THAT1510 THAT1512 Single-Supply Preamp with "T-Bias"

[mediatechnology]

This single-supply preamp design was presented by THAT Corp. at the 125th AES convention as part of one of their seminars. It uses an input biasing arrangement we're now calling "T-bias." Although the circuit was designed for single-supply operation from 10-36V, the T-bias network can be used in dual supply applications.

T-bias permits low-value input bias resistors to be used for low noise without sacrificing low frequency common mode performance or forcing an increase in input coupling capacitor size.

One of the keys to obtaining good noise performance from monolithic mic preamps is to have low-value input bias resistors. This is due to the relatively high input noise currents of the THAT1510, 1512, and similar devices from ADI (SSM2019) and TI (INA-series). The bias resistor values typically range from 1-10k ohm and connect (in dual supply apps) from each input to ground.

Lower values of Rbias however require higher-value input coupling capacitors (or closer matching) for phantom-powered and single-supply applications to provide acceptable LF common mode rejection. T-bias, using an additional resistor to raise the common mode input resistance, is shown in the following schematic.


(c) 2008 Wayne Kirkwood

In this single-supply application the mic preamp inputs and the Vref pin are raised to 1/2 supply by an MC33178 op amp rail-splitter.

R8 and R9 provide low-value input bias resistors. The noise current of each input flows through R8 and R9. In conventional single-supply applications these would connect to mid-supply or, in a split-supply application, ground. The common mode resistance that the input coupling capacitors would usually see (without R10) is the value of the bias resistor, 1k.

If R8 and R9 are made larger, the voltage noise across them is also made larger because it is the product of Inoise * Rbias. By adding R10, the common mode resistance of each input is made significantly higher. The noise currents at the junction of R8, R9 and R10 power sum and form across R10 as a significant voltage. This noise voltage, though much larger, appears in common mode and is attenuated significantly by the CM rejection of the THAT1512. With values of R10 ranging from 0-47k there is virtually no noise penalty.

Input bias current also flows at DC and the sum of both input's bias current form across R10. This causes a common mode offset to develop. The Ibias (max) limit of 14 uA/input (28 uA total) limits the maximum value of R10 depending upon input headroom requirements, supply voltage etc. It should be noted that this internal common-mode offset is essentially canceled by the 1512's differential to single-ended stage and does not appear on the output.

The common mode resistance at the preamp input can be made significantly higher using T-bias which reduces the value and matching requirements of the input capacitors. (The example schematic uses 50 uF/leg - these values could be made much smaller.) The overall common mode resistance of the preamp will be dictated by the phantom power resistors but these do not impact capacitor matching because they are at the input.

The differential input impedance of 2k at the 1512 input may be too low for some tastes and can easily be raised by making R8 and R9 larger.

In a future installment I'll post some noise comparisons showing the effectiveness of T-bias.

[JR.]

I think I have discussed this with you before but not sure if on a forum or in private correspondence.

IIRC Paul Buff mentioned something very close in a private conversation about his Transamp a couple decades ago. He was using something like 50-100k resistors for DC bias to bases inside the potted transamp. I need to do the math to see if how two resistors to a common impedance compares to a single shunt and two DC path resistors, but I would remind all there is an effective 150-200 ohm microphone impedance also in shunt between these two input nodes. Also your 1k+1k nominal may be a little low in parallel with the 2x 6.81k phantom power resistors.

This is a little obscure and not easy to wrap one's head around. Only the input noise current component enjoys any significant (but not total) cancellation. In your example the cross couple noise is attenuated by R8 or R9 divided by R10. But there is another path through the mic itself around the divider. This can be useful for bipolar device inputs. Less noise is ALWAYS better, even if it only a little less. I think it is inherent to this general topology but can be wasted by too low DC bias Rs (I guess that's your point). This is real and perhaps measurable but my guess is less than a dB with nominal mic termination. I guess you can demonstrate a larger effect with a higher impedance input than mic.

JR

[mediatechnology]

Thanks John -

I do agree that 1Kx2 might be a little low as a differential load. 1.5Kx2 or 2Kx2 raises Zdiff to a more acceptable level. There is one source I need to cite that places the desired Zdiff around 3K total. I think the 3K assupmtion is based on paralleled active inputs for splits...

I don't remember the TransAmp exchange WRT bias but I'd like to revisit the schematics posted elsewhere by CJ.

I think the modest noise benefit from this may not come from dynamic or transformer-coupled mic outputs in the 150R or less range but active mic outputs having a rising and higher Zdiff output impedance at lower frequencies. That's just a hunch.

There is also an open-circuit (unterminated) benefit (that of course becomes moot as soon as a low Z source is plugged in) or a benefit from applications fed from higher-than-prefered source impedance such as pads for line-level duty. Terminated with a pure 150 R source I don't think there's going to be much difference.

When low-value bias resistors are desired it definitely helps LF CMR and cap requirements.

Still not bad for less than a penny...

[mediatechnology]

I was just looking at the SSL 82E149 active mic preamp and noticed that it used T-bias:



The full pdf is here: https://www.ka-electronics.com/images/S ... _index.htm

I notice that the bias resistors form a ~3K differential load rather than the low-ish 2K of the 1510 circuit. Kinda hard to read the designators but it's R15-17. The common resistor is 100k, the bias resistors 1K54.

[mediatechnology]

FWIW the datasheet for the 1510 has been updated and now includes this bias scheme for the phantom power example.

http://www.thatcorp.com/datashts/1500data.pdf#page=6

Note also the addition of D5 and D6 to increase protection.

[mediatechnology]

I wanted to bump this thread since we're discussing THAT1510/1512 mic preamp servos once again. viewtopic.php?f=6&t=256

To make along story short, if you want to use the servo shown in the link above with T-bias, you have to provide a differential servo output.

The simple way to do this is invert the servo output and feed the opposite input with the inverted correction signal. This drives the node where the bias resistors meet - at the 22K - differentially.

If a differential servo is not used, then a portion of the correction signal makes it to the other input and appears in common mode.

Driving both inputs with an equal but opposite correction signal keeps the T-bias resistor at a constant potential since both legs are driven and the correction appears as a fully-differential signal.

[ricardo]

To make along story short, if you want to use the servo shown in the link above with T-bias, you have to provide a differential servo output.

Is that really important? Yes, there's a common mode signal but its VLF and only 250uV.
I'm looking at PCB Real Estate. Your original servo would be nice if there was a dual OP07. THAT 1570/1571 looks attractive if I hed wen 2 skul en unnerstod digits.
BTW, I'm selling miniature 4700uF 1% Unobtainium CGO caps. Send $500 in used bank notes to Cooktown Recording & Ambisonic Productions for a sample. First come 1st served.

Du.uh! I see the problem. The 22k attenuates the correction signal which has to be much bigger to have an effect so much bigger common mode.

THAT 1570/1571 goes into the lead.

[mediatechnology]

Du.uh! I see the problem. The 22k attenuates the correction signal which has to be much bigger to have an effect so much bigger common mode.

Correct. Most of the correction that you want to be differential becomes common mode due to the 22K in the middle. By driving both ends of the "T" with opposing polarities you eliminate the problem. The 1570 demo board has a jumper for T-Bias. The 5171 servo ouput is differential. I was the first to check to see if that feature worked. It does.