Update: 6/13/24.
A True Power Sum stereo detector using the THAT300 is shown here: https://proaudiodesignforum.com/forum/p ... 728#p17728
A Quadrature Power-Summed Compressor Sidechain For "The Uno Compressor" is shown here: https://proaudiodesignforum.com/forum/p ... php?t=1361
A THAT2252 replacement using a THAT4305 is shown here: https://proaudiodesignforum.com/forum/p ... 023#p17023
Original Post:
I've been kicking around some ideas for a compressor/limiter project and needed a replacement circuit for the discontinued THAT2252 RMS detector IC.
The THAT4301 Analog Engine® has an RMS detector, VCA and op amps which can be used to replace the 2252.
Its an elegant solution when all the pieces are needed.
The 4301 could be used, but a "discrete" RMS detector, made from a transistor array, would have a lower bill of materials.
The THAT4316 in surface mount has an RMS detector as well.
This is a DIY project so that rules out using surface mount.
There are lots of prior examples of log detectors available from Blackmer, Buff and others.
Most examples used 5 transistor arrays such as the CA3046 and CA3083.
The TI LM3046 is only available now in surface mount leaving the selection of five transistor arrays slim.
The THAT300-series, available in DIP, have four transistors per array.
At one time MicMix was going to become a dBx licensee and we obtained information on how to build a dBx 303.
I still have the 1970's-era hand-drawn copy we obtained from dBx.
The simplified schematic is shown below:

dBx 303 RMS Detector Section Simplified Schematic
Note: The actual dBx 303A Schematic including the discrete transistor VCA can be found here: viewtopic.php?f=12&t=860
The RMS detector begins at OA2 and uses five matched diode-connected transistors.
Is it really necessary to have all five transistors matched?
Can we use a four transistor array such as the THAT300?
Matching and thermal tracking are critical to maintain symmetry of the RMS detector.
A close examination of the circuit shows that for positive inputs, which produce a negative output at OA2, there is only one logging transistor. (QA2 pins 9,10, and 11.)
In order to obtain 2X log for positive inputs OA3 is used.
2X log for negative inputs is obtained by using two forward-biased junctions. (QA2 pins 1, 16, and 15 along with QA2 pins 13, 14 and 12.)
We only need one junction to obtain the log.
If an additional non-inverting gain of two stage were added for negative input polarity one matched transistor can be eliminated. (QA2 pins 1, 16 and 15.)

THAT2252 RMS Detector Replacement Using a THAT300 Transistor Array. Updated 2/07/17.
By sampling the log using one junction (QA2 pins 12, 13 and 14) and multiplying the result by 2 one transistor can become a 1N4148 diode.
Only 4 matched transistors are required.
(Looking at it from a 1970's perspective by using an "extra" diode-connected transistor we can eliminate an op amp. Now, op amps are cheaper than matched transistors.)
(Edit 2/07/17: I removed that diode to reduce the output swing requirement of the op amp. The figure has been updated. Now it looks very similar to the log Buff Dynamite detector: viewtopic.php?f=6&t=111#p6557 except for the X2 required for RMS.)
The circuit above does not include an output level-shifter or 0 dB trim.
Note that the peak currents in the charging capacitor are quite high and limited by the 4R7.
Be mindful of this current when laying out the timing cap return trace.
I used LME49720 for op amps A, B and C.
For op amp "A" bandwidth, low offset voltage and low bias current are important.
In terms of response characteristics this circuit appears to emulate the THAT2252 quite well.
The classic dBX sound can be obtained without using the discontinued THAT2252 RMS Detector, surface mount 4316 or the more expensive THAT4301.
Edit: 2/20/17. I'm posting later schematics below with links to circuit descriptions.
The second set of schematics show a conventional diode-based absolute value rectifier followed by a 2252-style log/RMS converter.
This circuit uses a single THAT300 transistor array per channel.

A Diode-based fullwave rectifier with current output.
The absolute value circuit shown above has current outputs which feed the log/RMS converter shown below.

A THAT2252-style RMS detector made from op amps and a THAT300 transistor array.
For higher performance a current rectifier can be used instead of a diode-based absolute value.
This current rectifier is an adaptation of the Roberts/Loft circuit used in the ubiquitous Loftec TS-1.
The output of the current rectifier feeds the log/RMS stage above.
The current rectifier uses two matched transistors in its current mirror (half a THAT300) and a THAT300 in the log/RMS converter.

A current rectifier based on the Loftec TS-1 and THAT2252.
The circuits above are discussed here: viewtopic.php?f=6&t=856&start=16
The following detector uses a current rectifier that is a hybrid featuring the TS-1's common base stage combined with the THAT2252's class A/B bias scheme.
This rectifier and RMS stage requires two THAT300.
It has the added advantage of being optimized for True Power Summing.

THAT2252 replacement using THAT300 and current rectification with class A-B bias optimized for True Power Summing.
Discussed here: viewtopic.php?f=6&t=856&start=25
Related reading:
Level Detectors, Absolute Value, Peak and RMS viewtopic.php?f=6&t=111
dBx 303 A Schematic viewtopic.php?f=12&t=860