[tomb]

This is limited to USB Audio Class 1, meaning it will not work at all with any modern USB Audio Class 2 DAC that does not include a UAC1 fallback option. Pretty noisy regulator for the power as well.

Other than that, sure. Pretty good.

In my system it would not work at all though.

Thor

Pretty noisy regulator? with Ripple + Noise in the millionths of volts? You have some really high standards. Good luck in finding better.

You are correct that it's limited to Class 1.

 

[Thorsten Loesch]

Pretty noisy regulator? with Ripple + Noise in the millionths of volts? You have some really high standards. Good luck in finding better.

You are correct that it's limited to Class 1.

LM317 - not even correctly applied for lowest noise. I call that "pretty noisy" by the standards I work towards.

Of course, it will regulate the voltage.

I'd like to put it on my AP2 if it is only 20dB (best case for 317) more noisy than my "baseline" designs or more like the 40dB I expect from the schematic.

Past that, Class 1 USB audio is appx. two decades out of date. Few DAC's designed in the last 15 years can use it and work correctly, bypassing either windows or apple mixer even for CD files.

No further comment.

Thor

 

[tomb]

LM317 - not even correctly applied for lowest noise. I call that "pretty noisy" by the standards I work towards.

Of course, it will regulate the voltage.

I'd like to put it on my AP2 if it is only 20dB (best case for 317) more noisy than my "baseline" designs or more like the 40dB I expect from the schematic.

Past that, Class 1 USB audio is appx. two decades out of date. Few DAC's designed in the last 15 years can use it and work correctly, bypassing either windows or apple mixer even for CD files.

No further comment.

Thor

The specs on the LM317 are written for the chip itself with the specified measurement conditions, a rudimentary application circuit, and "typical" results. You can easily exceed the 80 dB of quieting specified for the LM317, not even counting the rest of the application circuit and its effect on noise removal.

Back in the day, before Texas Instruments acquired National Semiconductor, this is how the data sheet for an LM317 looked:

Note the red-outlined column headings at top and the red-outlined rows near the bottom. There's a "Min," "Typ," "Max," and "Units" in the column headings, yet several are blank in the highlighted rows. The important point is that the "Max" column is left blank in all the RMS Output Noise and RRR column spaces. "Max" is left open for you to achieve higher values with your own circuit implementation.

Here's a further example. Note that in the rows for "Ripple Rejection Ratio," Cadj is specified at 0uF is 65 dB (Typ. only), and Cadj specified at 10uF is 80 dB (Typ. only). The addition of 10uf reduces noise by 15dB. A 10uf capacitor in electrolytic sizes is so small it's difficult to find in certain brand ranges. The reason they used that for measurement conditions is that National was demonstrating how capable the chip is for reducing noise - by just itself or with little else. Even the most rudimentary power supply circuits in audio have capacitors of 220uf, 470uf, or 1000uf and higher. That's not even discussing the fact that the conditions were specified at 10V as well.

As for Class 1 vs. Class 2 USB, you're off by at least 5 years in your "out of date" declaration with regard to audio, maybe even by 10 years in effective applications of high-resolution USB. The USB standard may have actually been there since 2000, but it takes years to develop follow-on successful chips and applications. I've been building and selling DACs (and amplifiers) during all that time. I'm not the DIY police and am not trying to be harsh. However, you shouldn't throw around terms like "pretty noisy regulator" or "not even correctly applied" if you don't have the evidence to back it up. Yes, there are circuits that can develop even less noise (AMB's Sigma power supplies come to mind), but the LM317 is far from a "pretty noisy regulator."

P.S. The last DACs I built and sold used a fully-differential DAC chip with high-speed, quad-channel isolation, a MOSFET-derived linear-regulated power supply, amorphous core output transformers, and USB synchronous-to-asynchronous conversion down to 192kHz. The only problem in making more was that Cirrus Logic killed the Wolfson WM8741.

EDIT: Not to put too fine a point on it, but I was responding to "Easy solution to isolate power and ground on USB," with emphasis on the "Easy." This is supposed to be DIY, btw. It's obvious the thread went way beyond that while you were trashing parts and implementations.

 

[Ghoostknight]

thanks again for your help

> Do you have a "easy" schematic for implementing super capacitors by myself?

A super capacitor is just a capacitor with a very large capacitance relative to size. So you apply it like any other capacitor. Observe voltage ratings. Or kind of treat as a battery that is constantly charged by the existing power supply and delivers extra current to the load when the power supply runs out of steam.

> for example, to make a box for the Ifi power supply or any other power supply to add it downstream to the chain

Just take a large enough number of low ESR Supercapacitors that wired in parallel will give you desired capacitance (say 10-50F per Ampere). If you use higher voltages than 5V you need to use series/parallel connection to get higher voltage ratings, for USB the 5.5V rated units off the shelf are fine.
Then wire with +of the whole bank to + of the PSU Wire and - to minus. There is no need for anything else.

hmm interesting, was i overthinking with limiting power draw of the capacitors itself? doesnt it trigger the protection of the power supply since it draws as much current as it can in the beginning (and maybe even the device that is powered cant get enough current as long the capacitors are charged?)

i see that ian canada designed these https://www.audiophonics.fr/en/powe...a-capacitor-conditioner-board-5v-p-14591.html to limit initial current draw and still allow the device to be powered from the beginning, is this just a concern if we are talking about the huge capacitances of 600F ?

if its as easy as you say i would just need to parallel for example these: https://www.mouser.de/ProductDetail/PowerStor-Eaton/PTV-6R0H305-R?qs=zW32dvEIR3ujAHpiST3XNg== , what ESR should i aim for? if i use 5 of them the ESR would be around 6,25mOhm

i think the ian canada module should be around 3,8mOhm in the end but it seems it would be much cheaper to just use 6 of the capacitors i posted (40€ vs 140€) how about the capacitance itself ? is the ian canada module overkill in this regard? (i think the main reason for the huge capacitance is that with increasing capacitance ESR also lowers)

The DIY DoodleBug has been around for a while:
DoodleBug USB Isolator and Power Supply

Oh i never heared of this and your website, looks great and i would actually enjoy building such a DIY Kit but Thor is right, most modern dacs wont be combatible if they offer more than 24bit/96khz, i would actually love a kit like this with a modern USB 3.0 or atleast 2.0 High speed isolator
while i also think the LM317 isnt top of the line anymore its probably better than the generic 20€ USB 2.0 Full speed isolator implementations

 

[Thorsten Loesch]

Hi,

> was i overthinking with limiting power draw of the capacitors itself?

