THAT 1512 based preamp - offset problem

[qwerty8787]

Hi all,

I'm building a preamp based on the THAT1512 IC, pretty much following the datasheet suggestions (http://www.thatcorp.com/datashts/THAT_1510-1512_Datasheet.pdf)

I'm trying to make a direct injection input for it, selectable via DPDT switch (commutes the +/- ic inputs from the +/- mic signals to the buffered DI input).

The DI input is a non-inverting opamp buffer with input voltage divider for attenuation. The output from it goes to a 10uF Non-polarized electrolytic cap, then pulled down by a 10K resistor.

The output of the 1512 is pulled down by a 20K resistor in another section (not present on the images below).

These are the schematic:

The problem is that in DI mode, at max gain (about 65dB) I have something like 15Vdc offset on pin 6 of THAT1512 IC in the worst case.

I managed to get about 3Vdc playing with pulldown resistor on inputs (removing R213) but it's still too much.

On the mic mode (SW102 switched) offset is about 200mV @ max gain, so it's good.


I can't understand what could be wrong, does anybody have any idea?

Thanks in advance! 

 

[SkunkFunk]

http://www.thatcorp.com/datashts/THAT_1510-1512_Datasheet.pdf

The input DC offset is amplified by the same gain as the audio signal in the 1512. They suggests placing a large capacitor in series with the gain pot, which sets the dc gain to 0 att all gains. Look for Cg in the datasheet, everything should be well explained.

 

[qwerty8787]

Thanks for the reply

If the problem is about the DC on the gain net, it should be also on mic mode. And it isn't.

DC blocking on gain net is suggested to avoid pot/switch clicking when changing gain, and I really don't have any clicking issue on gain changes fortunately.

I think the problem is about feeding the 1512 with a unbalanced/single-ended signal instead of differential/balanced. But in theory this kind of feeding shouldn't be a problem....am I wrong?

 

[SkunkFunk]

Thinking again, since the source resistances are different in DI mode, the input bias currents might create a voltage difference across the input legs. What happens if put a 10k resistor in series with the ground connection to the dpdt?

 

[qwerty8787]

I tryed before to disconnect the ground from DPDT lug and disconnect also R213 (10k) so that each side should see only the 20k resistors. I get about 3Vdc on output.

The strange thing is that I always get about 40mVdc to 80mVdc after C213, but before it I see 0V (got via OPA134 offset trim).

I really don't know where those mVolts comes from.

In mic mode I still have about 40mV, but on EACH 1512 input.
In DI mode I have different offsets between + and - input nets, so those offsets become differential and get amplified by 1512.

So the question could be: how can I null those input offsets, or at least make them similar so they're not differential anymore?

I provided a path to ground via pull-down resistors after DC-blocking capacitor (in either mic and DI mode) but I still get offset. What's wrong?

 

[qwerty8787]

Disconnected DPDT from ground and soldered a 10k resistor between DPDT and ground.
Also removed C213, C103 and 1512 IC.

Here are DCR and VDC (@ max gain, with power on, PSU @ +/-18Vdc) Measurements:

DI mode:

pin 2: 6,85k
pin 3: 6,80k (but unstable reading, raising from 6.70k, like there's some capacitance on that net. this value is taken after waiting about 30 sec)
between 2-3: 13,65K (raising)
VDC pin 2: 0 mVdc
VDC pin 3: -1 mVdc

MIC mode:

pin 2: 6,40k
pin 3: 6,41k
between 2-3: 3,65k
VDC pin 2: -1mVdc
VDC pin 3: -1mVdc


IC back on, measured again (@max gain):

DI mode:

pin 2: 6,86k
pin 3: 6,80k (unstable like before)
between 2-3: 13,65K (raising again)
VDC pin 2: -28 mVdc
VDC pin 3: -28 mVdc
VDC pin 6: -1.62 Vdc

MIC mode:

pin 2: 6,40k
pin 3: 6,39k
between 2-3: 3,65k
VDC pin 2: -46mVdc
VDC pin 3: -46mVdc
VDC pin 6: -1.5 Vdc

 

[qwerty8787]

Jumpered pin 2 & pin 3 to gnd:

VDC measurements (power on PSU @ +/- 18Vdc):

pin 2 & pin 3: 0 Vdc
pin 6: -1.7 Vdc

something really strange is going up here.

