Portable mic preamp (ENG)

Hi,

I'd like to build a portable, battery-powered mic preamp. It's intended to be used for ENG/film recording, so my requirements are :
- Low noise
- 90dB gain
- Built in limiter
- Stereo volume control
- Powered by NP1 regulated 12V battery

To sum things up, something similar to Sound Devices' Mix Pre, without the digital options.

I have done a first schem and layout, mostly based on THAT components to minimise the room (THAT1512 for the preamp, THAT2252/2181 for the limiter and THAT1646 for the balanced output) and to take advantage of their design notes' pre-calculated circuits.

Here are the schematic and layout :
https://dl.dropboxusercontent.com/u/28610725/FP_Schem.png
https://dl.dropboxusercontent.com/u/28610725/FP_Layout.png

Basically, I took the preamp circuit of the THAT1512, added an switchable RC HPF flat / 80Hz / 160 Hz, added 30dB gain with a non-invering opamp (IC3A), added a comp/limiter/volume control copied from THAT CORP"s design note dn116, and then added a THAT1646 for balanced output. I created the 48V from 12V using an LT1054 (44V actually), and would like to use a DC-DC converter like the MURATA - MEV1D1212SC for the +/- 12V supply rails. I left this part for the moment as I'd like to test its ability to supply the +/- 12V quietly enough. I replaced it with some perf board pads for doing some trial and error tests.

Any comment/critic is welcome ! I haven't built the circuit yet, waiting for some other opinions to correct my eventual stupid mistakes (I do a lot)

Of course, I'll share the "design" if I end up with a satisfying version !

Here are the datasheets of the components used :
http://www.thatcorp.com/datashts/THAT_1510-1512_Datasheet.pdf
http://www.thatcorp.com/datashts/THAT_2252_Datasheet.pdf
http://www.thatcorp.com/datashts/THAT_2181-Series_Datasheet.pdf
http://www.thatcorp.com/datashts/THAT_1606-1646_Datasheet.pdf
http://www.thatcorp.com/datashts/dn116.pdf
http://www.murata-ps.com/data/power/ncl/kdc_mev.pdf
http://cds.linear.com/docs/en/datasheet/1054lfg.pdf

Thanx in advance for your help !

Best regards.

Eric

I have little to add but that I've built a battery-operated mic pre using THAT's 1512 application note and it works well. Do you REALLY find the need for 90 dB of gain compelling? That's only needed to drive a pro console into clipping with a -56 dBm source signal - I find that a hard scenario to call "typical."

And 90 dB is a lot of gain, which will mean great care in your layout is required to keep your preamp quiet and stable.

I'll look at your schematic later and comment if I have anything useful. But your requirements seem like they might be a little extreme to me at first glance.

Completely biased here... But have you considered an Expat Audio Eden?

Hi, thanx for your answers !

Do you REALLY find the need for 90 dB of gain compelling? That's only needed to drive a pro console into clipping with a -56 dBm source signal - I find that a hard scenario to call "typical."

And 90 dB is a lot of gain, which will mean great care in your layout is required to keep your preamp quiet and stable.

Click to expand...

Well, it is a huge lot of gain, but it can be useful with ribbon mics that have a very low sensitivity like Beyer M160. Anyway the 30dB added bt the 5532 is switchable, the 30k feedback resistor of the 1k/30k non inverting network can be shorted to make it a follower. I've made some tests on a breadboard with the 1512+5532 and it worked quite well, no oscillations and noise still dominated by the AKG C414 I used for the test. 

Completely biased here... But have you considered an Expat Audio Eden?

Click to expand...

Looks like a fine project indeed ! I can't see any schematic on your website, but it looks like it's straight off INA163's datasheet fig. 5 + drv135. Kinda the TI version of my THAT "design" (I put some "" because it's more of a patchwork than of a genuine design...)
Thing is I have quite a lot of those THAT parts and would like to use them ! But thanx for the suggestion 

Best regards.

Eric

ricothetroll said:

Well, it is a huge lot of gain, but it can be useful with ribbon mics that have a very low sensitivity like Beyer M160. Anyway the 30dB added bt the 5532 is switchable, the 30k feedback resistor of the 1k/30k non inverting network can be shorted to make it a follower. I've made some tests on a breadboard with the 1512+5532 and it worked quite well, no oscillations and noise still dominated by the AKG C414 I used for the test. 

Click to expand...

Well, that sounds like a good result anyway - and BTW - I subtracted wrong - it's a -66dBm signal that'd clip the console with 90dB  of gain!

All I can say is more power to you - the more I see of these THAT chips the more they impress me.

this looks really nice! and if you make pcb's then i would be interested in couple of them
btw. when getting the 30dB out of the opamp does it lower it's bandwidth( ie. Bandwidth in unity gain vs 30dB amplified..)?

