0
\$\begingroup\$

My primary concern here is the correct way to apply power (and grounding) to an assortment of single and dual-supply ICs in a mixed environment for an audio application.

I constructed a portable audio mixer using different ICs to get (most of) the features I wanted. The tell-tale issue is that whenever the audio levels were increased enough to push the "VU meter" higher, all of the mute/active channel indicators would dim. This led me to the conclusion that my power handling was not right.

Due to this, and enclosure congestion, I am now working on redesigning the mixer with SMD components. I did seek out a “best practice” solution, but found nothing obvious (to me) on the subject of these "combined factors."


What I’m Looking to Learn

What is the best practice for powering ICs and managing audio grounding in this type of environment? Should:

  1. all single supply ICs be connected to the +/- rails, instead of +/0v in a mixed single/dual supply IC situation?

  2. I just pick one voltage (5v) appropriate to all ICs involved, and regulate to that right at the power input?

  3. I treat each IC circuit as separate devices within the whole project, and split and/or regulate each one as needed?

I've supplied the final state of the first build for your analysis and input. In can be viewed here. It may be a little messy, but it originally served as my "notebook" as I was breadboarding and updated as I was kludging.

Hopefully, I have provided a thorough enough detail on the current situation (additional details below).

Thanks!

P.S. I know this could be done much simpler - without all of the ICs. The constraints I have placed on myself for size and control options are due to the fact that I am trying to build this mixer to fit the same format as the instruments with which they will be used.


My Own Analysis

What I think I should be doing is bringing in the 9v power to the on/off switching circuit - positive always connected, and negative on a flip-flop controlled MOSFET like this. From there, regulate down to 5v, and split it with a TL072 to +/-2.5v in this manner. Then, build the circuits independently, with their own decoupling caps. Deliver the full 5v to all IC circuits, and connect the 0v/ground to ICs whenever a ground pin is present. Connect the audio jack sleeves and potentiometer grounds to 0v as well.


Design Additions

Since I am picking up so much space in the main compartment, I am adding a some functions with ICs having a max rating of 5.5v.


Current Component Considerations

The ICs I currently have and plan on working with at this point are CD4013 Flip-Flops (power MOSFET and switch IC control), CD74HC4316 Switches (muting), LM358 Amp (output), TL072 (ground reference), PT2399 Echo Processor, DG403 Switch (echo bypass), and AN6884 LED Driver. I am considering the IQS127D Touch Sensors to replace the tactiles for my mute switch and echo bypass controls.


Edit/Update 3/28

Below is my interpretation of how I should implement the suggestions I have received so far. Although power handling is my main concern, I have created a block diagram including a few passive components that might be of note. i.e. coupling capacitors and some resistors (for 0v, circuit interconnects, and those within the audio path). Some are left blank, but those related to power are labeled. I am considering 3 regulators, and blocks are color coded by the board on which they will reside. ICs at the heart of each block are already linked above.

Mixer Block Diagram #1


My Original Approach

I (mostly) applied the suggested applications from data sheets where available, making a few small adjustments.

The original build used a voltage divider to create a virtual ground reference with the option of either a 9vdc power supply, or a battery of 6xAAs. 10µF caps were placed at the power input for all decoupling (mistake?). Wires to the powered boards are about 7-8". All audio signals were grounded to the 0v rail.

Where a dual-supply IC was used, I connected the Positive, Ground, and Negative to +4.5, 0, and -4.5v accordingly. With single-supply chips, I connected to +4.5, and 0v. I think this was my biggest mistake. It made total sense at the time, but caused an imbalance yielding (if I recall correctly) an approx. +6/-3v supply. I "corrected" the issue with a potentiometer on the divider so that I could adjust it until the +/- rails were both equal.

When the original dimming issue occurred, I first reduced the resistor values of my divider (figuring current was being limited), then replaced the 10µF decoupling caps with 100s. Neither solution made a visible improvement.

\$\endgroup\$
  • 1
    \$\begingroup\$ I read the first two pages but gave up. Do you have a question hidden somewhere? If so, I suggest you remove 90% of this essay and focus on that. \$\endgroup\$ – pipe Mar 27 '17 at 14:24
  • \$\begingroup\$ @pipe My sincere apologies. In trying to avoid the criticism I often see here regarding how one tries to ask for help, and knowing the desire for as much information as possible regarding research, attempts to accomplish a task, approach taken, and intentions of the project, I tried to provide all of that. I've seen comments like "give us all of the information you have, and we'll decide what's important," so I did just that. Skip to the "sum it up" part - which is what I believe is most important, and I welcome any mod to edit out what is really irrelevant. \$\endgroup\$ – Jay Mar 27 '17 at 14:45
  • \$\begingroup\$ Does performance improve, if you use 6 batteries to provide +4.5 and -4.5? This should achieve a strong GROUND, and let you diagnose the dimming, etc. \$\endgroup\$ – analogsystemsrf Mar 27 '17 at 15:10
  • 1
    \$\begingroup\$ Absolutely, Jay - @pipe, he can't win - is he due condescending comments either way? \$\endgroup\$ – TonyM Mar 27 '17 at 15:52
  • 1
    \$\begingroup\$ @Jay I would suggest a short, focused "Executive Summary" section (clearly titled as such) with your main question in bold letters, or as bullet points if there are several of them. Then, a "Details" section (also clearly titled as such) as much structured and concise as possible. Otherwise, it's overwhelming. \$\endgroup\$ – Enric Blanco Mar 27 '17 at 18:01
1
\$\begingroup\$

