john curl said:
Thanks, John. Could you define R(S), R(g), C(L) and R(L).
Is one of these the internal device effective gate resistance?
Is this assuming a source follower configuration?
I don't see the Cgd in there.
Can we assume this is valid for both laterals and Hexfets?
What value do we get for an IRFP240 in a typical source follower arrangement?
Thanks again,
Bob
Workhorse said:
Hi Kanwar,
I've said it before and I'll say it again: the IRFP264N device is not a good choice for an audio amplifier. You've just shown why. It is built with a different process. Look at the data sheet.
Now look at the data sheet for the IRFP240 and you will see that there is no such problem.
BTW, lets hope IR and others continue to make the "old" style Hexfet-type MOSFETs.
I have heard that IR has sold their power components division to, I think, Vishay. This may include the Hexfets. Has anyone else heard about this?
Bob
"EL SEGUNDO, Calif. & MALVERN, Pa.--(BUSINESS WIRE)--Nov. 1, 2006--International Rectifier Corporation (IR) (NYSE:IRF) and Vishay Intertechnology, Inc. (NYSE:VSH) today announced they have reached an agreement for the sale of IR's Power Control Systems (PCS) business to Vishay. The PCS business includes IR's Non-Focus Products business and certain product revenue from its Focus Products business, including certain discrete planar MOSFETs, discrete diodes and rectifiers, discrete thyristors, and automotive modules and assemblies."
Bob Cordell said:
Hold on please Mr.Robert Cordell,
Just checked your IRFP240 old style hexfet...in real world
VGS 3.2V held constant...
VDS varied from 10V to 190VDC
following are the Idrain values with varying VDS
10V->55mA
50V->235mA
75V->389mA
100V->562mA
150V->964mA
190V->1350mA
Isn't these above mentioned figures are even worst than previously obtained figures with IRFP264N which is a Hexfet...
Now your old style HEXFET is also following the same destructive path...
In another test with 2SC5200 230V 16A Bipolar....with Vbe held constant at 0.58V with 320uA of base drive and sweeping VCE from 10V to 190V the Icollector showed a variationof just 25mA to 39mA only.....
Kanwar
Workhorse said:
Hold on please Mr.Robert Cordell,
Just checked your IRFP240 old style hexfet...in real world
VGS 3.2V held constant...
VDS varied from 10V to 190VDC
following are the Idrain values with varying VDS
10V->55mA
50V->235mA
75V->389mA
100V->562mA
150V->964mA
190V->1350mA
Isn't these above mentioned figures are even worst than previously obtained figures with IRFP264N which is a Hexfet...
Now your old style HEXFET is also following the same destructive path...
In another test with 2SC5200 230V 16A Bipolar....with Vbe held constant at 0.58V with 320uA of base drive and sweeping VCE from 10V to 190V the Icollector showed a variationof just 25mA to 39mA only.....
Kanwar
So if I understand you corrctly, you are saying that the IR data sheet is blatantly wrong. I'd like to see the deatils of your "real-world" test setup. Maybe there is something wrong or an effect you are not considering. Are you keeping the device at a constant temperature?? If you really measured it at 190V and 1350 mA, it was screaming hot, and indeed that is well beyond its rated power dissipation.
Bob
Get Real
The Device is mounted on heatsink and the device is subjected to 190VDC in real world, but only for a fraction of second[10mS to 100mS], thats why it doesnt get damage..
Why not ,Do a small test on your amp with hexfets,
Bias them at say 35mA with low rail voltage[20 to 30VDC] and then gradually increase or double the rail voltage with variac or so and see yourself the Idrain increasing in accordance with square law characteristics...of Vfets
Idrain is directly proportional to VGS applied as well as VDS across the Hexfet device...
The increase in Idrain due to the variation in VDS[incremental] is main cause of destruction of mosfets in an amp...This simple fact is always neglected and ignored by the designers so well that they just bias the Hexfet just like a regular temperature compensated bipolar amp is biased...
Bipolars doesnot show this much increase in their collector current with increase in VCE across the device.
Have a look at these patents 6,268,770 and 6,023,193
just the HEXFET amp
Get the pdf from:
http://www.pat2pdf.org
Kanwar
Bob Cordell said:
The Device is mounted on heatsink and the device is subjected to 190VDC in real world, but only for a fraction of second[10mS to 100mS], thats why it doesnt get damage..
Why not ,Do a small test on your amp with hexfets,
Bias them at say 35mA with low rail voltage[20 to 30VDC] and then gradually increase or double the rail voltage with variac or so and see yourself the Idrain increasing in accordance with square law characteristics...of Vfets
Idrain is directly proportional to VGS applied as well as VDS across the Hexfet device...
The increase in Idrain due to the variation in VDS[incremental] is main cause of destruction of mosfets in an amp...This simple fact is always neglected and ignored by the designers so well that they just bias the Hexfet just like a regular temperature compensated bipolar amp is biased...
