Noise shunt for 431 bias regulators?

[Deke609]

Recently, someone posted in the Monomour sub-forum about hearing hiss in his really sensitive tweeters. His solution was to replace the 431 bias regulator on the 5670 cathode with a resistor at the price of some gain which he didn't need, so all was good. See thread here: http://forum.bottlehead.com/index.php?topic=11953.30

This got me thinking. I don't hear any hiss from my Kaiju. But I wonder whether whatever noise the 431 reg might be injecting might be noticeable if it was gone -- e.g., "blacker background" or some such.  BUT I'm not keen on removing the regulated bias. I may well be wrong about this, but I would think that using resistor bias would make the 5670 less linear b/c the bias would vary with conduction through the tube (i.e., with signal). I don't want that. But I wonder whether some kind of noise shunt could be employed. I recall that the regulated DC filament supply in the BeePre uses a sophisticated noise shunt. I appreciate (in my limited way) that filament supply and bias are different beasts, but could something similar or simpler be employed to reduce any self-generated noise of the 431 bias regs? Just for example, maybe even something as simple as a film cap from cathode to ground in parallel with the bias reg?

This is more than idle curiosity on my part. I will be rebuilding my Kaiju and want to play around with reducing noise -- just to see/hear and to learn. I plan to regulate the B+ feeding the 300Bs using an SS reg that is reported to produce about 300 uV (0,3 mV) of self generated noise. I hope to be able to get that noise even lower by inserting the reg after two stages of LC filtering, and before the final third LC filter - hoping the latter will reduce the noise.  I think some experiments to reduce any 431 bias reg noise would complement these efforts.

All thoughts/input welcome.

many thanks, Derek

[Paul Joppa]

A bypass capacitor across the resistor is the standard approach. We used the 431 to avoid the sound of an electrolytic, so it's an interesting tradeoff.

The impedance of the 431 is very, very low, so a capacitor across it will have little effect unless the cap is huge. But comparing the resistor/cap to the 431 could be interesting.

An unbypassed cathode resistor increases the source impedance of the driver, but with the current-source plate load the gain is essentially unaffected.

[triode]

Hi Derek,

I'm the one who asked in the subject in MonAmour thread.

I did swapped the regulator to a resistor and the noise is significantly reduced to extend that I need my ear next to the tweeter to hear it.
This also true for my Stereomour, the hiss gone once I replace 431 to simple resistor. No bypass cap for both amp.

Beside reduce noise, I did not notice significant change in sound. Maybe a tad less top end, but it might be perception due to less hiss. Gain is not noticeable different.

My understanding, as Paul mention, due to C4S the bypass capacitor is less crucial. I might be wrong.

Regards


 

[Deke609]

Thank you PJ and Triode.

The impedance of the 431 is very, very low, so a capacitor across it will have little effect unless the cap is huge. But comparing the resistor/cap to the 431 could be interesting.

I think I'll compare the straight 431 bias reg with both: a resistor cap combination and a cap across the 431 reg - provided the latter is practicable. When you say "unless the cap is huge", are we talking something in the 10-30 mF range (which is doable), or something even larger (which is not)?  If something in the 10-30 mF range would potentially work to reduce the impedance even further, I thought I might "bypass" the large capacitance electrolytic with a much small film cap (say 1% of the 'lytic value), if only to reduce equivalent series resistance.

cheers and thanks, Derek

[Paul Joppa]

From the data sheet, the dynamic impedance of the 431 is about 0.22 ohms.So a bypass cap of 7000uFwould begin to be effective at 100Hz, and would reduce the noise at 1000Hz by 20dB - as longas its equivalent series resistance is less than 0.022 ohms.That might be audible.

[Deke609]

Oh excellent! Many thanks PJ. Very much appreciated. I think I can get 22 or even 33 mF Nichicon caps. I'll start looking at the ESR numbers. And a 10uF film cap in parallel should knock the ESR down considerably too.  I'll compare these bias schemes in my existing Kaiju and then decide on the approach for the rebuild.

cheers and thanks, Derek

[Paul Joppa]

Do the math for 10uF vs. 7000uF ...

[Deke609]

Aha, right.  It's equivalent series resistance -- so in series with the capacitance. So zilch benefit. Well ... I guess those 10uF Solen film caps bypassing some 22K uF electrolytic caps in the BeePre can come out!

Many thanks again, Derek

[Paul Birkeland]

Eh, it's not always that simple.  When you see switch mode power supplies you'll notice a big electrolytic cap bypassed with a small ceramic cap.  Old Dave can tell you about that...

