How to remove display noise on 5V rail

[T3sl4co1l]

Can you zoom in on those spikes?  Also does that not have a "hi res" mode or anything?  Very few bits in the acquisition there...

Only need a cutoff corresponding to somewhat longer than the time constant of that.  Even an RC filter might do, but CLC will do even better.

Okay, so 32 ohm load, and peak is what matters, and you have a square wave worst case so that's that.  No duty cycle, take peak not average.

Another thing to think about is short-circuit current, partly in case of actual shorts, and partly because loads can have weird impedances.  Opamps are usually rated on this, not very accurately, but say 30-60mA is a fine enough range for design purposes.

Then say delta less than 0.5V for say 20mA step load, and that'll be 0.5/0.02 = 25 ohms or less, so RLC around that; for 4.7uF that'd be under 3mH, which I don't think you're in any danger of hitting.   And then, enough ESR to keep it stable and that's fine.

Yeah, 7.5 ohms DCR is more than enough for 100uH + 4.7uF (Zo = 4.6 ohm), and at that load current, it's not much of a problem in terms of voltage drop. And even without those spikes being well resolved, some kHz is probably about right.

Tim

[Saimoun]

Can you zoom in on those spikes?  Also does that not have a "hi res" mode or anything?  Very few bits in the acquisition there...

I know Unfortunately not - It does have an oversampling which can help a tiny bit but still all you see is noise.

So I tried a few things:

• RCLC filter (if that's a thing  ), with 10 ohm, 4.7uF, 100uH (+1.5 ohm DCR), 4.7uF. That must have cut a bit, at least on the oscilloscope on the 5V supply the corners a tiny little bit more round. But listening to the audio output I cannot really hear much difference
• capacitor multiplier with 4.7k ohm and 4.7uF: that really works - the 5V supply actually looks clean on the oscilloscope, also the noise from the display is entirely gone, even at +11dB gain. The problem is the voltage drop which is quite high (drops to about 4.2V it seems). I mean the whole point is the make the audio louder, if I go through all this trouble to raise from 3.3 to 4.2... Maybe not worth it

I attached my circuit for the capacitor multiplier.

But now that the supply is clean I realized there is quite a lot of white noise (again it is not that bad - but the whole point of this is to reduce the noise so... ). I found out it comes mainly from the fact I used the Right channel of the DAC as the reference voltage HalfVCCA. The left channel is going through a 2nd order high pass filter (through the opamp). I tried using the TL431 - great IC! - and that solved that problem.

So new question is: is there a way to reduce that voltage drop in the capacitor multiplier? I guess the choice of transistor would matter? For now I used an NPN 2N2222A for testing.

[T3sl4co1l]

Sure, use germanium.

Or boost up the base voltage slightly.  Not trivial, as you already know regarding boost supplies.  Little current is needed at least, but still something.  And if you use any kind of converter or charge pump, you're injecting noise into that supply and it needs even more filtering.

I guess it would be fun, amusing, to use a photodiode, or a BJT in breakdown*, to try and source that extra couple 100 mV.

*This goes something like: E-B breakdown generates free charge carriers, and some light which is quickly absorbed nearby also creating free charge carriers.  The fraction of these which find their way into the B-C junction, basically act like a photodiode, thus generating an opposite voltage at very feeble current.  You'd need to do this with an RF transistor (Veb < 5V), and I'm not sure how much you actually get out of those types (since with the higher doping, diffusion distances are shorter; but the junction is thinner too, so maybe that doesn't matter?).

Is the DAC really noisier?  TL431 isn't especially quiet by itself... DAC shouldn't be any worse than its ref.

Tim

[Saimoun]

Regarding the DAC noise, I guess I could put a 100n (or even bigger) capacitor - do DAC's like capacitive loads?
I would say since this one is a R-2R type it's just a bunch of resistors so it should be ok, correct?

Regarding the photodiode/BJT breakdown, that sounds a bit out of my abilities 

And I guess changing the resistor or cap in the capacitor multiplier will have no impact on the voltage drop, right? (since it comes from the transistor)

[Saimoun]

I do feel like I'm going around in circles, always going back to: well I gotta make a boost converter 

Say I make a boost converter and then filter the output with a capacitor multiplier - can I count on the drop voltage to always be the same or should I add an LDO after?

