Phantom from a boost converter

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Rochey

Well-known member
White Market Member
Joined
Jul 2, 2004
Messages
2,697
Location
Dallas, TX - Formerly UK
evening all.

I’m on the road, sat in a hotel restaurant, chewing on dinner and on small circuits. I’m thinking about making a phantom power source from a simple boost converter - from a 5V or 9v source.

The boost converter I have in mind is running at 400kHz, way beyond audio range, but I still worry about noise.

I was thinking of adding a series inductor followed by a capacitance multiplier after it, before throwing it into a lm317 (with a zener to keep it in spec).

Overkill? Not enough? I’d like enough current to drive 4x mics

Cheers

R
 
I don't know if he is still lurking around but IIRC Andy Peters made one powered from 5V USB port, several years ago, I don't recall any details.
=====
Back about 15 years ago I made a phantom supply from a switching supply running from around 12v DC(?). I hung a variant capacitor doubler/tripler off the HF 12V square wave in the switcher. Because it was running at HF I could use tiny SMD caps in the charge pump. I followed it with a typical zener shunt regulator after getting well above desired 48V DC.

I don't recall using any extraordinary filtering, or any reported problems.

JR
 
 
Lots of products now use a SMPS for phantom power (and other voltages) Way back in the 70s, the Neve phantom power plug in module used a SMPS to generate 48V from the 24V supply. It can be done.

Cheers

Ian
 
It can (will) work, I have done it several times. LM317 is useless since its ripple rejection is very low at 400kHz (<30dB), a simple RC (150 Ohm, 220uF) will give you a better and more than enough filtering for the phantom power supply (and there will be an other stage of filtering inside the mic).
 
Rochey said:
I was thinking of adding a series inductor followed by a capacitance multiplier after it, before throwing it into a lm317 (with a zener to keep it in spec).
LM317's minimal current requirement is 15mA IIRC. That's a significant waste of energy. In addition it takes special precautions since its max input voltage is 40V. LM783 is a high-voltage version (up to 125V) but still requires 15mA for nominal operation.
Microchip LR8 would be  a better choice.
Anyway, Chris-V's comment is accurate. A linear regulator after the converter will not filter noise, but it would make the voltage constant. P48 standard is quite lax on that respect; +/-4V is close to +/- 8%, which a properly regulated switcher is capable of achieving, unless you want to use a free-running converter, which is not a bad option, in conjunction with a shunt regulator..
 
In my experiments I've found an LC filter to be overkill, more expensive, and more painful to design and select inductors.  Remember that at 400kHz, a simple 100 ohms into 100u low-pass filter is nearly 80dB more effective at filtering ripple than it would be at filtering 60Hz hum.  Also, ripple voltages tend to be specified at significant output currents (like >500mA), which you won't be approaching even with 4 phantom power loads, so as long as you keep the output C low ESR then the ripple pre-filter won't even be exceeding a few mV, even before the RC filter.

I also don't think a linear regulator is needed either:  most boost SMPS's regulate to within 1% via trimming resistors, so you aren't gaining any additional accuracy by using am LM317 afterwards.  I've had great success using a MAX1771 from a 9V supply to 48V:  in fact, this same module will do tube supplies as well at ~250V.
 
ruffrecords said:
Lots of products now use a SMPS for phantom power (and other voltages) Way back in the 70s, the Neve phantom power plug in module used a SMPS to generate 48V from the 24V supply. It can be done.

Cheers

Ian

Here is the Neve circuit, more for historic interest than anything else.
 

Attachments

  • Neve 3601.pdf
    386.5 KB · Views: 54
I went to sleep and woke up to this SLEW of replies! thank you very much gents.
That Neve schematic is fascinating! How that oscillator works without any caps is cooking my brain a little.

In addition to this awesome advice, I'd like to ask the collective brain to reason with me on some assumptions.

ASSumption #1:
Any noise generated in the phantom should technically be common mode, as it'll appear on both legs of the differential input to the mic pre - AC coupled of course, through the AC coupling caps. Any offset/difference between them will be caused by a difference in the 6K8 resistors typically used for phantom. At 1% matching on the resistors, the attenuation of noise could be calculated.

ASSumption #2:
I like the simple RC filter suggestion. Does having the series R have any impact on the circuit, other than dividing down (voltage divider) with the 6K8 resistors? I can't think of any bad things, as the phantom spec is pretty loose.