Well.

Let's suppose you split your 9V wall adapter into +4.5V and -4.5V rails using a virtual ground which we shall name VGND. The other option is to use standard split rails: +4.5V, -4.5V, and a real GND.

Now, we connect all the single supply (SS) chips between +4.5 and GND (or VGND).

All the SS chips' supply current is obviously drawn from +4.5V, and loops back into VGND. This includes your echo module. Therefore,

  • your VGND generator should be able to sink enough current to cover then entire supply current of your SS chips. TL074 can't do that, it's an opamp for very very light loads like a 10k resistor.
  • VGND is also the voltage reference for the dual supply (DS) chips. Feeding random variable currents into it (from your SS chips' supplies) will inject noise into VGND. It would do the same with a real GND, but the impedance of a ground plane is quite a bit lower than the output impedance of a virtual ground. This might make your layout more complicated if you don't want noise can enter your audio signal chain.

Now, the second point isn't set in stone. If everything is referenced to VGND, and its layout is good, then it will work just as well as a normal GND. However you should be very careful not to have two different references (can happen if a part of the circuit is AC-coupled). For example, if one reference is VGND and the other is a voltage divider between supplies, then as VGND wiggles around due to it being used as supply ground, the other will not follow, and the difference will be injected into your signal.

Note: When one of the chips pumps current into VGND, you can say "VGND has noise". But from the point of view of your circuits, VGND is fixed, since it's the reference. It's always 0V, by definition. The "noise" I'm talking about will appear on both supplies instead. And contrary to a normal design, where you could add filters into the supplies to isolate a noisy bit of circuitry from the rest, here it would be more complicated.

Also, if you use a standard virtual ground chip, the last I checked generated huge class-B distortion on the rails when AC current was drawn from VGND. It is a voltage follower opamp after all, with a class-AB output stage, and usually a very low bias.

Good capacitors are cheap. Virtual grounds are a headache. I would AC-couple everything, and use single-supply everywhere. Much simpler.

\$\endgroup\$
  • \$\begingroup\$ OK. Thank you!!! (whew, more to process) I'm not sure how to get standard split rails, but I'll to try to interpret. If I were to create VGND through the 072, and move all devices to the +/- rails, then VGND can remain only a reference where it is needed, carrying minimal/no load. At this point, the only things using VGND are the audio (jack, pan, volume), LM358 and switch ICs - none sinking current. Then I couple the signal through the switches, and the 358. I have the echo as a separate entity within the same box (no VGND is needed). Am I even close? \$\endgroup\$ – Jay Mar 27 '17 at 21:03
  • \$\begingroup\$ Please post a schematic of how you'd like to arrange the stuff. Hand drawn block diagram is fine as long as each block has supplies and ground. Also TL072 won't drive headphones if you have a headphone jack. \$\endgroup\$ – peufeu Mar 27 '17 at 21:40
  • \$\begingroup\$ Ok. Will do. Should I add to the question, or a comment here? Also, the LM358 is driving the phones. The 072, as I think I understand it, would only supply a stable reference. \$\endgroup\$ – Jay Mar 27 '17 at 22:19
  • \$\begingroup\$ You can add the images to the question. LM358 is wimpy for headphones, slow, low and asymetrical output current, high distortion, will sound like crap. Try NJM4560, RC4560 etc, costs 50c, works well, it's the industry standard opamp for driving headphones in pretty much any consumer equipment... \$\endgroup\$ – peufeu Mar 27 '17 at 22:29
  • \$\begingroup\$ I've created the block diagram, and updated the question. As for the LM358, it was one of 4 or 5 that I tested, and it was the one that sounded best, so I settled and used it. Output level was meh. Slight clipping (but comparatively tolerable) distortion was noted. Results were "good enough" but I welcome improvement, and I will take that recommendation and order a few 4560s in DIP and SOIC with my sensors. \$\endgroup\$ – Jay Mar 28 '17 at 16:39
1
\$\begingroup\$

The primary question here is, what is the best practice for powering ICs and managing audio grounding in this type of environment?