[mawyatt]

If current draw isn't high then look at a LDO with really good HF PSRR rejection. The LT3045 was mentioned on another thread by NiHaoMike, this has superb HF PSRR rejection.


https://www.analog.com/media/en/technical-documentation/data-sheets/30451fa.pdf

Best

[T3sl4co1l]

Cmult voltage drop varies with load because it's a resistor into a diode drop, more or less; with extra voltage overhead besides, this will be even more irrelevant.  You still need stable VREF for the DAC of course.

Tim

[Saimoun]

@mawyatt: thanks for the reply - LDO would work great but I need (ideally) 5V from a USB Vbus 

Cmult voltage drop varies with load because it's a resistor into a diode drop, more or less; with extra voltage overhead besides, this will be even more irrelevant.  You still need stable VREF for the DAC of course.

I'm confused - so it does vary but not much for me to worry about it? My idea would be to make a boost converter to like 5.8V and then put a capacitor multiplier to end on about 5V. Then use this directly as opamp VCC.
I think for the DAC I will just use a 3.3V LDO, then I know its power is clean (since clearly the DAC has no PSRR whatsoever).

What about adding capacitive load on the DAC's output?

[T3sl4co1l]

It varies in two ways, well three:
1. PSRR is only good at high frequencies.  That makes sense, it's a filter.  So, it's still one-to-one at DC.
2. The output impedance (change in supply voltage for change in load current) is a diode drop in series, which dominates high frequency impedance, and contributes to DC resistance.
3. Drop across the resistor, roughly proportional to load (it acts as R / hFE at the output).

PSRR at AC is your primary problem, as I understand it, so these are okay.  The op-amp won't care about varying supply voltage, as long as it's not banging its head into the rails on peaks.  The DAC needs a constant REF, else the output gets AM modulated by the supply.

If the DAC is buffered, it likely wants a series resistance (47 ohms or more?) before a big capacitive load.  If unbuffered, go nuts.

Tim

[Vovk_Z]

7.5 R may be too much, I would begin with a smaller inductance with a smaller internal DC resistance.

[Saimoun]

1. PSRR is only good at high frequencies.  That makes sense, it's a filter.  So, it's still one-to-one at DC.
2. The output impedance (change in supply voltage for change in load current) is a diode drop in series, which dominates high frequency impedance, and contributes to DC resistance.
3. Drop across the resistor, roughly proportional to load (it acts as R / hFE at the output).

I see, makes sense

@Vovk_Z: why would you think that?

I'm gonna try and make a boost converter anyways.

I did some measurements with a multimeter, and it seems peak current usage is about 33mA. That was at the highest volume, constant sound and with 0ohm load on both left&right channels. I tried with 32ohm headphones and the max was about 22mA.
EDIT: did one more measurement, this time by putting a 3.3 ohm resistor in series with the 5V supply and measuring the voltage drop with the oscilloscope.
The peaks go to 280mV, so actually about 85mA peak! The 33mA from the multimeter makes sense since those peaks are only about a third of the time.

So I guess I should design the boost converter to handle about 100mA peak, and about 10mA RMS (i.e. 33/5 since in normal operation sound is only played at most a fifth of the time)?

[Saimoun]

Hello

Just giving an update. So I found this IC which is a synchronous converter already set to 5V, then I'll put an RLC filter after that.
It's very efficient because it's synchronous and I only need an inductor - so less things to choose wrong (avoiding smoking issues, like https://www.eevblog.com/forum/beginners/my-boost-converter-is-smoking-d  ).

Also it's designed for 300mA output current so much closer to my usage.

The design I decided is:

Code: [Select]

(original 5V from USB Vbus)
5V -> LED Display
   -> 3.3V LDO  -> MCU & other digital circuitry
                -> synchronous boost converter to 5V -> RLC filter -> audio opamp
                                                                   -> 3.3V LDO    -> DAC

Maybe that sounds a bit crazy to go up and down like that, but this way the noise from the display is definitely isolated, and since the converter is after an LDO I am not limited in capacitance like I would be on the USB VBus.
Maybe the last 3.3V LDO is unnecessary, I'll make the design so I can easily test powering the DAC directly from the 5V after the RLC filter.

I attached the (google translated) datasheet of the synchronous converter, called ME2188.

