Folded cascode headphone amp

[roender]

Jonathan Carr, a guru in folded cascode topology

http://www.diyaudio.com/forums/showthread.php?postid=119516#post119516
http://www.diyaudio.com/forums/showthread.php?postid=117260#post117260
http://www.diyaudio.com/forums/showthread.php?postid=711676#post711676

 

[nelsonvandal]

So I guess Diamante is a very clever amp with it's current sources on the LTP and improved Wilson current mirror, trimable bias and DC-offset.

 

[lineup]

I have chosen ~0.6 Volt drop across those 300 Ohm resistors
Because 0.5 mA x 1221 Ohm = 0.6 Volt.
The BC560C, U7, will give some temp compensation.

Regards, Lineup

Hi Lineup,
is the LTP tail current =4mA?

My attached detail of my JFET input AD797:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1528754&stamp=1212581188
( From 2006, when there wasn't at all as much hype about this great Op-Amp )

Each of the U1, U2, U3 and U4 runs at 1 mA.
For simpicity

The sum, 4 mA is provided across the two 300 Ohm resistors.

600 mV across 300 Ohm = 2 mA


Regards, Lineup

 

[nelsonvandal]

My attached detail of my JFET input AD797:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1528754&stamp=1212581188
( From 2006, when there wasn't at all as much hype about this great Op-Amp )

Each of the U1, U2, U3 and U4 runs at 1 mA.
For simpicity

The sum, 4 mA is provided across the two 300 Ohm resistors.

600 mV across 300 Ohm = 2 mA


Regards, Lineup

I only get ~130 uA through U3 and U4 when simulating with theese values. But then I don't use a CCS to sink current from the cascodes like in AD797.

 

[Lumba Ogir]

jam,

The folded cascode is one of the best sounding gain stages I know.

What is the reason for that?

nelsonvandal,

If I remove the degeneration resistors I get increased speed

It does not make sense.

lineup,

He uses neither gate or source resistors, unless it is absolutely necessary!

Not very wise.

Uses 2SK117, almost identical data to 2SK170, except from that 2SK117 can take 50 Volt.

Not very identical.

 

[nelsonvandal]

jam,
What is the reason for that?

I don't know how Jam came to that conclusion, but I'm thinking of how monolithic opamps sound. AD797, ADA4841 and AD829 are very good sounding and the LT1469 is said to sound very good as well. I suppose ADA4899-1 also uses folded cascode, and it sounds great as well. It wouldn't surprise me if AD8599 also has this topology. I've been a bipolar fan, but after long time listening, there's some kind of musicality, or PRAT as some call it, that is lacking. When using JFET's as input stage something comes alive, but at the price of less clarity. That's why I'm exploring this topology now. And I really like the concept of fewer stages.

nelsonvandal,

It does not make sense.

It does to LTSpice. Dosen't this increase the OLG and BW?

Everything seems more stable when using these resistors, and I'm quite sure I'm going to use them. I don't think I want more OLG and feedback. When using diamond buffer output stage, I get low output impedance anyway.

I'd be happy if you develop your oppinion, so we can all be wiser. I think that's the main point with these discussion forums.

 

[jam]

Lumba Orgi,

The reason I think that the folded cascode sounds so good is in a word simplicity..............for the most part the folded cascode models as one stage rather than two..........the current mirror provides symetrical pull up and pull down..............even though the parts count may be relatively high the signal path is fairly simple. Just a guess mind you.

Just as nelsonvandal has pointed out some of the best sounding opamps use this topology. I have built discreet versions with excellent results.

Jam

 

[jam]

This is a folded cascode amplifier I designed and posted in the other forum, based on work by Scott Wurcer, Stan Warren and John Curl.

Jam

 

[nelsonvandal]

Do you have a link? Are there any schematic with the parts specified? Even these gurus use resistors on the LTP, just one LED and resistors for biasing the cascode, and there are degenerating resistors on the LTP (even though it's mosfets and not jfets). That's good news, since I want as low part count as possible and a stable circuit with some margins.

