
DNA
John Larkin
wrote:
Boris,
these puppies
http://www.directedenergy.com/Products/specs/deseries.htm
have very low inductive parasitics, but still switch in 2 ns.
(and they're *expensive*)
There's something about 2 ns!
John
Boris Mohar
<jjla...@highlandSNIPTHIStechnology.com> wrote:
>On Sat, 01 Sep 2001 20:28:59 -0400, Boris Mohar <bor...@sympatico.ca>
>wrote:
>
>>On Sat, 01 Sep 2001 17:15:02 -0400, Winfield Hill <hi...@rowland.org> wrote:
>>
>>>Bill yg wrote:
>>>>
>>>> sorry.. so the answer for my question is??
>>>
>>> It appears that there's agreement, the correct answer to
>>> your question, "How fast can one switch an FET?" is 2ns.
>>
>> The speed killer is the package lead inductance, primarily the source lead
>>inductance created by the bonding wire from source bond pad to the source
>>lead. This inductance of few nano henries develops a substantial voltage
>>drop caused by driving the gate very hard which opposes the gate drive.
>>
>>
>>
>>Regards,
>>
>> Boris Mohar
>>
>> Viatrack Printed Circuit Designs
>
>Boris,
>
>these puppies
>
>http://www.directedenergy.com/Products/specs/deseries.htm
Nice.
>
>have very low inductive parasitics, but still switch in 2 ns.
>(and they're *expensive*)
>
>There's something about 2 ns!
Drain-Source capacitance x Rds (on)?
Machone
Usually, the application has alot to do with the switching speed, if you
mean drain-source rise and fall times. By the way, how do you define
switching in a power application? The drain rise/fall is usually controlled
by the exterior circuit.
Machone
Bill yg <ygdr...@yahoo.com> wrote in message
news:9mr8ni$s5q$1...@news.ctimail.com...
William L. Bahn
1ns. On a recent process I used (0.5um AMI?) the propogation delay for an
inverter was 400ps and I know that the 0.25um and 0.18um process we are
using now beat that by quite a bit - but since the chips I am presently
working on do not need logic anywhere near that fast I haven't bothered to
look it up.
John Larkin wrote in message ...
Winfield Hill
>
> Winfield Hill <hi...@rowland.org> wrote:
>
>>John Larkin wrote:
>>>
>>> the guys who make intensified CCD cameras use mosfet shutter
>>> drivers that switch voltages around 300 with shutter times
>>> down to about 2 ns. But they won't tell me how they do it.
>>>
>>> Here's one: http://www.andor-tech.com/
>>> click on ICCDs. (My little digital delay generator board is
>>> just under the pagoda top of this camera.)
>>
>> Wow! I'm impressed, but I'll bet they're very expensive, right?
>> Actually I'm looking for an affordable CCD camera with 1us gate
>> time. Not necessarily with an MCP - I've got *lots* of light.
>
>
> I don't know pricing, but I'll try to find out.
> What are you photographing?
An exploding wire.
Robert
John Larkin wrote:
> Win,
>
> the guys who make intensified CCD cameras use mosfet shutter drivers
> that switch voltages around 300 with shutter times down to about 2 ns.
> But they won't tell me how they do it.
>
> Here's one:
>
> http://www.andor-tech.com/
>
> click on ICCDs. (My little digital delay generator board is just under
> the pagoda top of this camera.)
>
> John
I would start my search by looking at the ultra-fast MOSFET laser diode
drivers that switch in 40A for durations as short as 1ns. The pulse
width is programmed by a high voltage bias somehow. This should lead you
to high voltage gating by some sort of duality.
John Larkin
Robert,
do you have any links to these? All I know of are things like the DEI
and Avtech stuff, which are considerably slower. Kentech has fast
high-current pulsers - in the extreme - but they're mostly avalanche
stuff and can't fire at above a KHz or so.
I got 1KV into 50 ohms, 3.5 ns FWHM pulse, 500 KHz, using drift
step-recovery diodes, but that's not a very flexible beast.
John
Robert
John Larkin wrote:
> Robert,
>
> do you have any links to these? All I know of are things like the DEI
> and Avtech stuff, which are considerably slower. Kentech has fast
> high-current pulsers - in the extreme - but they're mostly avalanche
> stuff and can't fire at above a KHz or so.
>
> I got 1KV into 50 ohms, 3.5 ns FWHM pulse, 500 KHz, using drift
> step-recovery diodes, but that's not a very flexible beast.
>
> John
Their technique can be discovered. The MOSFET may be only peripherally
related. I have found at least one report where the high voltage
avalanche diode transmission line pulse generation has been duplicated
using a form of ferrite loading to "accelerate" the phase velocity. It
may be something along these lines. I will look.
Bob Wilson
>
>Bill yg wrote:
>>
>> sorry.. so the answer for my question is??
>
> It appears that there's agreement, the correct answer to
> your question, "How fast can one switch an FET?" is 2ns.
>
> If your question was, How fast can *I* switch an FET?
> or, How do I learn how to do this? I answered that too:
> use Google to read from numerous completed s.e.d. threads.
