Yet another "bright idea...."

[abbey road d enfer]

Actually the noise figure in the 1510 datasheet uses Rs = 200 ohms and those mic splitter transformers have 3 windings so if you wired two in series for a 2:1 step-down (using the "center tap" for phantom),  the lower Rs of the secondary might equate to lower noise with the 1510 compared to what you loose from the step-down.

I think you have your calculations wrong. The OSI for the THAT1510 is 500 ohms at 60dB gain, 700 ohms at 40dB. So using a 1:2 step-up is gonna give the best noise performance.
However the difference is moot; for a 20k BW the EIN at 200 ohms is 292nV and 282 at 800 ohms, supposing a perfectly lossless transformer.

 

[Bo Deadly]

Interesting. I didn't know what the OSI was. How do you know? What are you looking at in the datasheet? I didn't see voltage / current noise values.

 

[abbey road d enfer]

Interesting. I didn't know what the OSI was. How do you know? What are you looking at in the datasheet? I didn't see voltage / current noise values.

Noise current is indicated. Noise voltage is that at 60dB gain, where the contribution of feedback resistors is negligible. Everything is in the middle of page 3

 

[CurtZHP]

Made a few minor tweaks....

Dumped the pad switch.  Added a HPF between stages, but I'm not sure I did it right.  Lifted it from a previous build where the resistor was actually a level pot.

 

[Newmarket]

Pfff.

130dB CMRR at 50Hz, 3dB down at 5Hz and 80KHz, Faraday shield kills RFI stone dead, bullet proof against 48V phantom (no expensive preventative maintenance every few years to replace those leaky electrolytics at the input), galvanically isolated input (no hum loops). Work out the total cost of ownership for an electronic mic input over its life and I'll pay the 40 bucks up front for a transformer every time.

Okay - but I took PRR's point (and I'm happy for PRR to contradict if I'm wrong here) to be questioning the logic of using a real transformer together with a circuit designed to obviate the use of a transformer.
You can take some voltage gain at the transformer and design the subsequent circuitry to account for the increased impedance.
I'm avoiding discussion of transformer 'colour' here

 

[abbey road d enfer]

Made a few minor tweaks....

Dumped the pad switch.  Added a HPF between stages, but I'm not sure I did it right.  Lifted it from a previous build where the resistor was actually a level pot.

THAT1646 needs to see very low impedance AND galvanic connection at pin 4, which your circuit fails to provide. You need a buffer between the HPF and the 1646.

 

[CurtZHP]

THAT1646 needs to see very low impedance AND galvanic connection at pin 4, which your circuit fails to provide. You need a buffer between the HPF and the 1646.

So, even if I ditch the HPF, I should still ditch the coupling cap as well?

 

[abbey road d enfer]

So, even if I ditch the HPF, I should still ditch the coupling cap as well?

Sorry, I typed before turning the brains on. You need only very low source Z. Galvanic connection is established internally.

 

[CurtZHP]

Sorry, I typed before turning the brains on. You need only very low source Z. Galvanic connection is established internally.

So throw an op-amp buffer in between the HPF and the 1646 input, and Bob's your auntie's new boyfriend?

 

[CurtZHP]

Like so?

 

[abbey road d enfer]

Like so?

The 5532 also needs galvanic connection, so put the 100k at the non-inverting input, then the 1uF would be in parallels with the 0.1 for OFF (actually 1.6Hz @-3dB).
Note that with 0.1uF and 100k, the cut-off frequency will be 16Hz, which is fine for eliminating infrasonics, but not enough to clean up boominess of certain sources.

 

[CurtZHP]

The 5532 also needs galvanic connection, so put the 100k at the non-inverting input, then the 1uF would be in parallels with the 0.1 for OFF (actually 1.6Hz @-3dB).
Note that with 0.1uF and 100k, the cut-off frequency will be 16Hz, which is fine for eliminating infrasonics, but not enough to clean up boominess of certain sources.

It's 0.01uF, but I still think my value is wrong.  And I get the feeling 100K is too much as well.
By the way, when you say to put the 100K at the non-inverting input, do you mean move it entirely from its present position (so, 100K between NI input and ground)?  Or leave the end connected to the cap where it is and move the grounded end to the NI input instead?

It was my understanding, from an earlier build, that putting the 0.01uF cap in series with the 1uF would give me a total capacitance of around 0.01uF.  Then put the 100K resistor to ground downstream of that and you're supposed to get a rolloff in the neighborhood of 150Hz.

I really goofed this up!

Is this even the right place to put it?

 

[abbey road d enfer]

It's 0.01uF, but I still think my value is wrong.  And I get the feeling 100K is too much as well.

Nothing wrong there.

By the way, when you say to put the 100K at the non-inverting input, do you mean move it entirely from its present position (so, 100K between NI input and ground)?

Yes.

It was my understanding, from an earlier build, that putting the 0.01uF cap in series with the 1uF would give me a total capacitance of around 0.01uF.  Then put the 100K resistor to ground downstream of that and you're supposed to get a rolloff in the neighborhood of 150Hz.

That is correct. There's always several ways to skin a cat.
The thing is, in order to avoid possible clicks, there should be a high-value resistor across the switch (1-10 Megohms).

 

[CurtZHP]

Does that 1uF coupling cap even need to be there? 

 

[Bo Deadly]

So, even if I ditch the HPF, I should still ditch the coupling cap as well?

I would try to understand WHY it's better to do things one way or another as opposed to just asking people what method they think is best. The 1646 is a 5K inverting input. Two issues about this:

One issue is that your 10n capacitor is going to make a HPF of over 3KHz. Even your 1u is 32Hz which is actually a little high. Granted, there's not much gain in this part of the circuit so you probably wouldn't hear it but why not just use a larger cap? Even a 4.7u film cap is not terribly expensive. And, yes, you could just ditch the coupling cap entirely. You might need to sanity check the output offset but with the servo caps on the 1646 output, I don't think it would be a problem.