 

[lineup]

I only get ~130 uA through U3 and U4 when simulating with theese values.
But then I don't use a CCS to sink current from the cascodes like in AD797.

then you have not done like my schematic.
http://www.diyaudio.com/forums/attachment.php?s=&postid=1528754&stamp=1212581188

It is pure mathematics.

The voltage at U3 and U4 bases is:
= 0.5 mA x 1221 Ohm + the voltage drop across U7 (BC560C also)
= 0.6 + ~0.65 gives ~ 1,25 Volt

Drop across both 300 Ohms resistors will be 1.25 - Vbe of U3/U4
which will be same voltage as across R1, 1221 Ohm
Which is 0.6 Volt.
0.6 / 300 Ohm = 2 mA.

As the input transistors works at 1 mA,
the left over is 1 mA, for each of U3 and U4.

By adusting this resistor R1, it is possible to fine-tune the current through U3/U4. (this is why the odd value of 1221 Ohm)

 

[Lumba Ogir]

nelsonvandal,
the folded cascode arrangement does not place the JFETs into a more linear working region, does not decrease the highly destructive Miller capacitances, thus does not extend the dynamic bandwidth, does not provide a more favourable load nor isolation and does not improve power supply ripple rejection.
It does not achieve much of value.

No source resistors most importantly means strongly increased interelectrode capacitances compared to typical conditions like 2mA/side and Rs 100Ohm.

 

[Fanuc]

nelsonvandal,
the folded cascode arrangement does not place the JFETs into a more linear working region, does not decrease the highly destructive Miller capacitances, thus does not extend the dynamic bandwidth, does not provide a more favourable load nor isolation and does not improve power supply ripple rejection.
It does not achieve much of value.

No source resistors most importantly means strongly increased interelectrode capacitances compared to typical conditions like 2mA/side and Rs 100Ohm.

I agree. I was thinking this yesterday. I don't see a folded cascode as a "cascode".

 

[jam]

Yes...............but it sounds good, does not require much compensation, has a simpler signal path, has better slewing, good bandwidth............than the traditional topology.

Lower distortion does not mean better sound.

Jam

 

[roender]

Yes...............but it sounds good ...

Lower distortion does not mean better sound.


Jam

Very true!

 

[nelsonvandal]

Yes...............but it sounds good, does not require much compensation, has a simpler signal path, has better slewing, good bandwidth............than the traditional topology.

Lower distortion does not mean better sound.

Jam

I know there was a discussion a couple of years ago of how a folded cascode sounds. Some like it and some don't, but there was no real description of it, and how it compares to other topologies.

The opamps mentioned above AD797, ADA4841 and AD829 has a kind of dark and smooth sound, very enjoyable, but maybe a bit too "anti-aggressive". I don't think it's a "house-sound" of Analog Devices, since their JFET-input amps are rather the opposite, like AD8620, AD823 and AD744. The bipolar AD8397 also has a forward and a bit aggressive sound.

I've never heard a discrete folded cascode. How would you describe the sound of it, is there a "folded cascode sound"?

I think there is a difference in sound using bibolars or JFETs as input transistors. I find JFETs a bit grainy but warm, with more "body", maybe more musical with a natural flow. Bipolars sound "cleaner". This is my experience from monolithic opamps, but I've also changed from bipolar input stage to JFET input stage in discrete amps with the same result.

 

[Onvinyl]

Its thrilling, how different the perceptions are. I found the AD797 neither too agressive nor too lush. A good example for a too lush sounding opamp is the OPA627. A typical 'too aggressive' type is (for me) the op27 or, even more so, the OPA604.

I'd like to discuss the folded cascode in a more technical way. Roender gave links to some excellent earlier threads.
Does, for example, a folded cascode not fix the voltage over the cascoded device? Then the term 'cascode' may indeed be misleading.
Rüdiger

 

[TzeYang]

Err guys, I may seem to be hijacking this thread.

I've decided to ditch TINA Spice, for some odd reason whenever i simulate something the results are always inconsistent. I build something today and it'll work. I build something exactly the same tomorrow and it fails. Heck such buggy software. Hence I've decided to give LTSpice a try instead.

