That1580 mini mixer

[mediatechnology]

1) The 1206 may be a little noisier than the 1246.

So it looks like I will return to using the AD8273 due to better tolerances

2) If you look carefully at the AD8273 datasheet you will see in note 5 that the absolute tolerance is also ±20%.
For simple LCR switching I think you can come up with something simpler.

3) Tantalums as bypass are OK.

4) Soldering the 5171 by hand is a PIA. Amazed it works at all and am impressed with your SMT soldering skills.
Based on your resistance measurements its sounds like the RB gain leg is open.

5) What DC potential is the 5171 thermal pad referenced to?

6) You're really going to need a test oscillator, coupling transformer and pad to measure a mic input.

7) BTW maybe its just me but on the switcher board you might want to put ground in between the 48V and 3V3 rails. Having the 48V that close to 3V3 with those small clearances looks to be an opportunity for leakage/creep or "opportunistic" shorts between the highest voltage and lowest voltage rails. 48V to 3V3 is the path of greatest destruction.

[weroflu]

I think there was a miscommunication on my part. I have two 1580 preamp cards soldered. The first one was not passing signal, or maybe extremely weak. I could not find the problem so I decided to try soldering the second one, which is working but has the possible low bass problems (Im not even sure at this point).

In an effort to salvage the first one, and also test whether the problem was caused by the 5171 i removed the 5171, and attached fixed gain resistors. This turned out to be a bust. So I think this card will be retired. Though I am looking at the costs adding up, and wondering if I should try another 1580 chip on there. Im always paranoid about too many reflows/heat damage, but also cheap and trying to save as many parts as I can which is probably not a smart idea when testing.

I tested the 2nd preamp card again with the new summing card (that seems to work), and the signal sounds actually pretty good. At this point it's too hard to trust my ears so a proper testing jig is high on the list. Noise was not bad either, just some very faint high frequency that could have been ambient, or radiated who knows. The motorboating from the rfduino was very much attenuated compared to last time, barely audible. I have no idea what's improving things.

I'm going to try for a better LCR panning circuit.

For the smt parts, the supposed hardest parts are the easiest if you get it just right. For the passives I'm using paste and hot air mainly, graduated from tormenting myself with an iron. I tried 3 different paste syringes and could not get any of them to work. The last one, a small edsysn syringe, I tried again and it worked well. It turns out it was a simple fix. I had been letting the end dry out too much, even with a cap placed. Just a few primer squeezes and the paste came out more liquid and easier to control the dose. The more liquid paste seems best to me.

This https://www.indiegogo.com/projects/pcbi ... -holder--4#/
helped tremendously. I was using a cheap alligator clip vise and it was terrible.

The bga initially I did with hot air, even though it did supposedly work (after 3 attempts) since I got good outputs on the ltm8049 I wasn't happy. Got a 10$ hotplate, did a few thermocouple measurements around the perimeter and tried another 8049. It sat down perfectly, slid it off onto a larger pcb, and that turned out the easiest of them all. Alignment was very easy since there was no air or movement while the reflow took place.

Soic, Sop's are pretty simple to do with paste, just a thin line and some hot air. Then touch up with an iron.

The qfn's were a little tricky, and I'm still sort of working that out. For the 5171 I tinned the pads, cleaned, preheated the board with hot air, then applied tacky gel flux, then placed the part with one hand and used the other for hot air, it reflowed fairly quickly and self-centered very well. 5171 Thermal pad was left floating, I guess next rev. I will connect to Vee?

I ASSumed the same would be true for the 1580, but no! I guess due to the pcb thermal pad underneath it dissipated too much heat and it would not self-center, or just barely so (just enough to plant a seed of doubt in your head as to whether it actually worked). I switched methods: tinned the underside thermal pad, tinned the top pads, cleaned, fluxed and then heated the thermad pad from below with a blob-tinned iron tip for better conduction. It self centers fairly quickly like this. I meant to ask you if this is dangerous and is more likely to damage the part. My past failures with qfn were all due to not enough heat. Depending on your answer to this I might switch over to the hotplate for both the 1580 and 5171. I'm leaning toward this anyway because both hands will be free and it's easier to poke the chip and test for centering.

I have a signal generator on the oscilloscope, just have to get a transformer and make the correct pad.

The ltm8049's are not cheap. I got two as samples, and from here on it looks like it's out of pocket, or maybe I'll try the lt rep again. The preamp cards as well are not so cheap as they are 4 layer, so I need to be a little more careful with revisions. I think for the p/s board so long as it's working I'm inclined to wait for something to blow up before I make a change. For noise issues though I would like to see if I can fit some more filtering on board.

I need to do another run on those as I messed up the summin card connecter pad spacing. I moved the two large caps on the p/s board to the chassis which allows larger values, maybe 1000uF for the bulk input and 330uF for the phantom cap - I know it doesn't do much, and need to revisit phantom for a possible rc/other filter there, but first the preamp has to work properly. There is space on the chassis around the rfduino board but not sure what I can do there. I also meant to ask you about this alternate power scheme. I updated the shutterfly page with some more images. You can see the chiclet sized ltc3265 card there, small and almost no depth to it. At one point I considered using 6 of these as plug in sub-boards on the the preamp cards for the bipolar rails. It would mean bussing Input supply which is why i didn't do it, but I suppose if the dc supply was filtered well, bussing it would not be so bad. That would make the only radiated noise source the phantom supply.

