Has anybody used the ALfet lateral mosfets mentioned before - Mosfets/Modules - Lateral MOSFETs - Class D & can confirm that they are OK?
They come in single die 12A (£3) & double die 24A versions (£6.80) - any advantages to DD?
They come in single die 12A (£3) & double die 24A versions (£6.80) - any advantages to DD?
Nope, though they seem cheap enough and they're not rated as per the other sources in the UK, like Profusion and Magnatec. Those guys install the same chips from Semelab in plastic and T03 cases to also supply spares and OEM for pro. music and linear PA amp and Disco markets.
These ALfets may be a different process again, so you could ask a UK member like Nigel Wright, who might have used them in the field. I think he's posted in Lineup's thread, too. The double dies are great from the point of view of current sharing, but it's 2 die in parallel and twice the input capacitance, twice the bias current, so another matter for tweaking or even redesign, I suspect.
kean,
Thanks. Would you suggest that the more current you run in the VAS, the larger the cap should be? I have only tried 1uF and 22uF so far, but I hope to try some larger values soon.
Diff input pair situation
Hi,
Can you be more specific on the position for the cap for the diff input pair situation.
And if one use a VBE multiplier for biasing output mosfets...
thanks
Fab
Hi,
Can you be more specific on the position for the cap for the diff input pair situation.
And if one use a VBE multiplier for biasing output mosfets...
thanks
Fab
I wonder what this little GEM would be like with a lateral fet for the VAS. like the 2SJ79...? there you more than half the Ciss over the IRF part, but still a little higher than the Zetex..
FAB
the cap in the input is placed just under/over the devices and can be used with good effect on both Jfet and bipolar input stages, Nico Rass does this on his ELD design..
In the VBE design..(must say little expediences here as I primarily work with laterals) the cap must go right before gates resistors..
FAB
the cap in the input is placed just under/over the devices and can be used with good effect on both Jfet and bipolar input stages, Nico Rass does this on his ELD design..
In the VBE design..(must say little expediences here as I primarily work with laterals) the cap must go right before gates resistors..
There are a few factors which should be addressed directly, and may be just as likely to affect the sound as the capacitive loading which I mentioned.
First, the CCS is not ideal. It will present a DC load to the VAS. Because of this, the bias spreading resistor will act like a resistive divider network with the CCS being the shunt resistor. This will cause unequal current balance, independent of frequency. This could very well be a non-issue because of the bootstrap, which if I interpreted the layout correctly, will greatly improve CCS performance in the audio band (depending on the size of the bootstrap cap). I suspect Gm differences between the actual MOSFETs will matter more, judging by my SPICE models which come from Cordell.
Secondly, the feedback loop and VAS have no control over bias spreader accuracy. The feedback loop sees the output stage as a single device with a single transfer function. Therefore errors in the output stage must be dealt with as you would a no-feedback buffer.
Increased VAS current would equate to more available drive, however the bias spreader only sees the load (MOSFET gates), so the requirements may not change much.
The bias spreader network, in reacting to the drive current which must pass through it, will determine in part the symmetry of the output drive. A capacitor here may be chosen to be only large enough to compensate for the capacitance of the CCS and MOSFET gate. If it is larger, this correction will extend to frequencies not affected by this capacitive loading. I suspect this is the case for the 22uF cap. Only a 100n cap may be needed for the corrective effect, so the sonic requirement for a larger cap may indicate another factor at play.
If the CCS provides a DC resistive load, this is also something the cap will serve to correct, but it would take a large cap to cover the audio spectrum.
In engineering terms, either the cap only needs to be small, say 10n-100n, or large enough to cover the audio band, having an adequate RC time constant with the spreader resistor. In between is a matter of listening tests I think.
- keantoken
First, the CCS is not ideal. It will present a DC load to the VAS. Because of this, the bias spreading resistor will act like a resistive divider network with the CCS being the shunt resistor. This will cause unequal current balance, independent of frequency. This could very well be a non-issue because of the bootstrap, which if I interpreted the layout correctly, will greatly improve CCS performance in the audio band (depending on the size of the bootstrap cap). I suspect Gm differences between the actual MOSFETs will matter more, judging by my SPICE models which come from Cordell.
Secondly, the feedback loop and VAS have no control over bias spreader accuracy. The feedback loop sees the output stage as a single device with a single transfer function. Therefore errors in the output stage must be dealt with as you would a no-feedback buffer.
Increased VAS current would equate to more available drive, however the bias spreader only sees the load (MOSFET gates), so the requirements may not change much.
