Portable mic preamp (ENG)

[abbey road d enfer]

I just don't know how to select the higher of two signals

Analog OR

 

[ricothetroll]

Analog OR

Thanx a lot ! I didn't know this one... Goes straight to the schem

 

[ricothetroll]

Hi !

So I found some time to draw the schematic and do the layout for version 3. I finally decided to put both channels on a single board, to keep the audio at one place and isolate it from the switching PSU.

Here are the images :
http://img854.imageshack.us/img854/7172/fpschemv3.png
http://img254.imageshack.us/img254/103/fplayoutv3.png

Speaking about the PSU, do you think it would be efficient to surround it with a faraday cage in aluminium or iron ? I'm currently discovering those switched mode PSU, so I don't have any experience with them.

By the way I bought the service manuals of a well known brand of hi quality portable mixers, and here is their PSU schematic :
http://img14.imageshack.us/img14/8629/psu1.gif
http://img834.imageshack.us/img834/8199/psu2.gif
The PSU is a flyback type one that outputs 12,6V, -12,6V and +50,5V, further regulated to +/- 12,00V and 46,8V, with feedback control through . Would it be possible to DIY such a PSU ? I haven't the values of the flyback coupled inductors, nor for the feedback transformer (1:1). The PSU is implemented with SMD components and (at least) double sided PCB, so it might be complicated to make a small implementation out of it with through hole components and single sided PCB.

Any comments are welcome ! I'll remove those schems in a few days.

Best regards.

 

[ricothetroll]

Hi !
After a few very busy weeks, I finally have some time again to work on this project.
An idea came to my mind : what about putting a vactrol across the hot and cold input, and use it as an input limiter ? using it as the reduction cell of a feedback design, it could make the preamp virtually un-clippable ! I guess that it will interact with the mic's output impedance, it's not-necessarly-log response wil be compensated by the feedback topology as for fet compressors like 1176, and its noise will be balanced out.
I realise it's quite utopic, but I'd be glad to know at least why it wouldn't work
Best regards.
Eric

 

[abbey road d enfer]

Hi !
After a few very busy weeks, I finally have some time again to work on this project.
An idea came to my mind : what about putting a vactrol across the hot and cold input, and use it as an input limiter ? using it as the reduction cell of a feedback design, it could make the preamp virtually un-clippable ! I guess that it will interact with the mic's output impedance, it's not-necessarly-log response wil be compensated by the feedback topology as for fet compressors like 1176, and its noise will be balanced out.
I realise it's quite utopic, but I'd be glad to know at least why it wouldn't work
Best regards.
Eric

It would, provided the vactrol has the necessary low "on" resistance (or using several in parallels). The results would be source-dependant, and there would probably be some frequency response issues under GR. In fact, if the input used a xfmr, a "standard" vactrol could be used.
This is not very different than the basic principle of an 1176, after all. The only difference is that in an 1176, the impedance persented to the "vactrol" is pretty well controlled.
The fact that it would not be so well controlled when the source is a mic is what led designers to conclude that it's not right for a proper commercial application. But I think there have been some broadcast products using this technique.
I believe you could take an 1176, replace the input xfmr with a mic input xfmr, bypass the input attenuator and get rolling.

 

[ricothetroll]

Well you're right for the unpredictable effet with different mics. Maybe I should try to put a low Ron vactrol (VTL5C3) in series with the THAT1512's gain resistor. I'd be losing 10mA per channel but I might be able to source that current from a low voltage source like directly from the batteries... I'll try to draw a schem implementing this idea tomorrow !
Best regards.
Eric

 

[topcat]

A limiter pre  or as part of the gain stage is rather appealing...the Shure SM82 (US 4000370) Line Mic was a fantastic piece of kit and provided excellent results even with the most exuberant sports reporters.
I did try a vactrol/optofet in series with the gain resistor with a 2017 some years ago but never managed to get the required Ron for the normal state.
Any one seen the Aphex 207 mic pre schematic?  the data notes that it's transformerless. Does it use their 6266423 patent?
Most of the mic front end limiters i've encountered in broadcast equipment are based on shunting the feedback resistance of the gain stage with a Vactrol....
tc

 

[abbey road d enfer]

Well you're right for the unpredictable effet with different mics. Maybe I should try to put a low Ron vactrol (VTL5C3) in series with the THAT1512's gain resistor.

That would not be right, because the attack time would be the time the vactrol takes to recover from full light, which is loooong. You want the vactrol to be a shunt element, either on the source or the NFB path.

 

[ricothetroll]

That would not be right, because the attack time would be the time the vactrol takes to recover from full light, which is loooong. You want the vactrol to be a shunt element, either on the source or the NFB path.

