Recommended implementation for THAT1512

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Recommended implementation for THAT1512

Post by JR. »

mediatechnology wrote: Sat Jun 10, 2017 12:52 pm
Not sure why you need 2 inductors in gain leg, is the gain pot mounted at some distance in a noisy environment? In the past I have driven the gain pot cable shield to reduce capacitance effects.
To provide gain rolloff at AM RF frequencies is why I've tried them.
That explains one ... kind of.
The THAT chip probably has clamp diodes inside so added external ones (D1,D6) are at best redundant.
The internal diodes are not "strong" enough. That is what drove me to blow up lots of THAT1510s: http://www.thatcorp.com/datashts/AES790 ... eturns.pdf
in that case you may want to add small Rs in series with the base (inputs) to preferentially steer fault current to the external diodes

This design looks like 10uF phantom blocking caps with 10r already in series. Unless this is being used in a studio with a patch bay and operators who like to short (charged up) inputs for amusement, perhaps not needed.
I don't see any Cs to ground at input... high RF and input rectification could be an issue, what does THAT recommend in their app notes (I'm too lazy to look)?
I believe the current thinking is a delta-connected three capacitor network. Previously it was wye. (Or is it Wye?)
I would definitely put locations on the PC board for three RF caps.

I've had pretty good results in a modestly high AM RF environment with the flat phono preamp with a differential C (which also happens to be Cload) and input series resistors to prevent CM RF rectification.
Yup... between cellphones and wifi far more RF than the good old days.

JR
Cancel the "cancel culture", do not support mob hatred.
NOON
Posts: 9
Joined: Mon May 15, 2017 8:03 pm

Re: Recommended implementation for THAT1512

Post by NOON »

Thanks for the input so far.

Input and output C networks for RF aren't on this schem because I was going to mount them directly at the I/O rather than on this PCB. I will add them for clarity and future reference before posting an updated schem.

The 100R and 100uF in the 48V network is/was there for reduction of crosstalk between channels, seeing as they're all sharing the 48V 'bus' from the PSU. I see what you mean with it being common mode and the peak switching currents, but better to reduce it more if possible? Better to implement cap multiplier local reg like on your DC coupled preamp link?

Not back-grounding the 48V in this case because the nice looking switch I want to use is SPST only. Will consider other options if you think it's that important, or just adding the back grounding to schem for final update before removing it before I convert schem to PCB. R8 100k is there to drain 48V in place of back grounding.

Yep, C3/C5 are 10uF because they're the max value available in Wima MKS2 range of reasonable sized/priced film caps. Open to other suggestions if there's better options.

OPA2277 is about 50% more expensive than OPA2188 and is max +/-18V while the OPA2188 is max +/-20V. I'm intending to run the system at +/-18V so the 2188 seems like the safer and cheaper bet. Any reason not to use it? Main difference I can see is 25uV offset vs 10uV offset, but both seem to handily beat the OP07.

Thanks for the pointer on the infrasonic resonance. Simulating increasing the input RB to 20k did this!
Servo Sim.pdf
(27.7 KiB) Downloaded 625 times
I'll do some more research on the Deboo and get my head around it. Anyone that has more experience in this area and wants to suggest a good config and values for this servo circuit please do.

I was looking at using a diode bridge to save space on the board. I'll update the schem to put it between 4R7 input resistors to help with HF CMRR.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Recommended implementation for THAT1512

Post by mediatechnology »

Putting the RFI protection on the I/O board makes sense.

The cap multiplier would serve to isolate crosstalk if that is a concern. The 48V rail output impedance should be low enough without phantom currents causing crosstalk.

Since you only have a SPST switch the 100K arrangement makes sense for phantom switching/capacitor discharge.

Not sure I have a better suggestion if you're going to use film caps. I've toyed around with using polypropylene speaker cross-over caps from Parts Express (Dayton) to get higher values at reasonable costs but the 50 uF is bigger than a D-cell.

The OPA2188 ought to make a good DC servo.

Now that you are aware of the resonance it will affect your component choices. That sim shows a huge peak. Wow.