Bipolars doesnot show this much increase in their collector current with increase in VCE across the device.
Have a look at these patents 6,268,770 and 6,023,193
just the HEXFET amp
Get the pdf from:
http://www.pat2pdf.org
Kanwar
Workhorse said:
Hi Bob,
What about me, with 10 Ohm Rgate and 500Mhz & 1GHz Tek, but i see no parasitic oscillations....
I have seen your paper, its very well illustrated, but one thing confuses me that the paper states the title author as "Robert Cordell" while you are Bob Cordell....whats the secret
have a good day
Kanwar
Any chance of seeing the PCB layout ? Or photo's of amplifier innards that accomplish this?
I guess the common source FETs ft is divided by it's transconductance - like bipolars GBP/hfe
Thanks Kevin
Re: Get Real
Kanwar,
You didn't answer my admittedly blunt question. Are you saying that the IR data sheet for the IRFP240 is blatantly wrong? Please answer that question.
Thanks for the patent references; I'll give them a look.
10 ms? 100 ms? that's a big difference. The thermal time constant of the device chip is quite short. Look at the curves of effective thermal impedance vs time. Please provide more detail of your experimental setup. For example, when you measured the current under these pulsed conditions, how did you measure it? Did you measure it with an oscilloscope? If so, did you notice that the value of the current was flat with time during the pulse interval, or did you see it rising? Even 10 ms is a long time. I still think there is something wrong with your test, and that the problem could be instantaneous chip temperature; but it could also be something else.
If your findings were correct, many, many MOSFET power amplifiers would be blowing up.
Bob
Workhorse said:
Kanwar,
You didn't answer my admittedly blunt question. Are you saying that the IR data sheet for the IRFP240 is blatantly wrong? Please answer that question.
Thanks for the patent references; I'll give them a look.
10 ms? 100 ms? that's a big difference. The thermal time constant of the device chip is quite short. Look at the curves of effective thermal impedance vs time. Please provide more detail of your experimental setup. For example, when you measured the current under these pulsed conditions, how did you measure it? Did you measure it with an oscilloscope? If so, did you notice that the value of the current was flat with time during the pulse interval, or did you see it rising? Even 10 ms is a long time. I still think there is something wrong with your test, and that the problem could be instantaneous chip temperature; but it could also be something else.
If your findings were correct, many, many MOSFET power amplifiers would be blowing up.
Bob
The Real World
I didnot say that the datasheets were 100% wrong, but the output curves regarding Id vs Vds in their datasheets doesnot match with realworld results at all....
The Id was measured with a current probe using scope, the Vgs was held at constant level of 3.2VDC and the VDS is applied in a pulsed form[10mS precisely] with varing in magnitute of voltages...The current rise was a Step square, not a ramp....There was no oscillations at all...
Why Didn't you perform the simple test mentioned by me in previous post and check the results yourself just with the DMM itself........I am not forcing you to believe the what i said is right and nor i am imposing my statement on you, but it would be much better if you check your hexfets yourself first....
BTW: Tell me how many manufactures are using Hexfets in Class-AB amps reliably...
Idrain increases when Voltage across Drain to Source terminal is increased, even if Vgs is held constant......
Read this post 41, posted by Eva regarding Vgs threshold dependancy on VDS which inturns causes the Id to vary alot....
http://www.diyaudio.com/forums/showthread.php?postid=847285#post847285
Eva's Thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=40506&perpage=10&pagenumber=1
Kanwar
Bob Cordell said:
I didnot say that the datasheets were 100% wrong, but the output curves regarding Id vs Vds in their datasheets doesnot match with realworld results at all....
The Id was measured with a current probe using scope, the Vgs was held at constant level of 3.2VDC and the VDS is applied in a pulsed form[10mS precisely] with varing in magnitute of voltages...The current rise was a Step square, not a ramp....There was no oscillations at all...
Why Didn't you perform the simple test mentioned by me in previous post and check the results yourself just with the DMM itself........I am not forcing you to believe the what i said is right and nor i am imposing my statement on you, but it would be much better if you check your hexfets yourself first....
BTW: Tell me how many manufactures are using Hexfets in Class-AB amps reliably...
Idrain increases when Voltage across Drain to Source terminal is increased, even if Vgs is held constant......
Read this post 41, posted by Eva regarding Vgs threshold dependancy on VDS which inturns causes the Id to vary alot....
http://www.diyaudio.com/forums/showthread.php?postid=847285#post847285
Eva's Thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=40506&perpage=10&pagenumber=1
Kanwar
Kanwar's post made me paranoid so I thought I'd investigate further.
