LM3915 Issues

[MarcusS]

I didnt mean i want 5v rms for the first led to turn on.  That is how it is actually behaving right now with my ground reference issue. I want 80w rms into 6 ohms for led 10 and each led below that to be half power -3dB.  This is how an lm3915 is intended to be used in my understanding.

[StillTrying]

I will give it a go.  Im trying to get my head around why your circuit has a voltage divider for div low instead of that being connected to 36v gnd.

Yep, you could produce Vref Low from the 36V supply. But producing both Vref Low and the Sig_In from the same place(through their same proportion potential dividers) should mean that if the mid 18V is off a bit, it effects both of them - so the Vref Low voltage level should still run right through the very middle of the signal's waveform even at low volume levels. That was the idea anyway.

If you're looking at the last circuit I posted, R7 28K represents the internal resistor string inside the LM3915, I added it because of the small effects it has on the 2 potential dividers.

Changed speaker to 6R
Had to change the potential dividers R2 from 3K9 to 4K7, and R6 from 390R to 470R.
Plot looks good to me!
Blue is the 500Hz (peak 31V) reaching Vref High, Yellow shows the peak Watts in the 6R = 112W = 79 W RMS.
Shirley that's close enough!

I've not worked out the 3dB representations, I'm leaving that to you.

[Audioguru]

Tdk does make a ccg30-48-12s because they are soldered and working on my board in front of me.

Twice before you said it was a TDK CCG30-4812s which is not the TDK CCG30-48-12s that it is. The TDKCCG30-48-12s has a lot of output ripple and its output ground is isolated from its input ground but you do not need the grounds to be isolated.

[Audioguru]

Why don't you guys understand that the audio is AC, not DC so simply use a coupling capacitor to block the 18VDC but pass the audio.

[StillTrying]

the input of the LM3915 must be averaged at ground

Why must? But, I can see that connecting VrefLow to gnd and AC coupling Sig_In could a good idea.

Your LM3915 has no pin numbers.

And.

Why does your circuit have a "Vref Lo"?? Pin 4 of the Lm3915 should be connected to the circuit ground.
Why is your Vref so low? Make it about +10V (and adjust the input attenuator to match) then input offset voltages will not mess up the outputs.

Could be good ideas, but after a status bar taskbar crash followed by LT crashing I no longer have any .asc version!! 

[void_error]

I didnt mean i want 5v rms for the first led to turn on.  That is how it is actually behaving right now with my ground reference issue. I want 80w rms into 6 ohms for led 10 and each led below that to be half power -3dB.  This is how an lm3915 is intended to be used in my understanding.

Please be aware there is no such thing as W RMS, only V RMS and I RMS.
The LM3915 lights the LEDs up depending on voltage, not power, as the only input value is voltage.

Why don't you guys understand that the audio is AC, not DC so simply use a coupling capacitor to block the 18VDC but pass the audio.

I did previously suggest that here.

[MarcusS]

Turns out what was suggested doesn't work seems like the internal impedance between pin 4 and pin 2 is enough to cause all leds to turn on solid.

[void_error]

Pins 4 & 2 are meant to be tied together unless you're using a single supply with your incoming signal DC coupled and referenced to 1/2 of your supply voltage but that seems not to be the case here.
Leaving pin 4 floating is not a good idea because the internal resistive divider won't work.

[MarcusS]

Any more suggestions that might work.  I was kinda hoping that it compared pin 5 to pin 4 internally with the comparators to determine the scale.

[Chris Mr]

My suggestion would be that you read the data sheet and ask questions about things you don't understand.

You say "I was kinda hoping" - what's the point in Kinda hoping  you need logic, reason and a rigorous approach to tackle these things rather than guesswork.

The LM3915 needs pin 4 and 5 to be the signal input to generate the LED output.  See data sheet page 2 where there is a box that is labeled "signal source".  You have already put your scope on pin 4 (-12 on your circuit) and find that it is not as you previously described as "doesn't budge" but has audio on it - no wonder the display doesn't work.  All you have to do is hang two scope probes - one on pin 4 and one on pin 5 (the difference between the two is the signal source) - forget about what the LM3915 displays; the difference between those two points is what you are interested in.  Only when those two points have what you expect should you move on to look at what the LM3915 is displaying.

seems like the internal impedance between pin 4 and pin 2 is enough to cause all leds to turn on solid

See data sheet page 3 which tells you what the range of RLO can be (otherwise known as pin 4).  See data sheet page 7 for an internal schematic of the LM3915 that shows pin 4 is the bottom of a resistor ladder.