Yes. It does trigger protection circuitry. The 150F bank for the HP Bank takes 10 minutes to charge up, only after this can the HP Amp be turned on.

> i see that ian canada designed these https://www.audiophonics.fr/en/powe...a-capacitor-conditioner-board-5v-p-14591.html
> to limit initial current draw and still allow the device to be powered from the beginning, is this just a concern if we are talking
> about the huge capacitances of 600F ?

Well, if you cannot live without having instant turn after plugging in the PSU, sure.

I do not see this as necessary. With me gear is always plugged in and on, so the "startup" only happens if we have a power outage.

Otherwise, competently designed power supply should have relevant protection to not worry about that side.

> what ESR should i aim for?

As low as sensible, however DC connectors and wire will add a few mOhm, so there is a point where where lower is no longer better.

if i use 5 of them the ESR would be around 6,25mOhm

> how about the capacitance itself

As high as sensible. There is a point where more no longer improves things. I'd shoot for 1F per 50mA (maximum peak power draw).

Thor

 

[tomb]

thanks again for your help


hmm interesting, was i overthinking with limiting power draw of the capacitors itself? doesnt it trigger the protection of the power supply since it draws as much current as it can in the beginning (and maybe even the device that is powered cant get enough current as long the capacitors are charged?)

i see that ian canada designed these https://www.audiophonics.fr/en/powe...a-capacitor-conditioner-board-5v-p-14591.html to limit initial current draw and still allow the device to be powered from the beginning, is this just a concern if we are talking about the huge capacitances of 600F ?

if its as easy as you say i would just need to parallel for example these: https://www.mouser.de/ProductDetail/PowerStor-Eaton/PTV-6R0H305-R?qs=zW32dvEIR3ujAHpiST3XNg== , what ESR should i aim for? if i use 5 of them the ESR would be around 6,25mOhm

i think the ian canada module should be around 3,8mOhm in the end but it seems it would be much cheaper to just use 6 of the capacitors i posted (40€ vs 140€) how about the capacitance itself ? is the ian canada module overkill in this regard? (i think the main reason for the huge capacitance is that with increasing capacitance ESR also lowers)


Oh i never heared of this and your website, looks great and i would actually enjoy building such a DIY Kit but Thor is right, most modern dacs wont be combatible if they offer more than 24bit/96khz, i would actually love a kit like this with a modern USB 3.0 or atleast 2.0 High speed isolator
while i also think the LM317 isnt top of the line anymore its probably better than the generic 20€ USB 2.0 Full speed isolator implementations

Yes, I never said it was more than USB 2.0. However, its implementation is easily copied if you can find another USB digital isolator chip that's compatible at the resolutions you want. Plus, if you guys have some unreasonable prejudice against an LM317, you can substitute in a different regulator. That's not difficult (and not needed, IMHO). Again, I was going by the title of your thread and not reading in detail (TL, DR), which was my mistake.

Thor is correct that in your situation, modifying a high-quality USB hub is probably the way to go. The DoodleBug is essentially a knock-off of that in a ready-made DIY-package (but admittedly limited in resolution).

That said, I stand on my objection to Thor's characterization of the LM317 and its implementation in the DoodleBug. He mentioned a 7805 a while back and it isn't in any way superior to a properly implemented LM317. One might suggest that an LM317 isn't the best choice for voltages as low as 5V and he can probably make a case for that. Maybe I mistook his very dismissive statement when it might've just been a language issue.

On a different subject, I would be wary in the use of a super-capacitor in audio. They're more valuable in motor control situations and instances where the prevention of signal interruptions are at a premium. Complete signal interruption is not really an issue with audiophile equipment that's all directly-wired with decent cabling, anyway (and there are no motors to worry about). The issue with super-capacitors is that you are not going to get good ESR and ripple rejection values by themselves, certainly not compared to the reasonable power-grade quality of Nichicon UPW, UHE or Panasonic FM, FC type capacitors (as examples). Yeah, I suppose if you use enough super-capacitors you can simulate a battery, but at the price of adding more noise than is needed. You can probably still compensate for that with other circuitry, but you can't get super-high uf for nothing.

It also sounds like there are some misconceptions around about linear-regulated power supplies. They can't be made cheaply and they burn lots of heat. The reason is that they're designed to burn volts in order to regulate and quiet: it's inefficient, but simple, dependable, and with fewer parts. However, linear-regulation requires more expensive, hefty construction with attention to heat sinking and heat rejection. In commercial design, most Electrical Engineers would rather not involve the Mechanical Engineers that heat sink and heat rejection techniques require - that's more money, too. Plus, the entire issue of being highly-inefficient is politically abhorent (esp. in Europe).

All things equal, however, a linear-regulated power supply is going to be much quieter over more types of situations for "audiophile-grade" equipment than switching power supplies. Even if you can achieve similar noise performance with a switcher, the linear-regulated power supply is going to sound more organic. That includes DACs. More true-to-life sound is what we're looking for, after all. Class A operation in amplifiers is completely analogous. Class A is not popular in commercial amplifiers, for the very same reasons stated above. With DIY, however, we don't have to follow energy mandates and Class A is still the most dependable, superior-sounding amplifier operation regime. As with anything I've stated above, they are the general cases and YMMV.

 

[Thorsten Loesch]

Hi,

> The specs on the LM317 are written for the chip itself

I am not concerned with SPec's, but with the actual performance implemented in circuit.

You should peruse this series of articles:

Simple Voltage Regulators Part 1.1: Noise

In the design you recommend, the LM317 is not used in the lowest noise configuration possible (that would be a resistor and capacitor to ground, not using LED's that add noise).

I commented on both that LM317 is a fairly noisy part as it is and that way it is applied in this design adds absolutely unnecessary and avoidable noise, never mind that you add an absolutely unnecessary need for adjustment.

> As for Class 1 vs. Class 2 USB, you're off by at least 5 years in your "out of date" declaration with regard to audio, maybe even by 10 years in effective applications
> of high-resolution USB. The USB standard may have actually been there since 2000, but it takes years to develop follow-on successful chips and applications.
> I've been building and selling DACs (and amplifiers) during all that time.

I have not seen a shipped any new design UAC Product since around 2010. I specifically did not give any timeline, so I have no idea what you mean by me being "5 years off".

