[keithvv]

To the real EEs out there.

I wish to convert the input on a 55 N+ to accept balanced inputs (it will be at the end of a long run from the audio processor - hence  I will be using the balanced out option on the processor, through shielded twisted pair to the amp).

I have read of two ways to do this with an external box or "converter":
1) Use a step down audio grade transformer (e.g. Jensen). That's bulky and pretty darned expensive. Not too attractive an option.
2) Use an op-amp based buffer/converter stage. That gets into supplying another DC voltage level in the amp, another board, interesting (though currently beyond me calculations), and potentially polluting the signal by running it through another op-amp. etc. Another unattractive solution.

I have been digging a bit deeper into the designs of different amps and have seen where Nelson Pass and Norm Thagard provide a balanced input into their A75 amp design. It appears that they simply run the +Ve line into the "normal" input of the input stage, which is a differential amp (MOSFET); then run the -Ve line into the feedback (negative) side of the input pair; the ground line is connected to chassis. Of course, they use resistive coupling into the amp - not capacitive coupling ala the AKSA, so maybe it's simpler in that case.

First question: Is this feasible and reasonable to do with the AKSA 55 N+?

Second question: What adjustments must be made (if any) to the voltage splitting resistors R7, R8 and R9 to compensate for the additional negative input voltage appearing at the base of T2?

Third question: Do I need to attenuate the higher voltage swing signals before they get to the differential pair of T1/T2, or can they handle the voltage swing as is?

FInal question: Is there a better way to get the balanced input fed to the input stage?

Thanks in advance to any who have the time and expertise to help.

Cheers,
KVV

[AKSA]

Hi Keith,

Thanks for your post.

Yes, it can be done, but I've never done it, and a redesign of the diff input stage is required.  The details would take some time to figure out;  in fact I'd need to do homework on this.  The big issue is the treatment of the feedback signal.  

The Jensen balanced line transformer is the usual option in pro-audio, but it's expensive.  The IC buffer is an inelegant way of doing it, since the AKSA uses a balanced long tailed pair input and it is possible to directly inject a balanced signal anyway.

Can you point me at the Pass/Thagard article on the net?  Is is on the PassDIY site?  If I can read this up and doodle for a time, I should be able to come up with a workable balanced input.

The advantage would be apparent in just two situations:

1.  Noisy environment with very long interconnects between source and amp.

2.  Existing unused balanced output on the source.

It won't enhance sound quality per se (although folklore has it that it will!) but it may - repeat may - reduce noise.


Cheers,

Hugh

[keithvv]

Hugh,
Thanks for taking the time to reply (I never cease to be amazed at your energy and willingness to teach!).
The article is indeed on the Pass DIY site, the specific URL is
http://www.passdiy.com/pdf/A75p1.pdf
The topic is explained on page 4-5.
If time permits for you to look into this, marvelous indeed, and I'll be your beta site for any mods that you think might work, with full feedback (pun intended) provided to the forum.
Cheers,
Keith

[andyr]

Hi Keith,

Noisy environment with very long interconnects between source and amp ...

Hi Hugh,

As U know, I have unusually long ICs between GK-1 and electronic crossovers ... would  balanced ICs be any advantage for me?  Again, as U know, I don't actually have any "noise" ... possibly bcoz I use a "psuedo-balanced IC" with 2 wires for the signal & return, inside a shield (which, as Ginger recommends, is only earthed one end so does not form part of the 'return path').

Regards,

Andy

[AKSA]

Keith,

Try this:

1.  Use two 470nF or 1uF caps as series blockers for the inputs, both inv and non-inv.
2.  In series with inverting input, install a 2K2 resistor AFTER the input cap.
3.  In series on + (non-inv) input, place a 47K resistor (metal film).  Use existing input for + (non-inv) input.
4.  Connect - (inv) input to base of T2, the feedback transistor.
5.  No changes to existing circuit required.  

If Nelson is my guide, that will give desired balanced input.  - input and feedback are mixed at the base of T2.  There will be no effect on DC offset control.

The input impedance of the non-inv input will then be 47K, but the input impedance of the inverting - input will be around 2K.  This is an asymmetrical input load, and it could be evened by feeding the non-inverting input via a 1K/1K voltage divider to ground, and taking the existing input from the junction of these two series resistors.  That will make the inverting input also of 2K input impedance.

It should be noted that driving a 2K input impedance is tough going, and the only way to ameliorate this difficult input loading is to use either a low impedance source - 200R or less, fine for an IC based preamp - or use two emitter followers at each input to resurrect the high Zin as before.

Do please let me know the results if you go ahead with this - if this works as planned, it will become another input option for AKSAphiles!!

