Discrete line input circuit for summing box

[bjoneson]

I have a summing amp built based on 990 DOAs, in a VE configuration with 10K feedback resistors. I've built some simple line inputs based on INA2134s directly feeding bus resistors, but am interested in building some beefier (perhaps with a little more character) discrete inputs.

The line receiver portion is based on an API 325 line amp card configured per application notes for this purpose. A fixed gain of 1 for this stage. My understanding is that matching of the 47K series resistors R53 / R54 is critical for common mode noise rejection.

The output of that stage feeds a 10K log pot configured like a fader. The next stage provides and available 14.5dB of gain on the fader / pot, based on an API812 card.

Finally, the gain stage feeds an LCR bus selector switch feeding bus resistors.

Other than a just a "sanity check", have a few questions:

I'm planning to place 4 of these inputs per PCB.  The API312 card had 12.5uF supply bypass caps. Is it ridiculous overkill to use this size bypass cap on every DOA?

Likewise on the protection diodes. 1 set per PCB, or a set per DOA?

The coupling cap configuration on the gain stage I got from the 812 schematic is interesting to me. Why 2 caps in parallel, as opposed to a single cap twice the value?

Any recommendations on parts or different configurations for the LCR assignment?

These inputs will be driven by DA converters with *plenty* of signal 90% of the time. I may occasionally insert EQ or compression units between the converters and the line inputs. The primary purpose of the 14.5dB of available gain here is to provide the option to "drive" the summing amps a bit. Obviously this is dependent on the summing amps having a desirable behavior when driven hard. Am I barking up the wrong tree here?

Many thanks for any input / suggestions.

 

[ForthMonkey]

I'm not master but why do you use two DOA?You can build easier and cheaper line in.

I'm used this as mic preamp but it's fine for line,too,i guess.Just remove components after C8.(I said that i'm not master.Just an idea.)

http://users.otenet.gr/~athsam/mic_line_balance.htm

 

[bjoneson]

I'm not master but why do you use two DOA?You can build easier and cheaper line in.
http://users.otenet.gr/~athsam/mic_line_balance.htm

Provided I want a gain of more than unity, and adjustable to boot, making a line receiver adjustable requires simultaneous switching of the value of several resistors which have to be very closely matched.

I've done the "cheaper / easier" option with INA2134s. Literally, just an IC with an input socket, and bus resistors. And it works great. I want to go the discrete path for the challenge and potential subjective sonic advantages of DOAs. I'm not trying to optimize for low part count, or cost here.

 

[ForthMonkey]

I want to go the discrete path for the challenge and potential subjective sonic advantages of DOAs.

I got it!You're right.

 

[joaquins]

For your input you also want the HF impedance matched, so using a cap in parallel with R51 matching C49 value. You could use precision resistors or a trim as R51 to get the best CMRR but this still isn't the best implementation to get good CMRR since a slight impedance unbalance in the source degrades it badly. Still good for most applications if you are not in a bad environment, if your sources are well balanced or the wires are short.

For bypass caps I would use higher values, 47µF or 100µF, one each few inches on the board at least, or one for each opamp won't hurt, you are not looking for low part count or budget and a few caps won't influence too much but may be a big difference in THD. Protection diodes are fine one per PCB since they are there just to avoid reverse polarity supply.

Depending on your channel count and the DOA you are using for your bus you should try to match the optimal input impedance of the opamp, about 600Ω for the 990 IIRC, if you have 8 channels per bus your resistors should be 600Ω*9=5400Ω so go with standard 5600Ω.

About the two caps in parallel I don't have a clue, maybe lower parasitic components or form factor, but just saying. C64 and C63 probably could be taken out of there since it doesn't help much, there shouldn't be any DC there other than the bias of the opamp which should be good enough at the pot.

How much is plenty of signal? Probably peak is what you should look there, so you don't run out of headroom at the input amp before the converter does, since the converter may be ±15V symmetric signal you may have 24Vpp so the input amp won't be able to handle it. Inside a equipment you don't need a signal too high since noise isn't usually a problem but headroom is, when you get at the end of the unit you rise the level using two opamps driving signal with inverse polarity or a transformer so you have a higher output level to minimize noise effects in transmissions but once you get inside the next stage you attenuate the signal so you don't run out of headroom. You really want to avoid clipping. About driving 990 summing amp I don't know, 990 is designed to be rally clean, if you are looking for more distortion in your summing amp you probably want to get a different DOA. 2520 is usually a good choice there, still loads of driving capability but also more distortion before clipping. Once clipping you may have a problem in any case. API used ±15V rails but 1:3 output transformer to get close to +30dBu output, which will clip any internal stage, so internal path uses a much lower level to still get the headroom 20+dB.

JS