hmm well .. when you use the fast BJT's the output becomes less sensitive to the high frequencies of the amplifier. On the other hand, you start getting fatal oscillations in the output stage itself.
Unless you were to use an ultra fast driver transistor. (>100 MHz)
There may be some way around this, i just havent found it. Except going back to darlington.
If i was building an amp for my self i would use Sziklai output any day of the week. But for products, i wouldnt feel safe using it.
Customers would be calling 5 days a week complaining about their burn out amplifier. In that case i would play it safe, and work on getting good sound out of a darligton chain instead.
ChocoHolic said:
The topology I used was a basic standard Lin type, with bootstrapped VAS. This gives the upper half of the output stage the ability to be driven by voltages above the rail, but the lower half cannot. Thus if driving above the rail brought any great advantages, this would be shown on the Darlington and Sziklai plots in particular, as the upper half performing much greater swing ability.
I chose not to use any 'emitter' resistors because I was attempting to find out the very maximum swing that the various OPS configurations could produce.
Hi, LC
Usually, we use small device for differential pair (fastest, eg:100Mhz), bigger device for VAS (eg:50Mhz) and biggest device for final stage (slowest, eg:20Mhz).
I've been thinking about this. Imagine differential is at left, VAS at middle, output stage at right at drawing. If we use the fastest device for the most left device and the slowest device for the most right device, it should be prone of oscilation. (note : this is only idea, not tested to bode plot or SIM)
Why not try the non-usual configuration. Lets say putting slowest device for the differential, slightly faster for VAS and fastest device for the output stage. This should be more difficult to oscilate (I think....)
Say, MJE340 for differentials, MJE350 for VAS and C2922 pairs for output stage.
It's no use to use fastest device available without considering its position, when in the final schematic, the compensation caps will slow this fastest device like snail. Why don't use slower device in the first place?
Usually, we use small device for differential pair (fastest, eg:100Mhz), bigger device for VAS (eg:50Mhz) and biggest device for final stage (slowest, eg:20Mhz).
I've been thinking about this. Imagine differential is at left, VAS at middle, output stage at right at drawing. If we use the fastest device for the most left device and the slowest device for the most right device, it should be prone of oscilation. (note : this is only idea, not tested to bode plot or SIM)
Why not try the non-usual configuration. Lets say putting slowest device for the differential, slightly faster for VAS and fastest device for the output stage. This should be more difficult to oscilate (I think....
)Say, MJE340 for differentials, MJE350 for VAS and C2922 pairs for output stage.
It's no use to use fastest device available without considering its position, when in the final schematic, the compensation caps will slow this fastest device like snail. Why don't use slower device in the first place?
...hm, high VAS output should help more for the MosFet output.
But I think the strentgh of the Szilkai would be more effective, when the VAS cannot reach the rail at all (VAS out about 1V...5V less than rail).
For me this thread is exactly rolling up a lot of questions, that I never
examined detailed enough. Thanks to all for the helpful input!
....will curiously follow this thread tommorow evening again.
It's already late here.
Good n8 guys , - and lady of course (!!)
Markus
But I think the strentgh of the Szilkai would be more effective, when the VAS cannot reach the rail at all (VAS out about 1V...5V less than rail).
For me this thread is exactly rolling up a lot of questions, that I never
examined detailed enough. Thanks to all for the helpful input!
....will curiously follow this thread tommorow evening again.
It's already late here.
Good n8 guys , - and lady of course (!!)
Markus
The point of using Ft>30Mhz output transistors is that they require less base charge storage (and thus less charge removal before achieving turn-off) at high frequencies. The amount of charge stored in the base increases as base current is increased, so high hFE up to high frequencies is a very interesting feature for any bipolar transistor. The smaller the drive current, the better they will perform.
In a CFP, the only way for the base charge of the output devices to be removed is a bare BE resistor, so that topology benefits dramatically from the use of high-Ft devices. If they cause oscillation, clever compensation will cure it anyway.
In a CFP, the only way for the base charge of the output devices to be removed is a bare BE resistor, so that topology benefits dramatically from the use of high-Ft devices. If they cause oscillation, clever compensation will cure it anyway.
Eva said:
I can agree with you Eva. I built a CFP amp with 30Mhz outputs and drivers, and they oscillated. However, it was easily cured with 100 ohm driver base stopper resistors and 1nf B-E compensation caps. BTW, you were the person that told me to add the base stoppers in an earlier Forum thread. CFP is usable with fast devices, but I think you have to add compensation with them, vs darlington, where I've got away with using no compensation.
I've also worked with low voltage amps, (12V or less, battery type stuff) and found that CFP sounds louder at those lower voltages, and also much easier to bias, because of the only 2 B-E volt drops.
Hi Richie,
your test of output types is completely valid and you have specified how you carried it out. I would expect it to be repeatable by anyone who cares to re-do the comparison.
However I believe you have set up the test and display of the results in a form that masks the differences between the devices.