> I'm not the DIY police and am not trying to be harsh. However, you shouldn't throw
> around terms like "pretty noisy regulator" or "not even correctly applied" if you don't have the evidence to back it up.

Ok, my evidence is the noise levels of for example the iFi SMPS's which as an SMPS are MUCH, much quieter than a LM317.

> the LM317 is far from a "pretty noisy regulator."

To me 1uV is low noise. An optimally applied LM317 is around 10uV. If not applied optimally, like by following fashions that add unnecessary LED's noise will be higher.

> P.S. The last DACs I built and sold used a fully-differential DAC chip with high-speed, quad-channel isolation, a MOSFET-derived linear-regulated power supply,
> amorphous core output transformers, and USB synchronous-to-asynchronous conversion down to 192kHz. The only problem in making more was that Cirrus Logic
> killed the Wolfson WM8741.

The last DAC I designed only got released a few weeks ago. It is UAC2, naturally and UAC1 fallback is 48kHz only, as is common.

> Not to put too fine a point on it, but I was responding to "Easy solution to isolate power and ground on USB," with emphasis on the "Easy."
> This is supposed to be DIY, btw.

You recommended a design (of yours incidentally) to solve the Op's problem. The design is UAC1 only and limited to 12MBPS full speed.

The Op already owns a high speed USB2.0 isolator and wants to isolate things like a USB WiFi module SSD etc.
For this your project is inappropriate. Maybe you should read threads before jumping in and promoting yourself.

> It's obvious the thread went way beyond that while you were trashing parts and implementations.

You proposed something. I looked at it and gave it a commentary as to the implementation.

It is UAC1 only which makes it 100% useless to the OP and useless to many others, which you omitted to mention clearly.

It implements a old and noisy regulator in a suboptimal way that means that for example something using TPS7A4700 for example or similar would be substantially lower noise. Even just throwing the LED's out for a resistor plus capacitor would lower noise.

If we MUST use something like LM317 in the 21st century it may be a good idea to follow it with a denoise circuit.

Simple circuits reduce regulator noise floor

In the iFi iUSB Power shipped in 2013 I used a NE5534 based regulator with an LM317 used as series element (it gives good build in protection) and achieves well below 1uV 20Hz-20kHz noise, to give you another option to cheaply get sub microvolt noise while using the 317 for ease of implementation.

Further I suggest you do not expose your designs to the public if you cannot accept criticism.

Thor

 

[Thorsten Loesch]

Hi,

> Yes, I never said it was more than USB 2.0.

You should have said that it is USB 2.0 FULL SPEED ONLY, which limits transmission speed to 12MBPS instead of the actual USB 2.0 limit of 480MBPS.

> However, its implementation is easily copied if you can find another USB digital isolator chip that's compatible at the resolutions you want.

As I designed such an isolator (which went out of production as the isolation chip was discontinued), good luck finding such a chip.

> Plus, if you guys have some unreasonable prejudice against an LM317, you can substitute in a different regulator.

Yes. At that point we designed our own USB Isolator that has nothing to do with what you proposed.

> That said, I stand on my objection to Thor's characterization of the LM317 and its implementation in the DoodleBug.

By all means do. But in the 21st centrury we can expect sub microvolt levels of noise for audio.

Is it NECCESARY? Depends. In some cases it may.

> He mentioned a 7805 a while back and it isn't in any way superior to a properly implemented LM317.

Correct. First, at 5V a modern 7805 has lower noise than a LM317 (around 10dB lower at ~ 10uV & ~3uV).

Second, I mentioned I left it alone because the actual circuit follows it with a Supercapacitor Bank of 10F/5.5V with 15mOhm ESR using in effect the DC Cable from the supply as series resistor in a RC filter circuit that attenuates the regulator noise.

If we assume 0.15 Ohm resistance for the wires (it actually is more) this creates a filter that drops regulator noise by 20dB (ESR limited) above around 3Hz, so noise will be reduced to 0.3uV after the supercapacitor filter. QED

The result is that combination of a fairly noisy regulator (if not as noisy as a LM317) and a passive RC filter has dramatically lower noise together, than for example an LM317.

It's not the part itself, though why use it if better options exist, it is how it's used.

> One might suggest that an LM317 isn't the best choice for voltages as low as 5V and he can probably make a case for that.
> Maybe I mistook his very dismissive statement when it might've just been a language issue.

Thank you. Yes, I consider the LM317 subideal and as said, adding LED's does not help.

You could try adding a NJM5534 (I prefer JRC/NJR ones these to TI and other makes).

Connect the LM317 ADJ pin to 5534 pin 5. This directly is the output from the final NPN VAS Transistor collector. Set the 317 Resistor from Out to ADJ to give around 5mA VAS current (according to datasheet around 38mA is the current limit for this transistor). +in get's our reference voltage and -In is connected through a 100 Ohm protection resistor to the output. Unregulated power Input to 317 In and 5534 +V. I like to use 10R/100uF on the 5534 supply pin. Negative supply to ground.

Make a reference voltage with LM329 + LM334 set to 5mA current. Use a voltage divider to make a 5V reference from the 6.9V reference.

Place a large high quality electrolytic capacitor to ground to sink LM329 noise. The LM329 with a voltage divider to 5V will give around 5uV RMS noise to 100kHz. Adding the capacitor will lower this dramatically. A compensation capacitor from Pin 5 to Pin 8 will be needed. I'd start with 100pF and then optimise for step response.

Say we use 5.1k as lower resistor and 1.6k as upper resistor (giving 5.25V) and 100uF/6.3V as filter capacitor. This will give a -3dB Filter point of 1.6Hz, so at 16Hz noise will be attenuated by 20dB and more at higher frequencies. The main noise source in the circuit remains the 5534 which I estimate as 0.5uV in the 20kHz band.

Voila, using only parts that were old when I did my EE course and all date back to the 1970's and include the noisy 317 we have made a regulator with around 0.5uV output noise, beating the modern TI TPS7A4700 by neary 20dB. Fool proof, no adjustment and short circuit/overload proof.

> On a different subject, I would be wary in the use of a super-capacitor in audio.

Well, I have been leery, after testing, both subjective and objective I am convinced.

> All things equal, however, a linear-regulated power supply is going to be much quieter
> over more types of situations for "audiophile-grade" equipment than switching power supplies.