[mediatechnology]

indiegogo is a cool site. I got lost in there.

I don't have much experience soldering the QFNs.
I prototype with DIP adapters.

One thing I have become aware of is rising LF THD when there is a thermal void under the QFN.
The way you pre-tin it and heat from below should reduce the potential for a void.
I'm not sure that there is much harm from the thermal pad electrically floating but I wanted to make sure that it wasn't grounded since it is tied to Vee.

[weroflu]

For simple LCR switching I think you can come up with something simpler.

The only ideas I came up with are a dual op amp with discrete resistors as in 1206/8273, or 4 resistors off of the sp3t switch, two center resistors adjusted to whatever pan law I want. The latter foregoes a buffer amp but I think I'd have to switch over to virtual earth summing from the current passive mix bus plus gain.

Is there an advantage to having the center channel -3.5db down as opposed to 3db as this provides .5db in gain step resolution?

[mediatechnology]

The only ideas I came up with are a dual op amp with discrete resistors as in 1206/8273, or 4 resistors off of the sp3t switch, two center resistors adjusted to whatever pan law I want. The latter foregoes a buffer amp but I think I'd have to switch over to virtual earth summing from the current passive mix bus plus gain.

That sounds like a good plan. I would go with current summing.
That AD circuit saves a lot of parts but the resistor absolute tolerance of 40% is just too much.

Is there an advantage to having the center channel -3.5db down as opposed to 3db as this provides .5db in gain step resolution?

JR and ricardo can shed a lot of light on this subject.

My take on it is that constant voltage panning has a -6 dB dip in the center; constant power -3 dB.
Both have their merits.
A compromise is -4.5 dB.

When I worked for SSL -4.5 dB was what center was trimmed to.
That makes the constant voltage center bump +1.5 dB and the constant power dip -1.5 dB.

[JR.]

The only ideas I came up with are a dual op amp with discrete resistors as in 1206/8273, or 4 resistors off of the sp3t switch, two center resistors adjusted to whatever pan law I want. The latter foregoes a buffer amp but I think I'd have to switch over to virtual earth summing from the current passive mix bus plus gain.

That sounds like a good plan. I would go with current summing.

not sure I follow,,, do you mean virtual earth?

That AD circuit saves a lot of parts but the resistor absolute tolerance of 40% is just too much.

Is there an advantage to having the center channel -3.5db down as opposed to 3db as this provides .5db in gain step resolution?

JR and ricardo can shed a lot of light on this subject.

My take on it is that constant voltage panning has a -6 dB dip in the center; constant power -3 dB.
Both have their merits.
A compromise is -4.5 dB.

When I worked for SSL -4.5 dB was what center was trimmed to.
That makes the constant voltage center bump +1.5 dB and the constant power dip -1.5 dB.

ASSuming the bulk of the audience is listening in stereo -3dB pan law will sound more constant loudness while panning.. I have never embraced the -4.5dB compromise as being wrong for both scenarios.

JR

PS: I never tried BGA but have successfully prototyped processors in QFN package. The tiny class D chip I use in my tuner has a heat sink pad on the bottom, a little solder paste and hot air does the trick. You can always touch an iron to vias in the bottom pad to make sure solder on top has melted.

[mediatechnology]

not sure I follow,,, do you mean virtual earth?

Yes.

I think I recall from the SSL discussions that broadcast/mono compatibility drove the compromise to -4.5 dB.

[mediatechnology]

Looking at the 1286 LCR panner circuit I'm not so sure that the internal resistors' absolute tolerances do participated in the overall gain.
The first time I looked I didn't see that the reference terminal was looped back to the + input.
That puts the resistors in series with the noninverting input.
Though their absolute tolerance won't effect gain they will add noise.
I think it will work fine though.
I'm not sure why it's better than a dual op amp other than it saves 4 outboard gain resistors.

EDIT:

Wouldn't this be simpler and eliminate the 1286/pan buffer stages?



I realize that it may not have the "offness" of the shunt configuration but it is a lot simpler.

or 4 resistors off of the sp3t switch, two center resistors adjusted to whatever pan law I want. The latter foregoes a buffer amp but I think I'd have to switch over to virtual earth summing from the current passive mix bus plus gain.

It's what I meant by current summation - combining the LCR "pan" switch circuit with virtual ground summing.
The preamp output feeds a single build-out R and 1 pole 3 pos rotary.
This eliminates all those 1286s and tons of resistors and lots of power.
It also gives you an optional third bus. (Break where C sums into L and R.)

You could get fancy and eliminate the mechanical switch, add DG41X switches and put LCR assignment under remote control.
You'll still come out ahead on power (compared to the 1286s) and with CMOS switches use a series/shunt configuration to improve offness if its needed.