I didn't see that one coming !  :-[ That's what makes my theory fall apart...

A limiter pre  or as part of the gain stage is rather appealing...the Shure SM82 (US 4000370) Line Mic was a fantastic piece of kit and provided excellent results even with the most exuberant sports reporters.
I did try a vactrol/optofet in series with the gain resistor with a 2017 some years ago but never managed to get the required Ron for the normal state.
Any one seen the Aphex 207 mic pre schematic?  the data notes that it's transformerless. Does it use their 6266423 patent?
Most of the mic front end limiters i've encountered in broadcast equipment are based on shunting the feedback resistance of the gain stage with a Vactrol....
tc

Thanx for those very interesting links ! The Aphex approach looks OK, but as Abbey Road said, it could lead to quite bad frequency distortion and even clipping of the output circuit of the mic as it's forced to feed a much too low impedance than what it's designed for...

What would be easy would be to implement an LDR as a shunt across R6 in Samuel's "shared gain topology" :
http://home.datacomm.ch/gronerfamily/DIY/Shared_Gain_PreAmp/Shared_Gain_PreAmp_r1.gif
And use it with a LL1528 (that would give me a "free" 14dB gain when configured in 1:5).
Could this circuit, though more expensive, implemented with high end opamps, compete with Graeme Cohen's topology's specifications ?

Samuel ? 

Thanx a lot for your help !

Best regards.

Eric

 

[abbey road d enfer]

What would be easy would be to implement an LDR as a shunt across R6 in Samuel's "shared gain topology" :

It would certainly work, but would not prevent the first stage to distort when overloaded. Putting the vactrol in parallels with C2 would not have this limitation but the limiter's behaviour would much depend on the actual setting of the gain pot, in particular at the lowest gain settings.

Could this circuit, though more expensive, implemented with high end opamps, compete with Graeme Cohen's topology's specifications ?

If you wanted to include this in a Cohen topology, you would need two perfectly matched potentiometers. Any mismatch would ruin the CMRR. That's what THAT have done to a certain extent with the 5171, taking advantage of laser trimming for controlling the resistors..

 

[bigugly]

What would be easy would be to implement an LDR as a shunt across R6 in Samuel's "shared gain topology" :
http://home.datacomm.ch/gronerfamily/DIY/Shared_Gain_PreAmp/Shared_Gain_PreAmp_r1.gif
And use it with a LL1528 (that would give me a "free" 14dB gain when configured in 1:5).
Could this circuit, though more expensive, implemented with high end opamps, compete with Graeme Cohen's topology's specifications ?

Samuel ? 

Thanx a lot for your help !

Best regards.

Eric

Some food for thought. You can eliminate IC7 and the following 180R resistor. Connect the 750r directly to the inverting input of IC2. You could also eliminate the output transistor pair if you're not driving long cables and heavy loads.

 

[ricothetroll]

Thanx for your food

I did some further testing, and I came to the conclusion that two solutions seem to be suitable for the input limiter :

- Using Samuel's "shared gain topology" (trafo + non inverting opamp + inverting opamp) with a vactrol to perform a voltage controlled global negative feedback. The drawback of this is that it's quite expensive and seems to be noisier than the THAT1512, at least with NE5532 opamps (I have some OPA227 on the way...)

- Using the THAT1512 with Aphex's method (shunt load impedance, pat. US6266423) with a step up input transformer (LL1528) to decrease the vactrol's impedance seen by the mic, and taking advantage of the trafo's DCR to create the attenuation, compensated by the trafo's gain.

I did a comparison of the THAT1512 circuit with and without the trafo, to see if the DCR adds any significant noise. Here are the results, normalised for compensating the gain difference :

https://dl.dropboxusercontent.com/u/28610725/THAT1512_with_LL1528.wav (gain of the preamp = 74,5 dB, 47,4 dB added in software, R1=R2=24k)
https://dl.dropboxusercontent.com/u/28610725/THAT1512_without_LL1528.wav (gain of the preamp = 59,8 dB, +62 dB added in software, R1=R2=1k)
R1 and R2 are the THAT1512 input bias resistors, as seen on figure 3, page 5 of the datasheet
http://www.thatcorp.com/datashts/THAT_1510-1512_Datasheet.pdf
The input is terminated with a 220R 0,6W metal film resistor. LL1528 is configured 1:5 and has a 900R DCR on the secondary (in series for 1:5)

We can hear that the hiss (white noise) is pretty equivalent, but there is an added "vinyl crackling" with the trafo in the circuit. Though I'm not used to recognising to those noises, the "crackling" doesn't look like thermal noise to me. Could that be burst noise ? Or is it more likely some parasitics, magnified by the 14dB extra gain of the LL1528 ? The circuit lies unshielded on a breadboard. By the way there is a slight hum on the test without the trafo, most certainly due to the lack of shielding and the long test wires.

