The Saint said:
An externally hosted image should be here but it was not working when we last tested it.
The issue here to be solved in a good way, as I see it,
is to find the symmetrical complementary pairs of MOSFETs for input stage.
They are the key to make this amplifier unique and good.
Nelson Pass is not doing too bad,
with several famous amplifiers with MOSFET based inputs.
I think your concept will be intertesting to very many DIY-ers.
People simply like anything with FET - JFET, MOSFET ....
I am sure the result will be very good amplifiers.
There is not such a cancellation, charge has to be taken out of one capacitor and introduced into another. This takes double the energy to achieve. There are MOS devices suitable for these duties, like 2N7000 or BS170, but HEXFETs are power switching devices aimed only at a low on resistance. It's like building an input stage around MJ15022 and MJ15023 transistors.
You may be actually thinking into transconductance doubling. This does happen, but doest not improve things very much as MOS devices show ver low transconductances for drain currents in the low mA range. Note that transconductance tends to be proportional to the drain current squared, so if it's 0.8 at Id=1A, how much would it be at Id=1mA ?
You may be actually thinking into transconductance doubling. This does happen, but doest not improve things very much as MOS devices show ver low transconductances for drain currents in the low mA range. Note that transconductance tends to be proportional to the drain current squared, so if it's 0.8 at Id=1A, how much would it be at Id=1mA ?
Upupa Epops said:Hi Eva, exist also much better modern types, than these old ones... Look at pages of Zetex and Siliconix... BTW, NP knows it...
Oh, I know. These were only examples classic small-signal devices. But we are discussing 1A HEXFET power devices being employed as small signal transistors as in The Saint prototypes.
The Saint said:
Transconductance and capacitance are parameters useful only for linear operation analysis. I would be speaking in gate-charge and Rds-on terms if I were thinking about switching behaviour. Note that HEXFET datasheets doesn't even show low-current transfer characteristics.
By the way, what transconductance figures have you obtained from such an input stage? Have you done any open loop gain estimation? What unity-gain bandwidth-product figures are you managing?
I'm also curious on how low offset voltages are obtained from MOS devices whose gate threshold voltage varies wildly between 2V and 4V. How much devices have to be measured in order to find suitable pairs?
To me, these circuits look much like classic amplifiers where some switching MOSFETs were dropped in place of bipolar transistors, without much thinking about the consequences. Exotic-component lovers will find the exchange quite interesting, though.
Eva said:
Eva,
Either a) I don't understand your reply correctly or
b) you may be getting capacitances mixed up a little.
In a differential pair used in an amp with global feedback, mainly Crss is
seen at the IP (gate). This is due to the source following the gate very
closely and as such the capacitance is not 'seen'.
As such these Hexfets will usually impose less than 50pF or so on
the IP. What I am not so keen on is the capacitance modulation with
voltage swing. However this is another subject all together.
Cheers,
Terry
Electrical mass
Eva,
I think I concur with you. MOSFETs are fast but heavy - lots of capacitance and lots of non-linear capacitance. Great if you don't mind this drawback. The transconductance is hard to control too. Bipolars are "generally" slower and they are more easily damaged by extreme currents and voltages which is why they are not used so often in high-speed switching applications anymore, however, their gm is controllable and their phase response can be superior depending on what you're trying to achieve. My approach to audio amps makes FETs not very attractive to me. A "liquid midrange" is an outcome I like to avoid...I prefer "lucid".
Eva,
I think I concur with you. MOSFETs are fast but heavy - lots of capacitance and lots of non-linear capacitance. Great if you don't mind this drawback. The transconductance is hard to control too. Bipolars are "generally" slower and they are more easily damaged by extreme currents and voltages which is why they are not used so often in high-speed switching applications anymore, however, their gm is controllable and their phase response can be superior depending on what you're trying to achieve. My approach to audio amps makes FETs not very attractive to me. A "liquid midrange" is an outcome I like to avoid...I prefer "lucid".
Terry Demol said:
That's an over-simplification. MOS devices have a finite transconductance, and this means that gate capacitance has to be charged or discharged in order to get any change in drain current. That transconductance is very low for HEXFETs operating in the low mA range, far lower than in any small signal bipolar device, so at least a few dozens of milivolts of Vgs swing should be expected. Indeed, both sources (strongly coupled by input capacitances) will follow the input signal at the same time, thus nothing will be amplified.
I've just measured a transconductance of 0.00625 S in a randomly selected IRF520 HEXFET. The Ids interval considered was between 0.5 to 1mA and Vds was kept at approx 11V. I had to increase Vgs from 3.52 to 3.6 volts in order to get a 500uA increase in drain current. The transconductance value quoted in the datasheet is 4.1 S typical for Id=5.6A and Vds>=50V, but it tends to be proportional to Id squared, that's why these HEXFETs have little or no use in small signal circuits (except when you bias them with 5A or so
like Mr.Pass uses to do ).Indeed, this is very poor performance if we consider that the transconductance of a BC547 in the 0.5 to 1mA range is approx 0.03 S, while input capacitance is an order of magnitude lower.
Nelson Pass uses quite much in idle current therefore the gm will be rather much but at mA's how much do you get really and isn't it a hopeless task to match transistors? If you want to make an easy-to-build-product, avoid matching as much as possible. My advise here.
Notice also that the input stage is very important when it comes to noise and distortion for the whole amp. An input stage _must_ be good. A bad input stage can't save the amp even if feedback is used.
Notice also that the input stage is very important when it comes to noise and distortion for the whole amp. An input stage _must_ be good. A bad input stage can't save the amp even if feedback is used.
Per Anders,
Actually, of all the Europeans, the Germans, Dutch and ALL the Scandinavian peoples are unquestionably the most adept at English language, though it must be said we Orstrayans speak the lingo real bonza, mate, better than them Poms!
Perhaps you misunderstood me? I was comparing only the Chinese with the Indians from the POV of manufacturing pcbs. The language barrier is ever present with the former, though fast coming down. With the latter, where English is universally the second language, communication is NEVER a problem! In technical matters, and where business is done, communication is incredibly important, along with business ethics and expertise.
Cheers,
Hugh
Actually, of all the Europeans, the Germans, Dutch and ALL the Scandinavian peoples are unquestionably the most adept at English language, though it must be said we Orstrayans speak the lingo real bonza, mate, better than them Poms!
Perhaps you misunderstood me? I was comparing only the Chinese with the Indians from the POV of manufacturing pcbs. The language barrier is ever present with the former, though fast coming down. With the latter, where English is universally the second language, communication is NEVER a problem! In technical matters, and where business is done, communication is incredibly important, along with business ethics and expertise.
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.