Only if the switching supplies are designed by a mike oskar romeo oskar november as we say in NAFO.

I can design a SMPS which is quieter than MOST linear supplies.

> Even if you can achieve similar noise performance with a switcher,
> the linear-regulated power supply is going to sound more organic.

That is a subjective statement and one I do not necessarily share.

Output impedance, step response and noise can be made equal to best linear designs, arguably using linear add-on's (costing a few cent) to the otherwise "cooking" but
highly optimised SMPS.

Unless the linear power supply is aggressively designed for low coupling capacitance from mains (that excludes any transformers lacking shielding or 2-chamber construction) it will likely be worse than a well designed SMPS.

The result is that in many cases a truly well designed SMPS performs objectively and subjectively superior to "run of the mill" linear power supplies (especially those with LM317 ).

Thor

 

[Thorsten Loesch]

Hi,

> these: https://www.mouser.de/ProductDetail/PowerStor-Eaton/PTV-6R0H305-R?qs=zW32dvEIR3ujAHpiST3XNg==

They have fairly high ESR. I use CDA Brand, china made CHP Series,:

http://www.cda-cap.com/userfiles/20217303336828.pdf

These ones I use:

CHP5R5L504R-TW
CHP5R5L105R-TWX
CHP5R5L255R-TWQ
CHP5R5L755R-TW

A bank of 20pcs CHP5R5L504R-TW is what I use for the DAC, giving 10F with 15mOhm ESR. That is for a DAC drawing around 0.3A peak.

A bank of 20pcs CHP5R5L755R-TW is what I use for the HP AMP, giving 150F with 5mOhm ESR. That is for a HP Amp drawing around 5A peak.

Both at 5V nominal.

Thor

 

[Ghoostknight]

> was i overthinking with limiting power draw of the capacitors itself?

Yes. It does trigger protection circuitry. The 150F bank for the HP Bank takes 10 minutes to charge up, only after this can the HP Amp be turned on.

oh i see

> i see that ian canada designed these https://www.audiophonics.fr/en/powe...a-capacitor-conditioner-board-5v-p-14591.html
> to limit initial current draw and still allow the device to be powered from the beginning, is this just a concern if we are talking
> about the huge capacitances of 600F ?

Well, if you cannot live without having instant turn after plugging in the PSU, sure.

I do not see this as necessary. With me gear is always plugged in and on, so the "startup" only happens if we have a power outage.

Otherwise, competently designed power supply should have relevant protection to not worry about that side.

i probably wont turn off the devices much either, i just wanna make sure nothing bad can happen

1. if i leave the device connected and the supercapacitors are depleted (from internal leakage after 1-2weeks turned off for example) will something happen? my guess is that the device just cant turn on reliably since the supercapacitor sucks all the current up

2. i read that LDO`s (in my case probably the LT3045) dont like voltage applied to their output if they are not powered on, if i disconnect the power supply from mains and the super capacitors are still charged, is this safe?
2.1 or will the supercapactior discharge through the LDO?

3. is there a easy way (with a led) to show if the supercapacitors are fully charged? so i know when i can "safely" turn on the pi

4. i read that ian canadas supercapacitors modules are beneficial for SQ, even if they just power the RPI itself, i have a project ongoing where i modify a CM4 carrier board from waveshare and probably will exchange the regulators on this board, clocks of the raspberry pi and for example apply the usb cable mod discussed here
now i wanna add supercapacitors to the 5V input of the Pi and maybe to the other voltage rails (there is the 3,3V GPIO voltage rail, 3,3V for the clocks (but i will probably use the same 3,3V rail for both) and 3,3V and 1,05V for the VLI805 usb hub)
4.1(maybe one questions offtopic: can i leave both usb hub regulators without them having a effect on the remaining system? (3,3V is linear and 1,05V switching, im not sure if switching noise goes "upstream" too) if yes i would probably use the same 3,3V rail for GPIO/Clocks for the hub too (to keep cost somewhat low) plus a seperate LT3045 for the 1,05V rail

my idea is to implement the supercapacitors like this:
the RPI Case will have 3 inputs:
1. USB Device power (through the modded cables) where i probably will just use a cheapo smps power supply, since its not for dac use
2. 5V for Pi and LDOs to make the 3,3V and maybe the 1,05V rail, probably first the ifi ipower X 5V i already have, maybe later ian canadas 6,6V battery power supply with LT3045 afterwards to make 5V to go off the grid
3. clean 3,3V (probably from the battery power supply from ian canada) to power ians TransportPi Digi

prefereably i add supercapacitors to all rails, i know the ian canada modules are very good but it will be very pricey to implement them in the way i want and im not sure if its even worth it then

i know this is probably all overkill and i may regret spend "so much" money on this project since it will probably exceed my 300€ Aune X8 XVIII Dac and i dont know the outcome yet SQ wise but i kinda wanna try anyway since i hear improvements in many things with my room corrected studio monitors which kinda dont make sense "objectively"
i actually also tried this cheapo chinese supercapacitor device already: https://de.aliexpress.com/item/1005004825653850.html and heared a noticably improvement powering my stock RPI4 with it (combined with the Ifi ipower X 5V)
but mostly its A. about DIY and B. to have a very good (i dont dare to say "best" ) digital streamer which will be hopefully be somewhat futureproof and my new reference device and i will hopefully upgrade other devices in the future to match them pricewise a little better

any considerations about the project are very welcome

 

[Thorsten Loesch]

Hi,

> i probably wont turn off the devices much either, i just wanna make sure nothing bad can happen

Commercial Power supplies are designed to accept infinite time shorted outputs and not fail.

> if i leave the device connected and the supercapacitors are depleted (from internal leakage after 1-2weeks
> turned off for example) will something happen? my guess is that the device just cant turn on reliably since
> the supercapacitor sucks all the current up

I just had the case after going north for a week, my desktop system was unplugged.

When I came back I plugged it in, of course, the DAC was running on USB Power from the Laptop via the Hub and headphone amp was not turning on.

I went, did something else for a few minutes, when I came back everything was fine.

> 2. i read that LDO`s (in my case probably the LT3045) dont like voltage applied
> to their output if they are not powered on, if i disconnect the power supply from
> mains and the super capacitors are still charged, is this safe?
> 2.1 or will the supercapactior discharge through the LDO?

