will this 48 V pink noise source work?

[Chrisfromthepast]

Will this work?
Im trying to majorly simplify a pink noise source that I "designed" a few years ago.
I had a rail splitter, which is silly, so I am deriving audio ground from a little resistor ladder. Please let me know if I have done that wrong it would not be the first time.
I also am simplifying the output, which used to use 2 op amps to drive both pins, in one of the Doug Self configurations. I am attempting to impedance balance the output and just drive pin 2. Did I do this correctly? Impedance balancing is quite a complex concept for so few parts.

 

[ruffrecords]

You do not need the connection between C7 and R2. And you need to decouple R7 to GND. I am not sure you can connect the junction of R3 and Q1 to the plus input of the op amp.

Cheers

Ian

 

[Chrisfromthepast]

Thanks Ian. I did not understand about C7. Some designs seem to leave that out entirely.
I redrew the 1/2 supply stuff to be a little more clear and eliminate the additional gnd nets that could confuse the conversation.
Is this looking more correct?

 

[Chrisfromthepast]

C1 seems needless to me, if the bjt, and both op amps are biased at 1/2 V+ anyway.
If the bias is the same can I eliminate the coupling capacitor between gain stages?
Should virtual ground be derived once, servo’d and then treated like a separate rail? Or should it be derived separately for every gain stage? I imagine the answer is “it depends”. But like, in this circuit does it matter?

 

[L´Andratté]

+24V/-24V is too much for the opamps, +18V/18V is maximum rating of 2134.

 

[ccaudle]

Have you tried that circuit yet? Does the BC550 break down reliably? Is gain of 11 enough to amplify the shot noise into the range you want?

if the bjt, and both op amps are biased at 1/2 V+ anyway.

That node will be at the transistor reverse breakdown voltage, not at 1/2V+ (assuming it breaks down reliably). I'm not sure how reliable the breakdown is between devices, they are not really designed to be used consistently in that mode like a zener diode is. Might work fine, it's just outside my experience so I can't say for sure.

If the bias is the same can I eliminate the coupling capacitor between gain stages?

Yes, if the bias is the same (see above).

Do you want a well defined low frequency roll off? If you do that is a convenient place to put it. If you just want the noise source to go as low in frequency as possible you could leave it out.

Also keep in mind that op-amps amplify DC, so without a capacitor in the feedback network the first stage will be attempting to apply a gain of 11 to the DC value at the input. Assuming that the reverse breakdown is around 6V, the output will be trying to drive to 66V. It definitely cannot do that, so the output will be severely clipped to +Ve. A capacitor between R3 (second schematic reference, R4 in first schematic) and the GND connection will keep the DC gain at 1.

I am attempting to impedance balance the output.... Did I do this correctly?

Yes.

 

[ruffrecords]

I dislike the tern virtual ground because it seems to imply is is creating something akin to the 0V line of a bipolar supply. What you actually have is a single supply and you need to bias the op amps to about half the supply voltage. The 0V of the supply is signal 0V. The bias is just a bias, not a virtual earth. So you should derive the bias voltage once and decouple it to 0V. The feed it wherever it is need via highish value series resistors such as the bottom left schematic attached.

Cheers

Ian

 

[tinn]

Maybe this version will work,

Cheers Tinn

 

[Chrisfromthepast]

Have you tried that circuit yet

I have not. Usually that is my approach, but I have been away from my bench, and want to try something that works the first time. (not my usual approach)

Does the BC550 break down reliably?

I just know that I have stock of these. Im happy to try something else. the noise source, the quality of the filter, the

gain

all of these details will certainly be considered and tweaked.

But, first thanks for helping me to fix the glaring mistakes.

Working backwards from the XLR, the impedance balance looks ok, I think I can find capacitors that aren't excessively large or expensive at 63 V. I don't think I need to zener regulate, thats kind of why I wanted to get rid of the rail splitter and go as simple as possible.

+24V/-24V is too much for the opamps, +18V/18V is maximum rating of 2134.

I think? once I drop voltage from the 6K8's and the 2K2's there won't be enough to fry the Op amp, but I have not had enough sleep to quantify this maybe. lets try. Iq per channel (typ) (mA) is 4ma, so a dual package will draw 8 ma. lets pretend this is all the circuit will draw, and calculate the voltage drop across (2) 6k8 supply resistors to be 27V, leaving only 20V with which to burn up our op amp. at least, I've never killed an op amp in one of the op amp mic projects. I'm wrong alot.

so without a capacitor in the feedback network the first stage will be attempting to apply a gain of 11 to the DC value at the input

In series with the 100K like this? I sized the output for 5hz I thought.
This feedback loop looks like 1.5Hz if I did it right.

