G2 distorted audio using TRS and standard mic input

[w9ac]

I got permission from Abhi to publish the schematic of the Mic preamp board for the G2.

Laurence,

Thanks for posting the schematic. That's a good solution to stabilize the op-amp's Vcc/2 bus. It's strange not to see a ~22pF compensation cap between NE5534 pins 5 and 8 to prevent oscillations. It's not a unity-gain stage so it's probably forgivable.

It would be best to replace FL1 and FL2 with a bifilar-wound SMD common mode choke. Mount it physically close to the NE5534 input.

High mutual bifilar coupling between each L on a common core result in 2x the series inductance in common mode and complete inductance cancellation in differential mode to preserve input waveform integrity. Preserving waveform integrity is certainly much less of a concern in bandwidth-limited systems, but why not use the right choke and get superior RFI abatement?

I'm currently using the device linked below in my homebrew front-end audio designs. They're commonly used in datacom equipment where it's imperative to preserve a sharp-pulse digital waveform without fear of inductive overshoot and ringing.

Like all RF chokes, a peaking frequency occurs depending on ferrite composition and turns. Choking Z for the Bourns 475 type is 10K-ohm at 160m, diminishing to 2K at 10m. Optionally, adding a 513 type in series offers superior rejection through 6m. Price is $0.50 USD in quantities.

https://www.mouser.com/datasheet/2/54/dr331-777616.pdf

Paul, W9AC

[w-u-2-o]

I got permission from Abhi to publish the schematic of the Mic preamp board for the G2.

Thank you Laurence!

I also found that R8 was 10K on my board; that makes the gain lower than intended. I've increased it to 39K, which then means I have adequate gain from dynamic microphones without using "mic boost" in the codec chip.

- The schematic shows R8 as 15K which would result in a gain of 10 (20 dBv). That should be enough for any competent dynamic microphone and far too much for any condenser microphone. Since your board does not match the schematic perhaps some of the folks who are having problems might have improperly built boards. It also makes one wonder what sort of final testing these units go through.

- Increasing gain to nearly 30 dBv seems like far too much for even dynamic microphones unless the microphone in question is particularly inefficient.

- It's interesting that a phantom power connection is supported at J1 on Pin 4. People could therefore easily add a 48V phantom supply if they wanted to.

[echo_india]

Hi all,

Laurence, Scott and Paul thanks for looking into this further. R8 on my Mic Preamp board is 10k. I notice R15 which should be 2k2 ohms is bridged on my board - Top left of image. Thanks.

[w9ac]

R15 is a power switch LED current limiting resistor and isn't part of the audio input circuitry. If it's jumped out on the PC board it may have been moved elsewhere. A 2K2 dropping resistor is an unusually high value from a 3.3V source, even if the LED is deliberately "dimmed."

R8 changed to 10K means a slight reduction of mic amplification from 20 dB as shown on the schematic to 16.5 dB.

Paul, W9AC

[w-u-2-o]

This kind of variability in production units is crazy. Laurence's board has R8 per the schematic, you can see it in his photo. This is real quality problem.

[laurencebarker]

I had R8=10K. We don't know whether this is a quality problem, or a deliberate decision to change the resistance to 10K not reflected back onto the schematics. Similarly with the other component: there may well have been some experimentation in getting that power switch right (which I wasn't involved in).

[echo_india]

I am going to try and put a 22pf across pins 5 & 8 of the op amp without changing R8 value for now, and see if this improves things.

[w9ac]

I am going to try and put a 22pf across pins 5 & 8 of the op amp without changing R8 value for now, and see if this improves things.

It can't hurt and can only help, especially if the PC board layout is non-optimal. Low gain values create excessive phase shift beyond the specified phase margin of the op-amp. The 5534 is internally compensated for gains above A=3, 9.5 dB but becomes unstable as gain approaches unity. The gain value chosen by Apache shouldn't normally require additional compensation unless they tamper with gain during production as we've seen.

From the original Signetics datasheet:

"The 5534 is internally compensated for gain equal to, or higher than, three. The frequency response can be optimized with an external compensation capacitor for various applications (unity gain amplifier, capacitive load, slew rate, low overshoot, etc.)"

Let us know if there's any improvement.

Paul, W9AC

[echo_india]

From the schematic it looks like I'm missing a R3 and R6 on my board, both are 10k resistors which should be placed across the mic hot and cold pins and pin 4 from J1 header.

If you can see from the image there is no place holder for these resistors. I wonder if others that have issue are missing these resistors also.

