Hi,
WELL... 😉 I think we agree on the cascodes being overrated when used in the input diff pair of a 3 stage amplifier. Yet many many opamp topolies seem to feature them. In the BOSOZ however a cascode can make a faster more linear amplifier, however this is a totally diferent story; I know.
Hiraga used cascode in the monstre, but that's a FET input amplifier, does that matter?
Better skip my remarks about using a CCS as collector load, it will never work .
equally happely uncascoded🙂
gr,
thijs
WELL... 😉 I think we agree on the cascodes being overrated when used in the input diff pair of a 3 stage amplifier. Yet many many opamp topolies seem to feature them. In the BOSOZ however a cascode can make a faster more linear amplifier, however this is a totally diferent story; I know.
Hiraga used cascode in the monstre, but that's a FET input amplifier, does that matter?
Better skip my remarks about using a CCS as collector load, it will never work .
equally happely uncascoded🙂
gr,
thijs
Thanks Phase_accurate, you just made me realize what current feedback essentially is and why is has extneded bandwidth compared to valtage feedback: Miller effect reduced due to the 'poorly' grounded emmiters... isn't it?
But I don't understand why you agree with halojoy since he sad a cascode slowed the circuit down while you and I agree on the more extended bandwidth with cascodes.
Anyways, your idea of making a CCS that responses to the current through the output pair is interesting! another weekend maybe..
cheers,
thijs
But I don't understand why you agree with halojoy since he sad a cascode slowed the circuit down while you and I agree on the more extended bandwidth with cascodes.
Anyways, your idea of making a CCS that responses to the current through the output pair is interesting! another weekend maybe..
cheers,
thijs
I never said that I'd agree with Halojoy, I just like the simplicity of the basic circuit.
I assume the input capacitance of the output MOSFETS (and their associated miller effect) is the dominating factor determining the upper cutoff frequency.
I personally would rather go for an emitter follower driving the gates of the output devices in the first plac. Maybe you can then still add a cascode to the input.
But what about the nice little circuit you started with ........ ?
Regards
Charles
I assume the input capacitance of the output MOSFETS (and their associated miller effect) is the dominating factor determining the upper cutoff frequency.
I personally would rather go for an emitter follower driving the gates of the output devices in the first plac. Maybe you can then still add a cascode to the input.
But what about the nice little circuit you started with ........ ?
Regards
Charles
I agree with the KISS and ZEN principles. Simplicity is the attraction here, just like the Pass amps. Only two stages and VIOLA!
Remember that slew rate is dependent upon frequency and voltage swing. What are you using for these in your simulations? Amps are generally considered good with 100 v/us at 1Khz. Mr. Pass may very validly argue this point.
I think Fet's are probably the best choice for outputs as BJT's may require more current drive than the first stage can deliver. CCS limits the drive available
I like your poorly grounded bias arrangement better than the one I proposed in my schematics. Simplicity again.
How about trying small fet's in the input stage for higher input impedance while obtaining better current gain? 2SJ79 and 2SK216?
Remember that slew rate is dependent upon frequency and voltage swing. What are you using for these in your simulations? Amps are generally considered good with 100 v/us at 1Khz. Mr. Pass may very validly argue this point.
I think Fet's are probably the best choice for outputs as BJT's may require more current drive than the first stage can deliver. CCS limits the drive available
I like your poorly grounded bias arrangement better than the one I proposed in my schematics. Simplicity again.
How about trying small fet's in the input stage for higher input impedance while obtaining better current gain? 2SJ79 and 2SK216?
I think you are all right in different ways about cascodes. They can speed current up by reducing the Miller effect of the amplifying device BUT Halojoy is also correct in the sense that the shielding transistor does not have infinite bandwidth - it is no like a straight piece of wire. Otherwise darlingtons would have infinite bandwidth provided the VAS stage could drive them with enough current. Also, the more transistors the more distortion from semiconductor junctions.
I shall repeat an old axe of mine, 20kHz is the highest frequency you can hear. For a typical household-powered amp this equates to 5V/us or so. 100V/us is unecessary IMO.
KISS stands for "Keep It Simple Stupid". In general this means keep it as simple as you can whilst achieving the goal. Sometimes this can be overdone and you end up with a stupidly simple design. The great designs often look simple but it is the unseen aspects that are very complicated and lead to high performance.
The posted design looks a little too KISS to me. For example, the resitive bias chains for the input BJT emitters are attached to the power supplies, in fact to opposite power supplies. This means that the feedback from the output is mixed with psu noise which in turn means the amp is trying to correct the signal at the output as well as the signals on the psu. Is this really what the designer desires?
BAM
PS: Like the seasonal diyaudio logo, Pass et al!
I shall repeat an old axe of mine, 20kHz is the highest frequency you can hear. For a typical household-powered amp this equates to 5V/us or so. 100V/us is unecessary IMO.