> You can add the names Win Hill and Winfield Hill to your
> search if you want to read some of what I've written on
> the subject, and the responses to it. :-)
>
> Also read these IR app notes: AN-937, -944 and -957
> http://www.irf.com/technical-info/an937/an-937.htm
> http://www.irf.com/technical-info/an944/an-944.htm
> http://www.irf.com/technical-info/an957/an-957.htm
If this question relates to switchmode power supplies, then the REAL answer
is going to be different than just "whack the FET's gate as hard as possible
with the highest drive current you can, to get the fastest switching speed".
The reality is that a switchmode power supply designer has to carefully
balance increasing the switching speed (to get maximum efficiency) with the
problem of increasing EMI problems caused by sharper and sharper switching
edges. While there are many things one can do to attenuate EMI, quite often
it becomes necessary to limit switching speed to reduce the problem at its
source. A truly optimal design is one that dies not wring out the last 1 or
2 percent of efficiency, and thus eliminates the need for excessively
costly and clumsy EMI filtering and shielding.
Bob.
John Larkin
wrote:
Win,
low end Andor ICCD cameras run about $24K. The fancy one, with PCI
card, camera, my cute little DDG inside, and software is about $32K.
Plus you need a lens, which is no big deal.
I think somebody may make cheaper, low-res ICCDs.
How about gating a dirt cheap Russian night-vision viewer? Does
anybody know how they work? I think they might *not* use a
microchannel.
John
Win Hill
>
> Winfield Hill <hi...@rowland.org> wrote:
>
>> John Larkin wrote:
>>>
>>>>> the guys who make intensified CCD cameras use mosfet shutter
>>>>> drivers that switch voltages around 300 with shutter times
>>>>> down to about 2 ns. But they won't tell me how they do it.
>>>>>
>>>>> Here's one: http://www.andor-tech.com/
>>>>> click on ICCDs. (My little digital delay generator board is
>>>>> just under the pagoda top of this camera.)
>
> card, camera, my cute little DDG inside, and software is about $32K.
> Plus you need a lens, which is no big deal.
>
> I think somebody may make cheaper, low-res ICCDs.
Thanks. I'm not looking for low res, but 1us "low" speed...
> How about gating a dirt cheap Russian night-vision viewer?
> Does anybody know how they work? I think they might *not* use
> a microchannel.
I have one of those. Maybe that's exactly the answer...
Thanks,
- Win
Winfield Hill
Rowland Institute for Science
100 Edwin Land Blvd
Cambridge, MA 02142-1297
Roy McCammon
> An exploding wire.
You're having too much fun.
And you get paid too?
Robert
John Larkin wrote:
> Robert,
>
> do you have any links to these? All I know of are things like the DEI
> and Avtech stuff, which are considerably slower. Kentech has fast
> high-current pulsers - in the extreme - but they're mostly avalanche
> stuff and can't fire at above a KHz or so.
>
> I got 1KV into 50 ohms, 3.5 ns FWHM pulse, 500 KHz, using drift
> step-recovery diodes, but that's not a very flexible beast.
>
> John
The superfast MOSFET high voltage switches are most likely those of
Behlke Electronics GmbH. Many of these are fixed pulse width outputs-
several hundred ns, some are variable, and some are ultra-short like
5ns. These can be burst PRF to 1MHz but ~5KHz is the sustained maximum
depending on load. The HTS 30-08-UF is available with an ultrashort 5ns
option, can handle 3KV and will achieve <1ns rise times into light
loading like 1K ohm. You can check the specs at
http://eurotek.com:800/hts30-uf.htm . The Andor ICCD most likely uses
something like this. This is not a homebrew project- these things must
be designed by MMIC engineers and characterized to 20GHz. The 100ps max
trigger jitter and the fixed pulse widths suggest a strong positive
feedback design. The entire line can be viewed at http://eurotek.com:800
.
Win Hill
I have used these modules and they are amazingly rugged. Switching say
3kV in 5ns, with currents to 30A, one is sure to have various overshoots,
etc., and all kinds MOSFET mistreatment within the module seems likely.
But they keep right on working! It's not clear to me what kind of FETs
they are using inside the potted module.
Thanks,
- Win
Winfield Hill
Rowland Institute for Science
100 Edwin Land Blvd
Cambridge, MA 02142-1297
Robert
I think what is even more amazing is that it is a stack of FETs! They
are definitely relying on the packaging and probably extreme matching of
the FETs- as well as switching speeds too fast to cause damage.
Another lead along this extreme performance line is a UK firm Bournlea
Instruments and in particular Harry Kitchin at http://www.bournlea.com/
. The information is pushing four years old, but this firm has developed
very fast high voltage MOSFET pulse generators that switch 10KV in <2ns
at ~40KHz PRF to drive water cooled ferrite loaded
NonLinearTransmissionLine pulse sharpeners [ that they also developed ]
in support of UltraWidebandRadar antenna drives requiring switch times
on the order of 50ps.
John Larkin
Wow. But I would hate to be the guy committing to a 12-month warranty
for a potted 8KV mosfet-based switch.
John
Bob Wilson
Send me one and I'll un-pot it.
Bob.