Differential buffer without feedback?
 

Hello,

I am struggling to try to find out a principle to apply for a schematic I need. I am studying for a while my hybrid amplifier trying all known typologies. this amplifier is with symetrical input (differential 2 x 12AT7).

My goal is to use differential inputs (XLR) and make a connection to the 12AT7. Unfortunately, the feedback resistors (low for noise) are not compatible with the minimum input impedance for XLR. 10k?

So I investigate a single differential buffer. 2 inputs. 2 outputs. No feedback necessary, thus no room for resistor noise. Or maybe feedback ok, but idea to have optimum noise/input impedance?

I have seen some attempts here and there but no prototpe nor sims were done.

Do you have any guideline? Isn't it a common requirement? I feel not so good to use a 100dB AOP to have a x1 gain. I will go if I have no choice (well, you understand, it is philosophical... my Hybrid has no feedback neither... except final BJT buffer...)

Hope you can help!

best regards

Laurent

I found following document concerning fully differential amplifiers : http://focus.ti.com/lit/an/sloa054d/sloa054d.pdf

The schematic (attached) is a good starting point for a discrete design.

Back home yesterday evening, I LTspiced a version with +/-50V power supply. First version with current sources, second with resistors instead of current source.

Surprisingly, it is easy to have a low distorsion (lower than ppm) in differential mode with 4 V ouput signal (see application schematic).

Didn't think about coming office with models... no web at home. I will post another day.

I have now a differential buffer that suit half the requirement :
- differential in : ok
- differential out : ok
- gain=1 : ok
- low output impedance : ok
- high input impedance : not yet ok, depends on R1, R3
- low value resistors on signal path : not ok, compromize with previous point

I tried before night to degenerate the circuit to have a buffer without feedback. I have a buffer, but distorsion is not as low... as expected.

Hope you can react on that with your experience?

I would feel bad with 10k resistors for R1...R4, too high the noise. Not you? What about noi

If you put a unity gain buffer before each input of your circuit, you can have high input impedance even with low values for R1,2,3 and 4.

That's a way i am investigating. I need just two BJT buffers. Goal would be to have low distorsion ones at a few volts input level. Maybe Emitter follower loaded by a current source?

Will have to investigate that, not a lot of time right now. Will post simulation files ASAP.

or a pair of B1 as source followers.
or copy Cordell's jFET input LTP buffer.

Thank you Andrew for the tip. Right now, still trying to dig the schematic out. Any usefull link for B1? (now at page 9 of the B1 thread...)
Will have a look to Jfet fubber from Bob. Any link meanwhile?

I found it. Jfet follower loaded with jfet CS. ok. ".0007% distortion waveform at 1 Volt and 1 kHz." oh. nice.

have to find out if that is ok for DC coupling. no sure about vgs mismatch.

What? I forgot to say that? Yes, all the design is DC coupled from input to output with high voltage 12AT7 on the path (powered at 150V).

Some sparks to be expected. Spice doesn't warn about this.

Maybe some coupling cap on the path one day. Nobody's shocked yet.

BJT's arent' good for that buffer job?

I got it. I matched jfet with same temp and same id, then same vgs.

Possible without matching by hand..?

BJTs can be used for buffers. They can do an excellent job.
It just happens that the two simplest buffers that came to mind were both jFET input.

Go to Pass site and download the B1 manual.
Go to the DCB1 thread for a DC coupled version of the B1.
There is also a thread or two about using the BF862 as a buffer as well as a +gain amplifier.