Making balanced output from single end

That's what we call a ground-compensated or ground-cancelling output. It gives most of the benefits of a balanced output without the need for a balanced input at the receiving end.

Has never really caught on- I don't know why as if properly implemented a GC link is quieter than a balanced link and uses two opamps instead of three. (for the whole link)

Perhaps people don't like the idea of one pin of an XLR output being in reality an input.

'alexcp' has some interesting information about this type of active ground noise cancelling in the following thread: https://www.diyaudio.com/forums/ana...volume-controller-line-stage.html#post5497318
 
All this talk of how great GC is makes me want to try it for my next project...

An externally hosted image should be here but it was not working when we last tested it.

A couple of points about that diagram. The connection between the load and ground should be shown as solid; otherwise the load is floating and that does not represent an unbalanced input.

Secondly, R6 and R7 can be much lower as they are being driven from the receiving end ground, and thus noise reduced. In fact, you can remove R5 and make R6,R7 51 Ohms each. Not quite equal to R4 but may be near enough if you're using 1% resistors.
 
Hahah, that was assembled by the “xrk971 model 2000” pick and place machine
Hey,we´d LOVE to see that.
And a video showing it doing its thing would be even better.

and soldering was accomplished with a $10 Electric Single Burner 1100-Watts, Black room hot plate and skillet from the bottom and a $50 858D 110V Solder Station, Digital Display SMD Hot Air Rework Station Solder Iron Kit Heat Gun, Tweezers, Desoldering Pump hot air pencil from the top. I use MG Chemicals 4860P 63/37 No Clean, Leaded Solder Paste, applied with a syringe and hypo needle. Parts placed with good pair of techni-tool-00-sa-strong-sharp-precision-swiss-tweezer-00 Swiss Technitool tweezers and vision provided with SE Illuminated Dual Lens Flip-In Head Magnifier flip up magnifying goggles.
Those are 1210 resistors and 0805 caps - they are child’s play compared to 0402. The VSON10 packages look hard (but not really) have an underbelly pad and hidden pins. The solder surface tension self aligns them. Same with the rest of the parts. Trick is to apply just the right amount of paste (sparingly). Heat the skillet at medium to low heat and monitor PCB temp with IR thermometer. When it hits 130C and stabilizes, use 300C hot air pencil from above to liquify just the components of interest. Keep liquidus for at least 10 seconds and then go to next section. This way, parts only experience long term 120C to 150C and short term just long enough to melt and make a nice joint.


On complex jobs with lots of parts, I use a digitally controlled solder paste dispenser to lay down repeatable perfect size blobs even on 0402 and 0603 parts and skinny little 0.3mm pitch pins on a quad flat pack DSP chip. Your hand would cramp if you had to use a syringe on more than 30 parts.

It takes some practice of course - watch YouTube videos from SparkFun on SMT soldering and you can learn a lot. There is a lot of Pb-free solder paste out there - I prefer lead for its better flow and durability and lack of Sn-whiskers. Copper braid and thin Kester 959T Soldering Flux free flowing liquid flux with a regular soldering iron chisel tip is your friend for removing excess or cleaning up solder bridges.

In the end, it’s much faster than TH soldering as everything is on one side and there is no clipping of leads or bending of leads needed.

Those tweezers seem expensive - but very crucial as they are your fingertip extensions and cheap ones just don’t work. I have had mine for 20 years. You have links above to just about everything you need to do this.

I have been eyeing a binocular stereo microscope workstation for some time now. That would be the next step and enable 0402 assembly much easier.

Essentials to get started: tweezers, magnifier goggles, solder paste, hot plate. Add hot air pencil later. Copper braid and flux might be handy and useful for TH soldering as well.

Hope that helps. Maybe another Howto thread is in order?
You bet :)

Thanks for the detailed no nonsense data offered :)
 
Note also not simple impedance balanced, but a "hum sense" resistor and differential NFB. (as in post #37 here) So not the same thing. Sorry for the false trail.

Thanks PRR, those two pics give a clear description of a ground-compensated output. Looks like the dbx 160 came out in 1971, so, as is often the case, the idea is older than I thought.

