Matching - time and temperature
Well, I know what IRF recommends for testing HEXFET's -- microseconds of exposure. You can do this with a scope and an adjustable pulse generator. <p> I am wondering what all the fuss is about letting the devices warm up for "X" seconds. I can see why you might want to do this, but if the results for this method correlate closely with a shorter time period, what's the point? Has anyone ever graphed the data to see if the outcomes would be materially different if the periods change? Has anyone subject this to statistical analysis?
Well, I know what IRF recommends for testing HEXFET's -- microseconds of exposure. You can do this with a scope and an adjustable pulse generator. <p> I am wondering what all the fuss is about letting the devices warm up for "X" seconds. I can see why you might want to do this, but if the results for this method correlate closely with a shorter time period, what's the point? Has anyone ever graphed the data to see if the outcomes would be materially different if the periods change? Has anyone subject this to statistical analysis?
Since we do this all day long, I am very familiar with the
situation. The Vgs is temperature dependant, so as the
Mosfet warms up during the test, the value will wander.
The key thing is consistency between samples. They should
all be at the same temperature (room temp is fine) and they
should be tested for a consistent period. In our case we
simply use the time it takes for the meter to give a reading,
and then we go with that.
This is not a real big item; just use some common sense
and your results will be what you want.
situation. The Vgs is temperature dependant, so as the
Mosfet warms up during the test, the value will wander.
The key thing is consistency between samples. They should
all be at the same temperature (room temp is fine) and they
should be tested for a consistent period. In our case we
simply use the time it takes for the meter to give a reading,
and then we go with that.
This is not a real big item; just use some common sense
and your results will be what you want.
This is particularly true if they have the same lot code,
so that they were part of the same wafer.
It's interesting to note that the parts are all machine handled,
and as we test them out of the tubes, we can see the
repetition of characteristics in various areas of the original
wafer, where the picking machine has gone down the rows
and columns.
so that they were part of the same wafer.
It's interesting to note that the parts are all machine handled,
and as we test them out of the tubes, we can see the
repetition of characteristics in various areas of the original
wafer, where the picking machine has gone down the rows
and columns.
Nelson Pass said:
Interesting. What impedance are you referring to? Bias resistor,
source impedanc or reflected load impedance? I wouldn't mind
you elaborating a little bit on this topic.
Another related thing, since you mentioned the preference for
high beta in diff pairs, it seems to me that it would be preferable
to have high beta not only in diff pairs but in amplifiers in general.
I haven't really tried to verify this with theory, but my intuition
tells me that for the same overall gain, it would be better to
use fewer high-beta devices than more low-beta ones, since
we get a simpler topology. It seems reasonable that distorsion
and other errors should be more related to the number of devices
than the total gain. Do you have any opinion on this?
Nelson Pass said:
Nice to have someone that’s done LOTS of matching to ask when I'll finally do my matching
I have bought this fancy tester presented in Elector (well only the PCB, and I have collected the rest of the components myself). This have curve matching, and uses a low duty cycle (1 ms on time and 100 ms off time for each device) to not heat the device even if testing it at 10A (the testing current is continuously adjustable from 0 to 13A!).
But what I'm really interested in knowing if keeping different batches separate when matching?
The situation is that first i bought 100 IRFP044N for myself, and at a later point I agreed on helping a fellow DIY-Audio guy to get him matched sets too. So I ordered 100 more, but the ones I got were from a different batch.
Ok, should I keep the batches separated and just match up mine and keep them, and then match up the other batch for themselves and send them to him?
Or would we be better up if I use all the 200 to get as good matches as possible?
They are all IRF.Anders
In my view if you are not matching the devices at their operating temperature, you may as well not have matched them at all. Devices matched to within 1mV at 25°C may have quite different Vgs at 100°C. For that matter, Vgs vs. Id may vary. Match them at their working Id and Tj, and if you want to get truly identical devices, match them for Vgs variance with Id and Vgs variance with Tj.
Cheers,
jwb
Cheers,
jwb
not to worry, at 100 oC T(j) you won't have a device to worry about, unless you are pulsing the transistor for 20uS.
note that the data sheets from IRF show the derating which has to be taken into account for thermal factors -- if you know anything about statistics -- these factors are expressed to one decimal place -- I am not trying to be cryptic, but the point of matching to millivolts is lost in the noise when the slope is that imprecise.
note that the data sheets from IRF show the derating which has to be taken into account for thermal factors -- if you know anything about statistics -- these factors are expressed to one decimal place -- I am not trying to be cryptic, but the point of matching to millivolts is lost in the noise when the slope is that imprecise.
Testing..........
Just a thought....Anders.....it may be possible to heat a strip of aluminium and get all fets up to temp........of say 45-50deg.....just rest them on........get the temps stable and then test them with your tester.................you could do batches of 20...........use a digital cooking probe for temp........I'm not sure how you would keep the aluminium at a stable temp????
Perhaps you could heat something with alot of thermal mass..........like a dense conc block........that would take a long time to vary in temp.......especially if it was insulated from whatever it was lying on............(suppose you could heat block in oven......to warm up....it would take about 2 hours!)
Seems a common sense way of testing................?????
What do others think????
