An ideal line input stage

.

No one can tell you what you'll think is best. For CMRR nothing beats a transformer. For line level signals you usually don't need 100dB of CMRR. For unbalanced line level next to a transmitter maybe you want 100dB of CMRR. Maybe you'd rather have a little noise if you can only use a cheap transformer with a low saturation point.

Start building.

google instrumentation amp... you can make one from op amps and precision resistors, while dedicated instrumentation amps are not crazy expensive these days.

A good transformer is probably more expensive than a good instrumentation amp.

JR

Why no recommendations for THAT 1206? My understanding was it was as close as you can get to a transformer in an IC package.

Is $6 USD "expensive"?

I would either use an opamp like ssl does (lots of choices here) or use a pre-made dedicated line receiver like ina134, that 1200 series,  etc.

For CM performance the THAT 1200 1203 1206 are probably the  best performers.

But you also mentioned you need 1/4" and XLR simultaneously? In that case you might need a separate receiver for each.

Definitely use a THAT line receiver. You will need a summing stage if your going to use both inputs. Since your not using hi impedance inputs like instruments, you don't need Jfets or the TL072. You will have better noise performance with NE5532's.

The summing amp feedback resistor is 1K. Non inverting input to ground. You can add gain above unity by making the 1K feedback resistor bigger.

If you don't need separate gain for the unbalanced line in you can just use a simple voltage follower configuration.

[edit] you may want to add an on/off switch (not drawn) for each input that disconnects the output of the line driver or 5532 from its 2k resistor and grounds the open end of the 2K resistor.  That way no noise will be injected into the circuit when an input is not connected. Make sure the switch leaves the output of the line receiver or 5532 floating and not grounded with the 2K resistor.

Something like this:

80hinhiding said:

Thanks for the suggestions though, and taking time to sketch something out.

Click to expand...

No problem, and yea experiment, that's the best way to learn.

80hinhiding said:

edit: As far as I know, having a high impedance on an input, like a tube, isn't only reserved for instruments.  I'm not being smart here, I'm asking the experts more than telling, of course.

Click to expand...

Try this on a breadboard. Make eight stages of a TL072, that's four chips. Just make them unity gain non inverting voltage followers. One running into the other. Then put a big cap on the end of the chain like a 470uF.  put a 1/4" jack on the end.
Put a 47 ohm resistor to ground from the first non inverting input. So your just slugging the input to the whole chain.
Take the output and plug it into something you can monitor it with headphones. Turn the headphones up, listen to the noise/hiss. Then pop out all the TL072's and replace them with NE5532's Now listen to the hiss. Which op amp chain had more hiss?

I'm quite partial to the transistor gain stage  in Quad Eight  gear like the 312 or half of the 712. That's a good one. Unbalanced though.

If the common mode performance of your balanced line input stage is good, you can simply plug in an unbalanced signal to the balanced input and skip the switching and summing stuff. With some balanced line inputs, the resistive imbalance that happens when you stuff a single ended signal into one side of the balanced input and ground the other side can screw up the CMRR, but this depends on how the line input circuit is designed - not all circuits will do this, and it might not matter that much either.

It's at least worth a try using a 1/4" to XLR adapter or an adapter cable.

Many designs use a combined XLR/TRS socket for this kind of input. With a regular differential input stage you can just plug in an unbalanced jack, a balanced TRS or an XLR.

Cheers

Ian

Monte McGuire said:

If the common mode performance of your balanced line input stage is good, you can simply plug in an unbalanced signal to the balanced input and skip the switching and summing stuff. With some balanced line inputs, the resistive imbalance that happens when you stuff a single ended signal into one side of the balanced input and ground the other side can screw up the CMRR, but this depends on how the line input circuit is designed - not all circuits will do this, and it might not matter that much either.

It's at least worth a try using a 1/4" to XLR adapter or an adapter cable.

Click to expand...

Yes - you'll get some benefit anyway wrt CMRR from using a balanced output. The actual difference between what you get cf a balanced source being down to the impedance imbalance of unbalanced signal relative to the CM input impedance of the input.

Maybe you could add an unbalanced insert I/O and use return to connect unbalanced gear if needed.

Having a specific unbalanced input stage has a meaning only if it offers particular advantages. A THAT balanced receiver accomodates balanced and unbalanced signals beautifully as long as the source impedance does not exceed 5-10kohms. This is in view of the input impedance of the circuit. That makes them suitable for almost any source, except passive instruments, that need at least 500k. That's what DI's are for. Indeed, they would not be suitable for microphones, because their noise performance is not optimized for low-Z low-level sources either.
Debalancing a THAT 120x is only a matter of using the right connector.

Maybe have a look at input stage of  various discrete opamps, example with jfet: Forsell's Class A JFET
http://www.forsselltech.com/articles/9/#!/articles/9

> bare bones differential ....few discrete

It is basically a couple of crude op amps. In+ has a 100K feedback resistor and 100K on the op amp-ve input. In- is the same but also mixing in the output of the top op amp because the base of the bottom transistor is a virtual earth.

Cheers

Ian

Not to spoil the party, but I think it's important to separate the implementation of the amplifiers from the function of the overall circuit. One could re-draw a lot of differential stages with many different amplifier circuits, but the thing that makes the differential stage do it's job is basically gain and feedback arranged in a specific way relative to the input and output signals. How you get that gain is less important (to the differential function) than where you apply the gain and feedback to achieve the differential function.

