andy_c said:
I don’t know either, as I’ve never really studied this in depth, but I do know from my experience in building amplifiers with the venerable MJL21193/MJL21194 transistors, and having dabbled with the high fT devices, that the performance gulf between the two isn’t nearly as wide as between the fT specifications.
For instance, have a look at the Cob specification for one of the best high fT BJT pairs out there:
http://www.farnell.com/datasheets/82788.pdf
http://www.farnell.com/datasheets/82780.pdf
The Sanken 2SC3263 is the NPN with an fT of 60MHz. Its Cob at (Vbe=-10V f=1MHz) is 250 pf.
Its PNP complement, the 2SA1294 has an fT of 35MHz and its Cob (Vbe=10V f=1MHz) is 500 pF.
These Cob figures are identical to the 4MHz MJL21194/MJL21193 transistors under the same test conditions!!
Cheers,
Glen
Another thing worth noting is that while the Vce and Ic ratings are almost the same, the 4MHz MJL21193/MJL21194 pair is rated at 200W, while the super fast Sanken pair is only rated at 130W.
In other words, depending on output power, you may require two pairs of the Sanken devices were you could otherwise make do with only one pair of MJL21193/MJLl21194 devices.
With the fatster transistors, you'll then have double the Cob!
Cheers,
Glen
In other words, depending on output power, you may require two pairs of the Sanken devices were you could otherwise make do with only one pair of MJL21193/MJLl21194 devices.
With the fatster transistors, you'll then have double the Cob!
Cheers,
Glen
Yeah, but I guess the most important thing is total input capacitance of the EF. You see Cob directly, but Cib multiplied by (1-Av). On one hand, 1-Av is pretty close to zero, but Cib is pretty darned big. How does this compare to Cob for the different device types? Dunno. I haven't done the math as I'm listening to tunes while typing this. Priorities! 🙂
andy_c said:
Dunno either, but this is definately something worth looking into.
Right now, I'm going to have lunch.
Cheers,
Glen
OK, just a very quick calculation for now. It’s a darn pity that the Sanken datasheets don’t have any Cib graphs, but the On Semi ones do, so an easy quick comparison can be made between the 4MHz MJL21194 and the 30MHz version, the MJL3281.
Taken from the Cib and Cob graphs (@vbe=-10V), the values for the two transistors are as follows:
MJL21194 (4MHz fT NPN)
Cib = 19000pF
Cob = 230pF
MJL3281A (30MHz fT NPN)
Cib = 3800pF
Cob = 250pF
Cob for the 4MHz device is slightly smaller, but Cob is 4 times bigger.
The input capacitance of the emitter follower approximates:
Cin = (Cib(1-Av))+Cob
For an Av of 0.9, the input capacitance for each transistor is:
MJL21194
0.9 Cin = 2130pF
MJL3281A
0.9 Cin = 630pF
2130/630 = 3.4
OK, we really need some accurate Av calculations here, but at 0.9 we have a reduction of input capacitance equal to a little less than half the increase in fT (30/4=7.5).
Cheers,
Glen
Taken from the Cib and Cob graphs (@vbe=-10V), the values for the two transistors are as follows:
MJL21194 (4MHz fT NPN)
Cib = 19000pF
Cob = 230pF
MJL3281A (30MHz fT NPN)
Cib = 3800pF
Cob = 250pF
Cob for the 4MHz device is slightly smaller, but Cob is 4 times bigger.
The input capacitance of the emitter follower approximates:
Cin = (Cib(1-Av))+Cob
For an Av of 0.9, the input capacitance for each transistor is:
MJL21194
0.9 Cin = 2130pF
MJL3281A
0.9 Cin = 630pF
2130/630 = 3.4
OK, we really need some accurate Av calculations here, but at 0.9 we have a reduction of input capacitance equal to a little less than half the increase in fT (30/4=7.5).
Cheers,
Glen
G.Kleinschmidt said:
Of course, those calculations only have a jot of value in so far as the Cib @ Vbe=-10V value is close a approximation to whatever it may be when the transistor is biased on, which it probably isn't.
I think I'm having a braindead episode this afternoon.
and if the gain were 0.95 then the ratio falls from 3.4 to <2.7
The effective values are now <500pf & <1200pF.
What would be a reasonable gain value when the amp is cruising?
Does this hold fairly steady until the gain droop starts at very high currents?
Would the gain get even closer to 0.99 if the devices were carrying very low currents or does low current hFE start to have an adverse effect?
As Vce varies so does the Cib/Cob for both devices and in the same direction.
The effective values are now <500pf & <1200pF.
What would be a reasonable gain value when the amp is cruising?
Does this hold fairly steady until the gain droop starts at very high currents?
Would the gain get even closer to 0.99 if the devices were carrying very low currents or does low current hFE start to have an adverse effect?
As Vce varies so does the Cib/Cob for both devices and in the same direction.
AndrewT said:
G'day Andrew. There are lots of other variables involved, but we really need to establish Cib to get a reliable rough idea first. The values shown on the datasheets are measured with the base-emitter junction reverse biased.
Those quick calculations were just me thinking aloud, and they’re most likely way-off. however BJT’s do have very high transconductance, which proportionately reduces Cib, so it’s perfectly reasonable to assume that in an emitter follower configuration, Cib may be reduced to a value not making Cob insignificant. A thorough analysis would show exactly how much better the high fT devices really are, as it is obvious from the datasheets that Cob has not been effectively reduced in the high fT devices.
This is all a little over my head at the moment. Might have to dust off some of the old theory books!
Cheers,
Glen
yet another power electronics mag article on mosfet forward bias SOA and IXYS "linear" power MOSFETS:
http://powerelectronics.com/704PET23.pdf
http://powerelectronics.com/704PET23.pdf
The discussion on Spice simulators has been split here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=101810
Thanks to all for your patience.