Andy, I would doubt it, but heck, mon, I've been wrong before (once, in 1964, very traumatic..... )

Cheers,

Hugh

[keithvv]

Hugh,
Thank you kindly for the impromptu redesign...
I'll certainly give it a try - though it will be at least two weeks before I have the time to devote to making the changes - so a bit of silence on my part doesn't imply lost interest!
I'll post results soon thereafter.
Cheers,
Keith

[keithvv]

Hugh (and other interested parties):

I have now completed the testing of the balanced input design that you emailed directly to me (not design in the description in this thread). I'll not divulge the details of the circuit change - that is most certainly your perogative and no one elses!

It works great!
No noise, no trash. All sonics seem to be preserved. No undue hiccups in operation.

I used a solid state preamp with less than 20 ohm output impedance, so the reduction in input impedance for the AKSA turned out to be a non-problem.

The test was NOT blind, nor direct A-B, though the amp that was converted to balanced input was one that I had been listening to for months, so I was quite familiar with it and the speakers used. Of course, others may experience the YMMV syndrome, but I am definitely happy!
 

There is one testing "aberration" that Hugh may want to comment on. When I used the new (reduced) value for R1, the offset voltage - speaker output to ground - soared way out of spec; to -195mv from a near "perfect" value of -1.5mV. This worried me at first. I read the trouble shooting tips, and double checked the drawing for errors in the circuit I built up. I did swap out the R4 resistor with a new one in case I had somehow jarred loose the connection. I stopped short of adding resistance to R3 since before I went that far I double checked the emitter bias value (Test point 1 to Test point 2) and it had remained a rock solid 55 mV. So I trusted it, crossed my fingers and tested it with a live signal and with   speakers hooked up. Great results.

For the record, I then replaced R1 to its original value, and "presto", the offset bias went back to an in-tolerance value of +0.5mV without any other changes. My guess is that the offset voltage is somehow dependent on the R1 value. I'll leave that to the designers and the EEs to tell me if I should be concernced.

Hugh, thanks again for being an unbelievably supportive individual.

Cheers, and happy listening.
Keith VV

[AKSA]

Hi Keith,

Many thanks for letting us know that your balanced experiment worked out so well.  This was a great contribution for our community, and gives the AKSA one more string to its bow!!

I've since revisited Nelson Pass's comments on balanced input for the JLH 10W Class A - a highly regarded old design originally devised in 1969 and still popular today.  NP redesigns it for diff pair input and balanced line;  the very same article Keith referred me to.  It's ironic, because I'd also converted a JLH about two years ago for LTP input as an interesting paper exercise.

When you drive a LTP input amp in balanced configuration there is no way you can avoid the very low impedance of the feedback shunt resistor on the inverting side.  This resistor runs from the feedback transistor to the shunt cap, which is in normally connected to ground for conventional, unbalanced operation.

On the AKSA, this resistor is 2K2, so this means that the input impedance effectively decomposes to this resistor.  The base of the transistor is in fact a virtual earth since it hikes up and down alongside the input transistor's base, so this creates some issues because the input impedance at the inverting input resistor is very low.  So considerable drive from the source is required, and steps should be taken at the non-inverting input to render the input impedance identical.

Keith, the issue of offset control revolves around the input diff pair (LTP).  Both the input and the feedback transistor must draw their bias currents - on the order of 100 microamps each - from R1 and the feedback network respectively.  If we change R1, then this bias current drops rather less voltage and in order to keep the base potentials identical (that's what LTPs do best!) then the speaker output must be held at a lower potential, since this is the voltage source from which the feedback transistor is biased.  The Voltage Amplifier will make this adjustment to accommodate identical potentials at the bases of the LTP.  Ergo, offset is greatly thrown out of whack.

While a balanced input transformer such as a Sowter, Lundahl or Jensen is expensive, it does avoid this very low input impedance problem.  If we wish to restore the input impedance to the nominal unbalanced 47K, we need to put a buffer with an output impedance no more than about 50 ohms on the front of the balanced AKSA.  This adds more stages, of course, and contributes some distortion, but if your source has power to spare it's probably not a problem.

Keith appears to have sufficient drive on his source to move mountains, so his balanced AKSA is the simplest configuration possible.  Keith, would you say that the sonic qualities are identical to the original?

Cheers,

Hugh

[AKSA]

Keith,

I've upgraded my comments on the input impedance in balanced mode.  Look at the other thread to get my latest on this;  in short, I believe both input points, with R1 at 47K as you've found, offers the very best situation with virtually identical input impedances in the 43k range.

Essentially, you can very easily make a balanced AKSA with minimal changes.  My only suggestion is to increase the emitter degeneration resistors at the diff pair to 33R each to preserve gain around 31dB.  This is important as the feedback factor has influence on the sonics.

Cheers,

Hugh

[keithvv]

Hugh,
Thanks for all of the thought that you have put into the balanced in option.
I'll change out the 10R emitter degeneration resistors for the recommended 33R values, undo the previous mods (2.2K to the base of T2) and give it a go sometime this weekend.
I'll post results for all to review.
Cheers,
Keith