Stated another way, try to logically separate the circuit function of the differential amplifier circuit from the technology of the amplifiers used within that circuit. Given that clarity, you could use tubes, bipolar transistors, op amps, pulse width modulators, or lord knows what - an amp is an amp, and to make a differential circuit, you need gain and negative feedback. How you get there is a matter of taste, and if you separate the form from the function, you can have a wider palette to work with.

80hinhiding said:

Edit:
Here's a line receiver made from a NE5532 and external components.  This circuit is described as having 40-50dB CMRR.

https://m.eet.com/media/1177907/audio_line_receiver_fig3.jpg

That dB figure seems like it'd be noisey, but is it acceptable or even pleasing contribution in sound?  A mystery to me

Click to expand...

That's the 'SuperBal' configuration. It exhibits equal differential impedances on each input leg as well as equal Common Mode impedance. Whether that matters is a matter of opinion / application.

The 40-50dB CMRR figure ? - it's no different to a conventional stage (lose the inverting amp and the 10K from the non inv in goes to 0V instead of the output of the inverting stage).
The accuracy comes from the tolerance of the resistors shown and the value and tolerance of the source impedances feeding it.
40 - 50dB probably representative of using 1% parts. Improve by using tighter tolerance eg 0.1% and / or trimming.
See Douglas Self - Small Signal...p509 . Note lowering of resistors around the inverting amp stage.

Newmarket said:

The 40-50dB CMRR figure ? - it's no different to a conventional stage (lose the inverting amp and the 10K from the non inv inout to 0V).
The accuracy comes from the tolerance of the resistors shown and the value and tolerance of the source impedances feeding it.
40 - 50dB probably representative of using 1% parts. Improve by using tighter tolerance eg 0.1% and / or trimming.
See Douglas Self - Small Signal...p509 . Note lowering of resistors around the inverting amp stage.

Click to expand...

Agreed. These days, you can get pretty cheap 0.1% and better 2012 SMD resistors from Susumu that seem to sound good and are very stable. But, as you rightly pointed out, the ultimate CMRR also depends upon any source impedance imbalances. Having let's say a 0.05% 10K resistor gives you an error of ±5Ω. But, that same 5Ω error at the source will equally ruin the stage balance, since the 0.05% 10K resistor will be in series with the possibly rangy source. And, if you stuff an unbalanced source into a balanced diff amp, the output impedance will not be the same for the unbalanced signal and the ground leg, making the net source impedance imbalance much worse. So, in practice, you can only go so far with super precision resistors, unless you have a polite, well balanced source.

So, for polite, well balanced sources, most standard diff amps work well, and quality resistors will increase the CMRR to very high values. They don't cost that much today, so  why not use them. If you're making one-offs at home, a simple measure and match procedure can select closely matched parts pretty quickly using a quality DMM and not that big of a pile of resistors. However, with unbalanced single ended or poorly balanced sources, a design such as the Ingenius input stage will provide much better CMRR, but with its own set of tradeoffs.

80hinhiding said:

Is 40dB of CMRR on a line input receiver acceptably low in realistic application, or should I be aiming for much lower?  I was surprised to see a TL071 circuit here: http://sound.whsites.net/project51.htm that was described as having 40dB CMRR unless trimmed with a variable resistor.  Going from 40dB to 80dB.. he also used an NE5532.  I see SSL uses two NE5534 chips on line input.

Click to expand...

The CMRR figure is just the rejection ratio, not the actual noise level. CMRR is nice to have to clean up any small problems, but the goal is not to expect it to extract a signal buried in common mode hum and interference. So, for a mixer, you have to still use nice cabling and good system grounding to eliminate big hum and ground problems.

Good RFI filtering on the console inputs is also important, since CMRR quickly evaporates at high frequencies, and if RF hits an active stage, it can get demodulated and turned into a legitimate signal that CMRR will not alter. Cell phones, WiFi, and all the other RF junk in any metropolitan area will induce RF signals into your cables, even if they're high quality and well dressed. In my experience, simple RC lowpass stages added right next to the input jack can knock this down without messing up the audio passband response. In general, if you aim for a lowpass frequency of around 100-200kHz, you don't have to take even tiny HF rolloff if you don't want it.  SPICE simulations can be a big help in tuning these lowpass networks, and many capacitor vendors offer complex models of their caps, to make the simulations pretty good even at RF.

Another cute trick is to choose a small capacitor whose self resonant frequencies are lined up with the interference you want to knock out. For example, the Murata GQM1875C2E101JB12 is a small 100pF SMD cap that is designed for cell phone transmitters; they designed the electrodes to have self resonant frequencies at ~850MHz and ~1.8GHz, to minimize losses at the principal transmitting frequencies. Those nulls also help you at the input jack, since they will knock out these frequencies much better than just the bulk 100pF capacitance of a standard 0603 size cap. To get these benefits, you need to use a multilayer PCB with proper, RF quality ground and power planes, but this cap is not all that pricey, and the layout only costs you design time and a 2+ layer board, which is really the bare minimum anyway. Use that as the first lowpass cap, right next to the input jack pins, add possibly an additional RC lowpass stage or two to knock back FM and AM broadcast, and you'll solve this problem easily.

So, if you address RFI and hum / grounding problems, then you don't rely on CMRR for much anyway. Still, it's nice to  have CMRR in case something goes wrong. I can't tell you how much you need, but more is always better.