MY_REF technical description

[mauropenasa]

Free of charge to the little that have understanded something of my circuit, and to the DIYs members that have "bet" on my idea, investing the they money and their time. I thank George (Josef_K) for his help in the translation, and to have convinced me to publish this document.

GoodBye

Mauro

 

[Panelhead]

Thanks for sharing

Free of charge to the little that have understanded something of my circuit, and to the DIYs members that have "bet" on my idea, investing the they money and their time. I thank George (Josef_K) for his help in the translation, and to have convinced me to publish this document.

GoodBye

Mauro

Thank you for sharing your ideas and designs to some of us less gifted. The layout and features of the Reference A boards are at standard way above my usual cobbled up diy efforts.
Also, thank for sharing this paper on the design process.

George

 

[lumanauw]

Thanks alot, Mauro 😀

 

[SheldonD]

Mauro:

Thank you very much for the explanation of your circuit.

Very generous of you to create document to go with your amp design.

Sheldon

 

[lumanauw]

Hi, Mauro,

How to calculate
-Ccmp+Rcmp
-Cff
-Cbf
-Cf
And what is the criterion of each?
- "My-ampli" uses only 1 capacitor in the place of Ccmp+Rcmp. Is it possible just to use 1cap only instead of (Ccmp+Rcmp)? Whats the difference?

I'm thinking about making your design but with full discrete components. What do you say about this?

 

[inguz]

Wonderful document, and a fascinating design -- many thanks. And also to Josef_K for the very clear translation.

 

[mauropenasa]

Hi all,
I even add the Italian version...

Ciao

Mauro

 

[Panelhead]

Discrete output version.

I would love to use some dicretes here instead of the LM3886. Should work fine, but may not equal the Referance as it is.
I have a few matched pairs of 2SJ162/2SK1058 mosfets that might be perfect. Guess I need to study the schematics a little.
Since the LM3886 is just a current pump, this is easier to do than it would be if the output stage was a voltage and current gain.

George

 

[carlmart]

Re: Discrete output version.

I would love to use some dicretes here instead of the LM3886. Should work fine, but may not equal the Referance as it is.
I have a few matched pairs of 2SJ162/2SK1058 mosfets that might be perfect. Guess I need to study the schematics a little.
Since the LM3886 is just a current pump, this is easier to do than it would be if the output stage was a voltage and current gain.

You may clone the Musical Fidelity A-270, which Mauro has already mentioned. Though I don't know if those are the right MOSFETs for that project.


Carlos

 

[Ed LaFontaine]

Thanks, Mauro

I'm a little "late to the table". I benefit from your generosity.

 

[varun21]

Thanks Mauro. I'd been hunting for this for long. It is still way above my head but an awesome case study nonetheless. Planning to build one.

 

[Dave S]

Free of charge to the little that have understanded something of my circuit, and to the DIYs members that have "bet" on my idea, investing the they money and their time. I thank George (Josef_K) for his help in the translation, and to have convinced me to publish this document.

GoodBye

Mauro

Very interesting article!

I am hoping somebody can help me to understand how the LM3886 current pump can have gm~1 and output impedance of 500 Ohms.

Isn't the output impedance 1/gm, i.e. ~1

A pair of 2SK/SJ FETs also have gm of ~1 (maybe a bit higher). So how is this compound amp conceptually different to a "standard" topology LATFET amp?

 

[Joseph K]

Dave,

Here it is a good description:

Op-Amp Current Sources: The Howland Current Pump | EDN

Circuit gm is a function of the current sense resistor and the feedback ratio.
And it results indeed ~ 1 with the classic myRef parameters.
Output impedance is a function of the balance between the positive and negative feedback branches in the bridge. In a balanced case it can be ~ infinite, independently of the gm of the circuit.
But Rs, the current sense resistor provokes a dis-balance, which deteriorates the infinite output impedance in a measure that depends again on the concrete value of the bridge elements, their precision and the Rs.

Ciao, George

Re your suggestion about discrete current output is totally valid, it was done exactly this way in the 'original' musical fidelity amplifiers, which were a composite amplifier with an lm318 as control amp and a discrete fet current output stage.

 

[Dave S]

Thanks for the link George.

I found this:
http://file.scirp.org/pdf/CS_2013110810122254.pdf

and using equation 9 came out with a calculated Zout of 69K Ohms.

The equation reduces down when R1=R1 and R3=R5 and by inspection R4 cancels out i.e. the ouput impedance is NOT affected by R4 (0.47 Ohms) as long at the OLG of the amp is high.

I am still struggling to understand where the 500 Ohms figure comes from.

 

[Joseph K]

From the datasheet of the lm3886 one factor is missing: the dependence of open loop gain vs output current.

I suppose that the datasheet value is taken at low output current.
Anyway, it is not infinite, even from the datasheet.
It's between 5000-10000 at 1kHz, it's 1000 at 10kHz, 500 at 20kHz..

Mauro's values were observational; I would push the theory until it starts to resemble the real life values..

Anyway, a really good paper, thanks.