A couple of months ago I made a SPICE model for the Fairchild FQA12P20 (datasheet here). There's a parameter called LAMBDA for matching the variation of Id with Vds at constant Vgs to the datasheet value. I did this at the lowest characteristic curve (lowest Vgs) with Vds values of 20V and 40V. So I took this model and made an identical N-channel fake using the same parameters, but changing the sign of VTO, the threshold voltage. The idea was to use a model that I'd fit to some real data myself under known conditions. More info on the fitting technique can be found here.
I did a simple sim of a complementary circuit with a +/- 90V supply with the FETs biased as close to 100 mA as I could get (just for ease of reading the transient analysis graph). Devices were the FQA12P20 and the N-channel fake using the same parameters. Then I set up a transient sim, with the input signal being a sine wave with amplitude 80V at 1 kHz using an open-circuit load.
It turns out that the drain current of each of the devices is always less than or equal to the quiescent value. A little thought shows that the constant Vgs assumption of the FETs cannot be true. As the Vds of one device is increasing, that of its complement is decreasing. If Vgs of each device were equal, that says one device's current would increase while the other's would decrease. This is of course not possible due to the open-circuit load. The conclusion is that the Vgs of each device cannot be constant under these conditions.
Below is the plot of drain current for this circuit. It starts out at about 100 mA, then always stays at a value less than this.
A couple of months ago I made a SPICE model for the Fairchild FQA12P20 (datasheet here). There's a parameter called LAMBDA for matching the variation of Id with Vds at constant Vgs to the datasheet value. I did this at the lowest characteristic curve (lowest Vgs) with Vds values of 20V and 40V. So I took this model and made an identical N-channel fake using the same parameters, but changing the sign of VTO, the threshold voltage. The idea was to use a model that I'd fit to some real data myself under known conditions. More info on the fitting technique can be found here.
I did a simple sim of a complementary circuit with a +/- 90V supply with the FETs biased as close to 100 mA as I could get (just for ease of reading the transient analysis graph). Devices were the FQA12P20 and the N-channel fake using the same parameters. Then I set up a transient sim, with the input signal being a sine wave with amplitude 80V at 1 kHz using an open-circuit load.
It turns out that the drain current of each of the devices is always less than or equal to the quiescent value. A little thought shows that the constant Vgs assumption of the FETs cannot be true. As the Vds of one device is increasing, that of its complement is decreasing. If Vgs of each device were equal, that says one device's current would increase while the other's would decrease. This is of course not possible due to the open-circuit load. The conclusion is that the Vgs of each device cannot be constant under these conditions.
Below is the plot of drain current for this circuit. It starts out at about 100 mA, then always stays at a value less than this.
Excellent work Andy, how about taking IRF hexfets models as well and predicting their behaviour as well.....
Meanwhile, The test I performed were on Hexfets individually, not including them in the amplifier....The VGS was constant applied from a voltage source and the Drain current was monitored while varying the VDS accross it...
BTW: Have you figured out the condition when driving reactive loads....
Kanwar
Meanwhile, The test I performed were on Hexfets individually, not including them in the amplifier....The VGS was constant applied from a voltage source and the Drain current was monitored while varying the VDS accross it...
BTW: Have you figured out the condition when driving reactive loads....
Kanwar
Workhorse said:
Thanks Kanwar. I haven't done this with IRF FET models. The one I'm going to be using is the IRFP244. I didn't fit those curves myself, but used the IRF data. Here are the LAMBDA values for the IRFP244 and FQA12P20.
IRFP244: LAMBDA = 0.00320495 (from IRF model)
FQA12P20: LAMBDA = 0.0072 (from my fit to datasheet curves)
LAMBDA = 0 corresponds to flat curves (no variation of Id with Vds with Vgs=constant). So it looks like the IRFP244 is better than the FQA12P20 in this regard (assuming the IRF model is correct). That is, less variation of Id with variations in Vds will occur. So it would appear that the normal complementary FET circuit is self-correcting in this regard. Although to be honest I haven't quite figured out why the current would always be less than the quiescent value.
I've only looked at open-circuit loads because your post specifically mentioning this got me worried.
Re: The Real World
Kanwar,
I repeated your experiment and did not get the results you showed. I got results that conform with the data sheets published by IR, Fairchild, and others. I therefore think that you are wrong, and that there is something dreadfully wrong with your test setup. What it is, I don't know. Maybe parasitic oscillations that you are unaware of.
I took an IRFP240 and biased it at Vgs = 3.5, 3.75 and 4.0 V, and measured drain current with a 1-ohm resistor in the drain as a function of applied Vds. The source was directly grounded. The gate series resistor was 100 ohms, and there was a series R-C gate shunt to ground of 100 ohms and 150 pF. The whole arrangement was bypassed carefully. All probes to gate or drain were isolated with 100-ohm resistors. The HEXFET was mounted to a heat sink. All measurements shown here were done on a continuous basis (not pulsed), as the currents were not so high as to raise thermal concerns. To the extent that there are any thermal effects here, they will tend to slightly increase the drain currents at higher Vds.
