Reducing gain in mic input stage to match line level

Hi.

I'm adjusting this Studiomaster mixer to my needs. I would like to adjust input stages, to avoid line signal being reduced to mic level, and then amplified back to line level. In this way I'm hoping to reduce some input noise.

Input stage looks like this:
http://84.255.203.119/mixer/input.jpg

Line input is fed through 27k resistors, which reduce level by some 30dB (acording to simulation I did). Then signal is amplified by some 30dB back to line level. I want to remove 27k resistors, and go with full line level into transistors (for their balanced input and gain adjustment pot). I will not use mic inputs.

What would be the best way to reduce gain of this transistor stage by some 30dB, bringing it down to unity gain?


thnx

There is no way to get that circuit to be happy at unity gain.

I would be tempted to suggest maybe reducing the amount of front end pad to line up with preamps minimum gain trim setting, but you don't want to drop line input impedance too low or you may have drive issues with external inputs. While you could also tweak the pad by raising values of R62 and R63 that could increase input noise which is expecting low impedance at that point. You could reduce the current density of input devices to better match higher impedance input (for noise) but I'm reluctant to just throw out new values because there may be other interactions. Hint: making r48 and r49 larger will reduce current density and also reduce minimum gain (both moving in right direction for tweaked pad approach). You don't want to make the current so low that IC2b inputs get too close to v- (check data sheet for input range).

To remain true balanced with that complement of components, the existing topology is pretty effective.

JR

Agree with JR---it's an inherently unsuitable topology for your purposes. Might as well start with a clean sheet of paper, and knowledge of what voltage swings and source impedances you want to deal with.

John.
I did some simulations, and reducing front end pad from 27k to 5,6k puts gain trim to zero, increasing input level some 12dB. Problem is, that this pad is there to provide headroom for EQ boost. Reducing pad to 12dB leaves no headroom for EQ.

R48/R49 don't give much. Increasing them too extreme 100k reduces gain only by 1dB. So I guess I better leave them alone as they are.

I can very much reduce gain with reducing 15k R45. Reducing it to 1K gives 20dB overal reduction in gain. So this is interesting. But is this good? Can it do some bad? After reducing it there seem to be appearing some ringing with square wave in simulation, so it must be getting unstable. Even with 15k there is some slight ringing, but with 1k it seems too bad. What do you think? Changing this R45, do I need to change something else with it?


Anyway, I guess with tweaking this circuit within some reasonable limits I may get me some 10dB less gain reduction/gain boost. Still it is far from unity gain.

Probably best would be to make small daughterboards with balanced NE5532 line in stage (like SSL), attach them to channel boards, and thus pass by transistors entirely, and inject signal into channel board at C19/C20 capacitor pair, just before eq.

But... is it worhed? Is this topology (30dB down/30dB up) really critical regarding noise? How much improvement regarding noise could I expect by going in at unity gain with balanced ne5532 stage, bypassing transistors alltogether?

Asking again won't change the answers.

5.6k is IMO low for a line input.

Increasing value of R48 and R49 will indeed reduce minimum gain somewhat, but absolute minimum stage gain is dominated by pot R and R50, R46. My reason for mentioning them was to dial in input device current density for better noise performance with increased values of R62 R63 (to tweak pad) and resultant higher input source impedance. It didn't hurt that they also reduced gain.

As I mentioned already there will be interactions when you start changing values...

JR

John:
I was hoping for a simple solution. like one or two resistor change. Like you say there will be interactions, and I'm not sure if I want to do something on 24 channels, without even being sure that it is better. I guess I will just go for bypass boards.

MediaT:
Yes, I will defeat mic preamps in mixer, I will never need them. As you say, I could build line input buffers, and inject them just before C20. And I could rewire existing gain pot and use it as level trim. I could make nice SSL-like balanced input, like on ssl4000 82E01. Or even in 82E149, which is a bit more complicated.
In any case, I will have to use existing gain pot, which is 10k C (antilog?). SSL are using 25k linear pot so I need to see this....

