[squarewave]

The gain is maximum at 10 ohms which will be 60dB or so. But if you vary the series resistance with a potentiometer like you're looking at doing, the gain can be adjusted. I'm not sure what's confusing about that.

The 3300u cap makes an RC with the 10 ohms for a roll-off at 5Hz. But at lower gains, the series resistance is higher and so the roll-off is lower accordingly. If you switch the capacitor to a smaller size, the roll-off frequency goes up. At 1000u, it's 16Hz. But the recommended cap is oversized because even at 16Hz, a mediocre electrolytic capacitor might translate to distortion in the output.

I used 100u for the smallest cap in my circuit and my gain is trimmed to 67dB which equates to a roll-off of 200Hz or so. But as Abbey and I were discussing, the gain and roll-off interact with each other. As the gain is increased, the series resistance gets smaller and so the roll-off frequency shifts up.

[PRR]

> gain control network with a fixed 10Ohm resistor... though their example shows a 10K pot

One way you have 10 Ohms. The other way you have 10,010 Ohms. Put that in the gain formula, you have 1,000:1 or 60dB of gain change. (Maybe a little less.) Which IS what you want!

3,300uFd at 10r is 5Hz, or -1dB @ 10Hz, which is awful flat.

And IMHO, rooms rumble and I never want MAX gain with 5Hz-10Hz response. While "interaction is bad", the typical use-case makes maximum usable gain related to excess bass response, inversely, so some bottom-cut CAN be a good thing. I could be happy with 50Hz, so a few hundred to 1,000uFd may be fine.

Yes, electros distort. But if you have minimum resistance you have HIGH gain and input (and cap) level is a few milliVolts. Distortion is teeny, while hiss is high. OTOH at low gain the R is large and small nonlinearity of C is swamped-out.

[CurtZHP]

But if you vary the series resistance with a potentiometer like you're looking at doing, the gain can be adjusted. I'm not sure what's confusing about that.

That part I actually understood, believe it or not!

The 3300u cap makes an RC with the 10 ohms for a roll-off at 5Hz. But at lower gains, the series resistance is higher and so the roll-off is lower accordingly.

I think my confusion stemmed from trying to figure out where the pot factored into this.

I used 100u for the smallest cap in my circuit and my gain is trimmed to 67dB which equates to a roll-off of 200Hz or so. But as Abbey and I were discussing, the gain and roll-off interact with each other. As the gain is increased, the series resistance gets smaller and so the roll-off frequency shifts up.

So, the rolloff doesn't get as high as 200Hz until you are near maximum gain, right?  Somewhere in the middle, it's sort of variable, but still low frequency.  How am I doing so far?

It sounds like I want to select between two capacitors here:  something in the 1000uF-3300uF range for flat, and something around 100uF-200uF for rolloff.  I guess I could just try a few and "season to taste."

[CurtZHP]

3,300uFd at 10r is 5Hz, or -1dB @ 10Hz, which is awful flat.

That's how my math was shaking out, but when I started plugging in numbers for a certain rolloff, the capacitance figures were coming out astronomical!  Either I suck even worse at algebra than I thought (I should just give these problems to my kids to solve!), or I missed something.

[squarewave]

So, the rolloff doesn't get as high as 200Hz until you are near maximum gain, right?  Somewhere in the middle, it's sort of variable, but still low frequency.  How am I doing so far?

Right. With 10 ohms you get 60dB of gain. If you use a 100u cap, that's a 160Hz corner. But if you adjust the pot to say 32 ohms, the gain goes down to 50dB and now the 32 ohms and 100u makes a 50 Hz corner. So 160Hz might sound a little high but keep in mind that as you turn down the gain, the low-cut corner frequency comes down too (and you might be surprised at how high the low cut can be before you start to really hear it).

Arguably the interaction of the gain and low cut corner it's a bit of a dilemma. My "solution" was to use an ON-ON-ON DPDT toggle [1] so that I can switch between three cap values (only two of which provide significant low-cut). That provides just enough adjustment so that after I find a reasonable gain setting, I can try the two positions that provide significant low-cut and adjust by ear.

[1] An ON-ON-ON DPDT toggle is an unusual part but very useful. In the extreme positions, it behaves as normal. But it has a center position where one common is connected to one pole and the other common is connected to the opposite pole.

[CurtZHP]

It's a very intriguing idea.  Thanks for suggesting it.  And thanks for your patience in helping me pick my way through it.

