Discrete "diamond" I/V stage

[OliverD]

Hello all,

recently I've been experimenting a lot with the diamond buffer topology. It has some interesting properties which make it a good input stage for "current feedback" amplifiers. A well known implementation of this topology is Elektor's "titan" or "gigant" amplifier. The output of the diamond buffer is used as the feedback node.

I put "current feedback" in quotes to avoid lengthy discussions about the definition of the term "current feedback". For now, I'm referring to injecting the feedback signal (the output voltage) as a current into a low impedance node.

Coincidentally, injecting a current into a low impedance node is exactly what most current DACs like to do. So why not use a "cfb" amplifier as an i/v-stage?

This is what I came up with: A standard implementation of the diamond buffer circuit using two constant current sources, the input being tied to ground and thus providing a low impedance "zero volts" node at the output, an output stage driven from the buffer, and a "current feedback" resistor between the voltage output and the buffer output. This resistor sets the "gain" where Vout = Iin * R-IV. The input and cascode stages share a common heavily filtered voltage reference.

The circuit is very fast and has a low impedance output being able to drive about everything. With a careful layout, no phase compensation is needed. The power supply is critical - the original uses heavy RC-filtering, no-feedback pre-regulators, capacitance multipliers and individual bypassing for each stage. Reistor values can be calculated easily once the active devices are selected and optimal bias currents are found.

The circuit can be used in a parallel or in a balanced configuration. For the balanced configuration, both halves should share the same voltage references for further noise reduction.

The design is still in progress and at the moment no PCB layout is available. I'm looking forward to your comments!

[Extreme_Boky]

Hi Oliver D,

I also share your opinion about diamond transistor followed with a buffer. This is why I decided to try http://users.verat.net/~pedjarogic/audio/index.html.

Pedja used AD844 for the same purpose, with an option of using on-the-chip buffer OR external buffer.

Design also uses famous (we’ll see…) CS8412 / TDA1541A combination I want to try. From what I heard after mods I’ve done on some Marantz gear that uses TDA1541A/S2 – the expectations are high!!!

If you check the web-site carefully you'll notice some other options for I to V using MAX IC's.

Idea is great, we'll see how it works...

Regards,
Nick

[banana]

I guess the lower side of the output stage should be npn.

Have u consider the folded cascode config? Maybe something like the one in THS3001.

http://focus.ti.com/lit/ds/symlink/ths3001.pdf

[hjelm]

Nice circuit.
I do agree with banana though. Your lower part of the output stage must have at least one transistor wrong type and upside down.

How much gain do you have in the output stage, if you remove the feedback?

[Extreme_Boky]

It appears, after a quick look, that THS3001's pin 5 is not accessible like on AD844 to implement I to V resistor.


Extreme_Boky

[OliverD]

You are right guys. Lower side of the last two stages must be PNP and emitter upwards, of course. It was late when I drew the schematics.

[Jocko Homo]

These circuits sound better without the feedback loop. It really doesn't buy much wrt input impedance. Good to see some of you guys are thinking, instead of copying the '5532 out of the app note.

Jocko

[OliverD]

These circuits sound better without the feedback loop. It really doesn't buy much wrt input impedance.

I will give it a try both ways to see if I can confirm the "odd" sound of cfb.

Good to see some of you guys are thinking, instead of copying the '5532 out of the app note.

Nice to hear that from someone who knows better

[B Cullingford]

Hang a resistor off pin 5 of the AD844 (as in the spec sheet for the IC - or was it the other CFB ... not sure of the number , to implement an IV convertor.)
The output diamond buffer in the IC will buffer the output. IC op amp but no GFB. Distortion and offset are both very low, I tried it many years ago, maybe I'll finally get around to making it and using in my system, or maybe I'll do it discretely with eight current mirrors though four are really two buffers and two are to bias the input stage plus one jfet to set all the currents.

[Jocko Homo]

Only works with a '844. Or a '846 if you still have some in your junque boit.

Jocko