--
But, before jumping into making 24 PCBs.... Do you think I would gain much regarding noise, if I rebuild input stages? Now there is 30dB cut/boost through mic gain stage. How much noise would you say that such cut/boost introduces?

Now, as mixer is, if I turn on all channels, with input gains at minimum, there is some 3dB increase in output noise. And with all channels input gains set at normal operating level, there is another 3dB increase in noise. So it is 5-6dB increase compared to all channels turned off. I also plan to rebuild master bus amp, and then probably noise from inputs will stand out more.

You are on the right track to make measurements, but I would caution against just measuring the console wrt nil without considering the noise levels of the program sources you plan to mix. I don’t mean the noise floor of some Digital codec but real world noise sources like microphones and room ambient noise. With the exception of some purely computer synthesized sounds most program sources will have an inherent noise floor limited by thermal noise in microphones boosted by even a theoretical (perfect) preamp.

Just to do a little back of the envelope calculation a 150-200 ohm resistor amplified 50dB gets you up to the -80dB(?) range. Mixing together 4 channels of this at unity gets you up another 6 dB to the mid 70’s.

Assuming the console noise floor is also random and not distinctive sounding they will combine incoherently into a composite noise floor. If the console noise is 10 dB lower or more the final noise floor will be degraded less than 1/2 dB.

OK, these are not precise numbers and less is always better but don't kill yourself to perfect things that don't have a significant impact. I would be concerned about the phase shift and non-linearity in those paths that could impact signal quality in full measure. If using the stock mic preamp path degrades those parameters then dump it, or make it right

If the preamp stage is a source of error you could reconfigure the one opamp as a simple 4 resistor differential input. Not true balanced, but more than adequate for short line level interfaces.

JR

PS: "Good" mic preamps for decades have delivered noise figures better than 2dB (only 2 dB worse than theoretical perfect). High perforamnce designs are closer to 1 dB NF. The next performance consideration is the summing bus. While there is much focus on bus noise, the real issue IMO is phase shift and non-linearity. I look forward to the digital domain designers getting their act together to the point that this all becomes academic. On paper, digital combining looks pretty attractive. If present issues are simply inadequate hoirsepower, the technology will eventually catch up.

Thnx, John.
Noise fed to console from sources is small in my setup. Console is used only for mixdown, and 16 channels are fed from computer through 24bit converters, so not much noise is coming from there. Of course, there is noise in recorded tracks from mics/pres, and some old synths I have are quite noisy too. But many times I can clean noises in DAW a bit.

With 24bits recording, I wish for a lower noise. If I was recording 16bits, it would be just fine, some noise would even help as a kind of primitive dithering. But with 24bits it is tempting to lower noise some 10dB, or even more.

Now I decided to first remake summing bus. I feel that I will gain more regarding noise with better summing amp. Console is 24 channels, and busses are fed through panning resistors, which are equivalent to 16.66K resistor. Summing amp is done with half of NE5532, and has gain of 48. After fader there is another 3x gain. So together it is 43dB of gain, if I calculated properly, 33dB of which is at first stage before fader. That is a bit much for single opamp, is it? Maybe I can do something to improve there.

It won't be difficult to improve upon the ein of 553x series opamp with new improved models. I would look for decompensated opamp for improved phase shift/linearity in high noise gain application.

I don't follow your calculations so can't comment.

JR

PS: Without quantifying your sundry noise sources it's hard to say what your weakest link is, or if this will relieve that problem.

John.
Here are some numbers:

- (only post master fader) All channels down, master fader down gives -103dB noise (on digital recorder, clip level 0dB).
- (summing bus alone) All channels down, master fader at usual 0dB, gives -92dB noise on recorder.
- (full console, all on) All channels at usual working positions (driven to just before clipping), master fader at usual 0dB, gives -87db on recorder.

Here is schematics of my mixer:
http://84.255.203.119/smsm16-2schem.pdf

Additional noise comes in with putting channel faders up. From this I concluded that noise of channels is higher than summing bus noise.