This is a little more complex. If the input to the LDO is a short and the output has voltage, the fact that most positive LDO's have PMOS Pass devices means the PMOS Body diode will conduct. If there is a lot of energy on the output side and a short on the input, problems can materialise.

Using devices with bipolar pass devices is safe. I would have no qualms driving a super capacitor bank from a noisy and indestructible 7805, for the reasons outlined in the other posts. Just make sure you have a little series resistance either as parasitics from cables or as explicit resistor.

Or if I want to guild the lilly and impress the peons I might use my sub microvolt regulator made from LM317, NJM5534, LM329 & LM334... Not that it is strictly neccesary.

> 3. is there a easy way (with a led) to show if the supercapacitors are fully charged?
> so i know when i can "safely" turn on the pi

Yes, you can make a Voltage monitor from a TL431 device, schematics are in the datasheet.

> 4. i read that ian canadas supercapacitors modules are beneficial for SQ, even if they just power the RPI itself,

It is entirely possible, though it may seem far fetched to some. I have not tested this.

In my living room system everything is pretty tweak free, MI Box 4K running Kodi as source Topping D10S as DAC (I need to get around to swapping the output Op-Amp to my favourite LM6172) and a stock Breeze HiFI TPA3255 Amp (I need to swap out the fake 5532 in there for the OPA2604 and LM6172 I ordered recently) driving custom speakers.

I might want to add a supercapacitor buffered external DC to the Topping DAC as well and a supercap block to the PSU for the Mi Box.

> i have a project ongoing where i modify a CM4 carrier board
> now i wanna add supercapacitors to the 5V input of the Pi and maybe to the other voltage rails
> (there is the 3,3V GPIO voltage rail, 3,3V for the clocks (but i will probably use the same 3,3V rail for both)
> and 3,3V and 1,05V for the VLI805 usb hub)

Suggestion. Use Supercapacitors on the 5V input only. Also add a few Panasonic Os-Con on 5V.

Then use Panasonic Os-Con on all the power lines, as large values as you can fit/find/afford.

The best way to explain...

Supercapacitors are fairly "slow", so they work best for fairly low frequencies Audio Band mainly and even there more the lower portion. Think about changes that are counted in milliseconds.

Normal electrolytics are a little faster "faster", they are usually effective to around 100kHz, higher than that their electrochemical nature means they become progressively more non-ideal.

Solid electrolytics tend to have lower ESR compared to electrolytic and supercapacitors, the Panasonic Os-Con range are still the yardstick for those.

With Os-Con's they remain effective to well above a few 100kHz and loose efficiency only going on 1MHz. Similar high frequency performance with higher ESR is available from Tantalum Capacitors in a very small package.

Finally ceramic capacitors are mainly effective at higher frequencies, this depends upon the relative size.

You can see what I use in this picture:

The yellow capacitor ia 100uF/10V Tantalum with around 0.1 Ohm ESR.

The brown capacitors are 10uF/25V X7R with around 10mOhm ESR.

Here a drawn out schematic with all the relevant expected data stated of the whole power supply including the 10F Supercapacitor Bank.

The key points is that the system can deliver 11A for 0.15 seconds (with the output voltage dropping from 5V to 4.7V) and 2A for 3 seconds despite using a 1A power source (LM7805 etc) and expected noise on the output is around 5uV for 10Hz-100kHz bandwidth (the 1uV ripple effectively disappears).

All of this from a compact linear PSU that is cheap to buy e-bay for around 30 USD with the addition of around 30 USD worth of Supercapacitors and a few small components for maybe another dollar. I consider this to be very quick, easy and effective DIY.

Don't worry excessively about switching regulators in digital applications. Modern ones are very good. They typically switch at > 1MHz and are as low noise otherwise as many LDO's. If necessary (scope on PSU line) add extra SMD Ceramic capacitors in parallel to the existing ones mainly on the output to reduce the noise at the switching frequency.

> 4.1(maybe one questions offtopic: can i leave both usb hub regulators without them having a effect on the remaining system?
> (3,3V is linear and 1,05V switching, im not sure if switching noise goes "upstream" too)

Correctly designed, switching noise should be contained. Is the RPI "correctly designed"? I would leave the switching regulator, look at the local decoupling (and perhaps add more capacitors) and separate out the feed to the Hub Chip (cut the trace) and add an LC Filter. This usually needs a large value electrolytic capacitor on both sides to avoid turn on ringing.

> the RPI Case will have 3 inputs:
> 1. USB Device power (through the modded cables) where i probably will just use a cheapo smps power supply, since its not for dac use

If you have a hub on the RPI PCB, use this Hub and simply separate out the Hub Power and use this also to power the other devices.

Find a quiet SMPS and it will not harm to add an LC filter (say 33uH + 3 x 1000uF/6.3V Os-Con) to the input. This will cut off noise above ~500Hz with a second order slope and will have over 10,000 times (80dB) noise reduction at the typical 65kHz switching frequency many small to medium power switchers use.

FWIW, the iFi SMPS is based on generic switchers, but they use a CLCLC output filter with substantial inductance and capacitance and even an inductor in ground line.

DC feedback is taken from the output after all filtering, while two more feedback loops for AC start at the first capacitor. Finally there is a linear, analogue noise cancellation circuit that drops the overall noise to that of a NJM5534, which is the core part of the active noise cancellation. Result, a SMPS with an output performance (seen as "back box") that matches or exceeds that of linear supplies.

Most critical with the SMPS (and Linear PSU'd as well) is however leakage capacitance. The iFi power supplies use a very special and "non standard" winding for the SMPS transformer (immodestly I will claim this for me), which allows the SMPS to meet EMC regulations without any so-called "Y-Cap". From an overabundance of caution a 100pF capacitor is fitted (standard designs use 2,200pF). It would be better without, but it is commercial product that must pass agency tests under all conditions.
So key for all power supplies is leakage capacitance from mains to output.

All other problems like audio band or switching noise can be solved externally easily and cheaply if they become material, but this coupling capacitance usually cannot be changed.
> 3. clean 3,3V (probably from the battery power supply from ian canada) to power ians TransportPi Digi

I had a "Battery Power ueber alles" phase in the 90's. I got over it.

+/-48V 20AH Battery Pack for the Amp, +/-12V Analog plus 12V for digital and another 12V for MCU, Motors etc. for the CD Player. +/- 12V for Phono.

> prefereably i add supercapacitors to all rails

I would not suggest that. on the actual power supply rails inside equipment supercapacitors mostly make sense for Amplifiers.