On that basis, I think you are drawing a somewhat sweeping and invalid conclusion from your test.
your test of output types is completely valid and you have specified how you carried it out. I would expect it to be repeatable by anyone who cares to re-do the comparison.
However I believe you have set up the test and display of the results in a form that masks the differences between the devices.
On that basis, I think you are drawing a somewhat sweeping and invalid conclusion from your test.
Andrew, thanks for your comments. I agree there may be other reasons than swing to choose an output stage or device. I was just attempting to show swing as I was investigating designing an efficient an amp as possible.
Or did you mean some other thing by that statement? If so please enlighten me what is wrong with the method of presentation and even examination.
Or did you mean some other thing by that statement? If so please enlighten me what is wrong with the method of presentation and even examination.
Hm, so you also simulated a VAS, which cannot swing up to full rail level?
My guess was that this would be the strength of the sziklai, because the base drive of the output device is related to the rail and its voltage drop is not adding to the VAS drop anymore.
From this Sziklai should allow 0.7V more output swing than a darlington.
Sziklai with additional gain for me does not look worse in oscillation, but it probably cannot compete at all in terms of output impedance.
Especially when driven from a low impedance VAS, the normal Sziklai should offer much lower output impedances than the gain version.
....OK, there is another topic on my ToDo list.... "Playing arround with Sziklai"...
My guess was that this would be the strength of the sziklai, because the base drive of the output device is related to the rail and its voltage drop is not adding to the VAS drop anymore.
From this Sziklai should allow 0.7V more output swing than a darlington.
Sziklai with additional gain for me does not look worse in oscillation, but it probably cannot compete at all in terms of output impedance.
Especially when driven from a low impedance VAS, the normal Sziklai should offer much lower output impedances than the gain version.
....OK, there is another topic on my ToDo list.... "Playing arround with Sziklai"...
Hi Richie,
I meant exactly that.
The conclusion that all output stages except Nfet are equal in efficiency is misleading.
Your results appear to support this.
I believe there must be a better way to measure the OUTPUT STAGE without the limit from the driver stage or Vrail limits masking what you are claiming to be investigating.
I meant exactly that.
The conclusion that all output stages except Nfet are equal in efficiency is misleading.
Your results appear to support this.
I believe there must be a better way to measure the OUTPUT STAGE without the limit from the driver stage or Vrail limits masking what you are claiming to be investigating.
No Markus, the VAS could swing above the rail as it was bootstrapped, I already told you Thereby it can drive the device as high as it needs.
I wanted to design an efficient amp, but 0.7V extra drop of Darlington does not bother me, it's nothing in practical terms.
I don't think the worse output Z of the Sziklai with gain would worry me, it will still be pretty low. I'm assuming only small gains though - less than 2 (which would be plenty?).
EDIT: just seen your response Andrew. I tried to negate the VAS limit out of the equation, please see my answer to Markus above.
Thereby it can drive the device as high as it needs.I wanted to design an efficient amp, but 0.7V extra drop of Darlington does not bother me, it's nothing in practical terms.
I don't think the worse output Z of the Sziklai with gain would worry me, it will still be pretty low. I'm assuming only small gains though - less than 2 (which would be plenty?).
EDIT: just seen your response Andrew. I tried to negate the VAS limit out of the equation, please see my answer to Markus above.
our dreams?
N-ch output devices
A class AB all N-ch Vertical mosfet output stage properly implemented generally approaches the theoretical ideal of efficiency to a closer degree than either bipolar or darlington configurations. This is so when you consider that modern vertical power mosfets (like the IRF540) require less drive power and because they dont exhibit appreciable storage effects they can fully saturate, unlike bipolars which need hefty drive currents and antisaturation circuits to avoid large rail to rail currents. This is especially so when load impedances approach 2 ohms. More than a few manufacturers of autosound amplifiers have recently developed all n-ch output stages using vertical mosfets in their class ab output stage designs. (MTX Audio and JL Audio) Rockford Fosgates Punch 45 (circa 1988 ) Autosound Power Amplifier was the first successful implementation of a rugged, simple hi performance all N-ch output stage/driver design in terms of high volume consumer electronics product.
A class AB all N-ch Vertical mosfet output stage properly implemented generally approaches the theoretical ideal of efficiency to a closer degree than either bipolar or darlington configurations. This is so when you consider that modern vertical power mosfets (like the IRF540) require less drive power and because they dont exhibit appreciable storage effects they can fully saturate, unlike bipolars which need hefty drive currents and antisaturation circuits to avoid large rail to rail currents. This is especially so when load impedances approach 2 ohms. More than a few manufacturers of autosound amplifiers have recently developed all n-ch output stages using vertical mosfets in their class ab output stage designs. (MTX Audio and JL Audio) Rockford Fosgates Punch 45 (circa 1988 ) Autosound Power Amplifier was the first successful implementation of a rugged, simple hi performance all N-ch output stage/driver design in terms of high volume consumer electronics product.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.