Did I see a headphone amp in this gizmo?
If so add a 3.5mm stereo line input for playback or IFB.

[weroflu]

No headohone amp planned

re: cmos switches, I thought about those for phantom, pan and phase as was intended from the 5171 designers but I don't know what ground noise problems await so I went for analog. Plus, the intended use for this box is set and forget, set gain, pan, phantom and use it to record. There will be no gain riding or on the fly panning.

Metering maybe could be useful but the downsteam box will have meters so it's not really essential.

I'm still studying the panning you suggested. You were not the 1st, 5th or 10th person to tell me to use virtual ground. Initially my reason for staying with passive bus was my panning scheme was incorrect and I thought there would be 4 channels per bus which would be somewhat quieter with a noise gain of 4 vs. v.e. noise gain of 7. With 6 channels per bus I may as well go ahead with v.e. summing. JR, in some other posts I think brought up the point that it's not just a noise issue but phase errors as well from v.e. if I understood correctly. So that coupled with expected ground noise led me away from v.e. summing. Maybe it's possible to keep the summing topology modular/isolated to the summing card alone. I'll need an inverting stage before/after the v.e. summing.

There is one other issue. I actually wanted to burn a little power to increase efficiency on the ltm8049. Too low a current draw and I might be heating things up too much. Will check datasheet later. I really need to pick a number for the rails and stick with that, but it's going to require a new p/s board and parts since I used fixed voltages on the regulators. I wonder if I should put some mini trims on the ltm8049 fbx V output resistors and use TI matching TPS7A4700 and TPS7A33 regulators and mini trims there too.

On the subject of powering That preamps, do you think using the ltc3265 charge pumps would create any audible problems due to weak/slow current supply. I've heard various opinions on that topic. If I do wind up lowering the current draw I might switch over to using 2-3 ltc3265's for the rails, or one per preamp card.

I've been trying to learn basic measurements with the picoscope. Sweep and spectrum plot is working but the in-built signal generator is unbalanced 600ohm output. I didn't realize it would output down to 1mV, no pad needed. For now I'm just going to tie signal generator V+ to Mic input+, sig gen gnd to Mic input-, and float ground. I'm working on a test jig board that plugs into the backplane with a drv135 on it so i can just use a 3.5mm cable out from the test board into a preamp. It seems like there are a lot of very useful things I could put on this test board, any ideas? Maybe some resistors and a switch to simulate different microphone impedances.

[JR.]

No headohone amp planned

re: cmos switches, I thought about those for phantom, pan and phase as was intended from the 5171 designers but I don't know what ground noise problems await so I went for analog. Plus, the intended use for this box is set and forget, set gain, pan, phantom and use it to record. There will be no gain riding or on the fly panning.

Metering maybe could be useful but the downsteam box will have meters so it's not really essential.

I'm still studying the panning you suggested. You were not the 1st, 5th or 10th person to tell me to use virtual ground. Initially my reason for staying with passive bus was my panning scheme was incorrect and I thought there would be 4 channels per bus which would be somewhat quieter with a noise gain of 4 vs. v.e. noise gain of 7. With 6 channels per bus I may as well go ahead with v.e. summing. JR, in some other posts I think brought up the point that it's not just a noise issue but phase errors as well from v.e. if I understood correctly.

Probably not... The phase shift which I alluded to (in addition to noise), is related to loop gain margin and with only 6 stems not a problem for modern SOTA op amps. Further, passive summing requires make up gain in the same ballpark as the VE noise gain, so same-same....My strategy of summing current sources, did not have the noise gain component so less noise and less phase shift.

For only 6 stems, don't worry about it... I've done consoles with over 100 stems feeding the L/R bus.

JR

So that coupled with expected ground noise led me away from v.e. summing. Maybe it's possible to keep the summing topology modular/isolated to the summing card alone. I'll need an inverting stage before/after the v.e. summing.

There is one other issue. I actually wanted to burn a little power to increase efficiency on the ltm8049. Too low a current draw and I might be heating things up too much. Will check datasheet later. I really need to pick a number for the rails and stick with that, but it's going to require a new p/s board and parts since I used fixed voltages on the regulators. I wonder if I should put some mini trims on the ltm8049 fbx V output resistors and use TI matching TPS7A4700 and TPS7A33 regulators and mini trims there too.

On the subject of powering That preamps, do you think using the ltc3265 charge pumps would create any audible problems due to weak/slow current supply. I've heard various opinions on that topic. If I do wind up lowering the current draw I might switch over to using 2-3 ltc3265's for the rails, or one per preamp card.

I've been trying to learn basic measurements with the picoscope. Sweep and spectrum plot is working but the in-built signal generator is unbalanced 600ohm output. I didn't realize it would output down to 1mV, no pad needed. For now I'm just going to tie signal generator V+ to Mic input+, sig gen gnd to Mic input-, and float ground. I'm working on a test jig board that plugs into the backplane with a drv135 on it so i can just use a 3.5mm cable out from the test board into a preamp. It seems like there are a lot of very useful things I could put on this test board, any ideas? Maybe some resistors and a switch to simulate different microphone impedances.