CPU Cores invariably are supplied from switching regulators. These have some behaviours that are interesting. It means that you want to have super capacitors on the main DC Bus for a device or for the Amplification only. Everywhere else, for digital solid electrolyte capacitors like Panasonic (formerly Sanyo) Os-Con by all means ridiculously oversized are a better choice.
> i know this is probably all overkill and i may regret spend "so much" money on this project

Then look at ways to get the effect without spending that much.

> i actually also tried this cheapo chinese supercapacitor device already: https://de.aliexpress.com/item/1005004825653850.html
> and heared a noticably improvement powering my stock RPI4 with it (combined with the Ifi ipower X 5V)

Yes, the iFi PSU is a little marginal for the RPI, adding 10F will help for peak draw.

With an SMPS, it acts kind of like a gearbox. It will keep the output constant but increase input current draw. Worse, as the input voltage drops because the loading becomes high, where linear supplies will keep current constant, the SMPS will increase the current drawn. This can create an "avalanche" type behaviour.

Here a Supercapacitor Bank will show the greatest improvement.

For example, in a portable product I am working on, running Class D Amplifiers via a step-up switcher from a 3.7V Lithium battery, simply adding a few supercapacitors to the DC bus feeding the step-up switcher allowed much higher peak power with music (it did not effect long-term sine wave power) but the result is the product can play much louder acoustically (measured) and produces much higher power using an AES2-2012 test signal, which is made to resemble music.

Thor

 

[Ghoostknight]

> 3. is there a easy way (with a led) to show if the supercapacitors are fully charged?
> so i know when i can "safely" turn on the pi

Yes, you can make a Voltage monitor from a TL431 device, schematics are in the datasheet.

i will have a look, thanks

In my living room system everything is pretty tweak free, MI Box 4K running Kodi as source Topping D10S as DAC (I need to get around to swapping the output Op-Amp to my favourite LM6172) and a stock Breeze HiFI TPA3255 Amp (I need to swap out the fake 5532 in there for the OPA2604 and LM6172 I ordered recently) driving custom speakers.

i can also recommend the OPA1656, i was quite happy with it in the original D10 (i also tested OPA1612, OPA2156, OPA1692, OPA1642, LM4562, the 1656 sounded best to me followed by 1612)
i use right now in my Aune X8 a dual OPA627, it sounds quite different to the OPA1656 but i like them both (you can read my comparision between the two in the opamp thread, https://www.head-fi.org/threads/the-opamp-thread.432749/post-17281419 ), i probably will try the burson V6 vivid next, quite curious how (my first) discrete opamp compares
i actually didnt read much about the two Opamps you mentioned (OPA2604 and LM6172) are they worth trying out?

Supercapacitors are fairly "slow", so they work best for fairly low frequencies Audio Band mainly and even there more the lower portion. Think about changes that are counted in milliseconds.

Normal electrolytics are a little faster "faster", they are usually effective to around 100kHz, higher than that their electrochemical nature means they become progressively more non-ideal.

Solid electrolytics tend to have lower ESR compared to electrolytic and supercapacitors, the Panasonic Os-Con range are still the yardstick for those.

With Os-Con's they remain effective to well above a few 100kHz and loose efficiency only going on 1MHz. Similar high frequency performance with higher ESR is available from Tantalum Capacitors in a very small package.

Finally ceramic capacitors are mainly effective at higher frequencies, this depends upon the relative size.

does that mean normal capacitors are more effective in letting high frequency noise through to ground? while still "buffering" the power somewhat just not with the capacity the supercapacitors can offer? i guess then the oscons are probably fine or even better then and cheaper

my idea to adding supercapacitors to a few voltage rails was that probably a LDO after the 5V supercapacitor will probably reduce the effectivness of the supercapacitor properties but its probably true that the supercapacitors will be most beneficial on the 5V input, since it also has the most current demand

but i will trust you if you say the oscons are fine (or better), i saw that there are a few product lines of the oscons, do you recommend one ?
i also wanted to add some capacitors to the 4 voltage rails of the CM4 module itself since the switching 4 in 1 regulator cant be exchanged (it communicates over i2c with the kernel)
can you tell me in this case if it will be still beneficial to add capacitors -IF- i have to add a cable of approxiametly 10cm to connect each capacitor bank to the output of the inductor (each voltage rail of this chip has a LC filter on the output), will it still improve noise/power ? i can try to add fairly big wires to lower resistance but the solderjoints i can utilize are quite small

Correctly designed, switching noise should be contained. Is the RPI "correctly designed"? I would leave the switching regulator, look at the local decoupling (and perhaps add more capacitors) and separate out the feed to the Hub Chip (cut the trace) and add an LC Filter. This usually needs a large value electrolytic capacitor on both sides to avoid turn on ringing.

fortunaly the switching regulator has already a LC filter as far as it looks (there is conductor near the output) unfortunaly there is no schematic available, i guess i can add here also more capacitors to the output utilizing the inductor output solderjoint and some wires

I had a "Battery Power ueber alles" phase in the 90's. I got over it.

well for me the best benefit is really the off the grid property, i kinda just wanna try it out, tho i also believe modern voltage regulators (and AC/DC conversion circuits + AC EMI Filters) have fairly good noise rejection to not really worry about it

CPU Cores invariably are supplied from switching regulators. These have some behaviours that are interesting. It means that you want to have super capacitors on the main DC Bus for a device or for the Amplification only. Everywhere else, for digital solid electrolyte capacitors like Panasonic (formerly Sanyo) Os-Con by all means ridiculously oversized are a better choice.
> i know this is probably all overkill and i may regret spend "so much" money on this project

Then look at ways to get the effect without spending that much.

yea using oscons instead of supercapacitors probably help price-wise, you said add as much as possible, is there some capacitance i should aim for which is very good but still reasonable for all the fairly low current demand voltage rails?

 

[Thorsten Loesch]

Hi,

> i can also recommend the OPA1656

So can I, except...

A) it is a victim of chipageddon with zero stocks at reputable sellers and the overprices units on sale loose are likely fakes

B) the D10S already uses OPA1656 for I/U conversion and I dislike daisychaining the same Op-Amp too many times and I am looking for a specific sound I get from IC's like the LM6172, which is radically different in design (and performance) from common Op-Amp's build on differential circuits, as seen here from the internal design

> i actually didnt read much about the two Opamps you mentioned (OPA2604 and LM6172) are they worth trying out?