[CurtZHP]

I'm forced to take a break from this for the weekend.  (Probably a good thing...)
I left all my schematics and notes for it on my computer at work, and I'm not back there until Monday, but I think I have enough figured out to at least start breadboarding it.

[CurtZHP]

Another nagging question, that's probably the least of my worries at this stage...

I have a polarity switch right before the output XLR.  It simply flips the output pins on the connector.
I borrowed this from an earlier build that had an output transformer, and there was a 10K resistor across that switch.  (I think I had hijacked that idea from Jakob's G9...)  Is that resistor still necessary to prevent, or at least reduce, switch popping when switching polarity?  My schematic omits it.

[abbey road d enfer]

Another nagging question, that's probably the least of my worries at this stage...

I have a polarity switch right before the output XLR.  It simply flips the output pins on the connector.
I borrowed this from an earlier build that had an output transformer, and there was a 10K resistor across that switch.  (I think I had hijacked that idea from Jakob's G9...)  Is that resistor still necessary to prevent, or at least reduce, switch popping when switching polarity?  My schematic omits it.

THAT1646 has negligible DC offset so does not generate clicks, so resistors wouldn't change a thing. Note that the unit to which it is connected may have DC offset, which would generate clicks when switching.

[CurtZHP]

THAT1646 has negligible DC offset so does not generate clicks, so resistors wouldn't change a thing. Note that the unit to which it is connected may have DC offset, which would generate clicks when switching.

I'm not too concerned if it's a minor click.  I just don't want to be picking the remains of a tweeter out of the wall!

[john12ax7]

Pin 3 of U3 needs a DC connection.  You should put R9 after the switch so it is always connected.

[CurtZHP]

Pin 3 of U3 needs a DC connection.  You should put R9 after the switch so it is always connected.

I am so sorry, John!  That whole U3 thing is a mistake.  I posted the wrong version of the schematic.  That whole section has been removed.  I must have grabbed the wrong file.   

[CurtZHP]

So, after shelving this project for a time to tackle some other more pressing issues, I've finally had a chance to breadboard it.

We have this nifty breadboard at work with its own built-in 15V bipolar supply.  Got it mostly slopped together.  For now, I've omitted the phantom power, the polarity switch, and the protection diodes.  Just wanted to get as much of the basic signal path down and test the HPF.

Found out just how important the supply pin filter caps are!  It made a great white noise generator without them!

Right now, the only thing I'm still tackling is some hum that goes away when I grab the SM58 mic I have plugged into the input.  Obviously a grounding issue somewhere in the spaghetti.

I looked at the datasheets for the chips for any hints on specific component location requirements and didn't see any, although I've been told to keep supply pin filter caps as close to the pins as possible.

That, and the only reverse log 10K pot I have is scratchy as heck!

Hey, at this point, I'm just impressed it passes signal!

[squarewave]

Try separating grounds. Specifically, run a ground wire from the filter caps of the power supply to a new ground bus and try moving grounds that need to be quiet to it. Specifically, R7 and maybe the REF pin of the 1512 are good candidates.

However, it is also possible that your jumper wires are crap. I have some cheapo ones that do not make great connections. Sometimes I pull breakable header pins out and solder them to thick stranded wire like 22 AWG so that they make a really good ground connection. A little tarnish or broken wires inside can create resistance that will translate into noise. Breadboards are notorious for having problems like this. Sometimes just moving wires around or wiggling parts can cause noises to go away.

Regarding the HP filter, connect the 100u directly to the common of the switch so that you have either 100u or 100u+3300u so that the circuit is never open when being switched.

In practice, you probably won't get a really good noise floor with a breadboard. The THAT 1510 can be very very quiet but to get there you'll have to make a PCB and use shielded wires for low level / high impedance lines. Specifically, the lines from the XLR in need to be shielded and, because the pot and switch are probably relatively far away from the PCB, the lines from pins 1 and 8 of the 1510 should probably be shielded as well.

[CurtZHP]

However, it is also possible that your jumper wires are crap. I have some cheapo ones that do not make great connections. Sometimes I pull breakable header pins out and solder them to thick stranded wire like 22 AWG so that they make a really good ground connection. A little tarnish or broken wires inside can create resistance that will translate into noise. Breadboards are notorious for having problems like this. Sometimes just moving wires around or wiggling parts can cause noises to go away.

I did notice at least a change in the noise when wiggling a few things.  And I probably have power lines too close to input lines, etc.