I did try with using a 22pf across pins 5 and 8 of the op-amp and it did not improve anything on the audio quality on this balanced input.

I swapped out R8 10k ohm for 20k first and the audio gain did come up quite a bit although it did not improve on the distortion.

[w9ac]

From the schematic it looks like I'm missing a R3 and R6 on my board, both are 10k resistors which should be placed across the mic hot and cold pins and pin 4 from J1 header.

Ignore R3 and R6. They're +48V phantom power injection resistors. Apache must have decided to eliminate them at some point during production. No great loss.

Paul, W9AC

[echo_india]

Thanks Paul. Other variations is see from the schematic and my board is, on the schematic R4 & R5 are 1k5 resistors, and on my board they are both 2k2 resistors which are the audio inputs on the op-amp pins 2 & 3.

[w9ac]

Thanks Paul. Other variations is see from the schematic and my board is, on the schematic R4 & R5 are 1k5 resistors, and on my board they are both 2k2 resistors which are the audio inputs on the op-amp pins 2 & 3.

R4/R5 together with R8 set stage gain. Increasing the R4/R5 pair from 1K5 to 2K2 results in a gain reduction.

R4/R5 must always be the same value to maintain input balance and achieve some semblance of common mode rejection.

Paul, W9AC

[echo_india]

Thanks again Paul. Interesting, they must have made some changes to the component values in later releases for the Mic Preamp board. This schematic is dated April 27 2023. I purchased my G2 in the beginning of April and I guess this is why I see the mismatch with components values on my board to this schematic.

I will swap out R4 & R5 to what is in the new build and see how this goes. I will post after testing.

Leslie

[w9ac]

Thanks again Paul. Interesting, they must have made some changes to the component values in later releases for the Mic Preamp board. This schematic is dated April 27 2023. I purchased my G2 in the beginning of April and I guess this is why I see the mismatch with components values on my board to this schematic.

I will swap out R4 & R5 to what is in the new build and see how this goes. I will post after testing.

Leslie

I see no reason to change R4/R5 to match the schematic. Any necessary gain change can be effectuated by changing only R8.

Paul, W9AC

[w9ac]

I see no reason to change R4/R5 to match the schematic. Any necessary gain change can be effectuated by changing only R8.

More accurately, if R8 is changed to increase or decrease gain then R2 should also change in lock-step to the same value as R8 in order to maintain input symmetry. IOW, R2/R8 should be the same value for the same reason that R4/R5 should be identical in value. But, the value of the R2/R8 resistors are not required to be the same value as the R4/R5 resistors. That difference is what establishes stage gain.

I wish I could offer more to solve your problem. With a scope, the point of the NE5534's input and output clipping can be measured. In between, an applied waveform from an audio signal generator can be examined for symmetry and distortion. It may take that kind of measurment to know what's really happening in your G2.

Paul, W9AC

[w-u-2-o]

Perhaps a helpful point of clarification: gain is R8/R5 (assumes R5 always equals R4).

As an example, the schematic shows R8 = 15000 and R5 = 1500. Therefore the gain is 15000/1500 = 10 (unitless). In decibels this is 20log(15000/1500) = 20dBV.

It sure would be great if someone who is having a problem with this interface could start probing around with an oscilloscope.

[echo_india]

Thanks Paul, Scott for you help and tips on this. When I get a bit more time I will dig in a bit further and I will keep you posted. Thank you.

Leslie

[w9ac]

After the discussion, I decided to create a balanced-input preamp circuit based on my description above. It offers excellent common mode rejection across a broad spectrum without the expense of a high-quality transformer like those manufactured by Jensen , CineMag, and Ed Anderson.

The amplifier uses a Bourns DR331-475BE common-mode choke and on-board voltage regulation. The advantage of this design as an input stage is that the choke rejects common mode current while preserving differential-mode current. On common-mode currents, an inductance equal to 2*L is added in series with each one of the two bifilar conductors. However, inductance cancels to zero in differential mode. This means the choke has no effect on complex input waveforms while rejecting hum and RFI. Traditional use of series chokes causes complex waveform overshoot and inductive ringing.

The circuit is based on the Texas Instruments INA849 instrumentation-grade 3-element op-amp. Among other features, its input voltage noise density is only 1 nV/√Hz. Being a true instrumentation amplifier, it further reduces common mode current: separate controls can be seen for independent low and-high frequency common mode optimization. A pair of test points at the top are used for output measurements.

Paul, W9AC

[Trucker]

Paul, very nice work! That may turn into a profitable sideline for you. I am sure many G2 owners will be interested in this.
James
WD5GWY