Both are quite old designs I used a lot in the 90's and 00's. But they are a known quantity and quality to me and I can currently get cerdip originals at a good price. By current standards they are a little noisy and distortion is objectively a bit high.

I commented on the LM6172 already, it has a distinct sound quality, especially when driving easy loads (e.g > 1kOhm) where H2 becomes dominant. It is best used in line level application where you get 120dB SNR (re 2V) and it should use low impedance peripherals. Example would be CD-Player or DAC Output stages. Most famously perhaps I promoted the LM6172 in my modifications for the original Behringer digital EQ published enjoythemusic.com.

I also mentioned it in modifying the Marantz CD-63/67 where I preferred the LM6172 alone directly driving the line out over having the HDAM discrete Op-Amp's following the LM6172. We actually did a blind test of my modified unit against a stock unit and it was dramatically better. I also observed the LM6172 in some rather high end SACD Players... Seems some Sony engineers share my predilections in sound.

The OPA2604 (or the single 604) is an odd-ball. With it TI/BB tried to make an audiophile design. Much of the internals are very unusual and interesting. To sound best "Class A Biasing" must be applied and it must be applied to the negative rail, around 5mA to change the output from an interesting non-switching and non-symmetrical buffer to a simple emitter follower. And it should not drive difficult loads without buffer. It is best used in high impedance applications, e.g. buffering a volume control.

In many ways, as said, these IC's are outdated and the OPA(2)604 is discontinued.

> does that mean normal capacitors are more effective in letting high frequency noise through to ground?

Funny way of putting it.

We of course want to short noise to ground. That is one of the functions.

But mainly capacitors act as energy storage. If there is a current burst in the load (for example in CMOS logic when it switches) the idea is that local capacitor(s) "holds up" the power line avoiding the current being drawn along longer lines, where it would cause voltage drop and cause problems.

In effect "the closer" to a digital IC the capacitor is, the "faster" it needs to be. In this case "faster" means having low inductance and good energy release at high frequencies.

As a rule, the bigger the energy storage per volume is, the "slower" is the underlying energy storage mechanism. Film & high grade ceramic capacitors use the very basic system of a capacitor, very similar to old air capacitors. They have the fastest storage mechanism but they are relatively large for a small capacitance value.

So we get capacitors that use a chemical process as energy storage that allows a lot of more energy storage in the same volume. But by first converting first electricity into chemical energy and back is relatively "slow". This is independent from the parasitic inductance.

So film and ceramic capacitors are best used where "speed" is needed, or in other words high current pulses at high frequencies.

At the very opposite end are super capacitors which as a result are best used at the point where energy enters the overall system with other "faster" capacitors used as we move "inwards". The one exceptions are the power supply rails of power (including headphone) amplifiers.

> still "buffering" the power somewhat just not with the capacity the supercapacitors can offer?

Yes, plus, the extra capacity tends not to be required after a regulator seized right.

> i guess then the oscons are probably fine or even better then and cheaper

Where I recommend them - they are better in this application.

> but i will trust you if you say the oscons are fine (or better), i saw that there are a few product lines of the oscons, do you recommend one ?

It depends if you need SMD or through hole. I would go with the lowest ESR and largest physical size for a given capacitance and voltage in the correct geometry, if it can be fitted.

> can you tell me in this case if it will be still beneficial to add capacitors -IF- i have to add a cable of approxiametly 10cm to connect each
> capacitor bank to the output of the inductor (each voltage rail of this chip has a LC filter on the output), will it still improve noise/power
> ? i can try to add fairly big wires to lower resistance but the solderjoints i can utilize are quite small

Wires will reduce the effectiveness. Best to use the shortest possible wires. Can you make a Board that plugs into the CM4 connections and adds capacitance with the shortest possible tracks? Basically make a capacitor booster board.

> well for me the best benefit is really the off the grid property

Well, if you live in an off-grid house, of course.

> i also believe modern voltage regulators (and AC/DC conversion circuits + AC EMI Filters) have fairly good noise rejection to not really worry about it

Less faith, more reliable analysis is recommended.

> is there some capacitance i should aim for which is very good but still reasonable for all the fairly low current demand voltage rails?

Well, in the end its a question of size and cost. Set yourself a budget for all aspects, then stick to it.

Thor

 

[Ghoostknight]

A) it is a victim of chipageddon with zero stocks at reputable sellers and the overprices units on sale loose are likely fakes

oh i didnt know that, i bought mine like 3 years ago from a hungarian ebay seller (already soldered on a dip8 adapter)

In effect "the closer" to a digital IC the capacitor is, the "faster" it needs to be. In this case "faster" means having low inductance and good energy release at high frequencies.

Ah got it

Where I recommend them - they are better in this application.

Ok

It depends if you need SMD or through hole. I would go with the lowest ESR and largest physical size for a given capacitance and voltage in the correct geometry, if it can be fitted.

i probably go with trough hole for easier soldering and buy like a 50 pack to get a discount
i saw that 16V 470uF is probably the way to go, i also saw that the SEPG line is a little better than the SEPC line, tho the difference is quite small
and will probably add 5-10 to each voltage rail

a additional question:
i saw that the ian canada filter board (shieldpi pro) uses also little smd capacitors, probably something like 0,1uF ? is there a benefit to adding them too? (can you recommend one here too?)
and how should i solder them, a bank of 5x 0,1uF followed by the Oscons a little further away for example?

Wires will reduce the effectiveness. Best to use the shortest possible wires. Can you make a Board that plugs into the CM4 connections and adds capacitance with the shortest possible tracks? Basically make a capacitor booster board.

i think in the example of the 4 voltage rails of the CM4 itself i can make a board which sits directly above it with wire lengths of approxiametly 4-5cm, a other approach would be to solder the through hole legs directly, but this way i can only add one capacitor, maybe two

and one question just to make sure, if there is a LC filter right behind the regulator which consist of a Inductor+Capacitor is it alright to solder the capacitor bank directly to the inductor output pin, or will this mess with the LC filter properties (since i solder the wire to the capacitor bank inbetween the capacitor and inductor)?

