SDS Labs headphone amp hum

[Daroid]

I have finished building this amp, but in version 1.2: http://headwize.com/projects/showfil...stokes_prj.htm

I use a gain of 5 with OPA627BP with a opamp biasing current of 4.25 mA per opamp and around 70 mA biasing current for the mosfets which are IRF620 and IRF9620. The voltage regulators are LM317/337 supplying +/-15V.

One slight problem that is very annoying - hum!, there's slightly more in left channel than in the right, it's it not deafening loud, only moderate to faint. (The left channel is in the same side as where the LM317 is). If i turn up the volume (on the Alps RK27 100k pot) it will reach some point where the hum increases in volume and the goes faint/moderate again. At highest volume setting it give me a "blob" and some very faint sweeping frequencies (in the 5-10 kHz spectrum i guess).

DC offset is little to non-existant (atleast hardly measureable)
pot is grounded correctly with shielded cable and one place has been used for ground - the area surrounding the components on the PCB (see PCB layout: http://headwize.com/images/stokes4.gif) Inputs are also grounded...

Everywhere the parts list listed 4700 uF electrolytics, 1000 uF has been used instead - still, it takes some seconds to reach the +/-15V so there's plenty of power reserve.

What i HAVE tried to avoid hum:
- Disable the op-amp biasing by removing the lone JFETs (and current limiting resistors for it)
- Use some tricks from AMBs site: 0.1 uF ceramic discs were soldered to + and - of the op-amps and 72 nF was applied to either of the gain resistors in parallel (R11/R61)
- moving any ground wires farther away from the toroid doesn't make a difference.

nothing eliminated hum.

Suggestions and questions are welcome and much appreciated !

 

[amb]

First, what headphones are you testing with? Lo-Z and highly sensitive? When you're doing this test, do you have a signal source (e.g., CD player, etc) connected to the input? Without a source, as you turn up the volume it's bound to pick up some induced noise unless you short circuit the input to ground.

Second, the weird noises you describe as you turn the pot sounds suspiciously like oscillation. When I built this amp I found that the voltage regulators were oscillating, and I had to add a 0.22uF tantalum cap directly to each regulator's output to ground to solve that problem.

Also, the OPA627 is a single opamp. Sheldon's PCB design calls for a dual opamp. Are you using a Browndog adapter? If not, then all bets are off... If you are using an adapter, maybe the adapter/socket added enough inductance to the circuit to make the amp unstable. Try a more benign dual opamp first, like the OPA2132 or OPA2134.

Another suggestion is to add a series resistor (about 470 ohms) between the pot's wiper and the amp's input.

72nF as compensation caps? That's way too large a value, and will cause severe treble rolloff. Do you really mean 72pF?

Lastly, with all due respect to Sheldon, his pcb layout has the power transformer in close proximity to the opamp and an induced hum is likely if you listen with sensitive headphones. Unfortunately, there is no "fix" other than to redesign the pcb, or to remove the transformer and put it in a separate case, located away from the audio circuitry. Make sure all AC wiring is as far from the audio input and volume control wiring as possible.

 

[Daroid]

Hi,
Headphones used for testing puposes (i.e. doesn't matter if they break): senn HD56 - 32 Ohms, 106 dB/mW, and sony MDR-G56 - 24 Ohms, 104 dB/mW. Also tried HD25 briefly but not noticeably more hum than with the other 'phones.
Currently using the amp out of a digi 96/8 pad soundcard, I'll try with a CDP tomorrow.

An hour ago a small piece of bare wire had found it's way onto the underside of the pcb and shorted the LM337 directly to ground (all shops are closed now and no 337Ts left to use instead - my luck

). As soon as it's replaced i'll try your tip with the 0.22 uF tantalums on the regulators!

oops, should have mentioned that the PCB layout was modified to fit two mono-opamps instead of a dual, but they're still mounted in DIP sockets. Yes the 72nF was a mistake, it is 72 pF (3x222pF in series).

Thanks

 

[amb]

Quote:

Originally Posted by Daroid Headphones used for testing puposes (i.e. doesn't matter if they break): senn HD56 - 32 Ohms, 106 dB/mW, and sony MDR-G56 - 24 Ohms, 104 dB/mW. Also tried HD25 briefly but not noticeably more hum than with the other 'phones.

Yup, lo-Z and fairly sensitive cans...

Quote:

Currently using the amp out of a digi 96/8 pad soundcard, I'll try with a CDP tomorrow.

Just to rule out any problem with possible DC offset from your soundcard causing the noise, try just shorting the inputs to ground (i.e., connect the signal pins at your RCA input jacks directly to the ground tabs), and rotate your volume control to see if there is still a problem.

Quote:

An hour ago a small piece of bare wire had found it's way onto the underside of the pcb and shorted the LM337 directly to ground (all shops are closed now and no 337Ts left to use instead

Oops.

Quote:

should have mentioned that the PCB layout was modified to fit two mono-opamps instead of a dual, but they're still mounted in DIP sockets. Yes the 72nF was a mistake, it is 72 pF (3x222pF in series).

OK, but using three caps in series to drop the capacitance is not a great idea, especially for the compensation caps. It makes it much less effective due to increased series impedance.

If you're going to be adding tantalum caps, you might want to put one at each of the regulator's inputs too. Be sure the caps are rated 35V or higher here, because the next step down would be 25V, and that's a little too close for comfort. Also, tantalums are polarized so be sure to put them in correctly.

 

[Daroid]

Is it very important that they are tantalums ? The only tantalums i currently have is 4.7uF and 16V so that won't work. I have a bunch of 0.1uF ceramics (100V) if that would be just as good with two in parallel...

 

[amb]

No, I used tantalums only because of their size/capacitance ratio. Also they have better HF characteristics than the standard variety aluminum electrolytics. The datasheet of the LM317/337 regulators actually recommends a 1uF capacitor at the output pin of the regulator. Such a cap, either in film type or ceramic would be quite large.

 

[Daroid]

Hi, just to post an update.

The amp hum has disapeared completely and is completely silent (no hiss/white noise) to near maximum volume with a gain of five.
The voltage regulators didn't oscillate. The dead 337T was replaced with a new 337SP ($3.5

).
There were more hum in one side - and it was all caused by the fact that the R5 connects to ground where it does. Problems were pretty much solved by cutting that trace to ground and direct it to the pads where the bypass caps (C9 & C10) goes to.
Using that large area as a ground for everything yelded great improvements. Then afterwards soldered a large gauge wire to that area and to where it says "Amp" in "Headphone Amp 1.2.

You were right, many capacitors in series isn't a good idea, it also introduced noise here, so cutting capacity in parallel with R11 down to 222 pF (1 ceramic)solved the problem and reduced hum to inaudible levels. I needed to add an input cap which lowered hum even more (only 100 nF, so -3dB "threshold" will be below 2Hz).

This is one absolutely excellently sounding amp a great improvement on headphones which shouldn't need extra amplification. It has a mellow and fluid signature to it.
After building this i clearly understand the KISS philosophy and how this amp could use a lot of that when it was designed.

. Thank you for all your help

 

[amb]

Yeah, it also bothered me that there is that little "island" of ground in the PCB layout connected to the rest of the peripheral ground area through a mere little jumper. I think I'll try the same mod to my board and see if it makes a difference here.