Thanx again you all for sharing your knowledge !

Best regards.

Eric

 

[abbey road d enfer]

I did a comparison of the THAT1512 circuit with and without the trafo, to see if the DCR adds any significant noise. Here are the results, normalised for compensating the gain difference :

The input is terminated with a 220R 0,6W metal film resistor. LL1528 is configured 1:5 and has a 900R DCR on the secondary (in series for 1:5)

We can hear that the hiss (white noise) is pretty equivalent, but there is an added "vinyl crackling" with the trafo in the circuit. Though I'm not used to recognising to those noises, the "crackling" doesn't look like thermal noise to me. Could that be burst noise ? Or is it more likely some parasitics, magnified by the 14dB extra gain of the LL1528 ? The circuit lies unshielded on a breadboard. By the way there is a slight hum on the test without the trafo, most certainly due to the lack of shielding and the long test wires.

The 1512 is optimized for low impedance sources '200-800 ohms). With a 1/5 xfmr, the impedance is about 5kohms, so the noise factor is too high. Increasing the impedance in order to make the vactrol's job easier is detrimental to the noise figure. If you want to use a higher Z, you must use an input stage that is optimized for 5kohms; fortunately, any good old 5534 or one of the new über opamp will be perfectly happy with that.

 

[ricothetroll]

The 1512 is optimized for low impedance sources '200-800 ohms). With a 1/5 xfmr, the impedance is about 5kohms, so the noise factor is too high. Increasing the impedance in order to make the vactrol's job easier is detrimental to the noise figure. If you want to use a higher Z, you must use an input stage that is optimized for 5kohms; fortunately, any good old 5534 or one of the new über opamp will be perfectly happy with that.

Well I was aware of that but yet still wanted to try it, I though maybe the trafo's voltage gain would compensate for the added source resistance Nevertheless do you think those crackling noises are due to the In x Rs ? I'll give it a try without the trafo and with a 7k source resistance - account taken for the DCR of primary and secondary, (200+42) x 25 + 900.

I did some calculations that confirm that the trafo in front of the 1512 isn't a good idea, based on TI's noise calculator sheet :

- THAT1512 without trafo : EIN = 1.162 nV/Hz^0.5
- THAT1512 with trafo : EIN = 3.514 nV/Hz^0.5 => + 4,22 dB compared to 1512 without trafo :s
- Shared gain with NE5532 : EIN = 2.65 nV/Hz^0.5 => + 1.77 dB / 1512
- Shared gain with OPA2277 : EIN = 2.40 nV/Hz^0.5 => + 0.91 dB / 1512

I must do some other tests with the shared gain and 5532, the first time I found like a 5dB difference instead of the 1.77 dB calculated !

To be continued...

Best regards.

Eric

PS : shared experience : I ordered some OPA2277 instead of OPA2227... Damned those numbers look visually similar ! The specs does not

 

[ricothetroll]

Hi!

I just did a final comparison between THAT1512 and shared gain with OPA227 and NE5532 (LL1528 input trafo), input loaded with 220R. The conclusions are quite the same as calculated : THAT1512 is the most quiet, shared gain with OPA227 is a small dB noisier and NE5532 about 0.5dB noisier than OPA227. Those differences are subtle though. So for my application I'll go with the shared gain + OPA227, that will allow me to install a vactrol and perform a variable global negative feedback, to act as the limiting element.

The shared gain will be followed by a THAT2181 that will provide the channel VCA (fader) and an eventual +30dB extra gain. To make sure that this stage won't be saturated, the side chain will be switched between the output of the preamp if the VCA gain is < 0dB, and the output of the VCA if its gain is > 0dB. This shouldn't be too hard to implement with a comparator on the gain control pin voltage of the VCA, and jfet switches.

Schematic coming...

Best regards.

Eric

 

[abbey road d enfer]

I did some calculations that confirm that the trafo in front of the 1512 isn't a good idea, based on TI's noise calculator sheet :

- THAT1512 without trafo : EIN = 1.162 nV/Hz^0.5
- THAT1512 with trafo : EIN = 3.514 nV/Hz^0.5 => + 4,22 dB compared to 1512 without trafo :s
- Shared gain with NE5532 : EIN = 2.65 nV/Hz^0.5 => + 1.77 dB / 1512
- Shared gain with OPA2277 : EIN = 2.40 nV/Hz^0.5 => + 0.91 dB / 1512

