White follower DI circuit using JFETs

[Voyager10]

I've been playing around with a White follower circuit, inspired by the Tube DI box thread, but using everyday JFETs. Here's what I ended up with, as a phantom-powered circuit:



The idea behind the White follower is that Q1 is a source follower, with Q2 as its load, but R5 and C3 modulate the current through Q2 in opposite phase to the Q1 current. So it stays in Class-A throughout the voltage swing but without drawing excessive supply current, important for a phantom-powered device.

The circuit draws just under 3mA from P48 power, and the supply rail is +31.5V. It delivers 19V pk-pk (6.7V rms, +18.7dBU) into a 3K load before visible clipping. THD, as measured by REW, was 0.0087% at 500mV RMS out, rising to about 0.10% at 5V RMS, at which point my audio interface maxed out.

Input noise is -127dBV, or -129.5 dBV A-weighted, so the total dynamic range is somewhere over 140dB . To be fair, you would do just as well with an OPA1641, but it's quite fun that an 80-year old circuit topology does so well.

 

[MicUlli]

"Alte Besen kehren gut"
This nice example of old school circuits shows us once more that we don't need the newest OpAmps to get ultra high audio performance
Very low component count was always a way i have gone for years..
BR MicUlli

 

[PermO]

Super cool !

I am a bit confused about the ground / case connections...

I would think, J1 connects to ground only, and not case.
XLR 1 connects to case only, and not ground.

Am I missing something ?

 

[Voyager10]

Ah yes, my apologies, this was copied from another schematic.

For a DI box application you would need XLR pin 1 wired to the XLR shield and case as directly as possible.

 

[Wavelength]

But it doesn't create a differential signal on the output.

 

[Mondy]

Sort of an Impedance Balanced output? Wouldn't you need to know the output impedance of the circuit to set the value of R7 to match for this arrangment to be effective?

 

[MicUlli]

Sort of an Impedance Balanced output? Wouldn't you need to know the output impedance of the circuit to set the value of R7 to match for this arrangment to be effective?

No. R7 is only for compensating the source follower output impedance. But it is true that the input impedance of the connected stage has an influence on suppressing unwanted harmonics. Fortuately this is not that critical..

 

[ruffrecords]

Super cool !

I am a bit confused about the ground / case connections...

I would think, J1 connects to ground only, and not case.
XLR 1 connects to case only, and not ground.

Am I missing something ?

No, pin 1 of the XLR should go to the case and also to gnd (because it is phantom powered). The input jack only needs to go the gnd.

Cheers

Ian

 

[ruffrecords]

But it doesn't create a differential signal on the output.

It does not need to. I creates a balanced output.

Cheers

Ian

 

[Voyager10]

I should confess I've not tuned the R7 value for the actual built circuit, it was just a rough value from the simulator.

An "impedance balanced" output with a single driven pin is quite common and used in the Rode NT1, Neumann TLM193, MicUlli's circuit, several OPA164x designs, and more.

 

[merlin]

Don't forget to add a pair of protection diodes to the rails, from the negative side of C5. You don't want to blow your JFET with a voltage spike when you switch the phantom power on.

 

[Voyager10]

Yes, good point.

I guess I'm always careful with phantom power switching but it seems common in some places to use a jack patchbay for mic signals, where plugging in or out with the power on causes all sorts of momentary shorts.

Edit: D1 and D2 now added to schematic in post #1.

 

[Wavelength]

It does not need to. I creates a balanced output.