To put it another way, going from injecting a signal to the input to looking at LEDs as the output is too coarse for testing.  You need to know what to expect at every stage and check that is the case until finally looking at the LEDs.

Clearly, pin 4 (-12 on your circuit) should not budge at all as it is one of the inputs to both LM3915s.  Stick the 1,000uF on I suggested earlier and then measure the two inputs as described.

[MarcusS]

I have studied the datasheet in great depth the block diagram says you can compare pin 4 to pin 5.  That is not correct it is only correct if pin 2 is connected to pin 4.  In my instance where i have an audio ground that is different to 12v ground what can i do.  I did try 1000uf and it does not work.  What i tried this evening was to have pin 2 not connected to pin 4.  I had amplifier in on 5 and amplifier gnd on 4 and then vu meter powered from dc dc converter on pins 2 and 3.  Cant see on the block diagram where you cant do that and very confused why it refers to the comparayors as floating on page 1.

[Chris Mr]

Where does it say in the data sheet "you can compare pin 4 to pin 5" - which page and which paragraph?

[MarcusS]

It doesnt say in text the block diagram shows pin 5 goes to a buffer which connects to 1 side of a comparator and pin 4 connects to the other side of the comparator

[MarcusS]

All i want is someone to correctly tell me how to add an lm3915 to each output of the amplifier schematics i posted as a power indicator

[Chris Mr]

Let's take the block diagram apart:

There is a resistor network which is connected to 10 comparators - whose outputs drive LEDs.  Both ends of that resistor network are available on pins 4 & 6 BUT (time for first caveat) pin 6 must be either the same or higher voltage than pin 4 (data sheet page 4) and neither of them can go higher than +12v compared to pin 2.

The impedance of the resistor network is (6.63 + 4.69 + 3.31 + 2.34 + 1.66 + 1.17 + 0.83 + 0.59 + 0.41 + 1) K.  Notice that the lowest step is 1K and the next up is 0.41K.

The other connection to the comparators, the connection that's common to all comparators, is the signal input - pin 5 - which is buffered.

So, what we know is that, regardless of which bits you use internally, the comparators have the resistor network on one side and the signal input on the other.  The resistor network needs a voltage across it so that each comparator has a different input which they compare to pin 5 and output accordingly.

You need a voltage across the resistor network and guess what, those nice TI people provided one - but you don't have to use it.  The voltage reference police won't come around and put you in jail if you don't.  The reference is a 1.28v (nominal) between pins 7 & 8 and the current you take out of it is directly related to the LED brightness - here you have to comply as that is in LM3915 law.  Notice that they tell you in the data sheet that the reference terminal, pin 8, takes 70uA in case you don't connect it to ground.

It also says that pin 2 is the ground pin.

They are right, the resistor network is floating as long as it is within the caveat that the low end must be the same or lower than the high end and both must be between pin 2 and should never be more than 12v.

So when you put pin 2 at amplifier ground you went outside those requirements. Expect to have damaged the device - but there is always luck

[macboy]

If you want the output of the LM3915 to look good, you need to precede it with a rectification and peak hold circuit with a slow(ish) decay. The decay should be set to decay from full scale (10th LED lit) to minimum (no LED lit) in about 0.2 to 0.5 second, but you can adjust this to your preference. The LM3915 is very fast, so if you don't use a peak hold circuit, the LEDs will show the instantaneous level of the positive half cycle of the waveform. This gives very uneven brightness and usually makes it very difficult to tell when the highest LED ever lights because it only lights very briefly at the highest peaks of the waveform. Read the section "Tips on recitifier circuits". There is a striking and obvious visual difference between a directly connected LM3915 and one using a rectifier/peak hold.

If you are adding an op-amp based peak detector/rectifier circuit anyway, then it is easy to configure it to have a differential input for direct connection to any BTL amplifier such as yours.