Just a thought....Anders.....it may be possible to heat a strip of aluminium and get all fets up to temp........of say 45-50deg.....just rest them on........get the temps stable and then test them with your tester.................you could do batches of 20...........use a digital cooking probe for temp........I'm not sure how you would keep the aluminium at a stable temp????
Perhaps you could heat something with alot of thermal mass..........like a dense conc block........that would take a long time to vary in temp.......especially if it was insulated from whatever it was lying on............(suppose you could heat block in oven......to warm up....it would take about 2 hours!)
Seems a common sense way of testing................?????
What do others think????
Temps...............
I picked this idea up off Mr Rollins(digital temp probe)..............seems practical....surely all you need to do is establish a stable temp...within the range you are working the amp..........I would have thought a little temp probe would be ideal.
I can't see that there is any great science in doing this..........hmm
What do others think....
I picked this idea up off Mr Rollins(digital temp probe)..............seems practical....surely all you need to do is establish a stable temp...within the range you are working the amp..........I would have thought a little temp probe would be ideal.
I can't see that there is any great science in doing this..........hmm
What do others think....
Re: Temps...............
it isn't rocket science, but a Radio Shack sensor (I use several of these in my Durst photoprocessor) is covered with a plastic sheath so it won't respond very quickly -- they are pretty good in telling me the temperature of my photo-chemicals, but are insulated.
D3 said:
it isn't rocket science, but a Radio Shack sensor (I use several of these in my Durst photoprocessor) is covered with a plastic sheath so it won't respond very quickly -- they are pretty good in telling me the temperature of my photo-chemicals, but are insulated.
temps....
Well yeah....I am not fimiliar with radio shack item..........it is possible to get quite cheap sensors that measure.......body temp....I believe they measure temp of ear drum...........might work on this project..................
Otherwise ......... a reasonable cooking thermal probe ...(digital) should suffice.....................
Well yeah....I am not fimiliar with radio shack item..........it is possible to get quite cheap sensors that measure.......body temp....I believe they measure temp of ear drum...........might work on this project..................
Otherwise ......... a reasonable cooking thermal probe ...(digital) should suffice.....................
Hmmmm
What about my question??? Anyone with a comment on whether to keep the two batches separate or mix and match in both?
About the temp in measuring.... of course I have thought about the temp at witch I'll do the measuring. The thing is that I have absolute control over WHAT it is because of the low dissipation. ALL parts of the transistor have the same temperature.
The idea is that I'm not concerned with the temperature gradients in the devices at the normal power levels we use. With a heatsink temperature of about 50-60 deg C the chip temperature may very well be about 100 deg! With 30 W dissipated and an internal temperature resistance in the device on 1.3 we have a temperature rise above heatsink on 30 x 1.3 giving a temperature gradient on almost 40 degrees only there! And the rest of the path is probably more than 0.5 K/W too, so the temperature is rising FAST at higher power levels! This is also the reason for the limitation for dissipation levels we are able to use these devices at. Think of it, if we were to use the IRFP044 at 100 W the case temperature would have to be -30 deg for a chip temperature of 100 deg
! I don't thing anyone here is brave enough to take on a project that does this! Or maybe someone with access to LN2 (liquid nitrogen…).
So my idea was to use boiling water. That is one of the most constant temperatures I have access to! And to me this seems like a reasonable test temperature. I just have to find a practical way to implement it…….
And in addition I have Pt1000 temperature sensors just to make sure I have the correct temperature. And I could also use them to do a test at other temperatures. I think I'll do some more testing on different temperatures on the IRF9610's! Seems like it is worthwhile at these devices. And hopefully the working temperature on them isn't that high either
So, could anyone please answer my question about combining batches?
Anders
What about my question??? Anyone with a comment on whether to keep the two batches separate or mix and match in both?
About the temp in measuring.... of course I have thought about the temp at witch I'll do the measuring. The thing is that I have absolute control over WHAT it is because of the low dissipation. ALL parts of the transistor have the same temperature.
The idea is that I'm not concerned with the temperature gradients in the devices at the normal power levels we use. With a heatsink temperature of about 50-60 deg C the chip temperature may very well be about 100 deg! With 30 W dissipated and an internal temperature resistance in the device on 1.3 we have a temperature rise above heatsink on 30 x 1.3 giving a temperature gradient on almost 40 degrees only there! And the rest of the path is probably more than 0.5 K/W too, so the temperature is rising FAST at higher power levels! This is also the reason for the limitation for dissipation levels we are able to use these devices at. Think of it, if we were to use the IRFP044 at 100 W the case temperature would have to be -30 deg for a chip temperature of 100 deg
! I don't thing anyone here is brave enough to take on a project that does this! Or maybe someone with access to LN2 (liquid nitrogen…).So my idea was to use boiling water. That is one of the most constant temperatures I have access to! And to me this seems like a reasonable test temperature. I just have to find a practical way to implement it…….
And in addition I have Pt1000 temperature sensors just to make sure I have the correct temperature. And I could also use them to do a test at other temperatures. I think I'll do some more testing on different temperatures on the IRF9610's! Seems like it is worthwhile at these devices. And hopefully the working temperature on them isn't that high either
So, could anyone please answer my question about combining batches?
Anders
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.