Simple Perfboard Summing

Hey guys. I want to build a simple 8 Channel summer. I like the idea of audiox's Sum-12 but I don't need that layout so I'm designing my own on perfboard.

The schematic is here: http://www.groupdiy.com/index.php?topic=29082.0

Attached is my layout. Anyone see any errors? Also I really want to know how I can add gain control rotary switch to this circuit, anyone know?

EDIT: I already see I made an error on the power rails are backwards

And here's my summing network.

If you post the schematic it´s easier to check what you did on perfboard. The rails are defo wrong. Where are your coupling caps? Better don´t DC-couple 5532s without servo.

jensenmann said:

If you post the schematic it´s easier to check what you did on perfboard. The rails are defo wrong. Where are your coupling caps? Better don´t DC-couple 5532s without servo.

Click to expand...

This is audiox's schematic, I hope he doesn't mind me reposting it.

I thought C3, C4, C5, C6 are the coupling caps aren't they?

nope, caps you mention are for filtering +15, -15V rails. Coupling cap blocks DC in your signal path.

benlindell said:

Attached is my layout. Anyone see any errors? Also I really want to know how I can add gain control rotary switch to this circuit, anyone know?

EDIT: I already see I made an error on the power rails are backwards

Click to expand...

First sight at least 14 more ... (half of each opamp is wired backwards, unused opamp stages are missing connections)

For your gain control (as shown, circuit is +/-0dB unity gain), you might use the unused opamp stages (U1b) for a non-inverting buffer or makeup gain stage between the summing amp and following inverting stage. For control place a fader or fader substituting rotary switch with series resistors between the switches throw positions in front.

Shouldn't all the de-coupling and filtering caps connect to gnd? ie; looks like c5/6 connect between the +/- rails. I thought those are best connected at (or as close as possible to) the opamp +/- rail pins to gnd in order to filter out HF noise and such to gnd.

Cheers,
jb

I believe the idea is that coupling between the rails makes any noise seen on one rail 'common mode' (since at least part of it will couple to the other rail, which will then be reduced by the CMRR of the gain stages).

Matador said:

I believe the idea is that coupling between the rails makes any noise seen on one rail 'common mode' (since at least part of it will couple to the other rail, which will then be reduced by the CMRR of the gain stages).

Click to expand...

Perhaps you have your acronyms mixed up...  PS filtering reduces the amount of PS ripple voltage, which is kept out of the audio by PSRR (power supply rejection ratio) of the ICs. There is no relationship regarding noise common to both power supplies, in fact PSRR is often different for + and - supplies.

JR

Hey John, So does that indicate that an ideal solution to rejecting HF noise on the +/- PSU rails may be to de-couple to GND using a low value cap (like 100nF) to gnd directly at the opamp +/- pins?

And ripple (LF "noise") rejection caps are higher values (like 10uF) are installed closer to the PSU or star-ground point to keep those larger gnd currents further away from the audio sections?

Just really wondering why the 100nF may have been connected between the +/- rails here.

Best,
jb

Harpo said:

First sight at least 14 more ... (half of each opamp is wired backwards, unused opamp stages are missing connections)

For your gain control (as shown, circuit is +/-0dB unity gain), you might use the unused opamp stages (U1b) for a non-inverting buffer or makeup gain stage between the summing amp and following inverting stage. For control place a fader or fader substituting rotary switch with series resistors between the switches throw positions in front.

Click to expand...

Very n00b mistakes on the pins and not connecting the unused stages.

I'm almost wondering if using this design based on audiox's schematic has any advantages over tv's design here:
http://www.groupdiy.com/index.php?topic=30806.0

Without doing a rigorous analysis or writing a book on the subject of summing (i probably could), I see a few important differences.

In the second design the two summing amps (+ and -) are not single ended in the master, so any common mode input, carries through undiminished to the balanced outputs (this could be awkward if interfacing to an unbalanced unit following this sum stage). On the positive side the two paths are symmetrical so common mode noise will be passed without time errors so they should cancel out cleanly in a following differential stage.

The first design has the advantage of subtracting the CM component out by cross coupling the output of one inverted sum into the input of the other, but using this topology the signal from the + bus passes through one more opamp stage before subtracting out, so HF cancellation may be inferior.

An alternate topology that would preserve the benefit of canceling out the CM term, while improving symmetry of the + and - path (and using up that left over opamp), would be to not cross couple the + sum output to the - sum input, but use two simple differential amplifiers. The first differential amp takes the inverted output of the + sum inverter into it's - input, so it comes out correct polarity. The + input of the differential is fed by the inverted output of - bus. A second identical differential amp grabs those same two bus outputs but connected the other way around, again diff amp - input to - sum bus inverted output, so - bus is restored to proper polarity, diff amp + input fed from + bus inverted output.  The two resistors coming from the diff amps + inputs can be connected to output ground to forward reference the output to ground at that node. While this is not very critical when treated this signals as balanced.

This more symmetrical path should give slightly better HF CM performance..and improved performance should you have need for using one output single ended.  I think I might have mentioned this in the earlier thread.

JR
 

JohnRoberts said:

(this could be awkward if interfacing to an unbalanced unit following this sum stage).

Click to expand...

Who in their right mind would do that?

The design goal of my summing device is clear: to provide the cleanest/shortest symetrical signal path -with- some oomph added from class-a output stage. So you can also (over)drive your favorite transformer if you wish... But it was intended to go straight into a balanced line input.

Read the pdf. But I'm not around much these days.

