THAT1512 / Transformer 'blind test' mic preamp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

jamesk

Member
Joined
Jan 31, 2025
Messages
5
Location
United Kingdom
Hi,

This is my first proper post here, after my post in the introduction section.

I am working on two mic preamp designs. One uses a transformer input and NE5534 gain stage, the other uses a THAT1512 IC. Both versions use Douglas Self's Quasi-floating balanced output. The preamps have identical controls, with +48V and -20dB pad switches and a stepped 12-way gain switch. The aim of the build is to create two preamps that look the same externally but have completely different topologies, and then perform blind tests on various sources to see if people have a preference for one or the other.

Here is the schematic for the transformer version:
transformer preamp.png
and the THAT1512 version:
THAT preamp.png
This is the gain switch for the transformer version. The THAT1512 version uses the same PCB but with different resistor values to give the same gain.
gain switch.png
This is the PSU:
preamp PSU.png

The first issue I am having is understanding the behaviour of the output stage properly. When testing, I can adjust the balance between the hot and cold outputs to be equal at 1kHz, but when I increase the frequency, they go out of balance. I also simulated the circuit in LTSpice and observed the same behaviour. Can anyone explain the reason behind this and what the best approach to balancing the output levels is. It seems like a significant flaw of the design but since it has been used in commercial products, I assume I'm missing something!

Here is the output at 1kHz:
PXL_20250129_182917265.jpg

And at 30kHz:
PXL_20250129_182944626.jpg

Thanks in advance!
James
 
Not sure of the exact reason for the imbalance, but that circuit was designed for audio frequency (20Hz-20kHz), so I don’t think balance at 30kHz (or anything above 20kHz) was a design goal.
Thanks for the response. I should have said, I've shown it at 30kHz as it's very exaggerated and visible. The imbalance becomes obvious above 10kHz.
 
Another thing I don't understand is how the circuit is unbalanced in the SPICE simulation when the pot is set exactly half-way. My understanding was that it's there to adjust for tolerances in the resistors, but in the simulation, the resistors are all exact values.

I feel like I'm missing something really fundamental here.

Here it is at 1kHz.
spice balanced.png


Thanks,
James
 
A balanced signal exists only between the hot and cold terminals. The third terminal is a screen not a ground. The two voltages relative to "ground" do not have to be equal. It is the impedances that are balanced not the signals.

I also notice in the first schematic you have a pot with the wiper to the NE5534 -ve input, one end connected the the NE5534 output and the last one to ground via a capacitor. To ensure a fixed, known maximum gain is available you should include a resistor in series with the capacitor (But now I have just seen your gain switch schematic so I see you are already aware of this :))

Cheers

Ian
 
Why not use a THAT 1646 for the balanced output?
Good question. I did consider this, but my long-term goal is to build a mixing desk so using 1646s on all the outputs would be prohibitively expensive. I realise this is a far-off goal, but I want to build up the various circuit building blocks bit by bit, so one of the requirements is that it doesn't become stupidly expensive on a large scale.
 
Good question. I did consider this, but my long-term goal is to build a mixing desk so using 1646s on all the outputs would be prohibitively expensive. I realise this is a far-off goal, but I want to build up the various circuit building blocks bit by bit, so one of the requirements is that it doesn't become stupidly expensive on a large scale.
I see, but you're probably not saving as much as you think. If R18 is a good quality multiturn trim pot (which it should be) it's going to cost over half of the 1646's cost by itself. I was also initially put off by the cost, but finally started using these on all my preamp outputs and never looked back. They just work great.
 
+1

THAT 1646 needs few peripheral components, which not only translates into faster matching-free stuffing time but also less real estate on PCBs. And they are extremely well matched btw channels with worst case ~0.1 dB. Plus they get cheaper (in mid run) when you aim for a price break up front.

A classic alternative would be INA.
 

Latest posts

Back
Top