ccs-loaded sziklai

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
While designing an output stage i had to face the choice between EF or CFP, but when i finally dediced for the CFP one i had an idea:

Could the output stage driver collector resistor be replaced with a ccs? Why? This should give two nice features: Make the driver transistor run hotter (and have less largue signal distortion) and improove the turnoff speed.

Since i have never seen something like this in any design, is there any reason why this shouldn't work well (I assume it will need a bit of compensation).

With a mosfet output the gain developed by this arrangement would be incredible and if some brave one makes it stable it could provide super-low open-loop distortion.

Could someone give me... feedback? :D
 
darkfenriz said:
DC bias conditions will be not very well defined, moreover a typical ccs may need around a Volt to stay in active region, not really a problem with mosfets but hard with bipolar output devices.


I planned to get the voltage with a diode on the output emmiter or simply by powering the driver with slightly higher voltage (i usually power the small-signal stages from a cheap $15 transformer as this makes it much more stable).

I had always thought that the current depended on the collector resistance (the one from the power transistor R1-R33). I don't see this as an issue.
 
Not exactly CCS replacing collector resistor, but adding totem pole after collector resistor may improve turnon/turnoff of mosfet CFP
 

Attachments

  • rf3.gif
    rf3.gif
    18.8 KB · Views: 1,405
Hi Ionomolo
---simply by powering the driver with slightly higher voltage---

Without using a CCS but a simple resistor in a classical Sziklai configuration, powering the driver from a slightly higher voltage but clamped to the voltage of the power supply of the power transistor, should have the benefit to be able to reverse bias it, just like it is done in the standard four emitter followers push-pull by the capacitor in parallel with the resistor between the emitters of the drivers. It should give slightly lower HF distorsion.
http://www.diyaudio.com/forums/showthread.php?postid=1536092#post1536092
I think that Douglas Self had a short sentence about this possibility.
 
Originally posted by AKSA
It will work well, with very high OLG as you say and vanishingly low Zout. Use a 220R gate stopper, and choose around 5mA on the npn driver with a 9140 mosfet as output device.

Run it in Class A only.

Hugh

Thanks sincerely. I have a couple of IRFP9240 here that might work.

I have just noticed that in CFP configuration the Cgs is not bootstraped by the load as it happens on the source follower. With a capacitance arround 1.4 nF and a gate resistor of 200 Ohms this would have a -3dB corner at 560 KHz, which makes it comparable to the slowest bjt's. Since the swing of the bjt may be several times larger than the needed for full output at DC (assume 1.5V to get 6A at 4S transconductance) this may allow it to keep amplifying one decade further to 5MHz, which is enough bandwith as long as the phase shift does not make the whole thing crazy, but it does not allow much headroom for compensation.

I'm correct? Is there any workaround to increase output bandwith?
 
Since the use of mosfets will give 4V of margin for the ccs, would it be imprescindible to cascode?

I think that a ccs made with 2 transistors may become inductive with frequency, and a RLC network inside a circuit with high gain makes a power oscillator for sure. Are the diode-biased ccs's better in that sense? would cascoding solve that problem? does the problem exist at all?
 
I will give a look at the boostrap idea to check if it can go higher frequencies or perhaps compensate somehow the gate stopper + gate capacitance pole. I have no idea if this is possible so don't take it as a statement but only as a "thing i might check"

Thanks!
 
Ion,

Use a 5mA CCS to power the circuit earlier described, and run 1.5A through the device with 0.47R source degeneration to reduce transconductance (and hence OLG) slightly. A 1.5A CCS below the CFP is required to sink the standing current to ground, of course. Measured Zout is around 80 milliohms and 3dB points are at 18Hz and 75KHz, with Zout halved if you use double the number of mosfets top and bottom.

You do not need to worry about bandwidth. Measured and subjective speed is astonishing. The challenge is how you will drive the bipolar driver. I use a tube with cap coupling. Speaker is cap coupled too, using around 2,200uF.

Hugh
 
AKSA said:
Ion,

Use a 5mA CCS to power the circuit earlier described, and run 1.5A through the device with 0.47R source degeneration to reduce transconductance (and hence OLG) slightly. A 1.5A CCS below the CFP is required to sink the standing current to ground, of course. Measured Zout is around 80 milliohms and 3dB points are at 18Hz and 75KHz, with Zout halved if you use double the number of mosfets top and bottom.

You do not need to worry about bandwidth. Measured and subjective speed is astonishing. The challenge is how you will drive the bipolar driver. I use a tube with cap coupling. Speaker is cap coupled too, using around 2,200uF.

Hugh

75 Khz (I assume closed loop) is more than enough for an amp without global nfb, but far too low if some feedback is applied.

Did you measure the slew rate? Was yours built with mosfets similar to the IRFP9240?

I have all the requiered parts on hand, so i will do an compare it with the opamp driven bjt sziklai. I have just realized that i did ask a couple of npn's for it instead of one pnp and one npn! :bawling: I will have to look for a driver transistor locally.
 
Ion,

With just 80 milliohm Zout you do NOT need gnfb - not required. And with the OLG of the bipolar driver CCS loaded the local fb factor is probably about 70dB anyway.....

Any voltage amp you use in front of this circuit block would need only local feedback, such as an opamp. A tube would NOT need fb, of course.

Slew rate is better than 20V/uS. More than enough for audio. Don't get hung up on the figures. Build the damn thing.

Hugh
 
Hi Hugh,

Next time you are passing thru please let me know, lunch is on me.

I was thinking of using a mosfet bipolar Sziklai output stage in an amp that did not use a global feedback loop.
I am not sure if Steve McCormack's amp used this topology or a standard follower.
I guess a schematic of the output stage would help.

Regards,
Jam
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.