ASSumption #3:
Adding the feedback to the DCDC post filter is the way to go here. Pure and simple. Thoughts?

ASSumption #4:
A simple capacitance multiplier using a resistor, cap and NPN could add significant filtering in theory, so that no-one can bitch at me about adding noise to their precious phantom - and cost pennies. Any thoughts on why/if that would be a bad thing?

Thanks again

R
 
Rochey said:
I went to sleep and woke up to this SLEW of replies! thank you very much gents.
That Neve schematic is fascinating! How that oscillator works without any caps is cooking my brain a little.

In addition to this awesome advice, I'd like to ask the collective brain to reason with me on some assumptions.

ASSumption #1:
Any noise generated in the phantom should technically be common mode, as it'll appear on both legs of the differential input to the mic pre - AC coupled of course, through the AC coupling caps. Any offset/difference between them will be caused by a difference in the 6K8 resistors typically used for phantom. At 1% matching on the resistors, the attenuation of noise could be calculated.
yes, but CMRR is better at low-mid frequency than high, so worth scraping off high frequency noise with simple RC.
ASSumption #2:
I like the simple RC filter suggestion. Does having the series R have any impact on the circuit, other than dividing down (voltage divider) with the 6K8 resistors? I can't think of any bad things, as the phantom spec is pretty loose.
it's in series with 48V and current draw is modest... BTW 48V does not have to be very precise, most mics scrub off a lot of it. 
ASSumption #3:
Adding the feedback to the DCDC post filter is the way to go here. Pure and simple. Thoughts?
huh?  The one I did just hung off an existing DC-DC not a dedicated regulator for the 48V.
ASSumption #4:
A simple capacitance multiplier using a resistor, cap and NPN could add significant filtering in theory, so that no-one can ***** at me about adding noise to their precious phantom - and cost pennies. Any thoughts on why/if that would be a bad thing?

Thanks again

R
I like #4 but am not sure what NPN is doing (perhaps as a pass element for 48V regulator?)

JR
 
John,

Cap multiplier is in my head based on a recent eevblog video I saw. Capacitance is multiplied by the hfe of the npn.

In essence, it’s an emitter follower where the base is driven by an RC. Not really a regulator, because if he average input falls, so will the output.

I’m obsessed with removing the AC input from these projects and allowing folks to use power bricks. Lots of filtering required to get to nice analog rails though. I already have a +/- 15v board that I developed - needs work, but it’s pretty darn quiet. Just need phantom now.
 
Rochey said:
I went to sleep and woke up to this SLEW of replies! thank you very much gents.
That Neve schematic is fascinating! How that oscillator works without any caps is cooking my brain a little.
There is a cap across the secondary that defines the tuning frequency.

ASSumption #1:
Any noise generated in the phantom should technically be common mode, as it'll appear on both legs of the differential input to the mic pre - AC coupled of course, through the AC coupling caps. Any offset/difference between them will be caused by a difference in the 6K8 resistors typically used for phantom.
No. The CM voltage is transmitted to the input legs via a voltage divider constituted of 6.8k and the input impedance of the leg. This constitutes a balanced bridge. For it to be perfectly balanced, the ratio on each leg must be identical. R1/R2=Z1/Z2. In particular, with a xfmr, the impedance of each is constituted of the stray capacitance, and there is usually a huge difference between the hot and cold sides. With xfmr-less inputs, the tolerance of the input resistors and caps is what defines the difference. Using 1% resistors, worst-case is 4% unbalance, or about 28 dB CMRR.

ASSumption #2:
I like the simple RC filter suggestion. Does having the series R have any impact on the circuit, other than dividing down (voltage divider) with the 6K8 resistors?
Not against the 6.8k; it's against the current drawn by the mic.

ASSumption #3:
Adding the feedback to the DCDC post filter is the way to go here.
??

ASSumption #4:
A simple capacitance multiplier using a resistor, cap and NPN could add significant filtering in theory, so that no-one can ***** at me about adding noise to their precious phantom - and cost pennies. Any thoughts on why/if that would be a bad thing?
Not much against it, as long as you realize it takes a few volts to operate correctly, which will be substracted to the incoming voltage. Also beware that cap-multiplier needs short-circuit protection (death by Zener effect on the b-e junction).
 