As a base line, drain current at Vds = 10V were as follows:
Vgs = 3.50V => 4.4 mA
Vgs = 3.75V => 28.5 mA
Vgs = 4.00V => 127 mA
For Vgs = 3.75 V, Id as a function of Vds was as follows:
10V => 28.5 mA
15V => 29.3 mA
20V => 29.7 mA
25V => 29.9 mA
30V => 30.4 mA
35V => 31.2 mA
40V => 31.6 mA
45V => 32.0 mA
50V => 32.4 mA
For Vgs = 4.00 V:
10V => 127 mA
15V => 131 mA
20V => 136 mA
25V => 140 mA
30 V => 145 mA
You can see that the id dependence on Vds is very mild, as one would expect from a MOSFET. It is also in pretty good agreement with the data sheets. There is no evidence of the effect you describe and measure.
We all have at some time or another made measurements and results that seem out of whack with what we have learned or what is published in device data sheets. When this happens to me, I rigorously question my setup, measurement procedure and assumptions. I will often try to observe the effect with a different approach or setup. I will always suspect something wrong, even if I don't see it explicitly. I think you need to do the same before you insistently send others off on a wild goose chase.
It is difficult to speculate on what is wrong with your test setup, as you have not provided an exact schematic. My current suspicion is parasitic oscillations, but who knows? What value of gate series resistance were you using in your test setup (you were using one, weren't you?)?
Bob
Workhorse said:
Kanwar,
I repeated your experiment and did not get the results you showed. I got results that conform with the data sheets published by IR, Fairchild, and others. I therefore think that you are wrong, and that there is something dreadfully wrong with your test setup. What it is, I don't know. Maybe parasitic oscillations that you are unaware of.
I took an IRFP240 and biased it at Vgs = 3.5, 3.75 and 4.0 V, and measured drain current with a 1-ohm resistor in the drain as a function of applied Vds. The source was directly grounded. The gate series resistor was 100 ohms, and there was a series R-C gate shunt to ground of 100 ohms and 150 pF. The whole arrangement was bypassed carefully. All probes to gate or drain were isolated with 100-ohm resistors. The HEXFET was mounted to a heat sink. All measurements shown here were done on a continuous basis (not pulsed), as the currents were not so high as to raise thermal concerns. To the extent that there are any thermal effects here, they will tend to slightly increase the drain currents at higher Vds.
As a base line, drain current at Vds = 10V were as follows:
Vgs = 3.50V => 4.4 mA
Vgs = 3.75V => 28.5 mA
Vgs = 4.00V => 127 mA
For Vgs = 3.75 V, Id as a function of Vds was as follows:
10V => 28.5 mA
15V => 29.3 mA
20V => 29.7 mA
25V => 29.9 mA
30V => 30.4 mA
35V => 31.2 mA
40V => 31.6 mA
45V => 32.0 mA
50V => 32.4 mA
For Vgs = 4.00 V:
10V => 127 mA
15V => 131 mA
20V => 136 mA
25V => 140 mA
30 V => 145 mA
You can see that the id dependence on Vds is very mild, as one would expect from a MOSFET. It is also in pretty good agreement with the data sheets. There is no evidence of the effect you describe and measure.
We all have at some time or another made measurements and results that seem out of whack with what we have learned or what is published in device data sheets. When this happens to me, I rigorously question my setup, measurement procedure and assumptions. I will often try to observe the effect with a different approach or setup. I will always suspect something wrong, even if I don't see it explicitly. I think you need to do the same before you insistently send others off on a wild goose chase.
It is difficult to speculate on what is wrong with your test setup, as you have not provided an exact schematic. My current suspicion is parasitic oscillations, but who knows? What value of gate series resistance were you using in your test setup (you were using one, weren't you?)?
Bob
Are you trying to seek attention by these remarks ?
Hi Bob,
What about me, with 10 Ohm Rgate and 500Mhz & 1GHz Tek, but i see no parasitic oscillations....
I have seen your paper, its very well illustrated, but one thing confuses me that the paper states the title author as "Robert Cordell" while you are Bob Cordell....whats the secret
have a good day
Kanwar
Well any chance of seeing your modern day miracle ??
I guess not. Too much hype/talk......
Kevin
Hi Bob,
What about me, with 10 Ohm Rgate and 500Mhz & 1GHz Tek, but i see no parasitic oscillations....
I have seen your paper, its very well illustrated, but one thing confuses me that the paper states the title author as "Robert Cordell" while you are Bob Cordell....whats the secret
have a good day
Kanwar
Fanuc said:
Well any chance of seeing your modern day miracle ??
I guess not. Too much hype/talk......
Kevin