However, it should work if built properly.


p.s. if you really want a summing box that outputs a true balanced signal, plus DC-servoing, has some CM-rejection, and has balanced impedance at inputs, and is reasonably simple (read: three 8-pin dip chips for a stereo summer), use a so-called superbal topology. Google is your friend. You can easily pull it off if you study my pdf, discard the discrete output stages, and read about the superbal online.

Similar to my device (that has superior HF symetry but no CM-rejection //imo un-needed since the next device line input has it), you would connect summing resistors in the superbal config.

tv said:

The design goal of my summing device is clear: to provide the cleanest/shortest symetrical signal path -with- some oomph added from class-a output stage. So you can also (over)drive your favorite transformer if you wish... But it was intended to go straight into a balanced line input.

Read the pdf. But I'm not around much these days.

However, it should work if built properly.


p.s. if you really want a summing box that outputs a true balanced signal, plus DC-servoing, has some CM-rejection, and has balanced impedance at inputs, and is reasonably simple (read: three 8-pin dip chips for a stereo summer), use a so-called superbal topology. Google is your friend. You can easily pull it off if you study my pdf, discard the discrete output stages, and read about the superbal online.

Similar to my device (that has superior HF symetry but no CM-rejection //imo un-needed since the next device line input has it), you would connect summing resistors in the superbal config.

Click to expand...

Thanks for replying, I just found your PDF and have been studying it, it looks like it's much closer to what I want to do, especially with the easy provisions for adding a gainset switch. Is there any reason/advantage to using 15K summing network instead of 10k? Also you design doesn't require any modification for using 4 stereo channels as opposed to 8, correct?

Just omit 4 out of 8 to have 4 ch. with 10k you're on your own but it's flexible.

I killed my prototype because I fiddled with it too much while powered (went up with smoke)...

It should survive much abuse if it's properly built and safely mounted in a box...

JohnRoberts said:

Matador said:

I believe the idea is that coupling between the rails makes any noise seen on one rail 'common mode' (since at least part of it will couple to the other rail, which will then be reduced by the CMRR of the gain stages).

Click to expand...

Perhaps you have your acronyms mixed up...  PS filtering reduces the amount of PS ripple voltage, which is kept out of the audio by PSRR (power supply rejection ratio) of the ICs. There is no relationship regarding noise common to both power supplies, in fact PSRR is often different for + and - supplies.

JR

Click to expand...

Agreed, but the cap in question is between the two rails, not between the rails and ground.

PSRR is rejection from a rail with respect to it's ability to distort the audio signal through the gain stages.  CMRR is rejection of common mode signals between the differential inputs.  By coupling noise from one rail to another (through a cap between them), then such noise should be cancelled through both PSRR, as well as CMRR when the differential signals are summed.

Perhaps there's another use for such caps that escapes me?

Hey guys. Thanks for all the help and advice throughout this thread.

I decided to build one based on audiox's design, that I tried to make a layout for before. I took my layout and highlighted my errors and then just built it last week in one evening. I finally got a chance to put it together in my recycled chassis ($.99 on ebay!! much cheaper than par-metal) and it worked perfectly right away (except some of my input jacks are the wrong size so I only got to check a few channels of it at the moment). No noise or hum!!

I'll be using this one for the moment and I'm planning on a more elaborate summing mixer with that1200's and 2520's for gain with a nice big master fader and LED VU meters.

Here's another pic:

So I've been battling with my summing unit the last couple of weeks.

I'm getting a weird crackling noise that randomly appears BUT if I tap on one of the chips once or twice it disappears for a while then will come back again.

I've swapped chips from L to R and from 1 position to another and same noise from the right channel consistently. I just took it out and reflowed a few solder joints but it's still doing it.

Anyone one have any ideas as to what could be causing this?

> if I tap on one of the chips

Chips don't crackle.

Sockets can, but this is rarer in DIP (than tubes).

> reflowed a few solder joints

Don't blindly "reflow".

If a pin is tarnished and not taking solder, "reflow" just stirs the surface, does NOT fix the problem.

FIND the problem. BRIGHT light. Magnifier if chips are smaller than DIP or eyes are over 30.

Solder MUST "wet and cling" to BOTH metals being joined. Pin and pad. Wire and lug. Whatever. If more than two at a point (as in some turret work), all must wet and join.

You can't tell "wetted" if your solder joints are blobby. MINimum amount of solder to do the job. Imagine solder were expensive and heavy. (Hey, it IS expensive and heavy!)

If solder is beaded-up and not wetting some part of the joint, that's a clue. If a wire appears to have a wetted bead, but there's a thin line around the wire, either it's cracked in the solder (movement while cooling) or there is a film (flux or tarnish) around the wire. (One such case on an otherwise excellent CRT monitor drove everybody nuts until I found it.)

When in doubt, heat it up, and bang it on the bench. Get the solder off. LOOK and see if solder stuck to both sides. If not (or not sure), scrap for tarnish, solder-coat each half of the joint, look again, be satisfied, THEN add just-barely enough solder to join the two halves.

> right channel consistently.

Then the immediate problem is in that side. BUT if you have one bad joint, you probably have others. Find them now, rather than later when the talent is cooking.

These are borderline acceptable:
http://store.curiousinventor.com/media/images/how_to_solder/joint_examples_good.jpg?1248414209

Many are, IMHO, blobby. Some show no visible proof of wetting the pin. I would expect work like this to be 99% functional at final test. Since you may have 100 such joints, that means half your builds don't work first time. Finding the one bad joint takes more time than all the others. So it really pays to solder right EVERY time.