Rochey said:
I’m obsessed with removing the AC input from these projects and allowing folks to use power bricks. Lots of filtering required to get to nice analog rails though. I already have a +/- 15v board that I developed - needs work, but it’s pretty darn quiet. Just need phantom now.
All smps are not equal regarding noise. The ones I use (Sunpower) are very quiet, but still I use them in conjunction with a CLC filter.
 
Rochey said:
John,

Cap multiplier is in my head based on a recent eevblog video I saw. Capacitance is multiplied by the hfe of the npn.

In essence, it’s an emitter follower where the base is driven by an RC. Not really a regulator, because if he average input falls, so will the output.
I mistook multiplier for doubler (that I thought we were talking about... my bad).  Yes I grok C multipliers. The concern is that C multipliers use active gain to deliver apparent high capacitance, but there is no free lunch. Trying to filter out high frequency noise the gain bandwidth of the active device could make the multiplier less effective. For HF filters I prefer passive real poles.

I recall back in the 70's when I was trying to filter clock noise out of BBD audio paths with active op amp filters. The op amps lacked the HF gain bandwidth to effectively filter out the clock edges (in short the filter didn't filter). I would use a real pole in front of the active filter poles to remedy that.
I’m obsessed with removing the AC input from these projects and allowing folks to use power bricks. Lots of filtering required to get to nice analog rails though. I already have a +/- 15v board that I developed - needs work, but it’s pretty darn quiet. Just need phantom now.
As has been shared DC-DC has been used before for phantom with no apparent problems, of course as long as reasonably filtered. For another example the Peavey Mark VIII SR console used a DC -DC switcher for it's 48V rail...

JR
 
> How that oscillator works without any caps is cooking my brain a little.

I gotta believe you just need more sleep. Even in this barren age, if they don't teach EEs "blocking oscillators", there are plenty of examples in simple switchers (not chip-based "Simple Switchers(R)").

The BJT pulls-down on a coil. Current rises zero to infinity. But BJT current is limited by say a 13K base resistor. When coil current exceeds what the BJT wants to pass, collector voltage rises. Before the BJT burns-up, use another winding to switch the BJT off. As the coil current decays to zero, the BJT turns back on.

No caps!

Push-pull makes it symmetric and spreads the stress. (Since you need the turn-off winding anyway, it actually makes the coil cheaper by balancing DC.)

The raw waveform from this type inverter tends to be ugly and spiky. The cap pointed out by Abbey tends to reduce the rattyness. Also more-defines and shifts the frequency. But the thing will work without it.

https://www.americanradiohistory.com/hd2/IDX-Consumer/Archive-Poptronics-IDX/IDX/60s/69/Pop-1969-05-OCR-Page-0060.pdf#search=%22inverter%22
 
Rochey] ASSumption #3: Adding the feedback to the DCDC post filter is the way to go here. [quote author=abbey road d enfer said:
[/quote]
I think he's saying to move the resistive feedback network (which sets the output voltage) to after the RC filter, so as to not have the mic current draw effect the output voltage set point (aka. remote load sensing).
 
Matador, exactly, but I was going to use a capacitance multiplier rather than a simple rc.

I imagine the feedback would change less violently, given the RC created. It would also ensure that the boost converter would compensate for any drop in voltage across the npn emitter Follower of the capacitive multiplier

R
 
Rochey said:
Matador, exactly, but I was going to use a capacitance multiplier rather than a simple rc.

I imagine the feedback would change less violently, given the RC created. It would also ensure that the boost converter would compensate for any drop in voltage across the npn emitter Follower of the capacitive multiplier

R
The cap-multiplier is a low-pass filter. By including it in the NFB loop may alter stability and turn the system into an oscillator.
 
PRR - you are right. I was running on jet lag and lag of sleep. I appreciate you taking the time to talk me through it.

One question. I understand it’s an oscillator that’s created with the two transistors... but what I don’t understand is “which one starts?”

Yes current flows into the center tap. Got that. Voltage eventually exceeds the bax16 (D2 and D3) diodes and hits the collectors of both(??) npn transistors. Current also flows through the 13k resistors to the opposite base, causing current to flow.

I can’t picture why or how one side gets to start first. Being both identical, it simply cooks my noggin that they oscillate.
 

Latest posts

Back
Top