Cheers

Ian

Ian,
Sorry been at work doing a custom AES/EBU job...
Why doesn't it need to?
The whole idea behind a differential system is that any noise created will be unbalanced and when it reaches the receive end will be eliminated by the push pull nature of the differential signal.
Now if the above was transformer coupled before the XLR connector then sure the single ended to transformer signal would become differential.
But here the negative side of this interface is just terminated. Any noise on the positive side will remain at that the receiver.
Unless I am missing something?
Thanks,
Gordon

 

[Wavelength]

BTW gang here is my 9V based SS DI. It uses a folded cascode input in a pedal form. Gain is via a top pot and phase by switch. The stomp switch basically mutes the output. I am soldering that up now. Use the 1N914 for 2N404 germanium and the Red LED (1.7V) for 2N5087, R4 may need to be adjusted to get the right BIAS.
VCC = 8.5V using my shunt regulator, you can replace the pot with a 2.4K resistor. The input cap can be bypassed with a switch as I feel the problem with most guitar pedals is there is an output and an input cap which seems a little unnecessary. TS input TRS output.
Thanks,
Gordon

 

[Voyager10]

Here "balanced" means that the impedance between 2 and 1 is the same as the impedance between 3 and 1.

The idea is that if the connecting cable is sufficiently symmetrical (e.g. the two signal wires are twisted together evenly) any interference signal will affect both the pin-2 signal and the pin-3 signal to the same extent - provided that the impedance at the relevant frequency is the same.

The preamp receiving the signal measures the voltage difference between 2 and 3, so if the interference signal affects both equally, it won't affect that voltage difference, so the interference will be cancelled. It's not necessary for the signal on pin 3 to be the opposite of pin 2 for this to work; the magic happens at the receiver end, not the transmitter.

 

[Wavelength]

Here "balanced" means that the impedance between 2 and 1 is the same as the impedance between 3 and 1.

The idea is that if the connecting cable is sufficiently symmetrical (e.g. the two signal wires are twisted together evenly) any interference signal will affect both the pin-2 signal and the pin-3 signal to the same extent - provided that the impedance at the relevant frequency is the same.

The preamp receiving the signal measures the voltage difference between 2 and 3, so if the interference signal affects both equally, it won't affect that voltage difference, so the interference will be cancelled. It's not necessary for the signal on pin 3 to be the opposite of pin 2 for this to work; the magic happens at the receiver end, not the transmitter.

But the impedance would not be the same and would vary on the active terminal more than the fixed 47ohm side. We can basically say the 4K7 are out of the picture, right? and it's just the output Z of the active side vs the 47ohm/22uF Pin3.
Look at Jensen work on balanced receivers and yes a lot of his work had to do with basically making unbalanced reception work.

 

[ccaudle]

But the impedance would not be the same and would vary on the active terminal more than the fixed 47ohm side.

Sure, but that gets into discussing percentage of mismatch and impedance match vs. frequency, which is something amenable to numeric analysis. You need to know the common mode impedances of the receiver to accurately assess the effect on common mode rejection performance, but the big hurdle is getting over the thinking that balanced connections require symmetric signalling.

 

[Voyager10]

So the output impedance at pin 2 is determined by Q1's transconductance, and that will change as its drain current varies, but you need to get to several volts output before it becomes significant. (Changing transconductance would also show up as nonlinearity, i.e. THD, and that also stays low up to high output levels).

 

[Wavelength]

Guys,
I typically work in single ended land and not much in balanced. But I have done enough engineering to know that this is more single ended than differential. So lets see what happens if we apply this to typical inputs:
1) Transformer, so typically 10K:10K for input, some step up some step down lets just say 10K input. This would work, not so sure how well and the pullups are not really helping since they look to be grounded in the ac world. Low drive current so the bass response and speed are going to be low. Since this is driving only one side it would work, but I wouldn't give it a gold star.
2) THAT1200 (my fav, non transformer input) so this will look like 45K input impedance still only driving one side, the pull ups if the ground is common might effect the output swing if the REF pin is grounded. Why do we have the pull up resistors? The THAT1200 works both SE and Differential so this would work, but may clip depending on the effect the two pull up resistors have and the volt rails of the THAT1200.
3) INA/OPAMP differential receiver, again like the THAT1200 if the ground is common the two pull ups will effect the output signal. Since unlike the THAT1200 there is no build in servo to reference this may have some offset issues.

What I would like to see is something more balanced. You might find it easier to create a +/- supply to do that, though there are more complex ways that would work to create it.

 

[Wavelength]

Ops sorry I see the 2 4.7K are powering the circuit!