--
is there a LT3045 board you can recommend?
my plan was to go with boards from ldovr.com (https://www.ldovr.com/product-p/lt3045-1a0g.htm)
but there are also a few different boards on aliexpress like this one: https://www.aliexpress.com/item/1005003787293400.html

my plan right now is to add a single LT3045 3,3V board, replace the switching regulator on the waveshare board and the LDO for the usb hub which provide both 3,3V and also use it to power the clocks i have bought (but maybe i will wire the clocks later on to the clean 3,3V input for the hats)

the only regulator left (beside the all in one cm4 regulator which cant be easly replaced) will be the 1,05V switching regulator which is only powering the usb hub chip which shouldnt make a big difference if the switching noise stays contained (and i probably wont use the usb chip for dacs)

--

Can i hit you up with a little tip for the help? (like 25€) through paypal? just PM me your paypal (or bank) address if you want
since you save me quite a few bucks through DIY anyway!

 

[tomb]

Hi,

> Yes, I never said it was more than USB 2.0.

You should have said that it is USB 2.0 FULL SPEED ONLY, which limits transmission speed to 12MBPS instead of the actual USB 2.0 limit of 480MBPS.

> However, its implementation is easily copied if you can find another USB digital isolator chip that's compatible at the resolutions you want.

As I designed such an isolator (which went out of production as the isolation chip was discontinued), good luck finding such a chip.

> Plus, if you guys have some unreasonable prejudice against an LM317, you can substitute in a different regulator.

Yes. At that point we designed our own USB Isolator that has nothing to do with what you proposed.

> That said, I stand on my objection to Thor's characterization of the LM317 and its implementation in the DoodleBug.

By all means do. But in the 21st centrury we can expect sub microvolt levels of noise for audio.

Is it NECCESARY? Depends. In some cases it may.

> He mentioned a 7805 a while back and it isn't in any way superior to a properly implemented LM317.

Correct. First, at 5V a modern 7805 has lower noise than a LM317 (around 10dB lower at ~ 10uV & ~3uV).

Second, I mentioned I left it alone because the actual circuit follows it with a Supercapacitor Bank of 10F/5.5V with 15mOhm ESR using in effect the DC Cable from the supply as series resistor in a RC filter circuit that attenuates the regulator noise.

If we assume 0.15 Ohm resistance for the wires (it actually is more) this creates a filter that drops regulator noise by 20dB (ESR limited) above around 3Hz, so noise will be reduced to 0.3uV after the supercapacitor filter. QED

The result is that combination of a fairly noisy regulator (if not as noisy as a LM317) and a passive RC filter has dramatically lower noise together, than for example an LM317.

It's not the part itself, though why use it if better options exist, it is how it's used.

> One might suggest that an LM317 isn't the best choice for voltages as low as 5V and he can probably make a case for that.
> Maybe I mistook his very dismissive statement when it might've just been a language issue.

Thank you. Yes, I consider the LM317 subideal and as said, adding LED's does not help.

You could try adding a NJM5534 (I prefer JRC/NJR ones these to TI and other makes).

Connect the LM317 ADJ pin to 5534 pin 5. This directly is the output from the final NPN VAS Transistor collector. Set the 317 Resistor from Out to ADJ to give around 5mA VAS current (according to datasheet around 38mA is the current limit for this transistor). +in get's our reference voltage and -In is connected through a 100 Ohm protection resistor to the output. Unregulated power Input to 317 In and 5534 +V. I like to use 10R/100uF on the 5534 supply pin. Negative supply to ground.

Make a reference voltage with LM329 + LM334 set to 5mA current. Use a voltage divider to make a 5V reference from the 6.9V reference.

Place a large high quality electrolytic capacitor to ground to sink LM329 noise. The LM329 with a voltage divider to 5V will give around 5uV RMS noise to 100kHz. Adding the capacitor will lower this dramatically. A compensation capacitor from Pin 5 to Pin 8 will be needed. I'd start with 100pF and then optimise for step response.

Say we use 5.1k as lower resistor and 1.6k as upper resistor (giving 5.25V) and 100uF/6.3V as filter capacitor. This will give a -3dB Filter point of 1.6Hz, so at 16Hz noise will be attenuated by 20dB and more at higher frequencies. The main noise source in the circuit remains the 5534 which I estimate as 0.5uV in the 20kHz band.

Voila, using only parts that were old when I did my EE course and all date back to the 1970's and include the noisy 317 we have made a regulator with around 0.5uV output noise, beating the modern TI TPS7A4700 by neary 20dB. Fool proof, no adjustment and short circuit/overload proof.

> On a different subject, I would be wary in the use of a super-capacitor in audio.

Well, I have been leery, after testing, both subjective and objective I am convinced.

> All things equal, however, a linear-regulated power supply is going to be much quieter
> over more types of situations for "audiophile-grade" equipment than switching power supplies.

Only if the switching supplies are designed by a mike oskar romeo oskar november as we say in NAFO.

I can design a SMPS which is quieter than MOST linear supplies.

> Even if you can achieve similar noise performance with a switcher,
> the linear-regulated power supply is going to sound more organic.

That is a subjective statement and one I do not necessarily share.

Output impedance, step response and noise can be made equal to best linear designs, arguably using linear add-on's (costing a few cent) to the otherwise "cooking" but
highly optimised SMPS.

Unless the linear power supply is aggressively designed for low coupling capacitance from mains (that excludes any transformers lacking shielding or 2-chamber construction) it will likely be worse than a well designed SMPS.

The result is that in many cases a truly well designed SMPS performs objectively and subjectively superior to "run of the mill" linear power supplies (especially those with LM317 ).

Thor

Mostly TL, although I Did Read.

You are missing that the LM317 in the device I referenced allows the voltage output to be adjustable. We've found that voltage variances in the many PC/laptop/USB-cables can be significant when attempting to regulate prior to a USB-powered DAC. The LM317 circuit in the DoodleBug allows one to adjust that voltage to optimize for a 5.25VDC before connection to a DAC. A 7805 regulator can't create voltage greater than what it is supplied. Even then, the maximum is very limited, only to 5.2V under maximum conditions (7.5 to 20V supplied).

The USB standards themselves allows for voltages as low as 4.4VDC, even from "quality" hubs. The problem is that most quality DACs have onboard regulators for 4.75V analog and 3.3V digital, among other things. They're all designed and produced in theory for 5VDC supplies in mind, but potentially not very tolerant to lesser voltages for best noise-reduction performance.

Anyway, a case can be made that a slightly higher USB voltage from the power supply will result in less overall system noise. The LM317 gives you that flexibility.