I think I first heard about the idea in the mid-80s.
 
dBx has made WAY too many limiters.

Memory said "160" but that is mono. The 165 is the stereo, has the right knobs/meters, but the plan I found is simple single-ended.
DBX 165 - Manual - Over Easy Compressor / Limiter - HiFi Engine
Mine had a later manual (pamphlet instead of binder). This would be early-mid 1980s.

They boxed the same stuff several ways, and probably had several revisions of each model. I found the impedance-balanced plan in the 160:
DBX 161 - Manual - Single Channel Compressor / Limiter - HiFi Engine (manual covers 160 and 161)

Note also not simple impedance balanced, but a "hum sense" resistor and differential NFB. (as in post #37 here) So not the same thing. Sorry for the false trail.

Interesting, that dbx describes the circuit as featuring up to 40dB of ground noise reduction. An limit, no doubt, set by the use of 1% tolerance resistors for R83 and R82. This may have been due to the high cost of resistors under 1% tolerance at that time. It also might have been because they determined there was no practical benefit of greater reduction than that.
 
Interesting, that dbx describes the circuit as featuring up to 40dB of ground noise reduction. An limit, no doubt, set by the use of 1% tolerance resistors for R83 and R82. This may have been due to the high cost of resistors under 1% tolerance at that time. It also might have been because they determined there was no practical benefit of greater reduction than that.

1% resistors were pretty damn' expensive in 1971! However there are some E96 values in the schematic (eg 22.1k) so maybe they were using a premium type of resistor.

If they were using 1% then I suspect the 40 dB figure was a bit optimistic; at least two resistors are involved in setting the GCRR. (Ground Cancel Rejection Ratio)
 
Last edited:
1% resistors were pretty damn' expensive in 1971!
Oh, I mistakenly thought it was of mid-80's vintage. So, I guess those 1% values may have actually been state of the resistor production art back then. To do much better probably would have required hand sorting.

If they were using 1% then I suspect the 40 dB figure was a bit optimistic; at least two resistors are involved in setting the GCRR. (Ground Cancel Rejection Ratio)
Yeah, I noticed that dbx's performance description included the qualifier; "up to" 40dB. ;)
 
Founder of XSA-Labs
Joined 2012
Paid Member
A couple of points about that diagram. The connection between the load and ground should be shown as solid; otherwise the load is floating and that does not represent an unbalanced input.

Secondly, R6 and R7 can be much lower as they are being driven from the receiving end ground, and thus noise reduced. In fact, you can remove R5 and make R6,R7 51 Ohms each. Not quite equal to R4 but may be near enough if you're using 1% resistors.

Hi DouglasSelf,
Thanks for the helpful usage notes - I will try this.
Regards,
X

Btw, we are so lucky in this day and age that we can get 0.1% 25ppm metal thin film SMT’s for $0.66. I just ordered some Susumu’s for my headphone amp gain setting section and was pleasantly surprised.
 
Last edited:
Btw, we are so lucky in this day and age that we can get 0.1% 25ppm metal thin film SMT’s for $0.66. I just ordered some Susumu’s for my headphone amp gain setting section and was pleasantly surprised.

Last year I needed some Through-Hole 0.1% resistors and I too was pleasantly surprised when they were only ten times the price of 1%. Might have been Mouser or Digikey, I don't remember.

Is that other people's experience?
 

PRR

Member
Joined 2003
Paid Member
... Looks like the dbx 160 came out in 1971....

I think a decade later. 1980s.

HiFiEngine notes the 165 as 1981. They have a scan of the binder-manual; mine came with the later pamphlet manual. IIRC (I'm not doing well with that) there was a description of this fancy circuit in the manual, implying it was a new thing.

The 160 (basic single-ended output) may have come out in 1976:
Morrisound Recording >> DBX 160 & 165
dbx (company) - Wikipedia
160A | dbx Professional Audio
 
From the schematic you posted it looks as though the dbx 160 did have a GC output.