Now if I redo summing amp, noise will still remain more or less the same when channels are up, is it? If sum bus would be the problem, and channels would be quiet, then opening channels would give almost no change. Or if noise from channels and sum bus would be of the same contribution, then there would be 3dB increase (uncorrelated noise). But I see 5db increase with channels up, so channels must be weaker. Am I thinking in proper direction?

thnx
gnd.

OK, that looks pretty straightforward.. I'll look at it tonight for possible recommendations (I'm working now).

On your noise measurements what is the bandwidth? Wideband or unweighted noise measurements can be deceptive as wider bandwidth circuitry can measure higher while actually being quieter.

Also when measuring nominal noise levels, were all inputs terminated (mic pre's with 150 ohms, lines not as critical but technically should be terminated too?

JR

Noise is measured unweighted, as it hits recorder. I know that measures higher, and it is not what we hear with our ears, but it is what recorder sees and records.

Otherwise noise spectrum of my console, all faders up on working levels, more or less follows flat line at -135dB on spectrum analyzer. There is some 50Hz hum sticking out some 10dB (peaking at -120dB), otherwise it is flat line.

I didn't terminate mic inputs, because they are not terminated with my normal usage. I never use mic inputs, just line inputs. And they are connected in parallel to each other, both on all the time. Line inputs were disconnected from other equipment, when making measurements.

gnd

I did some additional testing with A-weightning. I don't think my software is properly calibrated, but comparing unweighted and A-weighted results, there is almost 3dB lower noise value when measuring with A-weighting.

So A-weighted numbers would be:
- All channels down, master fader down = -106dBA
- All channels down, master fader at usual 0dB = -95dBA
- All channels at usual working positions, master fader at usual 0dB = -90dbA.
(with 0dB = digital clip on recorder)

gnd

So you have nothing connected to the inputs during the noise measurements?

What kind of DAW/DAC do you use ? If it can drive a ~2k load with negligible distortion, you may want to consider a DIY 20dB pad between the DAW line out and the console line in. The advantage would be that you don't need to change anything on the channel strips.

Can you test if it makes a difference in noise if you connect a 150-200Ohm resistor across pins 2 and 3 of the mic in XLR with the master fader at 0dB and that channel at usual working position ?

JDB.

[quote author="gnd"]I did some additional testing with A-weightning. I don't think my software is properly calibrated, but comparing unweighted and A-weighted results, there is almost 3dB lower noise value when measuring with A-weighting.

So A-weighted numbers would be:
- All channels down, master fader down = -106dBA
- All channels down, master fader at usual 0dB = -95dBA
- All channels at usual working positions, master fader at usual 0dB = -90dbA.
(with 0dB = digital clip on recorder)

gnd[/quote]

OK, now we're getting somewhere.. If that's all 24 channels increasing output noise 6 dB that suggests that working backwards one channel is some 6-10 dB below the noise of the summing amp.

The next question is in a typical mix will you have all 24 ch at 0dB or only a handful at 0dB with most down 6dB or so? Of course throw in one or two at a few dB hot and we're back up there.

It might be useful to measure 1 channel at 0dB, 2 channels, 4 channels, etc.

I suspect you need to first get the typical number of channels you will be using down below the bus noise. Then perhaps address the bus.

But, as I've said before, bus phase shift and THD is IMO also a concern.

Since the channels have a direct out you might want to make measurements there to test tweaks to the input stage for improvement.

Think of this as archeology where you're peeling back layers of noise.

Sorry, more tomorrow, I'm a little busy.