Regarding the design:

• Tim - I'll take your advice and use the 22uH suggested in the datasheet 
• But again I'm struggling knowing what current the inductor should be able to endure.
  * I know it's probably the fault of the datasheet giving very little info.
  * I also tried putting the values in http://schmidt-walter-schaltnetzteile.de/smps_e/aww_smps_e.html (I know it's for asynchronous, so taking this with a pinch of salt), using 3.3V input, 5V output, 22uH, 0.1A output current, 120kHz (since the IC does PFM and I guess it's 320kHz at 0.3A), that gave me an indcutor ripple current of 0.43A. So a switching current of 0.38A.
  * So I found a 1210 inductor, 620mA, 22uH ±10%, with DCR= 270mΩ - that should be ok I think?
  
• For the caps I'll use 4.7uF MLCC at the input of the converter, and an electrolytic 47uF at the output.
• For the RLC filter I'll put a 1 ohm 0603 to avoid oscillations and then just reuse the same parts as the converter, 22uH and 47uF.

That's a lot of info - hope that's ok I'm thinking it might help someone that passes by this thread in the future


My specific questions:

• The IC recommends to use tantalum capacitor for the output - is it a bad idea to use electrolytic?
• It also recommends a 0.1uF ceramic on the output (before the big one I guess) - never saw that on a converter IC before... ?
• Am I in the ballpark with my 0.62A inductor or completely off?

Thank you!

[T3sl4co1l]

Cute, a 3-terminal sync reg!

So inductor ratings, low enough resistance that it's not annoying of course (within thermal ratings certainly, and any efficiency gain you want on top of that), and saturation current >= peak current used.  So, max DC (again, of the regulator!) plus peak ripple (or half of peak-to-peak or delta I).

In that size, multilayer ferrite is probably fine, otherwise most anything wound ferrite or composite powder should be fine.  Small chip inductors come in all sorts of styles, take your pick.  Raw-ass ferrite beads wouldn't be suitable (not that you'd find one with so much inductance, nor sold as such; multilayer FBs are made exactly the same way though, so, just that you won't know by eye in that case, don't let them get mixed ), and powdered iron toroids can be an issue as far as losses (not that they'll be a problem for immediate purposes either -- you'd be pondering those for 100s W capacity converters).  Shielded is usually better than unshielded; it kinda doesn't matter for tiny stuff like this, but is nice for low noise, and power stuff where not only does it reduce emissions but an open-core inductor with a lot of voltage on it might start cooking (induction heating) neighboring components.

Unfortunately the datasheet doesn't really tell you what or what things need to be what they are.  No internal compensation is listed.  PFM could be as simple as hysteretic control again, or it could be an independent oscillator skewed by operating conditions.  Also not obvious if it's synchronous rectified, or just free running (always driving LX high/low); efficiency seems reasonable to lowish currents so maybe it's the former.  It could also be a burst mode control, with consequent higher ripple.  It does say very low current at no load, suggesting some kind of burst mode.  But no way to really tell from here.  Such is the downside of shopping in China...

That they suggest tantalum, implies they want ESR.  Electrolytic can be used as well, just shop around for something comparable (0.1 to 1 ohm ESR?).  Ceramic can also be used with external ESR.  If it works fine with ceramic alone (do a step transient test -- again, a scope is extremely useful here ), well there you go.

Also, you probably want one a little step up from that, with only the 6V rating?  Since you were looking for 5.9 or so.  I suppose it doesn't matter as it's fixed 5.0V anyway, but more might be nice for the overhead you wanted.

Tim

[Saimoun]

Thanks again Tim and apologies for the late reply 

So inductor ratings, low enough resistance that it's not annoying of course (within thermal ratings certainly, and any efficiency gain you want on top of that), and saturation current >= peak current used.  So, max DC (again, of the regulator!) plus peak ripple (or half of peak-to-peak or delta I).

I'm still confused - the max DC current of the regulator, is that 300mA (output current) or 1A (LX max current)?

Or is it not possible to know because the datasheet is wonky (I mean it does say absolute max LXpin current 1A and then later on "limiting current" max 1.2A...)?


Ok about the capacitor - thanks

And as for the 5V output - because here the regulator is after the 3.3V LDO I do not need to remove the Display noise anymore, but only the switching regulator's noise, so the RLC filter should be enough.



Say I was to change that with a non-chinese part, i.e. a regulator with:
* 5V fixed output
* synchronous
* max output current about 150mA
* small and cheap

which part would you recommend?