[MarcusS]

I didnt put pin 2 at amplifier gnd i put pin 2 at 12v gnd. It was pin 4 that was connected to amplifier gnd and pin 5 to the positive spk out.

[StillTrying]

Right.
After the crash I rebuilt my rubbish simulation .asc from the .gif posted here, although my LM3915 was just an empty box, I hoped the voltages supplied to it would give at least some sort of result. (I don't know where the OP is up to with his real-life version).

So I found a LM3915 model, connected it and ran it, and what happened, - well with increasing input the LEDs came on in order! - although the Refs would need tweaking. It's impossible to plot these 10 LEDs switching at 750Hz all on the one plot - so I did just that.

I can see that the LED currents are switching on in order (by the colours) with increasing input signal amplitude even if no one else can!
I won't be simulationing any further with this rubbish version but if anyone wants the .asc to play with I'll add it here.

[StillTrying]

I didnt put pin 2 at amplifier gnd i put pin 2 at 12v gnd. It was pin 4 that was connected to amplifier gnd and pin 5 to the positive spk out.

You confuse me with all these gnds, there's only one.

It might be time to test your LM3915, connecting it's VrefLO to it's gnd, it's VrefHI to it's 12V supply, and a Pot across it's 12V with the pots wiper going to it's Sig_In would do.

Meanwhile, I have another (better) go at getting the 18V speaker output safely into the LM3915's input.

[MarcusS]

Think I have found the solution.  But I need to do some serious testing on both channels to make sure it is all well behaved.  It involves decoupling the audio into the vu meter via a polyester film capacitor calculated high enough in value to not affect the frequency response of the vu meter in the audio band and then by doing so the dc offset is removed allowing me to take the floating output gnd of the dc dc converter and link that to the main amp ground thus having a stable reference.

Tying the dc dc converter output gnd to main amp ground did not work without the right capacitance as either the audio signal was high passed into the vu meter and thus not responding to audio frequencies and if no capacitor is used at all the dc offset (that is present when the output of the dc dc converter and amplifier are connected) makes all the leds turn on solid. 

The capacitor value is critical and obviously related to the impedance of the voltage divider between pin 5 and 4 to determine the correct value.

[MarcusS]

Thinking about it now when I first posted the question the answer was in the question when I asked if anyone had done anything with decoupling capacitors.  Turns out that its on the input of the vu meter that needs it, not decoupling the dc dc converter gnd to the amplifier ground.

[void_error]

 
Finally.

It involves decoupling the audio into the vu meter via a polyester film capacitor calculated high enough in value to not affect the frequency response of the vu meter in the audio band and then by doing so the dc offset is removed allowing me to take the floating output gnd of the dc dc converter and link that to the main amp ground thus having a stable reference.

That's called AC coupling the LM3915 input. It's exactly what everyone's been telling you. And yes, reference it to the power amp's ground.
Also, take into account what macboy suggested if you want it to look nice.

You might have an even better/cheaper option: buffer the signal from your amp's input and feed it into the LM3915 via an AC coupling cap, and optionally use the peak/hold circuit in the datasheet. The result will be identical as the amp has a fixed voltage gain over the audio range.

[StillTrying]

The capacitor value is critical and obviously related to the impedance of the voltage divider between pin 5 and 4 to determine the correct value.

You might have to put the AC coupling capacitor at the top of your voltage divider so that it's not effected by the horrible change in impedance of pin 5, - about 200M for +ve voltages and 20K for -ve.
I'll look back in 3 pages time to see where you've got up to. 

[MarcusS]

It was the Ac decoupling cap at the top of the divider that fixed it that allowed me to tie the dc dc converter output ground to main amplifier ground.  Something I didn't think I could do due to it misbehaving.  My original tests had the decoupling occurring across the dc dc converter rather than the input.

[StillTrying]

Now that the multiple gnds problem is solved. 

With the high signal voltage level on the speaker outputs (+/- 14V peaks), and the LM3915's very high input impedance, it should be possible to produce a simple peak/envelope detector out of a few passive components.

Here's using both speaker outputs so that both +ve and -ve peaks are detected, I don't know if the component values are very good - I've only simulated it with the 500Hz speaker signal.

I think this has quite a good response for such a simple arrangement myself.