There must be something wrong with the calc.
The 5532's noise voltage is specified at 5nV/sqrtHz. Provided with a source impedance of 7 kohms, the EIN should be around 7nV/sqrtHz.
Reflected to the input of a 1:5 xfmr, that computes at 1.4. The 5534 has a better (guranteed noise performance at 3.5, which would put its noise factor very close to that of the 1512 (provided the right xfmr).
I don't understand the 2.65 figure

 

[ricothetroll]

There must be something wrong with the calc.
The 5532's noise voltage is specified at 5nV/sqrtHz. Provided with a source impedance of 7 kohms, the EIN should be around 7nV/sqrtHz.
Reflected to the input of a 1:5 xfmr, that computes at 1.4. The 5534 has a better (guranteed noise performance at 3.5, which would put its noise factor very close to that of the 1512 (provided the right xfmr).
I don't understand the 2.65 figure

Well, from TI's noise calculator, I obtain :

  - 624,6 nV/sqrtHz output noise of the first stage (non inverting, Rs = 6k95, R2/R1 = 22k/470, G=47,81)

  - 77,3 nV/sqrtHz for the second stage (inverting, R2/R1 = 22k/2k4, Rs = 0, G = -9,16)

  - Total gain = 5 x 47,81 x 9,16 = 2189,7 = 66,8 dB

  - Total noise = 624,6n x 9,16 + 77,3n = 5,80 uV/sqrtHz

  - RTI = 5,80u / 2189,7 = 2,65 nV/sqrtHz

I put the calculator here :
https://dl.dropboxusercontent.com/u/28610725/Noise%20Calculator%203_05.xls
...I just replaced "Opamp V noise" and "Opamp I noise" by the 5532's values, an the values of the resistors and bandwidth.

Please correct lme if my calculations are wrong, it's the first time I do it !

Best regards.

Eric

 

[abbey road d enfer]

There must be something wrong with the calc.
The 5532's noise voltage is specified at 5nV/sqrtHz. Provided with a source impedance of 7 kohms, the EIN should be around 7nV/sqrtHz.
Reflected to the input of a 1:5 xfmr, that computes at 1.4. The 5534 has a better (guranteed noise performance at 3.5, which would put its noise factor very close to that of the 1512 (provided the right xfmr).
I don't understand the 2.65 figure

Well, from TI's noise calculator, I obtain :

  - 624,6 nV/sqrtHz output noise of the first stage (non inverting, Rs = 6k95, R2/R1 = 22k/470, G=47,81)

Well, I couldn't use the calc (I'm using a Noptebook) but using the 5532's noise specs, I get 5nV/sqrtHz for noise voltage and 0.7pA/sqrtHz multiplied by 7k that gives another 5nV/sqrtHz; combined quadratically, that computes to 7nV/sqrtHz. THe other resistances are negligible.
Taking into account the 1:5 gain provided by the xfmr, I get 1.4nV/sqrtHz.
Put differently, I get about 350nV/sqrtHz at the output of the 1st stage. I don't know how the calc finds nearly twice...

- Total noise = 624,6n x 9,16 + 77,3n = 5,80 uV/sqrtHz

I'm not sure; are you using arithmetic summing or quadratic? (not that is changes much since the 2nd stage noise is swamped by the first).

https://dl.dropboxusercontent.com/u/28610725/Noise%20Calculator%203_05.xls
...I just replaced "Opamp V noise" and "Opamp I noise" by the 5532's values, an the values of the resistors and bandwidth.

I couldn't find how to do it on a notebook perched on my knees...

 

[ricothetroll]

Well, I couldn't use the calc (I'm using a Noptebook) but using the 5532's noise specs, I get 5nV/sqrtHz for noise voltage and 0.7pA/sqrtHz multiplied by 7k that gives another 5nV/sqrtHz; combined quadratically, that computes to 7nV/sqrtHz. THe other resistances are negligible.
Taking into account the 1:5 gain provided by the xfmr, I get 1.4nV/sqrtHz.
Put differently, I get about 350nV/sqrtHz at the output of the 1st stage. I don't know how the calc finds nearly twice...
Quote

You need to add the thermal noise of the source resistor. The calculation is nicely summed up on fig. 47 (p16) on the OPA211 datasheet :
http://www.ti.com/lit/ds/symlink/opa211.pdf
e1 and e2 could be neglected because the gain doesn't apply on it and e1 because its value is small.

I'm not sure; are you using arithmetic summing or quadratic? (not that is changes much since the 2nd stage noise is swamped by the first).

My bad, I forgot to use quadratic summing ! The result is 5,7nV/sqrtHz.

 

[bruce0]

WRT Vactrol based limiter.  Here is a schematic for a rather old battery powered in line mic preamp (no phantom) using such a limiter.

Anyone know what that output transformer is - couldn't find the number anywhere?