What Degrades Channel Separation in a Stereo Power Amp?

[Ron AKA]

The reason I ask, is I am refurbishing a Hafler DH-200 power amp I assembled in the 70's (vinyl & tape days). The specifications are still reasonable for today's CD specifications, except for Channel Separation. Hafler says it is 66 dB at 1 kHz. That is fine with the 30 dB or so you get with vinyl, but a lot short of the 110 dB that a CD player can deliver.

This amp has a common transformer/rectifier/filter capacitor arrangement. After that the left and right boards are separate and located on opposite sides of the amp. I'm wondering if there are some simple things I could do to improve separation?

I looked at the circuit layout of a later Hafler design (DH-220), and it claims 85 dB separation. The only difference I can discern that could have an effect are 0.1 uF film filter caps on the power supply rails of each channel. Would that make a difference? Would be easy to add. Anything else?

[jcx]

even -66 dB shouldn't be a problem but the most likely issue would be common impedance coupling in the ground system

hierarchical gnds, slightly modifying a less well done "pure "Star gnd" can help - some times even just a brute force heavy gnd braid (how about Cu roof flashing?) gnd conductor

you need a "gnd schematic" that also shows physical implementation, location, order of "gnd" connections for ps, chassis, signal, boards, speaker terminals

[DF96]

Channel separation is set by unwanted coupling between channels. Given that many people are happy with vinyl, my guess is that you will not be able to tell the difference between 66dB and 85dB except on test signals (not normal music).

Adding capacitors in the right place may help. Adding them in the wrong place, or with the wrong grounding, could make it a lot worse as you might introduce coupling.

[Ron AKA]

Originally Posted by jcx

you need a "gnd schematic" that also shows physical implementation, location, order of "gnd" connections for ps, chassis, signal, boards, speaker terminals

On page 17 of this document you can see the physical layout of the DH-200 amp. I've compared it to the DH-220 and can't see any difference in the grounding -- or at least to my unskilled eyes. Any thoughts on weaknesses in this layout?

[Ron AKA]

Originally Posted by DF96

Adding capacitors in the right place may help. Adding them in the wrong place, or with the wrong grounding, could make it a lot worse as you might introduce coupling.

This National Semiconductor article on Audio Amplifier Power Supply Design has a circuit with many additional caps that are not found on the Hafler design. Any thoughts on which ones may be most helpful to add?

[jcx]

have you measured the crosstalk? - a soundcard, R divider attenuator for the amp out to soundcard line in and RMAA is one quick way - if the crosstalk is flat with frequency then it likely isn't going to be fixed with capacitors - because gnds in the amp may be involved you may want isolation xfmr for the soundcard - even cheap Radio Shack ones wil be OK for mid band audio - ESI Juli@ soundcard has balanced inputs which would help the measurement if you have a older PC

if you don't measure it then you'll never know if your changes fixed anything


you do have pretty good documentation – did I miss the chassis safety gnd?

I am hesitant to give advice on unsafe mains amps – if it Class II the xfmr must be rated tested for 4000 V HiPot pri-sec and the mains entry/xfmr pri wiring needs additional insulation/separation, redundant fixing so that if any soldered connection comes loose it still has one more separation system preventing contact with any secondary circuit or case metal – paint doesn’t count as an insulation system

just sayin…

[Ron AKA]

Originally Posted by jcx

have you measured the crosstalk? - a soundcard, R divider attenuator for the amp out to soundcard line in and RMAA is one quick way - if the crosstalk is flat with frequency then it likely isn't going to be fixed with capacitors - because gnds in the amp may be involved you may want isolation xfmr for the soundcard - even cheap Radio Shack ones wil be OK for mid band audio - ESI Juli@ soundcard has balanced inputs which would help the measurement if you have a older PC

if you don't measure it then you'll never know if your changes fixed anything


you do have pretty good documentation – did I miss the chassis safety gnd?

I am hesitant to give advice on unsafe mains amps – if it Class II the xfmr must be rated tested for 4000 V HiPot pri-sec and the mains entry/xfmr pri wiring needs additional insulation/separation, redundant fixing so that if any soldered connection comes loose it still has one more separation system preventing contact with any secondary circuit or case metal – paint doesn’t count as an insulation system

just sayin…

Don't have the equipment (or more honestly the smarts either) to measure separation. I'm an ME not an EE!

The way I see the ground configuration is that the primary of the transformer is floating, as the plug is non polarized, and could be plugged in either way. The center tap of the secondary goes to the chassis ground and is the circuit board ground. I don't see any evidence of Underwriters Lab, or CSA safety certification, so it may not be as safe as it should be. Perhaps Hafler got around that by selling it as a kit. Does anyone run a three wire grounded supply to the amp, effectively connecting the chassis to the house ground system, and real ground. Or does that just create more problems? It would seem to give the supply fuse a chance of keeping the chassis from getting hot in case of a failure of the primary transformer insulation.

Based on specs the DH-200 separation is 66 dB at 20 Hz, 1 kHz, and 60 dB at 20 kHz. The DH-200 is better at 75, 85, & 65 dB respectively.

[jcx]

if forced to "do something" that might help crosstalk without changing holes, mounting positions I would use 10 oz or heavier Cu roof flashing (unless you can get electrical bus bar stock) to make a super low resistance gnd for all of the ps cap gnd , speaker out gnd and RCA input gnd - one single sheet, as wide as can be fit

should be right up a ME's alley - cutting metal, bending, drilling, tinning with flux, torch...

(swapping speaker gnd/out binding post positions to put the 2 gnds in the middle saves some notching the sheet metal, is better electrically too)

[ingenieus]

Something I read somewhere - if the coils in the Zobel output network are too close to each other it leads to increased crosstalk. When they cannot be far apart due to space limitations they have to be at a 90 degree angle.

[jrockhead]

Originally Posted by Ron AKA

Don't have the equipment (or more honestly the smarts either) to measure separation. I'm an ME not an EE!

You might have all the stuff to do it. Got a computer running XP? Got a decent sound card in it? Can you download a program called Rightmark Audio Analyzer and install it ?

If answered yes to all these questions you can measure separation and a whole bunch of other stuff.