160A | dbx Professional Audio says it appeared mid-1970s

This is where I got 1971 from:

Top 20 Best Compressors Of All Time - Page 9 of 20 - Attack Magazine

I found that convincing because the schematic has a very '1970' look to my eyes. However it is a bit confusing as one minute they say 160A then 160.

It would be nice to try and track this bit of history down. I imagine some of the people concerned with the dbx 160 are still with us.
 

PRR

Member
Joined 2003
Paid Member
> I imagine some of the people concerned with the dbx 160 are still with us.

That legacy moved to THAT Corp. Many of the 1970s people must be retired, but the THAT Corp people would/should know your name, share history, give you contacts.
 
You can connect the input of this circuit to a balanced or an unbalanced source and the output to a balanced or an unbalanced load. It's basically an indirect voltage feedback (in Ernst Nordholt's terminology) differential amplifier with a weak and slow output common-mode loop. I built it for a local radio station that has a mix of consumer, semiprofessional and professional equipment. If the clipping behaviour matters, it may be a good idea to give the feedback differential pair a slightly higher tail current than the input differential pair, even though that degrades the distortion cancellation between them a bit.
 

Attachments

  • schema.pdf
    34.9 KB · Views: 296
You can connect the input of this circuit to a balanced or an unbalanced source and the output to a balanced or an unbalanced load. It's basically an indirect voltage feedback (in Ernst Nordholt's terminology) differential amplifier with a weak and slow output common-mode loop. I built it for a local radio station that has a mix of consumer, semiprofessional and professional equipment. If the clipping behaviour matters, it may be a good idea to give the feedback differential pair a slightly higher tail current than the input differential pair, even though that degrades the distortion cancellation between the

That's an interesting amplifier.
I have a few questions about it:
Why the use of diodes at the BE junctions of the bjt transistors (D1-D4, for example)?
How did you decide on the resistor values for Q5 and Q6? I guess their base voltage are pretty well defined Q13+R13 to set that transistor into linear region
Are the degeneration transistor in both differential pairs Q1-Q4 necessary for stability?
 
The diodes are meant to prevent avalanche breakthrough of the emitter-base junctions when the amplifier is driven into clipping. Transistors tend to have lower hFE at low currents and increased 1/f noise after emitter-base avalanche breakthrough.

Do you mean the emitter resistors of Q7 and Q8? It must have been a compromise between noise (the bigger, the less noise) and voltage headroom (the smaller, the less voltage they take). The noise contribution is now of the same order as that of the input stages together. I designed it 16 years ago, so I don't remember it in more detail.

R1...R4 are needed for signal handling. Q1 and Q2 with R1 and R2 have to handle the full input signal without clipping. As long as they don't clip, their distortion is (ideally) cancelled by the equal distortion of Q3, Q4, R3, R4 in the feedback.
 
  • Like
Reactions: 1 user
Transistors tend to have lower hFE at low currents and increased 1/f noise after emitter-base avalanche breakthrough.
Can dramatically increase the noise in fact (20dB or so for prolonged breakdown or repeated peaks over time). You can partially reverse this by passing heavy forward current through the EB junction to attempt to free the lattice defects created in breakdown, but its not going to get the datasheet performance back. Low noise BJTs need those diodes from the moment the board is first powered up. Low noise bipolar opamps will include them.

Great resource here: https://www.uio.no/studier/emner/matnat/ifi/INF5460/h16/undervisningsmateriale/f8-1p.pdf
 
Can dramatically increase the noise in fact (20dB or so for prolonged breakdown or repeated peaks over time). You can partially reverse this by passing heavy forward current through the EB junction to attempt to free the lattice defects created in breakdown, but its not going to get the datasheet performance back. Low noise BJTs need those diodes from the moment the board is first powered up. Low noise bipolar opamps will include them.

Great resource here: https://www.uio.no/studier/emner/matnat/ifi/INF5460/h16/undervisningsmateriale/f8-1p.pdf
Hugh thanks for another se class a simple design. Ive built your 39 watt se class a amp and i love its sound. Youve clearlly dont your homework thanks again.