THAT1512 / Transformer 'blind test' mic preamp

[dmp]

Another thing I don't understand is how the circuit is unbalanced in the SPICE simulation when the pot is set exactly half-way. My understanding was that it's there to adjust for tolerances in the resistors, but in the simulation, the resistors are all exact values.

I feel like I'm missing something really fundamental here.

The top opamp is inverting while the bottom opamp is non-inverting. The gain is different, Rf/Rin vs 1+Rf/Rin

 

[Newmarket]

The top opamp is inverting while the bottom opamp is non-inverting. The gain is different, Rf/Rin vs 1+Rf/Rin

The cross coupling (see R32, R33) takes care of that.

Below is as implemented in DDA AMR24

 

[amplexus]

I’ve used both the DRV and THAT chips with no output coupling caps with no issues. Just make sure you have the caps to the SENSE lines and you should be golden.

 

[dmp]

The cross coupling (see R32, R33) takes care of that.

Then why do the top and bottom have different pk-pk magnitudes?

 

[Newmarket]

Then why do the top and bottom have different pk-pk magnitudes?

idk. I might look to measure differentially with a floating "True RMS" DMM for confidence in the measured result.

 

[dmp]

It's in the spice simulation results

 

[Wavelength]

I have used both the DRV and the That parts. I would say hands down go That for the output. I direct coupled both but had issues with the DRV in protos so I talked the company I was designing for and the final design went with the That. I use the 1626 in a couple of 500 series modules I designed. The parts are great, another real nice sounding part is the That 1200 series for line inputs, stellar sounding. Used a set on an Audio Preamp board I am designing now.

 

[soulsonic]

I realise that I've made a mistake on the IC board and connected the input RF filter to signal ground. I can remove those capacitors from the board and wire it to chassis ground, directly on the input XLR instead. Is that correct?

Thanks,
James

Yes, I believe so. I made this same mistake on mine...

 

[jamesk]

The cross coupling (see R32, R33) takes care of that.

Below is as implemented in DDA AMR24

This is interesting. It's exactly the same circuit as in Self's book. I've just been having a look at the schematics for the desk and it's full of circuits from that book. I know he worked at Soundcraft but did he work at DDA when they made that, I wonder.

What's very interesting is that it says to calibrate the balance at 1kHz, which implies that the balance does vary with frequency, as I'm observing.

I have used both the DRV and the That parts. I would say hands down go That for the output. I direct coupled both but had issues with the DRV in protos so I talked the company I was designing for and the final design went with the That. I use the 1626 in a couple of 500 series modules I designed. The parts are great, another real nice sounding part is the That 1200 series for line inputs, stellar sounding. Used a set on an Audio Preamp board I am designing now.

Hmm, that's annoying to hear that about the DRV as it's much cheaper than the THAT. Maybe I'll end up going with the quasi-balanced circuit.

Yes, I believe so. I made this same mistake on mine...

Thanks, I'll do that!

I just want to check the overall grounding arrangement. I have the power supply in a separate steel enclosure. This will be connected to the preamp enclosure by a 5-pin XLR lead: +/-17V, +48V, signal ground and chassis ground. At the preamp end, signal and chassis ground will be connected to the two separate wires. These will then be joined at the power supply end to a star ground point on the enclosure, where the 0V from the power supply board also connects. Does this arrangement sound sensible or should I change something? I have also got separate ground connections on the PCBs for signal ground and the power ground from the decoupling electrolytics on the power input. Should this power ground be connected to the signal or chassis ground?

Thanks,
James

 

[DaveDC]

As Ian correctly pointed out, the voltage balance of the two output signal legs is entirely irrelevant when driving a balanced input, since that input is only sensitive to the *difference* between the two signals. There is absolutely no need to tweak them for balance. The output impedance of each leg should indeed be matched so as to not degrade the common-mode rejection ratio (CMRR) of the balanced receiver downstream. But this only matters for the traditional “diff amp” style of balanced receiver, the one with two input resistors and a single opamp. Any imbalance in the impedances of the driving signal legs (which are in series with these input resistors) equates to an imbalance of the diff amp’s resistor “bridge,” which in turn reduces its CMRR. This is why the “impedance balanced” output stage works so well: it does not degrade the CMRR of the following stage. However as Bill Whitlock has been telling us for years, an input stage with high common-mode impedance like a transformer or an “instrumentation” style diff amp (INA163 type, with two unity-gain buffers in front of a regular diff amp) is not degraded by mismatched source impedances.

The true purpose of these cross-coupled output stages is to emulate the action of a transformer when the balanced output signal is converted to unbalanced by grounding the negative leg. Doing this with a transformer doesn’t change the balanced signal level, but rather just causes the ungrounded leg to double in voltage. These cross-coupled opamp circuits do the same thing. One caveat is that power supply voltage and signal level choices need guarantee that there will enough output voltage swing in the opamp to not clip when the opposite leg is grounded.

There is a pretty big difference when designing balanced drivers for stand-alone gear compared to internal signals within a console. For stand-alone gear, you have no knowledge of how or to what the user will connect your outputs. So the driver has to handle all kinds of contingencies, like short circuit and over voltage protection, etc. Within a console, it is possible to tailor the driver to match the receiver, since both are under the designer’s control. For a console, probably the best bet is to ditch the cross-coupling and go with one inverting and one non-inverting amp, both fed from the same preamp.

BTW, the imbalance at high frequencies you are seeing is probably due to capacitive imbalance from either the 22pF cap tolerances or your board layout. Or your opamps are too slow and are running out of gain at high frequencies.

 

[jamesk]

As Ian correctly pointed out, the voltage balance of the two output signal legs is entirely irrelevant when driving a balanced input, since that input is only sensitive to the *difference* between the two signals. There is absolutely no need to tweak them for balance.

Thanks for your detailed response. I realise that the signal is differential but if the two signals are different amplitudes, the larger one will clip earlier, reducing the headroom compared to if they were matched. It's this that I'm concerned about. In a typical signal, though, the high frequency components will be much lower in amplitude anyway, I think I'm right ins saying, so maybe this isn't an issue.

The true purpose of these cross-coupled output stages is to emulate the action of a transformer when the balanced output signal is converted to unbalanced by grounding the negative leg. Doing this with a transformer doesn’t change the balanced signal level, but rather just causes the ungrounded leg to double in voltage. These cross-coupled opamp circuits do the same thing. One caveat is that power supply voltage and signal level choices need guarantee that there will enough output voltage swing in the opamp to not clip when the opposite leg is grounded.

Yes, this was why I went for this stage, and it is performing it's role of doubling the output voltage when connected to an unbalanced input correctly. Since the ultimate aim is to use it as an output stage on a desk, it will need to cope with balanced and unbalanced inputs, particularly on the inserts.

I think I've been a bit unclear about where the stage would feature. It's intended as an output stage on the desk to connect external equipment. The circuits and PCBs here are just to test the mic preamps, and this output stage, and aren't intended as part of a desk as they are. I haven't got a detailed plan yet but I think the mic preamp will just connect to the next stage single-ended. If there's any balanced signalling within the desk, I will follow your suggestion of using a straightforward balancing arrangement, given that connection to unbalanced equipment isn't an issue.

In terms of the frequency-dependent imbalance, I don't think it's down to the capacitor tolerances or layout as I'm getting the same behaviour in the SPICE simulation. I'll investigate your suggestion regarding op amp choice by trying a different model in LTSPICE. I'll see if a higher speed op amp displays the same behaviour.

Thanks again for your response,
James