[mcandmar]

May not be applicable to this application, but i experimented with 431 regulators to bias the 1st stage of my S.E.X amp and ultimatly gave up on the idea as the noise was excessive for headphone use.   The compromise i settled for for a HLMP6000 diode and resistor string in its place.  Frequency resposne was not as good as the 431, but was still an improvement over a resistor and ran quiet.  I documented most of it here https://forum.bottlehead.com/index.php?topic=5316.100

[Deke609]

Many thanks PB and Mcandmar.

@PB - I forgot all about that stuff! - despite having delved into it early last year when I was trying to figure out why the 300Bs in the BeePre have such huge bypass caps.  I think i understand it a bit better this time round - the EEVBlog video helped a lot. Thanks. I'm guessing that I don;t really need to worry about stuff in the 100's of kHz or even in the MHz range -- unless it could cause oscillation.  The 300Bs have grid stop resistors, so I'm hoping that they'll kill or at least tame any possible oscillation.

@Mcandmar - very interesting and helpful. I see that I am walking an already worn path.  I'll try to do some similar measurements of my experiments -- my grasp of how to use my O-scope is still quite shaky, but I'll give it a go. I may just go with ARTA since I've been impressed with how low a noise floor people seem to get when using it.

For the cap bypass experiment, it looks like the higher the capacitance the lower the ESR. The big 22K uF Nichicons I had in mind don't look all that impressive ESR-wise - but maybe I calculated it incorrectly. Working backwards from the reported dissipation factor, I calculated an ESR of 18 mOhms or so -- so under PJ's stated cut-off limit of 22 mOhms, but not great. However, I found some big Kemet 0.1F (a tenth of a Farad!) 40V caps that have a reported ESR of 7 mOhms. Kemet doesn't provide reactive inductance figures, but reports a maximum impedance @ 10kHz of 7 mOhms -- so I figure reactive inductance wouldn't be an issue (I don't think I can hear much above 14 kHz) - and I figure I could add a bypass cap to deal with any reactive inductance at higher audible frequencies.  Which I'll probably do if a can find a really low esr small value film cap (or perhaps a small ceramic cap?).

Yes, I am overdoing it. But (1) I am interested in this and using huge value caps should make it easier to identify/hear any changes in the sound profile, and (2) if there's an audible "improvement", the expense will be worth it. The big caps are a lot cheaper than the exotic iron and fancy interstage and parafeed coupling caps I've already sunk money into chasing even "cleaner" sound.

cheers and thanks, Derek

[Deke609]

Last night I installed a bypass cap across each A-side TL431 filament regulator in my Kaiju. I am pleased with the results. I haven't taken any measurements, just listened, so my impressions are subjective and potentially the product of "comfirmation bias".  But it sounds "better".  More clarity across the entire frequency range. The additional clarity is particularly noticeable with cymbals and vocals, but also with bass and bass drums.

I ended up going with a 25 VDC rated cap instead of a 40 VDC cap b/c, although the 40 had slightly better datasheet specs, the 25 VDC cap is about 1/2" slimmer.  Here's a link to the caps I used: http://www.mouser.com/ProductDetail/KEMET/ALS40A104KF025?qs=Ces2wY188b349PJKsEMT0w%3D%3D

Key specs: 100,000 uF, max 9 mOhm ESR @ 100Hz, max 8 mOhm impedance @ 10 kHz. [Edit: I just noteced that the ESR and max impedance specs are quoted for a temp of 20C, so the numbers are presumably higher when installed in the amp]

If anyone is inclined to try this, the following info may be useful - all of which is clearly set out in the manual and verifiable with continuity tests. +ve and -ve cap terminals can be wired directly to the A-side board.  Kreg(A) is connected to breg(A) and there is at least one unused breg(A) solder connection.  Likewise, -reg(A) is connected to the first two closest Cc solder holes. 

I haven't yet tried replacing the A-side TL431 reg with a 931R resistor. Perhaps the same results could be had this way for a lot less money and chassis space.  I've ordered them, but it may be a while before I try them. I have some big Kemet power supply film caps that are mounted above the reg boards, so getting at the boards and the 9-pin sockets will require a bit of dismantling. But I will definitely try the resistors before the Kaiju rebuild.
 
cheers, Derek

[Deke609]

On the off chance that you need to drop the 12V to 5V to get a directed 5V gracefully for your A/D converter power flexibly...

Nope.  But thanks.