John mentioned that you could use a non-inverting servo which would allow you to use a simple RC passive integrator at the input to the op amps. That might be simpler than the Deboo.
NOON
Posts: 9
Joined: Mon May 15, 2017 8:03 pm

Re: Recommended implementation for THAT1512

Post by NOON »

OK, been reading and simulating and reading and simulating... and come up with this servo design.
THAT1512 PREAMP SCHEM2.pdf
(61.79 KiB) Downloaded 662 times
Simulations say it should do the trick without affecting the LF response. (Used an SSM2017 model for preamp sim and OP07 for servo, closest I could find and pretty similar for sim purposes)
ServoSim.jpg
Could add another pole at the output, but the OPA2188 should be fast enough to handle it and 2M2 feed resistors should be enough to isolate noise etc.

Is the input impedance switch better off on the other side of the blocking caps? If a mic is drawing uneven currents on the 6k8s then there could be DC across it.
Any further comments on the design?
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Recommended implementation for THAT1512

Post by mediatechnology »

You might want to simulate it with a higher source resistance maybe 150R.

I would leave the termination resistor on the left-hand side of the coupling caps.
Two reasons I can think of are that its not going to increase the differential cutoff frequency and the DC from current imbalance will tend to equalize more.
NOON
Posts: 9
Joined: Mon May 15, 2017 8:03 pm

Re: Recommended implementation for THAT1512

Post by NOON »

Good point re: leaving the impedance resistor on the outside of the blocking caps, don't know what I was thinking!
Did a more complete and accurate sim, was originally working on a 'half' circuit with a generic opamp model, only changed to SSM model at last stage and didn't bother updating input circuitry.
Less than 1dB bump around 2.5Hz with 1u cap and 1M resistor. Previous sims were showing little difference using a 470n cap, but it's a significantly bigger bump with the 'full' circuit.
ServoSim2.jpg
I simulated moving the impedance setting resistor to the other side of the blocking caps, see what it does to the LF repsonse!
ServoSim3.jpg
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Recommended implementation for THAT1512

Post by mediatechnology »

I simulated moving the impedance setting resistor to the other side of the blocking caps, see what it does to the LF repsonse!
With 10 uF coupling caps its pretty LF-challenged with the bias resistors alone.
You might want to consider increasing R3 to reduce the matching requirements for LF CMRR.

The non-inverting integrator that JR suggested that you applied in differential form has the desired passive pole at the input.
Forgot to mention that I noticed that improvement earlier.
NOON
Posts: 9
Joined: Mon May 15, 2017 8:03 pm

Re: Recommended implementation for THAT1512

Post by NOON »

I'm keen to use the WIMA MKS2 film caps for the input and output and they only go up to 10uF. Simulations show no significant drop off or phase change at 20Hz, will see how it goes in the real world and experiment with R values if necessary. I've done a PCB design and it's off getting a small batch made up for testing, will report back when I've got some made up.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Recommended implementation for THAT1512

Post by mediatechnology »

Simulations show no significant drop off or phase change at 20Hz, will see how it goes in the real world and experiment with R values if necessary
I understand why you want to use films.

The primary degradation from low C values in this range is the increased source impedance the mic preamp "sees" when looking out its inputs back to the source. Though the source impedance might be a 150 Ohms, the capacitors, which have an effective value of 5 uF, add about 1.5K. Beware of rising 1/f noise with low-value input caps.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Recommended implementation for THAT1512

Post by JR. »

mediatechnology wrote: Wed Jun 28, 2017 4:51 am
Simulations show no significant drop off or phase change at 20Hz, will see how it goes in the real world and experiment with R values if necessary
I understand why you want to use films.

The primary degradation from low C values in this range is the increased source impedance the mic preamp "sees" when looking out its inputs back to the source. Though the source impedance might be a 150 Ohms, the capacitors, which have an effective value of 5 uF, add about 1.5K. Beware of rising 1/f noise with low-value input caps.
I don't know if 1/f is best characterization for the noise but increasing level with LF so not unlike 1/f. The mechanism is the device ein current times the effective impedance. The nominal 2k input termination will be in parallel with the 1.5k so somewhat lower. If I haven't already mentioned using lower value blocking caps will require better matching to keep LF CM rejection decent.

JR
Cancel the "cancel culture", do not support mob hatred.
Post Reply