Anyhow, I have a folded cascode circuit in mind that i would like to try simulate. However I lack the proper spice models. Does anyone here have PN4392 2N5088 2N5087 MJE243 MJE253 BD139 BD140 spice models? Could not really find any reliable ones on the internet.

Also, NelsonVandal, how did you simulate the following with your circuit:

1) Generate Square Waves for stability/slew rate test.
2) Frequency Response.
3) DC Offset measuring.
4) Open loop gain vs bandwidth graph.
5) Generate Sine Wave and phase response.

TINE Spice I admit, is ridiculously easy to use, comparing to LT Spice.

Thanks

 

[Onvinyl]

Hi!

Here is the 139:
* BD135/137/139 NPN EPITAXIAL SILICON TRANSISTOR

*------------------------------------------------------

* MEDIUM POWER LINEAR AND SWITCHING APPLICATIONS

* Based on BD135-16

* BVcbo: BD135 45V / BD137 60V / BD139 80V

* BVceo: BD135 45V / BD137 60V / BD139 80V

* BVebo: 5V

* Complement to BD136, BD138 and BD140 respectively

*------------------------------------------------------

.MODEL BD135_137_139 NPN (

+ IS =2.3985E-13

+ BF =244.9

+ NF =1.0

+ BR =78.11

+ NR =1.007

+ ISE =1.0471E-14

+ NE =1.2

+ ISC =1.9314E-11

+ NC =1.45

+ VAF =98.5

+ VAR =7.46

+ IKF =1.1863

+ IKR =0.1445

+ RB =2.14

+ RBM =0.001

+ IRB =0.031

+ RE =0.0832

+ RC =0.01

+ CJE =2.92702E-10

+ VJE =0.67412

+ MJE =0.3300

+ FC =0.5

+ CJC =4.8831E-11

+ VJC =0.5258

+ MJC =0.3928

+ XCJC =0.5287

+ XTB =1.1398

+ EG =1.2105

+ XTI =3.0 )

*-------------------------------------------------------

* FAIRCHILD CASE: TO-126 PID:BD135-16

* 2000-03-20 CREATION

* BD136/138/140 PNP EPITAXIAL SILICON TRANSISTOR

*------------------------------------------------------

And here the BD140:

* MEDIUM POWER LINEAR AND SWITCHING APPLICATIONS

* Based on BD140-16

* BVcbo: BD135 -45V / BD137 -60V / BD139 -80V

* BVceo: BD135 -45V / BD137 -60V / BD139 -80V

* BVebo: -5V

* Complement to BD135, BD137 and BD139 respectively

*------------------------------------------------------

.MODEL BD136_138_140 PNP (

+ IS =2.9537E-13

+ BF =201.4

+ NF =1.0

+ BR =23.765

+ NR =1.021

+ ISE =1.8002E-13

+ NE =1.5

+ ISC =7.0433E-12

+ NC =1.38

+ VAF =137.0

+ VAR =8.41

+ IKF =1.0993

+ IKR =0.10

+ RB =1.98

+ RBM =0.01

+ IRB =0.011

+ RE =0.1109

+ RC =0.01

+ CJE =2.1982E-10

+ VJE =0.7211

+ MJE =0.3685

+ FC =0.5

+ CJC =6.8291E-11

+ VJC =0.5499

+ MJC =0.3668

+ XCJC =0.5287

+ XTB =1.4883

+ EG =1.2343

+ XTI =3.0 )

*-------------------------------------------------------

* FAIRCHILD CASE: TO-126 PID:BD140-16

* 2000-03-20 CREATION
+MJC=0.415

 

[Lumba Ogir]

Nelson,

I find JFETs a bit grainy

Not when used adequately.

 

[jam]

nelsomvandal,

I would not compare ic's to discreets as there are huge differences in sound. I once built a discreet opamp which ended up sounding far better than the ic it was based on.
Lumba Ogir is correct about the use of fets, they sometimes need to be biased slightly higher than a bipolar for optimum results.
I would go ahead and build a discreet folded cascode amplifier and experiment with the circuit to see what it can do.

Jam