Masterpiece

[Stefanoo]

Originally Posted by MiiB

Best way is to accept the Caps and stick to the bipolars... Reason why caps have a bad reputation is when they are used in a zero crossing position. In the paradise there's no zero crossing, as the caps are always charged. When used like this the drawbacks of a cap is very very minute, and The cap can of course be bypassed with film, but in reality the MHz loss in the El-co is not unwelcome. quite confident that you'll have a lot better result with the Bipolars, than with the much too hard to get Jfet critters.

you can also boost gain by adding current gain in the mirrors. but you have to watch dissipation of the Lower mirror transistor.

MiiB,

thanks for your reply.
If I use 16 JFETs for the balance at the input I get to a dumb 59dB.
However I am sure I can squeeze out 2-3 dBs, I am convinced that this arrangement won't sound any good.

One other thing I have noticed is that frequency response, of the SE Jfet with no caps, is very narrow for some reason. and significantly drops down at 80KHz, and I was unable to trace this problem down nor understand why it would do that also because on simulation it doesn't show having this issue at all.

Going back to bipolars and capacitors.
It sounds really nice. However I Can say from tests I have conducted that ELCAP has a much bigger impact than what you think it doesn.
Simply enough, just changing the type and/or arrangement significantly changes the overall sound, which obsiously highlights the fact that cap there is introducing somehow a clear sonic signature as expected.

Unfortunately it looks like in order to do it much simpler with JFETs, you will have to find devices with at least 5-6 time the transconductance of the K170/J74 and I am not sure is a possible tasks

How come devices like those don't exist? I can't imagine why they wouldn't make them or made them at a certain point?
There are a ton of BJTs but very limited JFETs.
Clearly JFETs is an evolution of the BJT came after than and I Don't understand why this technology is only used on IC but not available discrete.

Maybe some custom military or NASA parts...I am sure they must have something like that.
I remember talking to a hi-end manufacturer in my country who was explaining to me that in prder to create his very fine electronics he had to haunt down parts from NASA (don't know if it is true) devices with extremely high gm and linearity which allowed him to only have one gain stage.

What do you guys think of it?

[Joachim Gerhard]

There are much steeper J-Fets from Interfet for example. This one has 80mS and there is even one with 350mS.InterFET Corp. Page Selector - IF9030 datasheet
Again the problem is that there is no matching P-Fet.
Just to be safe i designed a good sounding phono stage with only high Gm N-channel J-Fets some time ago. It must not be parallel symmetric, there are alternatives.

[Stefanoo]

woowww

interesting. What do you think then to modify it into just a N-Ch desing?
Would that be worth it?

I would be really interested to know what you think..

[Joachim Gerhard]

I would do a differential stage with only high Gm N-channels, sure. That P-channel thing is simply impossible to solve. Linear Systems simply does not make them available.
The disadvantage is 6dB more noise with the same amount of input devices. Bob Cordell solved it by paralleling 4 LSK389 and got the noise down to 0.7nV/qHz. That is ok but i would go for 0.5nV/qHz. That is possible with the Interfet parts without paralleling that much. I know you are not a fan of paralleling but i simply have no better idea when you do not want the noise to be audible at the listening seat unless you take the super high Gm parts. Bob also converts in the input stage from balanced-In to unbalanced-out.
I would do the same or you need really tight matching of better then 0.1% of all parts.
If you do not do that you throw your common mode rejection out of the window.
There is also the risk that your double RIIA cancels in part in the micro details.

[Joachim Gerhard]

Here is one clever way to make a balanced-in, balanced-out stage with only N-Channels AND only one set of RIAA parts so it can be done :schematics

OHH, sorry, it has some coupling caps and a transformer, something you do not like too.
Coming up with a way of eating the cake but also keeping it is not easy but that is what i would like to see.

[Stefanoo]

Originally Posted by Joachim Gerhard

I would do a differential stage with only high Gm N-channels, sure. That P-channel thing is simply impossible to solve. Linear Systems simply does not make them available.
The disadvantage is 6dB more noise with the same amount of input devices. Bob Cordell solved it by paralleling 4 LSK389 and got the noise down to 0.7nV/qHz. That is ok but i would go for 0.5nV/qHz. That is possible with the Interfet parts without paralleling that much. I know you are not a fan of paralleling but i simply have no better idea when you do not want the noise to be audible at the listening seat unless you take the super high Gm parts. Bob also converts in the input stage from balanced-In to unbalanced-out.
I would do the same or you need really tight matching of better then 0.1% of all parts.
I you do not do that you throw your common mode rejection out of the window.
There is also the risk that your double RIIA cancels in part in the micro details.

so basically it wouln't be a full balance topology, am I right?
Could you post a link to the schematic by BC you are referring to in this discussion?

Also, what would be the disadvantages of using a super high gm device if this brings you less noise, less parallel device and more gain?
On your comments you make is sound like a possible alternative, rather than a better solution.

[Stefanoo]

Originally Posted by Joachim Gerhard

Here is one clever way to make a balanced-in, balanced-out stage with only N-Channels AND only one set of RIAA parts so it can be done :schematics

OHH, sorry, it has some coupling caps and a transformer, something you do not like too.
Coming up with a way of eating the cake but also keeping it is not easy but that is what i would like to see.

I posted while you were posting.
Ok got the schematic, very interesting however, like you can imagine, I dislike transformers on the signal path.

Also, why would you want to convert to unbalance out to then re-convert it balance out?

If you have a decent matching, I am assuming you could get a cancellation of 2nd harmonic by keeping it balance throughout, while this way you just end up summing up, thus having more unwanted distortion.
Am I saying something wrong?

[Joachim Gerhard]

The high Gm parts have only the disadvantage to be very expensive but that did problem did not hinder you to buy the super expensive Teflons.
There is a super high Gm N-Fet with 0.3nV/qHz that could do the job with only pair.
I have somewhere on MPP designed a hypothetic phono stage with that parts, but that topology was more similar to the Pink Triangle Pip.
You have to browse the Intefet web page to find it. Sorry, i do not have enough time today to find it.


Bob Schematic is in Jan Diddens Linaer Audio 4 so i can not simply scan it and post it here.
I recommend to buy that issue anyway. There is a lot of interesting material about phono in there, also MM stages and tubes.

[Stefanoo]

oh wow...thanks for the hint...very interesting.
I will look for that issue and see if it is avaialable for purchase.

[Stefanoo]

I mean, how expensive is it? Is it like a new part that can be had easily or has to be browsed on ebay risking fakes and stuff?