A brief look at the schematic suggests that basically this should be a well working implementation. As others I do very much question the need for +90 dB gain. Going from +60 dB to +90 dB does not provide any useful SNR improvement, all you do is shift the levels around. You could easily do this in post production. As this is a portable application, I see some improvement in quiescent current consumption. These thoughts, along with a few others, listed below:

* I'd skip the second gain stage and instead arrange the volume pot such that the THAT2181 may provide another +10 dB gain. Overall +70 dB should be really enough, and saves one opamp.
* C7 and C29-C31 should not be left floating when switched out, or they may accumulate charge and cause a click when switched in. Use a 100k (C7)/1M (C29-C31) to ground.
* Note that a 1st order HPF is only mildly effective to supress low-frequency rumble. For your application I'd consider at least a 2nd, perhaps even a 3rd order filter.
* The opamps in the side chain could probably all be TL071/TL072. Less money and less quiescent current.
* I doubt that the THAT1646 driver is necessary--as far as I understand your application has short cable runs and is battery powered, so no ground loops. An impedance balanced output from IC3B should do very well, and again saves money, power and space.
* R3 could/should be 10x (perhaps even 100x) larger. This will mitigate the effect of C2/C3 mismatch on low frequency CMRR.

Samuel

With +/-12VDC supply, I'd decrease R8 to 3K9 for VCAs bias current. The VCA bypass setting probably will connect to 0V reference voltage, not left floating.

Hi,

Thanx for all your answers ! I'm already thinking of a second version

I'm also giving a second though to the limiter use : what's the point of having a limiter if the preamp circuit clips at about the same level as the recorder that follows ? I think I'll try to get a +/- 20V supply for the THAT1512 and the THAT2181, and run the side chain a a lower level with a lower PSU.

To be continued...

Best regards.

Eric

ricothetroll said:

I'm also giving a second though to the limiter use: what's the point of having a limiter if the preamp circuit clips at about the same level as the recorder that follows? I think I'll try to get a +/- 20V supply for the THAT1512 and the THAT2181, and run the side chain at a lower level with a lower PSU.

Click to expand...

That'd be uncomfortably close to the maximum ratings of the THAT1512, and improves the headroom by nothing more than a few dBs. What's more effective is to arrange for some gain within or after the VCA (e.g. by adjusting the level pot for +10 dB as suggested above).

Samuel

You could save some space (and money) using a 4301 instead of the 2252/2181 combo.

That'd be uncomfortably close to the maximum ratings of the THAT1512, and improves the headroom by nothing more than a few dBs. What's more effective is to arrange for some gain within or after the VCA (e.g. by adjusting the level pot for +10 dB as suggested above).

Click to expand...

You're right, and it adds lots of complications to get a +/- 20V supply... I'll stick with +/- 15V !

Looking at the THAT2181's datasheet, figure 4 page 3, it seems like the noise isn't proportional to the gain of the VCA : -98dBV at 0dB gain, and -76dBV at 30dB gain, so while the gain increases of 30dB, the noise only increases of 22dB ! So, that could mean there's no point using the 2181 as an additional gain stage, isn't it ? (I have the feeling that I'm missing something here). I just hope this noise figure isn't to the input, that would mean the noise contribution of the 2181 becomes HUGE past 0dB....
http://www.thatcorp.com/datashts/THAT_2181-Series_Datasheet.pdf

You could save some space (and money) using a 4301 instead of the 2252/2181 combo.

Click to expand...

I'm actually drawing another schematic without the RMS detector : after performing some simulations it appears that it's nearly impossible to get fast enough attack (<500us) combined with not-too-fast release (100ms) times. I'll stick then to the good old feedback topology, something like this :
https://dl.dropboxusercontent.com/u/28610725/2181%20comp_schem.pdf (optimal threshold not defined yet)
https://dl.dropboxusercontent.com/u/28610725/2181%20comp_sim.pdf
The ratio seems high enough and the attack/release can be easily trimmed to the best sounding values.

Best regards.

Eric

ricothetroll said:

Looking at the THAT2181's datasheet, figure 4 page 3, it seems like the noise isn't proportional to the gain of the VCA : -98dBV at 0dB gain, and -76dBV at 30dB gain, so while the gain increases of 30dB, the noise only increases of 22dB !

Click to expand...

That is a consequence of the current sharing of the gain cell. Current increases in the log converter while it decreases in the antilog and vice-versa.

So, that could mean there's no point using the 2181 as an additional gain stage, isn't it ? (I have the feeling that I'm missing something here).

Click to expand...

Probably; the suggestion is to take advantage of the VCA's capability to deliver gain as well as attenuation, not adding anything.

I just realised I forgot the NOT when writing that sentence :
"So, that could mean there's no point NOT using the 2181 as an additional gain stage, isn't it ?"

ricothetroll said:

Looking at the THAT2181's datasheet, figure 4 page 3, it seems like the noise isn't proportional to the gain of the VCA: -98 dBV at 0 dB gain, and -76 dBV at 30 dB gain, so while the gain increases of 30 dB, the noise only increases of 22 dB! So, that could mean there's no point not using the 2181 as an additional gain stage, isn't it ? (I have the feeling that I'm missing something here). I just hope this noise figure isn't to the input, that would mean the noise contribution of the 2181 becomes HUGE past 0 dB...