JR

[quote author="jdbakker"]What kind of DAW/DAC do you use ? If it can drive a ~2k load with negligible distortion, you may want to consider a DIY 20dB pad between the DAW line out and the console line in. The advantage would be that you don't need to change anything on the channel strips.
[/quote]

DAW DACs are emu1820m and Alesis AI3.
But I don't know about 20dB pad.... Isn't it that reducing level before console will require more gain in console, and will only increase noise. To reduce noise, I'd have to increase signal before console, to be able to lower channel gain.

gnd

[quote author="gnd"]But I don't know about 20dB pad.... Isn't it that reducing level before console will require more gain in console, and will only increase noise. To reduce noise, I'd have to increase signal before console, to be able to lower channel gain.[/quote]
My mistake. I meant to say "a DIY 20dB pad between the DAW line out and the console mic in". I suspect that R62/R63 dominate the noise contribution of pad+pre, so having a pad with lower impedance would lower the noise floor, too (assuming the EMU can drive it without distorting).

JDB.

Jonk, luckily you reminded me, that I have direct outputs. :grin: Now I tested single channel with/without 150ohm termination between pin 2 and 3 on xlr.

I had a bit problem with direct out cable, because it is unbalanced output, and it was noisy. Referencing ground to negative pin of balanced input in recorder didn't help either. In the end I remembered trick from forselltech.com summing bus, and i connected like this:
- tip (signal) from console direct out to (+) pin on recorder balanced in
- shield (ground) from console direct out to 1k resistor and then to (-) on recorder
- cable shield was disconnected at console, and connected at recorder side.

Thus I got quiet cable, measuring the same low noise as recorder input.

Here are noise figures with 150ohm termination on single mic input, and line-in from DAW plugged in (added is also reading at mixer output, with master fader at 0dB, all other 23 channel gains/faders up in usual positions and without 150R termination on them):
- minimum gain position, noise from single channel is -108dBA / -104dB unweighted, (master -92dBA / -90dB uw)
- working gain position (approx 30dB gain?) noise is at -108dbA / -105dB uw, (master -92dbA / -90dB uw)
- max gain position (60dB gain) noise is at -96dBA / -94dB uw , (master -85dBA / -83dB uw)

And without termination:
- min ch. gain -108dBA / -104dB uw (master -92dBA / -90dB uw)
- working gain position -107dbA / -105dB uw , (master -92dBA / -90dB uw)
- max gain -82dbA / -80dB uw (master -72dBA / -69dB uw)

Unplugging line-in from DAW (improves channel numbers by some 0.5 dB at working gain, and 5dB at max channel gain.

I guess I'd have to measure master gains with all mic inputs terminated, but there may not be much difference with terminated mic inputs. Measurements on single channel at working levels show only slight improvement (less than 1dB) with 150R termination.

As a permanent solution I'll have to do something with those mic XLR inputs anyway. Maybe terminate them by 150R resistor, or terminate them alltogether by cutting them out of channels. But this depends on modifications that will be made to inputs.

gnd

[quote author="gnd"]

DAW DACs are emu1820m and Alesis AI3.
But I don't know about 20dB pad.... Isn't it that reducing level before console will require more gain in console, and will only increase noise. To reduce noise, I'd have to increase signal before console, to be able to lower channel gain.

gnd[/quote]

All things equal sure, but the mic preamp is optimized for lowest noise with a low Z input and moderate gain. Hopefully the line input will also deliver good S/N but why assume when you can measure?

Perhaps a good experiment would be to measure the stock line input compared to DIY pad that terminates mic in with 200 ohm instead of 2k or so.

I still don't want to lose sight of the fact that true real world noise floor is not defined by 24 bit theoretical digital words, but it's good to understand your gear's limits and noise contributions.

JR

[quote author="jdbakker"]
My mistake. I meant to say "a DIY 20dB pad between the DAW line out and the console mic in". I suspect that R62/R63 dominate the noise contribution of pad+pre, so having a pad with lower impedance would lower the noise floor, too (assuming the EMU can drive it without distorting).
JDB.[/quote]

But there is already such pad done with line input. Those 27k resistors and 1k R62/R63 reduce line level to mic level, and that is then fed to mic preamp.

But I see what you mean. External pad would lower input impedance to mic pre.

Please see my noise measurements at channel direct out. May be that inputs are not so noisy after all. Of course, direct out level is some 10dB lower than master out, so all noise figures for channels must be reduced by some 10db.

gnd

gnd