THAT2252 RMS Detector Replacement Using A THAT300 Array

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

Thanks John for uploading the compander.
I'll go find that broken link in the OP and fix it.

I got out my crayons and sketched the schematics of the various test circuits.

RMS Detector

The first schematic is the 2252-style clone sent to me by Gary Hebert.
The 2252, unlike the original 303, performs absolute value ahead of the RMS stage.
The following schematic is the RMS portion after the absolute value rectifier.

Image
A THAT2252-style RMS detector made from op amps and a THAT300 transistor array.

The input is a current.
Q1 and Q2 provide a 2Xlog output for negative input currents.
D1 is a reverse polarity clamp.
Q3 Rt and Ct are the log-domain filter.
Q4, op amp B and the 1MΩ current source provide level shift.
All transistors should be matched and in thermal equilibrium so a THAT300 array is used.

Rt is not true current source but for practical purposes can be thought as one since the delta-V at the emitter of Q3 is quite small.
Both resistors should be equal value and set the timing current which is √(It1*It2).

The Absolute Value Circuit

The absolute value circuit ahead of the RMS detector can be done a number of ways and can be either voltage or current output.
Since the 2X logging stage uses a current input, a current output is preferred.

The first absolute value stage is a conventional fullwave rectifier:

Image
A Diode-based fullwave rectifier with current output.

Halfwave rectification for positive inputs is performed by D2.
D1 clamps the op amp for negative inputs.
The sum of the currents at the output represent the absolute value of the input voltage.
The negative-going current output of the absolute value can connect directly to the current input of the RMS stage.

The second example shows current rectification as it was originally done in the THAT2252 and Loftec TS-1.

The actual circuit is the TS-1's rectifier. The 2252 is almost exactly the same.

Image
A current rectifier based on the Loftec TS-1 and THAT2252.

I've redrawn the TS-1's rectifier in a "reinterpretation" that I can more easily follow.
I shifted the focus of the current mirror stage, Q3 and Q4 so that Q3 looks more like a rectifier.

Q1, a common base stage provides additional non-inverting voltage gain as a "helper" to improve bandwidth of op amp A.
Q1's collector load is 15K+470R.
The base is biased to -7.5V.
The emitter resistor is 470R making the added inside the loop gain about 30X.

The op amp inputs are held at -3.75V by virtue of the voltage divider formed by the 30K and 10K resistors.
This permits Q2 and Q4 to sink current for loads held at or near ground.

Negative inputs are rectified by the base-emitter junction of Q2.
This sets up a nearly-identical collector current in Q2 which provides the negative polarity rectified output.
The collector current for Q2 (when loaded by the following stage) flows from the output to the inverting input held at -3.75V.
The 470Ω base resistor pre-biases Q2 to reduce rectification deadband.

For positive inputs, the output of op amp A and the collector of Q1 swing more negative.
The emitter base junction of Q3, a diode-connected transistor, provides rectification for positive inputs.
The Q3/Q4 stage is where I took "artistic liberty" in the schematic layout to emphasize Q3's base-emitter junction's roll as rectifier.
Q3 and Q4 are a current mirror with Q3 also a rectifier diode.
As the current in Q3's base emitter junction increases a nearly-identical mirrored current is setup in the collector of Q4.
The action of the current mirror serves to invert polarity for positive inputs.

When the collector currents of Q2 and Q4 are summed, the final result provides the absolute value.
The output of the current rectifier is a current that can be directly connected to the RMS stage.

Q3 and Q4 require matched devices in thermal equilibrium.
For a stereo detector, a single THAT300 could be used and share between channels.
A low-cost dual NPN transistor pair such as the NST45011 could also be used.

Advantages and Disadvantages.

The diode-base absolute value circuit has about 10 dB or so less dynamic range but does not require matched transistors in the rectifier.
The current rectifier has more dynamic range but requires matched devices.

I may have missed a few subtleties but really wanted to get this posted before this session times out. LOL.

Related reading:
A Discussion About True Power Summing for Stereo Compressors https://www.proaudiodesignforum.com/for ... p?f=6&t=61
Compressor Attack Release Signature Comparisons https://www.proaudiodesignforum.com/for ... ?f=6&t=281
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by JR. »

mediatechnology wrote:Thanks John for uploading the compander.
I'll go find that broken link in the OP and fix it.

