How much ripple rejection in PSU for high gain micpre's?

[volki]

hi all,

i'm about to build a 44V unipolar + 20V bipolar PSU for some channelstrips i have racked up. the 44V (less than 200 mA) go to discrete modules with max. gain of 75 dB, the +/- 20V (about 800 mA each) to op-amp modules with max. 60 dB gain.

my concern is somewhat related to using different types of voltage regulation in the PSU. here are my my questions:

- how much ripple rejection (or better, overall noise) of the PSU would be reasonable for the different types of modules for practical applications (i.e. taking other preceding noise sources into account, like mic's, recording room itself, etc.) ?
- i figure that 75dB max. gain would call for more rejection than 60dB, provided all other things being equal?
- would I be fine with Vregs such as L78/79xx for all app's (65 dB rejection), or should i go with LM3x7 (80dB with reg leg bypass cap)?
- ripple rejection is just one factor of overall Vreg noise... are 78/79xx considerably worse in broadband noise performance, too?
- are unipolar powered circuits less sensitive to PSU noise than bipolar powered ones, or is that nonsense? ;-)

why don't i go with LM3x7 in the first place... well, i have this 2x17,5V xformer here, which would give me around 23,5V DC after the rectifier bridge - obviously still good for L78/79xx to regulate down to 20V, but NOT for LM3x7, which i understand need 4 to 5V difference for proper regulation?

thanks for any input!

cheers,
volker

 

[Samuel Groner]

How much ripple rejection (or better, overall noise) of the PSU would be reasonable for the different types of modules for practical applications (i.e. taking other preceding noise sources into account, like mic's, recording room itself, etc.)?

I figure that 75 dB max. gain would call for more rejection than 60 dB, provided all other things being equal?

Answering this questions is impossible without knowing the actual PSRR of the circuit at hand. As these are console strips you can be quite sure that PSRR is not that shabby as otherwise making the console quiet would be a pretty difficult task. I would be surprised if a standard linear PSU wouldn't do the trick. Chances are that radiated hum and noise will be much more of a problem unless the PSU got its own rack and/or a big hunck mumetal screen and proper diode snubbers.

Would I be fine with Vregs such as L78/79xx for all app's (65 dB rejection), or should i go with LM3x7 (80dB with reg leg bypass cap)?

Ripple rejection is just one factor of overall Vreg noise... are 78/79xx considerably worse in broadband noise performance, too?

I use the fixed versions for auxiliary use only as the adjustables are--as you say--better with respect to noise and ripple rejection. You'd need to hack the fixed versions anyway as you won't get them with the voltages you need.

Are unipolar powered circuits less sensitive to PSU noise than bipolar powered ones, or is that nonsense?

Unipolar requires a few more thoughts but I don't think that they are inherently more sensitive. A precise answer is difficult though as it is a bit of comparing apples with oranges.

Well, i have this 2x17,5 V xformer here, which would give me around 23,5 VDC after the rectifier bridge--obviously still good for L78/79xx to regulate down to 20V, but NOT for LM3x7, which I understand need 4 to 5V difference for proper regulation?

Why guessing about drop-out voltage? It's all in the datasheet. Both drop at 3 V and show somewhat increased ripple rejection at 4 V or 5 V more IIRC. However you forgot about ripple voltage (about 1 V for properly chosen smoothing capacitors) and mains voltage drop (10%). With those figures you'll end up at 20.3 V. You'd need a discrete ultra-low drop-out regulator to catch this. You can bet on less mains voltage drop and/or sufficient regulation of the transformer but personally I'd go with a different part.

BTW, I just recently posted an Excel sheet for this sort of calculations: PSUCalculator.xls

Samuel

 

[volki]

samuel, thanks for your thoughts. especially the hints in your last paragraph, i hadn't thought of that!

cheers,
volker

 

[JohnRoberts]

At the risk of giving you an oversimplified answer.. yes it varies with different topologies and component choices, but most noise sensitive circuit block (mic pre, sum bus, etc) designers use simple PS filters (series R big cap to ground) in addition. GP 3 terminal regulation will reduce/remove large ripple voltages but RC will knock down residual noise pretty well.

JR

 

[Samuel Groner]

Check the label in cell 13C. I think it should be "VA" not "VAC".

Indeed, thanks for the hint. Took the chance to elaborate some of the comments in the spreadsheet as well.

Samuel

 

[volki]

one more question: an alternative Xformer i have is 2x 28VAC secondary. to regulate 800 mA down to 20VDC, each Vreg would have to dissipate close to 20W. even with a big heatsink, i'm fearing my TO220 packages might get too hot to deliver the needed current (datasheet of LM3x7 specifies only guaranteed 400mA for dropping voltages greater than 15V).

i figure it would be best to put a dropping resistor (around 10R) btwn rectifier and Vreg, so as to distribute voltage drops and hence power dissipation. adding a capacitor to ground btwn rectifier and dropping R would then form a pi filter together with the bypass cap at the i/p of the Vreg.
now:
- should i place the big charging cap (approx 10000uf) between rectifier and dropping R, or btwn dropping R and Vreg?
- or should it be 4700uF - 10R - 4700uF - Vreg?
- or, even, would the additional filtering allow for a smaller charging cap?

sorry if this has been asked before, i couldn't seem to dig up any answers to this one in the archives...

again, thanks a bunch,
volker

 

[Samuel Groner]

Perhaps the best place would be between transformer and bridge. This will greatly reduce peak currents in the transformer and hence reduce stray field (I've measured 6 dB reduction for a 1 V drop resistor)--not sure how good the other possible configurations are with this respect. Another advantage is that you can get away with lower voltage rating of the capacitors.

Samuel

 

[JohnRoberts]

[quote author="Samuel Groner"]Perhaps the best place would be between transformer and bridge. This will greatly reduce peak currents in the transformer and hence reduce stray field (I've measured 6 dB reduction for a 1 V drop resistor)--not sure how good the other possible configurations are with this respect. Another advantage is that you can get away with lower voltage rating of the capacitors.

Samuel[/quote]

Indeed the resistor can be effective after the bridge too as long as it is before the reservoir capacitor. This is useful to throw away extra voltage with less power loss than putting the resistor after the cap. I have also used this topology to power limit the current drawn from a wall mount transformer to prevent it's thermal fuse from opening in use.

JR

 

[bcarso]

One added note: be wary of the actual waveform of your local power company and immediate environment. Take a (careful!) look at the waveform out of the wall and note the peak voltage. This, less all the other voltage drops and bulk cap ripple is what your regulator has to work with.

As we move forward into more and more consciousness of, and requirements on, how we draw power from mains, the situation may improve and we may have more sinusoidal power again. But the prevalence of cap-input filters on PSUs and high-conductance diodes at present means often the AC is heavily clipped, so you don't get your anticipated paek voltage charging your own cap-input circuits.

 

[volki]

thanks for all your input everybody!!! :thumb: