Cascoding the folded cascode

[nelsonvandal]

It's been discussed, but I haven't seen it done so I had to try it. In spite of the messy appearance it's stable, low noise and measures well. Measured with RMAA THD is lower compared to the standard arrangement and mostly it's the 2nd harmonic that's reduced. I like the sound of it and think it's an improvement. It makes the sound cleaner and a little bit leaner.

[nelsonvandal]

These are the RMAA results. Leftmost is the standard circuit (AD823-like), the cascoded folded cascode in the middle and loopback rightmost.

[WuYit]

Hi,
I`m sorry to say that there´s no sensible reason to do that.

[AndrewT]

Looks like a nice headphone amplifier.
160mW of ClassA into your 300r phones.

Are j1 & 9 required?

[Zen Mod]

input jfets are voltage starving , for my liking

didn't look further ....

[nelsonvandal]

Originally Posted by AndrewT

Are j1 & 9 required?

Not required but wanted. I've tried all four input stage concepts - standard, CFP, cascoded and CFP+cascode - and this it what I prefer.

WuYit, yes it's probably esthetically displeasing but I like it. The standard common base VAS sounds a bit rough and muddy. The AD797-like arrangement and this sound better to my ears. I find it has about the same effect as cascoding the VAS in a LIN-type of amp.

Zen Mod, I can try it with a zener instead of the LED, but I't will be hard to tell if it's to the better. The amp adds very little to the sound (when I evaluate amps I usually add them to the signal path). What voltage would be appropriate?

[MiiB]

the SK170bl runs best at app 5-7 mV and with app 12 V sving.....(my liking)

[jcx]

some curious choices there

cfp around high current, low noise input jfet, setting the ~10mA Idss jfet Id at 1.5 mA, adding ~2-3x degeneration re the naked fet's 1/gm - both seem to say you actually don't trust the linearity/square law cancellation of the input jfet's themselves

the diff pair's "cascode" jfets add greater input Vgd 1/gm modulation than the drain R would without the folded cascode - I usually expect a added cascode to reduce the Vswing on the cascoded Q - leaving them out entirely would be better by that criteria


having used the cfp front end you don't use it where it would reduce distortion the most - in a Class A biased output stage


I would move the cfp to the cascoding jfets where the gm boost makes the cascode actually work to reduce input Vds change with common mode signal

and I would use cfp for the output - when used in Class A

and +/-12 V isn't enough for dynamic headroom and linearizing Vbias - I wouldn't want output device Vcb to go below 5 V - as long as it's a descrete circuit why not at least +/-24 V supplies? - with those jfet ccs on the output diamond buffer you need the extra bias V beyond usable output Vswing

at +/-12 Vsupply you could more easily just use an op amp

also I hate selecting parts so I would never use jfet ccs when a "ring of two" bjt feedback ccs just works, on less voltage, with greater output Z, the 1st time with the calculated values - no trimming

[nelsonvandal]

Originally Posted by MiiB

the SK170bl runs best at app 5-7 mV and with app 12 V sving.....(my liking)

Thank you!

What is the appropriate voltage over a BJT, SC2362? I use it in the same manner as the SK170 in the ground channel input stage.

[WuYit]

Nelson,
no, it is theoretically displeasing. The role of J1/J9 is exactly the same as that of T5/T2. Why devices with opposite characteristics? One of them is not suitable. Consequently, why don`t you want to cascode the input stage cascode as well?
There´s nothing that would make the standard common base VAS rough and muddy (when the voltage gain is kept at a reasonable level).