Improve Ripple Rejection on Scully Regulator

[bjoneson]

Been working on lowering the noise floor on my Scully 8 track machine. FFT shows single largest component at 60Hz (with decreasing harmonics). Some more analysis revealed ripple rejection at the regulator circuit was pretty poor. I did get about a 10dB drop in the noise floor by replacing and enlarging the main filter caps on the unregulated rails in the main PSU (were originally 2900uF, brought them up to 10,000uF. Transformer and diodes are up to inrush).

Trying to figure out if I should just scrap the existing regulator circuit and replace with a stock LM338 circuit, or if there's anything worth tweaking in the existing topology in terms of modern components that might help.

Schematic attached.

 

[joaquins]

  Is the problem on both sides or just on the +25V side?

  I'd probably add an RC filter between C16 and R35 to have lower noise supply for the control circuit, as well as increasing C17 and C19. Increasing the gain of Q13 also could improve the rejection, but over doing so could bring stability issues.

  For the negative side increasing C22 could help but as it's following the possitive side it won't be any good if you have issues there.

JS

 

[ruffrecords]

Assuming the rectifier is full wave, there should be no 60Hz component in the raw dc, it should be 120Hz plus harmonics. Either the rectifier is part blown or your hum problem is of an entirely different kind.

Also check/replace C19 to ensure you have a ripple free reference voltage.

Cheers

Ian

 

[bjoneson]

Gah... you are of course exactly right. Filter caps did significantly effect that 60Hz component, so suspect you're right. Something very amiss with the unregulated supply. Will do some more measurements and scope that out.

Many thanks,

Bob

 

[JohnRoberts]

Do you measure 60Hz on the DC regulated voltages?  Noise vector may be elsewhere, or additional.

JR

 

[bjoneson]

It's measured at the line output. All channels (8) exhibit similar noise characteristics. I feel like the filter caps having reduced that significantly points to the power supply as a primary vector, but I'm definitely interested in alternative theories. I'll take some more measurements and post more data tonight.

Many thanks,

Bob

 

[radardoug]

Is it measured in repro or input? Connected to heads? What level is the hum relative to +4?

 

[bjoneson]

It's measured in both repro and input monitoring. Peak was about 40dB below reference (+4dBu), before replacing filter caps, and about 50dB below after replacing filter caps.

 

[bjoneson]

Here's the FFT from before the filter caps change. (-18dBFS=+4dBU). This is 1k monitored from input to output.

https://www.dropbox.com/s/mmuzaiyxar0j0hs/SCULLY-CH1-LOOPBACK-20161127.png?dl=0

 

[ruffrecords]

That is definitely much too high. I suggest you ac couple to the raw dc (via a film capacitor and 100K bleed resistor) and look at its spectrum. If there is any 60Hz there then you have a rectifier or transformer winding problem. If its all 120Hz and above then the problem lies elsewhere.

Cheers

ian

 

[bjoneson]

Makes sense. Thanks again for all the great info. Will let you know what I uncover.

 

[bjoneson]

Ok, got some more probing / measurements done. I was able to use the AC coupling function on my scope to view the ripple on the unregulated supply, and judging by the result, don't need an FFT.

This is the positive unregulated rail (spec'd at +35V) DC measurement is +32.7V. Scope is set to 5ms, 0.2V / div. I think that means the ripple is on a 60Hz interval? A little over 3 divs (15ms p-p):

https://www.dropbox.com/s/rl6oude8dxvilc4/20161130_160635.jpg?dl=0

This is the negative unregulated rail (spec'd at -35V). DC measurement is -39.3V. Same scope settings as above.

https://www.dropbox.com/s/8k96u1dln0gsx0n/20161130_160952.jpg?dl=0

The difference in shape is intriguing to me, in addition to the voltage variance. I think it's safe to assume this likely reflects a rectification issue?

Schematic of the main unregulated supply is here:

https://www.dropbox.com/s/rx3eobhqtnz8myf/20161130_161045.jpg?dl=0

The 2900uF filters have been replaced with new 10,000uF caps.

Out of curiosity what does the .33uF cap across the rectifier do?

I've got spare diodes around (have a boneyard of Scully parts). Absent other thoughts / ideas, will probably swap those out next.

 

[Audio1Man]

C202 (+35VDC) is bad or ? The waveform should be similar to C201 (-35VDC). Is the cap connected?
Duke

 

[bjoneson]

I think the weird behavior is due to something being wrong with the rectifier. Caps are both charging or there would be pure AC on that rail.

 

[bjoneson]

Holy crap.... well.... this is a bit embarrassing. Got to looking closer tonight, and one side of the transformer secondary was disconnected (just one end and center tap were still connected). Needless to say.... "there's your problem".

I can't thank folks enough for their willingness to lend an ear here. I did gain some "common sense" wisdom / knowledge out of this experience.

 

[bjoneson]

Much more sane FFT now. (120Hz dominates noise, and MUCH lower):

https://www.dropbox.com/s/3gforwyjsuqxg2g/SCULLY-CH8-LOOPBACK-201611230.png?dl=0

 

[ruffrecords]

Out of curiosity what does the .33uF cap across the rectifier do?

Regular rectifier diodes do not turn off instantaneously. This results in very small spikes at twice the mains frequency. They easily propagate past the regulator and into your audio path. The 0.33uF (usually a film cap) snubs these spikes at source.

Cheers

Ian