Eli, thehoj,
I think both of you may be looking too closely at the pattern of the capacitance in the charts and not drilling down into scaling of the charts. If you look at the Fairchild Fqpf2n80 it isn't using a logarithmic vertical scale. It's a linear scale. But the voltage scale IS logarithmic. That's downright deceptive advertising to this set of eyes. It deludes one into thinking it measures much better than it does.
Here's another example of a mosfet:
http://www.mouser.com/ds/2/408/TK2Q60D-482885.pdf
It doesn't look nearly as good in the charts but actually it's much better in almost all of the parameters related to capacitance. The one exception is that the Toshiba has it's transition at 11 volts vs about 6 volts for the Fairchild for Crss. If you used a logarithmic scale for the capacitance it would no longer look at all like a mild rise above 6 volts. It would look pretty extreme just like the Toshiba.
So the question you have to ask yourself is "Do you feel lucky, punk". Oops, got carried away. No, you have to ask yourself if that 5 volt difference between the transition point in the Fairchild and the Toshiba matters. The Toshiba beats the Fairchild in spades above 11 volts for Crss. It only deviates +/- 0.5 pf all the way to 100 volts. It wouldn't doubt that it would still beat the Fairchild above if it was shown. The Fairchild only shows to 50 volts. Are they trying to hide something?
I think both of you may be looking too closely at the pattern of the capacitance in the charts and not drilling down into scaling of the charts. If you look at the Fairchild Fqpf2n80 it isn't using a logarithmic vertical scale. It's a linear scale. But the voltage scale IS logarithmic. That's downright deceptive advertising to this set of eyes. It deludes one into thinking it measures much better than it does.
Here's another example of a mosfet:
http://www.mouser.com/ds/2/408/TK2Q60D-482885.pdf
It doesn't look nearly as good in the charts but actually it's much better in almost all of the parameters related to capacitance. The one exception is that the Toshiba has it's transition at 11 volts vs about 6 volts for the Fairchild for Crss. If you used a logarithmic scale for the capacitance it would no longer look at all like a mild rise above 6 volts. It would look pretty extreme just like the Toshiba.
So the question you have to ask yourself is "Do you feel lucky, punk". Oops, got carried away. No, you have to ask yourself if that 5 volt difference between the transition point in the Fairchild and the Toshiba matters. The Toshiba beats the Fairchild in spades above 11 volts for Crss. It only deviates +/- 0.5 pf all the way to 100 volts. It wouldn't doubt that it would still beat the Fairchild above if it was shown. The Fairchild only shows to 50 volts. Are they trying to hide something?
Fenris, good catch. I missed that. I think in a lot of cases the Toshiba will work better if you don't have high source/drain voltage requirements. But if you do then you have to pay the piper.
I think the lesson is not to be profligate in having more negative bias potential than your application requires. Just use a mosfet and source/drain voltage with enough oomph to completely turn off the biased tube, and perhaps a bit more. Any more than that and one probably has to choose an inferior mosfet to handle it.
I think it would work in some limited applications. It biggest problem seems to me to be its 200 volt max drain source voltage. That's definitely limiting. One has to account for the voltage required to completely turn off the valve. Minimum in the negative direction is usually -100 volts or -40 or so for turning off some specific valves like el84 and similar.
In the positive direction you need voltage sufficient for the whole envelope of the ac signal. You also need some reserve so you don't come close to flat topping the output signal. I did some simming using LTSpice and I needed +150v just to completely avoid those problems just driving a 6v6. That's not a hard tube to drive either. So adding the negative and positive direction together comes uncomfortably close to the limit of 200 volts. I think there are better devices out there for that. In a follower application the linearity is still important but not nearly as much so as in an amplification stage.
In the positive direction you need voltage sufficient for the whole envelope of the ac signal. You also need some reserve so you don't come close to flat topping the output signal. I did some simming using LTSpice and I needed +150v just to completely avoid those problems just driving a 6v6. That's not a hard tube to drive either. So adding the negative and positive direction together comes uncomfortably close to the limit of 200 volts. I think there are better devices out there for that. In a follower application the linearity is still important but not nearly as much so as in an amplification stage.
Last edited:
Member
Joined 2009
Paid Member
Well, I depend on others with more direct experience on this. Tubelab is my "go-to" reference on the choice between bi-polar and gated in a follower application and he is a believer in mosfets there. I think the use of zeners for protection on gated devices and esd safe practices will eliminate most of the static problems. I'm certainly no expert but that is my opinion. Everyone's got em, eh.
I've tried a bunch of different FETs but I've settled on using IRF830 for pretty much all (high voltage) purposes. You can buy 50 for 15 euros from far east.
Gyrator plate loads, B+ regulators, CCS tails / loads, grid driver followers... I use it for all. I'm fully satisfied with the high frequency operation of my amps, by ear and by measurement.
Just curious, do you guys hear the differences between the FETs, with at least 4 or 5 mA bias current let's say?
IRF820 and IRF830 have different capacitances, but I couldn't hear any difference using them for all of the above mentioned purposes in my amps. Could be I'm not so good on the high treble hearing - I do still hear grasshoppers, very clearly!
Gyrator plate loads, B+ regulators, CCS tails / loads, grid driver followers... I use it for all. I'm fully satisfied with the high frequency operation of my amps, by ear and by measurement.
Just curious, do you guys hear the differences between the FETs, with at least 4 or 5 mA bias current let's say?
IRF820 and IRF830 have different capacitances, but I couldn't hear any difference using them for all of the above mentioned purposes in my amps. Could be I'm not so good on the high treble hearing - I do still hear grasshoppers, very clearly!
I've been successful recently driving a triode-strapped 13GB5 well into the grid current region with a BJT Sziklai pair attached to an input stage having high output impedance (>100K). Works like a charm. There is zero evidence of load change effects at the grid current threshold when viewing THD analyzer residual.
Member
Joined 2009
Paid Member
I too like the bipolar transistor option, I don't like static sensitivity of gated devices. I have never tried a follower grid drive, I may just have to put this on the 'list'
If you use 800V+ FETs, and solder the mandatory 9V zener to protect the gate, and the stopper resistor for the gate - before first applying power to the circuit, you won't have problems with gate oxide failures.
The same number of parts are required for a bipolar transistor, in a driver application. An ordinary PN diode like a 1N4148 will do - we need to prevent the base from ever going more than 3-4V negative, referring to the emitter. And a stopper is mandatory for an emitter follower, since they will also oscillate in the MHz region without 47-100Ω very close to the base.
FETs fail short circuit if the maximum G-S voltage is breached, but the damage to bipolar transistors is more subtle. Sometimes a negative Vbe event causes increased noise, other times Hfe is degraded. Both of these degradations are most unwanted in an audio amp - as is self-oscillation. Please remember the Rs and Ds!
+1
I too used to prefer BJTs over MOSFETs until I got it into my head that they were actually very robust once you added the gate protection zener. What's more they don't suffer secondary breakdown, so they can withstand more punishing start-up conditions.
Nothing wrong with your hearing. All the way back there on pg 1 (#8) I started that for solid state there is no need to worry over different types of transistors. These are high gain devices, and so in-circuit performance is pretty much independent of the actual device. Unless you need some special property, so long as the transistor is able to function at the proposed signal frequencies, then it really doesn't matter.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.