Mid-boost circuit loading down everything else?

[sircletus]

So I'm tinkering with filters and boost circuits at the moment and ran into something curious (well, curious to me).  I have a textbook, low-Q midboost with a fixed 6dB of gain at 1kHz (pic attached) build around two sections of a TL074; one for the inverting boost, the other an inverting, unity gain amp.  The circuit is switched into place with a relay.  Right before this circuit, there are a switchable second-order HPF and LPF filters, each using another section of TL074.  This is running on a single 30V rail, btw.

Both filter sections function perfectly, but something odd happens when I put the mid-boost in the chain: the sine wave I'm using for a test signal suddenly drops in level by about 50% and the positive crest is sliced off at about a 45º angle - this happens to the signal at the INPUT of the circuit.  I assume this is some kind of loading effect caused by the mid-boost circuit.

I ran a sim in LT Spice and everything looked kosher, thus my surprise.  Should I maybe move that inverting buffer from the output of the mid-boost and put it before?

 

[abbey road d enfer]

I assume this is some kind of loading effect caused by the mid-boost circuit.

The input impedance of this circuit is about 1.2k. The preceding stage using a section of TL074 does not have the capability to drive it correctly.

I ran a sim in LT Spice and everything looked kosher, thus my surprise. 

I suspect you made the sim with a zero-impedance generator... Try with a higher source impedance.

Should I maybe move that inverting buffer from the output of the mid-boost and put it before?

It wouldn't fix anything because the inverting stage would not be capable of driving properly the mid-boost circuit.
You must increase the resistor values and decrease accordingly the capacitor values. If you 10x the resistor values, the resulting impedance will be correct for the preceding stage.

 

[sircletus]

Ah, right.  TL07x wants to see, what, minimum 2k load?  Thanks so much for the response.

Oh, and thanks for the advice on setting a higher source impedance in my sims.

 

[PRR]

> setting a higher source impedance in my sims.

The default signal generator in most SPICEs can power all of Las Vegas on a busy night. 10,000 Amps? No problem, zero sag.

The real TL07x degrades THD below 2K load, and will not deliver over ~~30mA.

Why are your impedances so low? Make all caps 10 or 100 times smaller, all resistors 10 or 100 times larger. Let's see.... 120K, 250K, seems reasonable for TL0, then 0.033u (33,000pFd) becomes 330pFd. Maybe a bit small, but a factor of 33 would work. 39K etc, 1,000pFd etc.

 

[abbey road d enfer]

Oh, and thanks for the advice on setting a higher source impedance in my sims.

The circuit you're using is a derivative of the Baxandall EQ; its proper operation depends on a very low source impedance. So its normal to start the sim with the default "infinite-power" signal source, but it is also good practice to evaluate the load that represents the input impedance, either by graphing the input current, examining the level-drop when increasing the source Z, or by directly measuring the input impedance, by replacing the voltage source with a current source and measuring the voltage across it.
The basic issue is that in AC analysis, LTspice linearizes the small-signal parameters and doesn't care about such trivial limits as an opamp delivering 1000 V in a 1ohm load.
Running a Transient analysis would often unveil that kind of problems.

 

[sircletus]

The circuit you're using is a derivative of the Baxandall EQ; its proper operation depends on a very low source impedance. So its normal to start the sim with the default "infinite-power" signal source, but it is also good practice to evaluate the load that represents the input impedance, either by graphing the input current, examining the level-drop when increasing the source Z, or by directly measuring the input impedance, by replacing the voltage source with a current source and measuring the voltage across it.
The basic issue is that in AC analysis, LTspice linearizes the small-signal parameters and doesn't care about such trivial limits as an opamp delivering 1000 V in a 1ohm load.
Running a Transient analysis would often unveil that kind of problems.

Fantastic!  Much obliged.  Oh, and this was originally going to employ an LME49740