The bias spreader network, in reacting to the drive current which must pass through it, will determine in part the symmetry of the output drive. A capacitor here may be chosen to be only large enough to compensate for the capacitance of the CCS and MOSFET gate. If it is larger, this correction will extend to frequencies not affected by this capacitive loading. I suspect this is the case for the 22uF cap. Only a 100n cap may be needed for the corrective effect, so the sonic requirement for a larger cap may indicate another factor at play.
If the CCS provides a DC resistive load, this is also something the cap will serve to correct, but it would take a large cap to cover the audio spectrum.
In engineering terms, either the cap only needs to be small, say 10n-100n, or large enough to cover the audio band, having an adequate RC time constant with the spreader resistor. In between is a matter of listening tests I think.
- keantoken
Last edited:
This could be a good candidate for the bias spreader cap -
UES1H330MPM Nichicon Aluminum Electrolytic Capacitors - Leaded
JC mentioned he used the Nichocon Muse in his Blowtorch preamp, but in the power supply I think. It is a bipolar and designed for audio signal circuits.
UES1H330MPM Nichicon Aluminum Electrolytic Capacitors - Leaded
JC mentioned he used the Nichocon Muse in his Blowtorch preamp, but in the power supply I think. It is a bipolar and designed for audio signal circuits.
Kean...Thanks for the very nice description....
Funny thing with caps....one can dimension and say that this is large enough..and then by increasing to twice the size all music aspects grows a magnitude..
quess that this one must be adjusted by ear...
The voltage need not to be any higher than 10V...But I agree that the nichicon ang also Elna are good choices..Wonder if Rybucon that used to manufacture the BG's have any caps similar to what they used to produce for BG..
Funny thing with caps....one can dimension and say that this is large enough..and then by increasing to twice the size all music aspects grows a magnitude..
quess that this one must be adjusted by ear...
The voltage need not to be any higher than 10V...But I agree that the nichicon ang also Elna are good choices..Wonder if Rybucon that used to manufacture the BG's have any caps similar to what they used to produce for BG..
Last edited:
I began with 1uF film cap
Switched to 10uF black gate - sounded awful, didn't want to wait for ages to see if it got better after burn in.
Settled with 4.7uF film cap - this will do for now
Have to say didn't really notice much of a difference with this mod - treble sounded good already
I often use Rubycon ZA/ZL - used to be called ultra low impedance, now called general purpose . . .
However, changing from irf9610 to zvp4424a had a mind blowing positive impact on sound. Had to add a few pico farads of miller cap to suppress 50MegHz 2V oscillation ( on one channel ) - used about 2" of a tightly wound CAT 6 pair for this tiny capacitance ( both channels )
The sound is much much much cleaner after this mod - can't work out if this is just due to more OLG, better PSRR etc, or because irf9610 is inherently noisy - perhaps a bit of both ? ? ?
I have some 2sj79s' so I can try sometime but as far as I could work out from the data sheets 4424a gave the best transconductance at 10mA.
Downside for me is that the DC drift just got worse with this mod.
Mikelm,
Sad to hear the cap didn't work for you. Maybe give another type a go some time though just to be sure, and try 22uf?
Very glad that your results at least agree with mine re removing the 9610. The zetex fets really work wonders don't they?
What are the size of the resistors in your feedback network and what size feedback shunt cap are you using? My DC offset is fine even with the zvp3310 which is only a 650mW device. Perhaps it's time you made a fet based CCS?
Sad to hear the cap didn't work for you. Maybe give another type a go some time though just to be sure, and try 22uf?
Very glad that your results at least agree with mine re removing the 9610. The zetex fets really work wonders don't they?
What are the size of the resistors in your feedback network and what size feedback shunt cap are you using? My DC offset is fine even with the zvp3310 which is only a 650mW device. Perhaps it's time you made a fet based CCS?
FB = 500 / 47R
I know that you have no real DC offset problems with AC coupling but once you have done all your other mods for filtering etc if you try complete DC coupling you might decide you would be happy to have the offset problems . . . .
. . . you might even decide you prefer a film cap across the bias resister My boards a due today - they are designed to accommodate DC offset measures so I will begin that soon.
Last edited:
Mike
the SJ79 may have lower trans-conductance, but at the soma time also low capacitance (90pF vs app 150pF).. I dont think that the sound suffers by a lower OLG... and in my Sims its dead stable without miller cap .. and with limited lead compensation 220R+22pf
(sorry for the double post, on vacation and mobile devices seem to live a life on their own)
the SJ79 may have lower trans-conductance, but at the soma time also low capacitance (90pF vs app 150pF).. I dont think that the sound suffers by a lower OLG... and in my Sims its dead stable without miller cap .. and with limited lead compensation 220R+22pf
(sorry for the double post, on vacation and mobile devices seem to live a life on their own)
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.