I got out my crayons and sketched the schematics of the various test circuits.

RMS Detector

The first schematic is the 2252-style clone sent to me by Gary Hebert.
The 2252, unlike the original 303, performs absolute value ahead of the RMS stage.
The following schematic is the RMS portion after the absolute value rectifier.

Image
A THAT2252-style RMS detector made from op amps and a THAT300 transistor array.

The input is a current.
Q1 and Q2 provide a 2Xlog output for negative input currents.
D1 is a reverse polarity clamp.
Q3 Rt and Ct are the log-domain filter.
Q4, op amp B and the 1MΩ current source provide level shift.
All transistors should be matched and in thermal equilibrium so a THAT300 array is used.
While that looks like the 2252 it lacks some typical RMS aspects. The classic RMS math is square root of the integral of signal squared.

The two Vbe drops in series performs 2xLOG which is equivalent to signal squared when compared to a single Vbe at the decode side. That squared current in Q3 shows up in the faster charging of integration cap Ct, but to finish the RMS equation a square root needs to be performed. This is generally accomplished by biasing the second Vbe (or the first) with the integrated squared current. Both diode drops on the decode side appear to be fixed currents, no doubt representing 0dB at equilibrium.

I won't be a purist or pedant about true or real RMS vs ave because my development work while developing my last peak/VU console meters revealed very little difference (I couldn't see any on my bench) between RMS (actually computed by my microprocessor) and simple average for complex audio signals. (I ended up removing the RMS code because it did not appear to add any value other than perhaps for marketing)
Rt is not true current source but for practical purposes can be thought as one since the delta-V at the emitter of Q3 is quite small.
Both resistors should be equal value and set the timing current which is √(It1*It2).
RT sets the 0dB current, for 0V output.
The Absolute Value Circuit

The absolute value circuit ahead of the RMS detector can be done a number of ways and can be either voltage or current output.
Since the 2X logging stage uses a current input, a current output is preferred.

The first absolute value stage is a conventional fullwave rectifier:

Image
A Diode-based fullwave rectifier with current output.

Halfwave rectification for positive inputs is performed by D2.
D1 clamps the op amp for negative inputs.
The sum of the currents at the output represent the absolute value of the input voltage.
The negative-going current output of the absolute value can connect directly to the current input of the RMS stage.
A subtle issue with the simple diode FW rectifier, is that the op amp DC offset between that op amp and logging op amps can cause a low level error. If the DC errors are in the same direction they partially cancel, but opposite direction Dc offsets add. Back in the day using TL07x op amps with poor DC specs could be a factor.
The second example shows current rectification as it was originally done in the THAT2252 and Loftec TS-1.

The actual circuit is the TS-1's rectifier. The 2252 is almost exactly the same.

Image
A current rectifier based on the Loftec TS-1 and THAT2252.

I've redrawn the TS-1's rectifier in a "reinterpretation" that I can more easily follow.
I shifted the focus of the current mirror stage so that half of it looks more like a rectifier.

Q1, a common base stage provides additional non-inverting voltage gain for op amp A.
It's collector load is 15K+470R.
The base is biased to -7.5V.
The emitter resistor is 470R making the added Inside the loop gain about 30X.

The op amp inputs are held at -3.75V by virtue of the voltage divider formed by the 30K and 10K resistors.
This permits Q2 and Q4 to sink current for loads held at or near ground.

Negative inputs are rectified by the base-emitter junction of Q2.
This sets up a nearly-identical collector current in Q2 which provides the negative polarity rectified output.
The collector current for Q2 (when loaded by the following stage) flows from the output to the inverting input held at -3.75V.
The 470Ω base resistor pre-biases Q2 to reduce rectification deadband.

For positive inputs the output of op amp A and the collector of Q1 swing more negative.
The emitter base junction of Q3 provides rectification for positive inputs.
It is in this stage where I took "artistic liberty" in the schematic layout to emphasize Q3's roll as rectifier.
Q3 and Q4 are a current mirror.
As the current in Q3's base emitter junction increases an identical mirrored current is setup in the collector of Q4.

The action of the current mirror serves to invert polarity for positive inputs.
When the collector currents of Q2 and Q4 are summed the final result provides the absolute value.
The output of the current rectifier is a current that can be directly connected to the RMS stage.