 

[PRR]

> something really strange is going up here.

What is the input bias current of the 1512?

What are the resistances it flows through?

Are they balanced?

What is the gain for unbalanced DC input?

Some wild guess: 20uA current, your switching puts different combinations of 2K, 20K, and 10K on inputs, easily tenth-Volt common-mode and half that much offset, times 100 is essentially full supply slam.

 

[Tekay]

470p between pin 2 and 3 or 470p from pin 2 and 3 to ground. Check the datasheets at Thatcorp. I got some weird offsets when I forgot these caps.

 

[qwerty8787]

470p between pin 2 and 3 or 470p from pin 2 and 3 to ground. Check the datasheets at Thatcorp. I got some weird offsets when I forgot these caps.

Actually they are in the circuit (speaking of C1 and C2 on figure 5 in datasheet), and they are before DC blocking caps just like suggested.

 

[qwerty8787]

> something really strange is going up here.

What is the input bias current of the 1512?

datasheet says it's about 5uA typical

What are the resistances it flows through?
Are they balanced? 

In my circuit, from 1512 going back we find R111 and R112 each @ 20k to ground.
In mic mode we find R103 and R104 @ 2K, connected to ground by R105 @ 10k.
In di mode there's only R213 @ 10k on positive net to ground but I managed to remove the short to ground in DPDT and place between another 10k resistor to balance it.

What is the gain for unbalanced DC input?

nice question, on datasheet I can find anything about unbalancing.
The only thing noted is about offset gaining proportionally to gain, but in the worst case we talk about 250mVdc on output, not some volts...

Some wild guess: 20uA current, your switching puts different combinations of 2K, 20K, and 10K on inputs, easily tenth-Volt common-mode and half that much offset, times 100 is essentially full supply slam.

the strange thing is that shorting to ground each input I get again that strange offset. Maybe then it's an output thing?
I managed to disconnect almost everything after pin 6 but I still get huge offset.
Tried to change IC too.

I will try to lift pin 6 from socket to be sure it won't be connected to anything and repeat the measure

 

[SkunkFunk]

Since + and - are connected through a pretty low resistance in mic mode (2*2k), an offset volgate on one of them will affect the other one as well, nulling out much of the mismatch. Nothing in the datasheet mentions that the bias currents should be the same for both legs and they can be as high as 14µA Unfortunately, you can neverexpect to get a chip that lives up to the typical ratings. If you replace the 2k-2k-10k Y-configuration with two 10k, one for each leg, do you get the same high offsets as in DI mode? Or have you tried another IC?

 

[qwerty8787]

Maybe I solved it.

I tried to shunt XLR input pins to gnd, and offset in mic mode seemed to be good (about 200mVdc @ max gain)

Same thing in DI mode, I putted the signal generator in input and the output offset seemed good.

probably the problem was that there wasn't any source impedance BEFORE dc-blocking capacitors.

In DI mode the 10k between DPDT and gnd SHOULD be used, or differential dc voltage of about 40mV at inputs will screw output offset.

I will do further tests but I think now i'm fine

Thanks to anyone helped me, really appreciated 

 

[SkunkFunk]

Great that it worked better! If you're keeping the new 10k resistor, how about put a capacitor similar to C213 in parallell with it to eliminiate the increased thermal noise and give similar source impedances for AC frequencies a well? Just a thought.

 

[qwerty8787]

Yes, it should be good to put it! Thanks  8)

The problem is that I already made the pcb, and if the only problem with this version is gonna be this resistor/cap missing, I won't make a new one.
I don't have too much time, this little project is my thesis in 3-year-degree Electronic Engineering, and I'm gonna graduate probably in February so I've got to go ahead quick 

butI'll keep changes in schematic so next time I need to etch one, I'll have the updated version.

By now everything seems good, this IC sounds really clean!  :