Lowering power supply noise - looking for ideas

[SeDeuced]

I recently did some upgrades (in red) that significantly reduced the hiss I was getting from this. I know it can be improved even more, but I'm not sure which direction to go - more filtering, shunt regulator, finesse regulator or what.

I've been reading threads non-stop for the past three days and my head is spinning. I'm a newb at ckt. design, so any help would be greatly appreciated.

This supply feeds rails for an array of 5532s running on +/- 5VDC. The op amps are the input buffers for a 16 input audio switching matrix. There is a similar array of 16 5532s that buffer the output and run on +/-18VDC rails.

What is the easiest way to clean up the power?

Thanks for all comments.

[!]

Look at output on a scope. Determine where the noise is getting in. This is a schematic, what does the actual PSU look like? Is it a professionally designed PCB? Does it have a ground plane? Is it in a shielded metal case? Is the whole product it powers, old with possibly degraded mechanical controls or dried up electrolytic decoupling capacitors around the opamps?

You can get cleaner power replacing the LM78xx regulators with LM317 and LM337 adjustable, set to the right voltages of course.

On the other hand I can say that the design as shown, or similar enough w/o any significant changes, has powered lots of opamps for many people over the years with no hiss resulting from dirty power. If the PSU was old, possibly the capacitors were bad and it wasn't the capacitance increase that helped as much as just having properly working capacitors swapped in but I may be barking up the wrong tree since we don't know the other details, just that what you have pictured should be fine for a hiss free result if there are no other problems independent of the PSU schematic.

[SeDeuced]

Thanks for the reply. Some additional details. Reading products reviews from when the unit was new (circa 1989), reviewers mentioned this patch bay being somewhat noisy and that the signal needed to be hot in order to stay well above the noise floor.

I'm trying to improve it using modern parts, a better supply design and a "performance-first" objective versus a "use whatever is cheap to cut costs" objective.

The PSU appears to be laid out very well - ground plane, filter and decoupling caps close to input and output of regs. You can see the actual board here. I've replaced all of the caps on the pcb to eliminate any cap issues. The input op amps are actually under the I/O jack pcb (2nd to bottom photo). The op amps shown to the far right are the output op amps.

It is encased in a metal enclosure and ckt. gnd. floats from frame/earth ground. Hope that this helps.

My scope is not capable of measuring the noise, so it's difficult to pinpoint the actual source. Changing to LM78xx had a very noticeable impact on the hiss, so my guess is that LM 317s might be better and a super or shunt reg. design might drop it to a barely perceptible level.

I've read posts that recommend 2200 uF of filtering per amp of current. This design has nowhere near that much filtering. Is this part of the problem - too little filtering?

[sofaspud]

You could parallel a big filter cap on the bottom of the board while listening to the noise. This is a diy mod/refurbish that you're doing? Have you considered a hotter signal, perhaps including a little ps voltage boost on the input?

[infinia]

Hi
Determine if the noise is due to supplies 1st. Replace fixed voltage regulators with either discrete (like lab supplies) or adjustable 3 terminal regulators. The band-gap references in adjustable devices are not created equal. Brands sometimes matter. next step after above is passive filters not for novices.

[SeDeuced]

Sofaspud - Correct, this is a diy mod. The input signal is from a guitar pickup (~ 200 - 300 mV). I increased the amp gains to unity. Originally, the input gain was .25 which got boosted back up by a factor of 4 by the output amps. I was considering running the input amps from the 18v supply, like the output amps. The only issue is that the switcher between the two amps can only handle a signal up to 10v.

infinia - what brand have you found to work the best?

[AndrewT]

@ some outputs you show 18 & 5 which I presume are +5Vdc and +18Vdc.
But, some show +5A and -5A, what are these?

5532 on +-5Vdc is getting close to the lower limit of Supply @ >=+-3Vdc.
Compare the 5V performance to the 15V performance and realise you are throwing away a lot yet using a 5532 for higher quality.

If you intend to use two transformers then it would make sense to use a 9+9Vac for the +-5Vdc supplies and use an 18+18Vac for the +-18Vdc supplies.
Using a centre tapped transformer with a full wave rectifier for a single +5Vdc supply is nonsense !

[sofaspud]

That's 10Vp-p I assume. The 5532 isn't r2r so there is some wiggle room. Of course, the acknowledged first step in dealing with noise is "find the source." My suggestion was made in the context of replacing the regs with better devices (btw, I think infinia made a subtle reference to NatSemi's parts and Bob Pease "Czar of Band-Gaps"). A LM317 could set the input voltage a bit higher at the same time. If the noise is inherent to the design, the solution might not be one thing but rather a little bit here and a little bit there.

[Bob Richards]

Based on my experience, it's all about grounding technique. Tie the high current surge ground connections together independently from all other grounds that should go to a "star center", then have one wire at least a half inch long go from the high surge current ground junction to the "star center" ground point. High current surge grounds are the usually the transformer center tap, and the largest filter caps.

The star center can be the place to connect to the chassis so the chassis can act as an Rf shield, or some say noise is even lower when the chassis connection is near the audio input jack(s). Since I always float my input jacks, that wouldn't make any difference. Also, tie the 3rd prong earth ground green wire from the AC line cord to the chassis in a different place. That made a big difference and I'm not sure I can explain why.

[infinia]

I would investigate running all op amps from the same clean supply rather than mudding it back up again with fixed regs. s/b an easy jumper mod and some caps my need to pulled or changed depending on where they are. Look at the possiblity 15/-15V instead of 18, this will also make sure the regulators have enough headroom, but then they may need more heatsinking.
3T regulator brands I like Linear Tech and/or National Semi > hint don't forget to bypass the adj pin.