Two second issue is that the noise gain of an inverting stage is higher with higher impedance. So the HPF capacitor looks like higher impedance at low frequencies. So the cap either needs to be large (which means you don't get filtering) or you need to buffer the filter.

Personally, I would do it the way I just did in the schematic I posted a the top of this thread. If you use a high quality audio grade cap in the CFP gain control network, you can implement low cut very effectively. The only issue with doing this would be distortion from the large electrolytic capacitor but I was not able to measure any distortion using a Nichicon UES cap. And the high impedance at low frequencies does not increase noise. And because it limits the gain at low frequencies (as opposed to removing low frequencies later), you can get more signal into the circuit without clipping.

 

[CurtZHP]

I would try to understand WHY it's better to do things one way or another as opposed to just asking people what method they think is best. The 1646 is a 5K inverting input. Two issues about this:

One issue is that your 10n capacitor is going to make a HPF of over 3KHz. Even your 1u is 32Hz which is actually a little high. Granted, there's not much gain in this part of the circuit so you probably wouldn't hear it but why not just use a larger cap? Even a 4.7u film cap is not terribly expensive. And, yes, you could just ditch the coupling cap entirely. You might need to sanity check the output offset but with the servo caps on the 1646 output, I don't think it would be a problem.

Two second issue is that the noise gain of an inverting stage is higher with higher impedance. So the HPF capacitor looks like higher impedance at low frequencies. So the cap either needs to be large (which means you don't get filtering) or you need to buffer the filter.

Personally, I would do it the way I just did in the schematic I posted a the top of this thread. If you use a high quality audio grade cap in the CFP gain control network, you can implement low cut very effectively. The only issue with doing this would be distortion from the large electrolytic capacitor but I was not able to measure any distortion using a Nichicon UES cap. And the high impedance at low frequencies does not increase noise. And because it limits the gain at low frequencies (as opposed to removing low frequencies later), you can get more signal into the circuit without clipping.

Seems you've told me the "why." 
I just need to settle on how.

Going back over your earlier post, I don't know how I missed that suggestion about doing the rolloff before the preamp chip.  Giving it another look.  If that works, I can then eliminate the 5532 buffer, which means less active electronics in the signal path.  (I wasn't too keen on that anyway...)

 

[abbey road d enfer]

Personally, I would do it the way I just did in the schematic I posted a the top of this thread. If you use a high quality audio grade cap in the CFP gain control network, you can implement low cut very effectively. The only issue with doing this would be distortion from the large electrolytic capacitor but I was not able to measure any distortion using a Nichicon UES cap. And the high impedance at low frequencies does not increase noise. And because it limits the gain at low frequencies (as opposed to removing low frequencies later), you can get more signal into the circuit without clipping.

Unless I missed something, it seems that the cut-off frequency is gain-dependant; if the cut-off frequency is let's say 80Hz at 60dB gain, it will be 8Hz at 40dB and 0.8Hz at 20dB.
Nothing beats a good 2nd-order 2S&K.

 

[Bo Deadly]

Unless I missed something, it seems that the cut-off frequency is gain-dependant; if the cut-off frequency is let's say 80Hz at 60dB gain, it will be 8Hz at 40dB and 0.8Hz at 20dB.
Nothing beats a good 2nd-order 2S&K.

Yes, that's true. As you increase the gain, the low-cut frequency will shift up.

However, in practice is that really a problem? This circuit is just a high gain boost and a line driver. So why would you want 20dB? Why plug a line level signal into a line level output? Unless you're micing drums you probably need more then 50dB even for an LDC mic. So the practical gain range is relatively limited. My circuit maks between 45 and 67dB in 6 steps. I used an On-On-On toggle to switch between 3 caps one of which is large enough for all frequencies. So when I setup the pre, I turn off the low cut, adjust the gain and then add one or the other low cut steps to knock out any rumble. If there's some flexibility about the output level (my peak LED clips at +17dBu so for my pro-sumer gear there definitely is) I could fine tune the low cut frequency with the gain control.

Yeah, it's kinda strange. But considering the circuit is very simple and performs well accordingly, it's a pretty nice compromise IMO.

Also, if the principal factor in adjusting the gain is to get good level without clipping and low frequencies are reduced as you increase gain, these two interacting parameters actually complement one another. Meaning it's naturally resistant to clipping as you increase gain. So it's more forgiving about gain setting.

 

[abbey road d enfer]

Yes, that's true. As you increase the gain, the low-cut frequency will shift up. 

However, in practice is that really a problem? This circuit is just a high gain boost and a line driver. So why would you want 20dB?

OK, 20dB is a little extreme, but we have a different estimation of gain needs. I found out that

Unless you're micing drums you probably need more then 50dB even for an LDC mic.

  I am consistently using between 30 and 40dB of gain, as well with LDC's for vocals and piano, SDC's for rhythm guitar or dynamics on drums, guitar amps or Leslie.

So the practical gain range is relatively limited.

Indeed.

But considering the circuit is very simple and performs well accordingly, it's a pretty nice compromise IMO.

My main concern is I don't like interactive controls.

 

[CurtZHP]

As I understand it from looking over "More Analog Secrets Your Mother Never Told You" from THAT, their suggestion of putting a 3300uF cap in the gain control network with a fixed 10Ohm resistor doesn't see a 3dB drop until below 5-10Hz.

Even though their example shows a 10K pot in that circuit, their formula never mentions it, so that's a little confusing.

And even if it's a simple matter of adding another cap for higher rolloff, that cap's gonna be huge! 

I simply have to be completely misunderstanding this.