Click to expand...

That plot is very likely RTO, not RTI/EIN. As for most gain stages, EIN is best at highest gain. Possibly the distortion increases a bit at high gain (not clear from the datasheet).

Samuel

Hi !
So, here's version two : I replaced the RMS/feedforward limiter with a peak/feedback one, more effective on peak reduction, cheaper and that allows me to limit the signal after the gain of the VCA. I've also replaced the THAT1646 output with a 5532, and the first order filter by a 3rd order Sallen-Key one. I've also added a limiter LED indicator.
https://dl.dropboxusercontent.com/u/28610725/FP_Schem_v2.png
Comments are welcome, of course ! I'll do the PCB routing soon, maybe after a few remarks to correct the remaining mistakes
Thank you all again for your help !
Best regards.
Eric

Hi Eric,
Great project!
I did a similar project some years ago and found that using a multiple time constant in the limiter was really useful as some users didn't seem to adjust the gain at each location if ever. A simple stacked  time constant such as that described in http://downloads.bbc.co.uk/rd/pubs/reports/1967-13.pdf  or the Aphex US5483600 Wave dependent compressor patent, or Studer 1.914.539 limited mic amp...or  the 'deluxe' approach adopted in the Studer 1.914.519 or 1.915.700 (from the studer ftp site) or ntp etc.... I used the simple approach and it really paid a dividend in better quality audio bought back by journo's.  Tried more complex approaches such as setting the base gain of the VCA using the average level of a 10S sample....but didn't improve the quality of the actuality significantly.
Might be worth moving D2 to the other side of C1 too...it's always easier to spot minor stuff in other peoples work than glaring errors in my own work!

Cheers
tc

ricothetroll said:

So, here's version two : I replaced the RMS/feedforward limiter with a peak/feedback one, more effective on peak reduction, cheaper and that allows me to limit the signal after the gain of the VCA. I've also replaced the THAT1646 output with a 5532, and the first order filter by a 3rd order Sallen-Key one. I've also added a limiter LED indicator.
https://dl.dropboxusercontent.com/u/28610725/FP_Schem_v2.png

Click to expand...

Looks like a good step forward! A few notes:

* C37 doesn't need to be in the signal path if the HPF is switched in. Mostly a cosmetical detail of course, but I'd move it to the HPF bypass branche.
* I think it would be good practice to add a 100k each from C15 and C37 to ground, such that their DC level is well defined (noted in a previous post already).
* Alternatively to the above two points you can merge C15-C37 into one cap, which is in series with R10.
* I think C16 is unnecessary, and in fact if you want to sense "true" peak level you can't AC couple.
* C17 is a very heavy load for a TL07x--attack time will be limited by the opamps output current. You can/should scale the impedance levels up (say 100x to 1k, 47 nF, 1M).

And a thought for the layout: C17, R22, R29 and T1 inject a lot of distorted current into ground. Make sure these nodes are carefully routed, and don't share common impedance with the audio reference.

Samuel

A compensation for the soft knee D9 forward voltage drop by injecting an offset at IC2b-pin6 might be a nice to have.

Hi,

Might be worth moving D2 to the other side of C1 too...it's always easier to spot minor stuff in other peoples work than glaring errors in my own work!

Click to expand...

Ooooh yes right ! didn't see that one ! Thanx also for the links, some very interesting stuff there ! I actually implemented the auto release ladder a la GSSL, I'll see if I keep it or not while performing tests.

* C37 doesn't need to be in the signal path if the HPF is switched in. Mostly a cosmetical detail of course, but I'd move it to the HPF bypass branche.
* I think it would be good practice to add a 100k each from C15 and C37 to ground, such that their DC level is well defined (noted in a previous post already).
* Alternatively to the above two points you can merge C15-C37 into one cap, which is in series with R10.
* I think C16 is unnecessary, and in fact if you want to sense "true" peak level you can't AC couple.
* C17 is a very heavy load for a TL07x--attack time will be limited by the opamps output current. You can/should scale the impedance levels up (say 100x to 1k, 47 nF, 1M).

And a thought for the layout: C17, R22, R29 and T1 inject a lot of distorted current into ground. Make sure these nodes are carefully routed, and don't share common impedance with the audio reference.

Click to expand...

Thanx Samuel, I'll implement those mods will in the third version !

A compensation for the soft knee D9 forward voltage drop by injecting an offset at IC2b-pin6 might be a nice to have.

Click to expand...

It's done by the threshold trimmer, that comes through R19 (maybe I didn't understand your remark correctly)

I may have one last question (for the moment ) : for the stereo linking, the best way would be to do it as in the GSSL Turbo version, that is to say each channel is processed separately and we apply the higher of the two SC signals on both VCA. My problem is : I can't find any schematic of that particular mod to get inspired (due to API patent violation IIRC) and I just don't know how to select the higher of two signals (excepted with a comparator and two fet switches but I'm sure there's a better way )

I'm already routing V3... 

Best regards.

Eric