KISS stands for "Keep It Simple Stupid". In general this means keep it as simple as you can whilst achieving the goal. Sometimes this can be overdone and you end up with a stupidly simple design. The great designs often look simple but it is the unseen aspects that are very complicated and lead to high performance.
The posted design looks a little too KISS to me. For example, the resitive bias chains for the input BJT emitters are attached to the power supplies, in fact to opposite power supplies. This means that the feedback from the output is mixed with psu noise which in turn means the amp is trying to correct the signal at the output as well as the signals on the psu. Is this really what the designer desires?
BAM
PS: Like the seasonal diyaudio logo, Pass et al!
Cascoded Jfet inputs
Jfet are mostly lmimted to voltages like 25-30 volts
Therefor it is necessary to cascode them for higher
supply voltages. A protection.
From this comes a "tradition" to think that cascode
should be always good or necessary. How wrong!
Bipolar smallsignal transistors is availeable
for much higher voltages than JFETS
I plan on buy me some 100red of a ra- transistors.
ra = rausch arm (German) = low noise
they have 120 volts specification.
and low noise and high Ft MHZ
so I do not have to cascode - and I do not want to!
Bad Habits, based on lack of knowledge
Monkey see monkey do
/halo - still happy simply uncascoded
Jfet are mostly lmimted to voltages like 25-30 volts
Therefor it is necessary to cascode them for higher
supply voltages. A protection.
From this comes a "tradition" to think that cascode
should be always good or necessary. How wrong!
Bipolar smallsignal transistors is availeable
for much higher voltages than JFETS
I plan on buy me some 100red of a ra- transistors.
ra = rausch arm (German) = low noise
they have 120 volts specification.
and low noise and high Ft MHZ
so I do not have to cascode - and I do not want to!
Bad Habits, based on lack of knowledge
Monkey see monkey do
/halo - still happy simply uncascoded
Hi all,
Phase_accurate:
Maybe that would be better! Could you draw such a circuit?
dshortt9:
A BJT output pair can very well work. A single output BJT will have too little gain but a darlignton or CFP with gain of 1000 will only need 3mA drive for 3A output.. that could work...? As for FET for the input, I don't have good models for that so I don't try.
traderbam:
Good point! Didn't think of that before, but than again, the input BJT pair draws less than 1 mA each, so a simple 1KOhm/1000uF will take care of that? The circuit as it is now has no such things and is only conceptual. But still very good point, it is a inherent feature of the design.
but about cascodes?
gr,
Thijs
Phase_accurate:
Maybe that would be better! Could you draw such a circuit?
dshortt9:
A BJT output pair can very well work. A single output BJT will have too little gain but a darlignton or CFP with gain of 1000 will only need 3mA drive for 3A output.. that could work...? As for FET for the input, I don't have good models for that so I don't try.
traderbam:
Good point! Didn't think of that before, but than again, the input BJT pair draws less than 1 mA each, so a simple 1KOhm/1000uF will take care of that? The circuit as it is now has no such things and is only conceptual. But still very good point, it is a inherent feature of the design.
but about cascodes?
Hmmmmmm, a cascode will be more linear than a non cascode, even it the AC current goes though two instead of one transistor!
gr,
Thijs
With a BJT that is operated as a grounded emittor amplifier sans bootstrapped cascoding, you certainly have the famed Miller effect to contend with. Note the changes in the Vcb (collector-base voltage) will also result in modulation of Cob (collector output capacitance) and Cre (reverse transfer capacitance). Even if a constant Ic (collector current) is maintained, the Vbe (base-emittor voltage) will vary according to the changing Vce (collector emittor voltage), known as the Early effect.
Bootstrapped cascoding maintains a fixed voltage relationship between base, emittor and collector, thereby avoiding the aforementioned problems and removing some major sources of distortion.
regards, jonathan carr
Bootstrapped cascoding maintains a fixed voltage relationship between base, emittor and collector, thereby avoiding the aforementioned problems and removing some major sources of distortion.
regards, jonathan carr
You go cascoding your input pair
/halojoy will not need it - at any time
- still his amplifier does very good - belive it or not!
halomatics - SCHEMATICS for fanatics
I guess you need it.jcarr said:
/halojoy will not need it - at any time
- still his amplifier does very good - belive it or not!
halomatics - SCHEMATICS for fanatics
The slew rate of a sinewave in volts per *second* at the zero crossing point is 2 x pi x F x volts peak to peak. So if the outer limits for a domestic amplifier rails are + / - 100v rails 😎 then at 20kHz a sinewave (which it will be if it is coming from a cd player) will equate to 25.13 V/uS. But of course your tweeters and your ears will probably have caught on fire by then. For more normal domestic situations the slew rate will be a lot less as traderbam says. 25 years ago I had a listen to an Amcron / Crown DC300A driving some RTR electrostatics with a direct cut disc and a Dynavector straingauge cartridge. The DC300A only had a slew rate of 0.5 V/us IIRC (long time ago!) and it sounded just fine.traderbam said:
- Status
- Not open for further replies.