 

[Chrisfromthepast]

Maybe this version will work,

Cheers Tinn

The capacitor you added next to R4, this is the proper place to block DC in the feedback path? I dunno why I wanted it inside the loop. Can you help me understand?

 

[ruffrecords]

Maybe this version will work,

Cheers Tinn

That should be OK but you still need to disconnect C8 from R13. The positive end of C8 goes only to pin 3 of the XLR; nowhere else.

Cheers

Ian

 

[ccaudle]

In series with the 100K like this?

No, that is not between R3 and gnd (I just checked, that is what I wrote in the previous post).


If you put the capacitor between the op-amp output and R4 you break the feedback loop at DC, so the output will saturate at either 0V or positive supply.
You want to keep the feedback at DC, but the voltage divider which reduces the feedback voltage to the inverting input (which is how you set gain) becomes open circuit at DC, which means at DC all of the output voltage is applied back to the inverting input, which results in unity gain.

 

[tinn]

it should be Unity Gain,
at the input you may have about 5 ..7 V and you want the same DC plus noise ant the output

 

[Chrisfromthepast]

To emulate a mic, unity is fine. I guess Im not going to need gain to get up to a decent condenser mic level if the 1/2 supply is at 6V, and the filter is going to knock off 3db/octave over the 12 octaves I care about. what is the total loss? Im lost on that calculation, its less than 6db Id guess, and depends on the weighting? Calibrated feature creep towards "mic level" of -26dbu, -10dbV and +4 dbu is inevitable...

 

[Chrisfromthepast]

Does this address all the concerns we have discussed so far?
We still need to talk about the noise source and the filter.

Thank you folks so much for helping me understand how to make this thing!

 

[Chrisfromthepast]

Maybe this version will work,

Cheers Tinn

Thanks for catching the inverting op amp thing. That makes more sense now.
Im down to make the Op amp's PSU bypass cap larger. For a source with very predictable crest factor, does it matter as much as usual?

 

[MisterCMRR]

What purpose does R16 serve? If you're trying to reduce DC offset due to bias currents of the op-amp, it should match R11 - but why care about the offset anyway. I'd replace R16 with a short or much lower value like 1 k-ohm. Incidentally, is the "X" at the bottom of R13 a feature of KiCad? It makes possible confusion about a connection there worse than just letting the lines cross without a dot.

 

[Chrisfromthepast]

Thanks for helping me.
For clarification, why not connect pin 3 to the v+ node through a 2k2 resistor? I thought this whole impedance balance thing required matched parts on each leg so that the real and complex impedances were exactly the same. Is 2k2 so large compared to 150r to make it negligible?
Sourcing the same current through both legs is important for transformer based mic preamps as well right?

 

[Chrisfromthepast]

What purpose does R16 serve?

Its just a value higher than the resistor divider per Ian’s advice. I have no idea how to choose R16’s resistor size. 1k to 1M is quite a large range of values. Almost no current flows through this resistor right? What is the practical difference between 1M and a short?

the "X" at the bottom of R13 a feature of KiCad? It makes possible confusion about a connection there worse

Yes, I can confirm that little X has made things worse in this discussion. I’ll redraw as soon as I can find a laptop.

In addition, if I were the creator of this schematic, I would certainly reannotate all of the schematic symbols' REF DES to be more linear and have a more "TOP DOWN / LEFT RIGHT" flow to them sequentially in order to make things easier to read. As an example, starting off with C9 and then the next capacitor is C1 just kind of makes you wonder, you know???

In the interest of trying to discuss many iterations of a similar design, I made attempts to keep components that did the same “thing” with the same values, even as I added or moved parts.
The “input” is on the left along with the original parts numbers. I think renumbering the parts for every schematic update would have made for an even more confusing thread.
Thanks for your patience.

 

[JohnRoberts]

I dislike the tern virtual ground because it seems to imply is is creating something akin to the 0V line of a bipolar supply. What you actually have is a single supply and you need to bias the op amps to about half the supply voltage. The 0V of the supply is signal 0V. The bias is just a bias, not a virtual earth. So you should derive the bias voltage once and decouple it to 0V. The feed it wherever it is need via highish value series resistors such as the bottom left schematic attached.

Cheers

Ian

In electronics, a virtual ground is a node of a circuit that is maintained at a steady reference potential, without being connected directly to the reference potential.

"Virtual earth" is common nomenclature for inverting summing amplifiers because that bus is actively held at 0V by negative feedback.

"virtual" implies active circuitry to provide low impedance reference voltage. V/2 supplies often get low impedance using large electrolytic capacitors between v/2 and real ground.

Sorry about a veer into semantics.

JR