MPP

[Stefanoo]

Thanks Stefanoo. That looks like it works. You use emitter resistors, i asume to lower distortion and input offset.
I think we can do two things :
One is to implement this input stage into the Paradise, what you have done.
I think a lot of people whould apreciate this option.
This could be then done as a small aditional board. As far as i can see a lot of the existing Paradise can be used.

The other is to make a fully balanced stage. For that we would need a totally new layout. Maybe we can put that version into the Masterpiece thread.
The problem is that it is not that easy to follow this new input stage with a balanced transimpedance RIAA. What i would do instead is make a shunt feedback RIAA with the OPA1632. This is a fully balanced chip. It is used by Rowland and also by Mark Levinson in their most expensive stuff. I do not like to open a discussion if that equipment sounds good. I can only tell you that this chip sounds excellent and is very hard to beat with a discrete design.

wonderful idea, we could open the balance version on the Masterpiece if you would like.
Yes the emitter resistors were added to lower input offset and likely lower distortion by adding a little local feedback.
Anyway I will post a version same topology but with the K170 and J74.
It looks like gain-wise it doesn't even need paralleling, which is pretty amazing and I wonder if this will work in practice the way is designed here.
It seems to good to be true

 

[Joachim Gerhard]

That is nice way too. I think the gain comes also from the cascode. On the other hand the Paradise is setup for an input bias of 10mA and this JFets with 8mA Idss are close to this.

 

[Joachim Gerhard]

I think when we use emitter resistors 22 Ohm is a bit high because of the Johnson noise they ad. Maybe 10 Ohm is a compromise.

 

[MiiB]

Hey, that is my fet version of the paradise.. Mikkel built it to me as a way to avoid in cap in the signal path, It's not as detailed and open as the original, but solves two issues, one is the cap (that I don't consider a problem), the other is the dc at the input, somthing the balanced input also would solve. To me the jfet's sound a little wailed and unfocused compared to the bipolar's, but then again could be my preference playing a role.

 

[Joachim Gerhard]

The transimpedance version with Fets ?

 

[FdW]

Here you can learn about peaking in mirrors and how to reduce it ( Barry Gilbert ).

I think I can improve on this this evening.

 

[Joachim Gerhard]

Ok, fine.
The peaking is one thing. The other is that the mirrors should have a high output impedance and the offset at the input from plus in to minus in should be low.

 

[MiiB]

we have tried both, current and voltage input.

 

[Bksabath]

I think I can improve on this this evening.

Just to say I am paying attention
Paper mention small cap on Q1 ?

I see that a number of things can be done with the conveyour topology Integrators and differentiator and Filters Is there a possibility to make servo?

 

[Joachim Gerhard]

Yes, a servo can be made. We simulated the non overshot mirror but it had more distortion. Frans is working on a better solution.
This needs a little study so please be patient.

 

[Stefanoo]

Hey, that is my fet version of the paradise.. Mikkel built it to me as a way to avoid in cap in the signal path, It's not as detailed and open as the original, but solves two issues, one is the cap (that I don't consider a problem), the other is the dc at the input, somthing the balanced input also would solve. To me the jfet's sound a little wailed and unfocused compared to the bipolar's, but then again could be my preference playing a role.

too bad you didn't post it back then uh!?

 

[Stefanoo]

I think when we use emitter resistors 22 Ohm is a bit high because of the Johnson noise they ad. Maybe 10 Ohm is a compromise.

fully agree.
Do we want to persue the Jfet version or the BJT's?

 

[Joachim Gerhard]

Stefanoo, you may not imagine what we do in the back round in our "real" time.
It must be hundreds of circuits i tried, also MiiB, Frans, Hesener and other contributors here. We are in contact with most of them. This is already an obsession.
Usually we filter out what is not doable for DIY or is under NDA.
We can make an all out approach in the Masterpiece, i have no problem with that.
We can throw in expensive parts, use high temperature RF material for the PCB, gold plate the PCB, do something special mechanical. There we can grind our ax.
I think the Paradise already serves 95% of the DIY comunity and the balanced input option will make it even better i hope.

 

[Joachim Gerhard]

For the Paradise i would like to go BJT.
It is nearly imposible to find enough low noise P channel J-Fets.
If the new input circuit gets half as popular as the Paradise R3 we are in trouble.
In the masterpiece you can decide. Maybe J-Fets serves your taste better.
I usually prefer them in tonality.

 

[Joachim Gerhard]

By the way in the Masterpiece i would not like to make a transimpedance input also not the differential circuit i have shown. Both have a 3dB noise penalty.
When we use your high Gm Interfet -Fets i can show you a balanced circuit from the La Boheme stage. It is symmetrical and need only one gender ( N- channel ).
Or we could use the input stage of the Rauschfrei. That works with P and N-channel.

 

[Joachim Gerhard]

By the way in the Masterpiece i would not like to make a transimpedance input also not the differential circuit i have shown. Both have a 3dB noise penalty.
When we use your high Gm Interfet -Fets i can show you a balanced circuit from the La Boheme stage. It is symmetrical and need only one gender ( N- channel ).
Or we could use the input stage of the Rauschfrei. That works with P and N-channel.

 

[Joachim Gerhard]

Here is the La Boheme :
revised headamp
I think this is a nice way to make a balanced input with only one gender of J-fets ( n-channel). We could then do the RIAA active-balanced-shunt like i have shown with the OPA1632.

 

[MiiB]

Funny way of drawing a cascoded LTP... I like the RIAA, between the legs..

 

[Joachim Gerhard]

Yes, it is only a long tail pair. The way it is drawn is unusual though.
Yes, the RIAA is passive. Unfortunately that comes with a 14dB overload penalty at 20kHz
( Self at.all. ).
I prefer my shunt solution. Gives also much less distortion.

 

[Stefanoo]

Here is the La Boheme :
revised headamp
I think this is a nice way to make a balanced input with only one gender of J-fets ( n-channel). We could then do the RIAA active-balanced-shunt like i have shown with the OPA1632.

This is extremely interesting! I will have to crank this on the simulator.