A ground plane carries parasitic resistance (and a tiny amount of inductance) , the main idea is to design the ground plane in such a way that the parasitics don't affect the design. One of the best ways to accomplish this is to make one layer of the PCB a solid plane. If you don't, then you need to figure out how much parasitic resistance and current, then find out if the voltage from the parasitic resistance will affect the design.

Should all single supply ICs be connected to the +/- rails, instead of +/0v in a mixed single/dual supply IC situation? If so, how does this affect things like the flip-flop output to the switch inputs for my muting circuits?

The same thing (about parasitics) goes for the power rails, they also have parasitic resistance. If you run traces next to each other you will have small amounts of mutual inductance (nH to pH) and capacitance (

Amplfiiers have a Power Supply Rejection Ratio (PSRR) which tells you in dB how much rejection of noise into the output signal of the amplifier that might be on the power supply rail, usually this is +60dB and 90dB to 120dB is not uncommon. So you if you can come up with noise requirements for your signal, you can also come up with noise requirements for your power rails. Then you can determine what a switching current will do to the source and how much it will drop the rails and affect the signal. Trace parasitics will also come into play so if your really worried about it, run a quick worst case spice scenario. Also consider what power filtering capacitors will do to the power rail regulation.

Should I just pick one voltage (5v) appropriate to all ICs involved, and regulate to that right at the power input?

See above post, then consider if breaking out the sources with your analog on one and switching or digital loads on another regulator is worth the cost.

Should I treat each IC circuit as separate devices within the whole project, and split and/or regulate each one as needed?

Regulation is an art, not a science and very much design dependent. It really depends on your noise requirements. That being said its generally good to have a regulator for your analog components and one for your digital components. There are also more costly regulators that have better regulation and less noise.

\$\endgroup\$
  • \$\begingroup\$ Thanks for all of this. I am going to need some time to process it. I've been using single-sided boards, but might have to switch to 2-sided for a ground layer. I'm OK with a small amount of noise (and distortion) since this is a one-off "writing" mixer, and not for recording, but will try to keep it as low as possible for learning reasons. Looks like the LM358 has a PSRR of 65-100dB, btw. I was considering a 5v Regulator to supply the whole unit, but might break out the analog/digital now - sticking with 5v throughout to compensate for dropout as batteries drain. \$\endgroup\$ – Jay Mar 27 '17 at 19:15
  • \$\begingroup\$ These days, you can get professional double sided plated through holes 10x10cm board for 20€ on pcbway, seeedstudio and others. There is no more any reason to mess with FeCl3 in your basement... And LM358 would be an option in 1985, yeah. Myself, as a cheap opamp, I like OPA1652. It's 1.62€ at Mouser, has ludicrous performance, RR output, FET input, low noise, and it sounds very, very good. A 50c opamp like 4560 or 4580 or 5532 will beat LM358 on measurements and listening tests, too. \$\endgroup\$ – peufeu Mar 30 '17 at 19:44
0
\$\begingroup\$

From looking at your schematic, it appears that you've defined a virtual ground at mid supply using resistors. This works just ok as a reference (for maybe an instrumentation amp or ADC/DAC). However, it works absolutely awful for a power system.

Instead, split your battery supply in two. Connect the middle of the supply to Gnd.

For example, (2) AA's in series to generate +3V, and (2) more AA's to generate -3V. Ground comes from the mid point in between. Add more series batteries for higher voltages.

Like: (-3V) [Battery]+ [Battery]+ (Ground) [Battery]+ [Battery]+ (+3V)

Do not insert any resistors in the ground return path. You want this path to be as low-resistance as possible.

After you fix the supply, get rid of R101, R102, R105, R106 and R110. You will also end up saving some power.

Good luck.

\$\endgroup\$
  • \$\begingroup\$ Thank you for your suggestion. I used the virtual ground since I will primarily be running the mixer from a wall adapter into a switched power jack - mostly using battery power only when I am away from mains. At the time, that was all I knew to accommodate both options. I had considered a "center-tapped?" battery supply, but the adapter seemed to negate that - at least based on my level of knowledge. I will carry this information for future projects because it was something I was curious about. \$\endgroup\$ – Jay Mar 27 '17 at 18:18
  • 1
    \$\begingroup\$ The circuit as shown is will not operate properly with the ac adapter. If you want to use an adapter, then the adapter could supply the +V. Then you will also need to use an on-board switching supply to generate the -V rail. And because wall-wart supplies tend to be very noisy, then you should consider adding post-regulation with linear regulators for both rails. \$\endgroup\$ – Chris Knudsen Mar 27 '17 at 19:04
  • \$\begingroup\$ Yes. I am leaning toward multiple internal linear regulators. \$\endgroup\$ – Jay Mar 27 '17 at 19:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.