Regarding the HP filter, connect the 100u directly to the common of the switch so that you have either 100u or 100u+3300u so that the circuit is never open when being switched.

I'm a little confused here.  The common of the switch currently goes to the IC.  Are you saying to put the 100u cap between the switch and the IC?

In practice, you probably won't get a really good noise floor with a breadboard. The THAT 1510 can be very very quiet but to get there you'll have to make a PCB and use shielded wires for low level / high impedance lines. Specifically, the lines from the XLR in need to be shielded and, because the pot and switch are probably relatively far away from the PCB, the lines from pins 1 and 8 of the 1510 should probably be shielded as well.

My input and output connectors are attached via shielded cable.  The pot and HPF switch, however, are not.  As for distance, I've got them about as close as I safely can, but I can certainly replace the wiring with shielded.  Also, a few of the resistors on the input side of the 1510 still have most of their lead wires intact, so there's a lot of extra bare wire just sticking in the breeze, acting like little antennae.

[squarewave]

I'm a little confused here.  The common of the switch currently goes to the IC.  Are you saying to put the 100u cap between the switch and the IC?

You want the 100u to always be in the circuit so that when you're switching the switch, it's never just an open circuit. In this particular case it actually wouldn't matter much but it's generally bad form to put a switch in the signal path and not have it bypassed in some way so that it never creates an open circuit.

So I mean disconnect the 100u neg side from the switch pole and connect to the common terminal that is connected to the IC. The terminal disconnected would just be unconnected. So the 100u is not switched at all - it's permanently in the circuit. When the switch is down, the coupling capacitance is 100u. When it's switched up, it's 3300u in parallel with 100u which is vaguely 3400u.

[CurtZHP]

You want the 100u to always be in the circuit so that when you're switching the switch, it's never just an open circuit. In this particular case it actually wouldn't matter much but it's generally bad form to put a switch in the signal path and not have it bypassed in some way so that it never creates an open circuit.

So I mean disconnect the 100u neg side from the switch pole and connect to the common terminal that is connected to the IC. The terminal disconnected would just be unconnected. So the 100u is not switched at all - it's permanently in the circuit. When the switch is down, the coupling capacitance is 100u. When it's switched up, it's 3300u in parallel with 100u which is vaguely 3400u.

Ah, like this?

[squarewave]

Yes.

[CurtZHP]

So, I completely disassembled the breadboard mess and started over, making things a little neater, and moving the IC's closer together.  Also replaced the pot and HPF switch wiring with shielded cable.

MUCH quieter!

Still have to find a non-scratchy pot.

I tried the HPF switch, and it switches cleanly (no pops or clicks...), but under current working conditions, I can't really hear a difference yet.  Might have to experiment with a few things on this.  I even tried it without the switch, just swapping the two caps in the circuit and listening.  Granted, I'm listening to myself talking into an SM58 through a little battery powered headphone amp and a pair of AKG headphones (which, in my experience, require more power to drive to a decent level).

But, the rest of the circuit seems to be fine for now.

Once I rig up a temporary phantom supply, I'll test that.

[squarewave]

I tried the HPF switch, and it switches cleanly (no pops or clicks...), but under current working conditions, I can't really hear a difference yet.

Remember that as you increase the gain, the low-cut frequency will shift up. That "problem" has been discussed repeatedly in this thread. The resistance of the pot + 10R and cap is the low cut filter. So if you set the gain to the max you have 10 ohms and 100u which is a corner of 160Hz. But if you back off the gain even a little bit so that the resistance is say 100 ohms, that frequency drops to 16Hz.

So this actually is a problem for your circuit because you're not going to get the benefit of the low cut until your gain as at maxium. In my circuit, I limited gain to a narrow stepped range and provided for 3 possible cap values instead of 2. This allows the low cut to be effective with most of the upper gain settings.

To correct the problem you could use a reverse log 1K potentiometer. With 1K, the minimum gain becomes 15dB instead of 0dB but who cares? When do you need +15dB of gain in a mic pre? You only need 10K if you also wanted to use it as a line input. But that would be quite pointless with this circuit (just a buffer?).

G9 bright tube glow on power up

Started by Smoke Preamplifiers

FET power amp suddenly bright and crackly

Started by Butterylicious The Lab

Bright tube heater glow when applying power?

Started by [email protected] The Lab

This forum is way too bright white but there's a cure

Started by Kingston Brewery