/Hugo
http://www.diyaudio.com/forums/showthread.php?s=&threadid=101810
Thanks to all for your patience.
/Hugo
What I am curious about is that I don't think that you people appreciate that the EFFECTIVE INPUT CAPACITANCE in an output device is very much smaller than what is on the data sheet. This is because when an
active device is used as a follower, the input capacitance is BOOTSTRAPPED automatically which reduces the Cbe to almost nothing, and the Ccb is not that much and is not subject to Miller multiplication. The finite beta of the bipolar device is the real drive concern. This can get up into amps, under worse case conditions. That is why I use high current mostfets as drivers, rather than wimpy (1A) mosfets.
active device is used as a follower, the input capacitance is BOOTSTRAPPED automatically which reduces the Cbe to almost nothing, and the Ccb is not that much and is not subject to Miller multiplication. The finite beta of the bipolar device is the real drive concern. This can get up into amps, under worse case conditions. That is why I use high current mostfets as drivers, rather than wimpy (1A) mosfets.
john curl said:
John, did you even read the previous posts on this topic? We all know that Cib is bootstrapped. What about Cob? Cob hasn’t been reduced in the high ft devices. That’s interesting to note, isn’t it? It is the same for a 4MHz MJL21194 as for a 60MHz Sanken. And of course the collector-base capacitance isn’t subject to the miller effect in emitter follower applications. Duh.
Do you have any theoretical figures or have you ever measured the effective input capacitance of high fT BJT’s?
john curl said:
Hi John,
This means that you still want to use Mosfets at the output also , but haven't because you cannot some how safeguard them from direct short circuit protection....😉
You arenot using them because you are unable to cope with their destruction during shortcircuit...I think that's old thing, but why haven't you conducted any further experments to eradicate this flaw...
Meanwhile...I have just finished another All-N-channel Design, which neither uses output coil, nor the Zobel network and is still happy to drive any complex load, safe into shortcircuit,doesnot oscillates with long connecting wires of any length.......
Kanwar
john curl said:
Impertinent question: why not settle for 300 watts rms and go for the 180Vce Sankens instead ?
I fancy the Shelby GT500 Super Snake, but i figured the amp arms race would have come to a halt.
May I just remind you that Cbe of a BJT is quite predictable and follows from semiconductor physics, and is AFAIK not dependent on any other factors the manufacturer can toy around with. Due to the low voltage swing of Vbe, the junction capacitance can be ignored so Cbe can be considered to equal the diffusion capacitance, that is
Cbe = hfe/(2*pi*fT*vT)
At room temperature that is approximately
Cbe = 20*hfe/(pi*fT)
That is, hfe (aka beta) and fT are the only twp parameters that affect Cbe.
Cob is different, though, and seems to depend on more device dependent factors, so there it is necessary to consult a data sheet.
BTW, I don't quite see what you mean by bootstrapping here. Cbe is the capacitance between base and emitter, and since Vbe depends on IC only and is not dependent on whether we use the transistor in CE, CC or CB, I don't quite see what you mean.
Cbe = hfe/(2*pi*fT*vT)
At room temperature that is approximately
Cbe = 20*hfe/(pi*fT)
That is, hfe (aka beta) and fT are the only twp parameters that affect Cbe.
Cob is different, though, and seems to depend on more device dependent factors, so there it is necessary to consult a data sheet.
BTW, I don't quite see what you mean by bootstrapping here. Cbe is the capacitance between base and emitter, and since Vbe depends on IC only and is not dependent on whether we use the transistor in CE, CC or CB, I don't quite see what you mean.
Christer said:
The base-emitter capacitance, in the emitter follower configuration, is bootstrapped because the emitter 'follows' the base. The effective input capacitance is therefore reduced.
Same goes for the gate-source capacitance of a MOSFET when configured as a source folower, but the degree of reduction is much less as MOSFET's have much lower transconductance than BJT's.
BTW, thanks for the Cbe formula; that's all I need to know.
Cheers,
Glen
But
Vbe = vT * ln(Ic/Is)
so it only depends on Ic (as long as we only bother about first order effects). From this we can derive that a change dVbe (meaning a delta rather than differential here) is
dVbe = vT*ln(1 + dIc/Icq).
Either we consider dVbe so small that we ignore it, and then Ve follows Vb in whatever configuration we use the transistor (although in a CB connection it would make more sense to think of Vb following Ve). If we take the ac component of Vbe into account, Ve no longer follows Vb in a proportional way, and that does also not depend on how the connect the transistor. It only depends on Ic. A small signal analysis gives a simplified view, but one that essentially agress with this.
Of course Vb increases with Ve, so one might perhaps say that Vb is bootstrapped, but that is Vb, not Vbe. This also happens in whatever configuration we use. Of course, in a CE connection with no degeneration, Ve will be fixed, but otherwise there is no difference.
Vbe = vT * ln(Ic/Is)
so it only depends on Ic (as long as we only bother about first order effects). From this we can derive that a change dVbe (meaning a delta rather than differential here) is
dVbe = vT*ln(1 + dIc/Icq).
Either we consider dVbe so small that we ignore it, and then Ve follows Vb in whatever configuration we use the transistor (although in a CB connection it would make more sense to think of Vb following Ve). If we take the ac component of Vbe into account, Ve no longer follows Vb in a proportional way, and that does also not depend on how the connect the transistor. It only depends on Ic. A small signal analysis gives a simplified view, but one that essentially agress with this.
Of course Vb increases with Ve, so one might perhaps say that Vb is bootstrapped, but that is Vb, not Vbe. This also happens in whatever configuration we use. Of course, in a CE connection with no degeneration, Ve will be fixed, but otherwise there is no difference.