Q3 and Q4 require matched devices in thermal equilibrium.
For a stereo detector a single THAT300 could be used and share between channels.

Advantages and Disadvantages.

The diode-base absolute value circuit has about 10 dB or so less dynamic range but does not require matched transistors in the rectifier.
The current rectifier has more dynamic range but requires matched devices.

I may have missed a few subtleties but really wanted to get this posted before this session times out. LOL.
Added degeneration resistors in series with the current mirror device emitters would reduce the sensitivity to Vbe matching.

Being able to capacitor couple the input resistor eliminates input offset DC performance errors affecting low level accuracy/dynamic range. (perhaps not a concern with better modern op amps.).

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

While that looks like the 2252 it lacks some typical RMS aspects. The classic RMS math is square root of the integral of signal squared.
True.

Image

Since the VCAs being used with this detector presumably have a 6 mV/dB scale factor (THAT218X) the resulting response is RMS.
The square root is effectively taken downstream from the detector in the VCA by virtue of it's 6mV/dB response.
But the RMS representation, at 6mV/dB is still in the log domain.
If the audio input of the VCA were a fixed DC reference, the VCA output would represent the DC RMS value in the linear domain.
That would be useful in a measurement instrument.
I won't be a purist or pedant about true or real RMS vs ave because my development work while developing my last peak/VU console meters revealed very little difference
People seem to prefer the way RMS sounds. I'm not going to argue with success.

Related reading:
Compressor Attack Release Signature Comparisons viewtopic.php?f=6&t=281
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by JR. »

We've been around that tree before, I think we discussed this exactly in another thread about RMS. DBX also made a big deal about RMS being special for tape NR a much bigger business for them than dynamics back in the day (they claimed RMS introduced less companding errors from tape's less than ideal transfer function).

2xVbe modulated by audio signal current is indeed 2x LOGx or x^2. Then subtracting two constant current biased Vbe is not remotely dividing it by 2, or taking a square root. It gives the correct divide by 2 answer at 0dB, while it is probably a more useful gain law for their VCAs (-6mV/dB).

The charging current at Ct is arguably X^2-(RT/15V) but discharge current is simply -RT/15V for a roughly linear -dB/sec decay (for useful range of the side chain voltage). I suspect that makes a useful sounding fast attack/slow release side chain characteristic.

If it sounds good it is good, enjoy.

Sorry, disregard the grumpy old man... :lol:

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

Then subtracting two constant current biased Vbe is not remotely dividing it by 2, or taking a square root.
Are you talking about the role of Q3 and Q4?
They neither divide it by two or take the square root.

Image

The output of the detector is the mean of the sum of the squares at 3mV/dB scaling. (At Q3 emitter or Q4 emitter.)
Q4 provides a -Vbe level shift.

If the scale factor is defined as 6 mV/dB then the 3mV/dB mean of the sum of the squares is divided by two which is a log-domain square root of the mean of the sum of the squares.
Nothing electrically divides by two in the detector, certainly not Q3 or Q4.
A 3/6 change in scaling ratio divides it by two.

Q4 and associated components simply drop the final output by one Vbe. (And buffer the timing network.)

An actual THAT2252 has its log converter biased up by one Vbe at the noninverting input to offset itself from the absolute value current sinks.
(To give the rectifier current output compliance.)
The 2252's level shift then has a two Vbe offset to subtract: One from the log converter and one from the rectifier/log domain filter.
In the 2252 V2 = Vbe; V3 = 2Vbe.
See: http://www.proaudiodesignforum.com/imag ... ematic.JPG

The 2252 clone shown above has op amp A's non inverting input at ground.
It doesn't need to be biased up by one Vbe because the current rectifier ahead of it can work into a virtual ground.
Q4 (V3) only needs to offset one Vbe (Q3).
As a consequence, and to make the 2252 clone compatible with the THAT2252, only a single Vbe offset is required.
In the 2252 clone V2 = 0; V3 = Vbe.

There is no square root operation performed by Q3 or Q4.
If I look at the output with 3mV/dB scaling it's not giving me RMS.
If I use a 6mV/dB scale factor after the detector's output (either electrically or by definition) then a square root is performed on a 3mV/dB "mean of sum of squares" measurement and I get linear dB/log-scaled RMS readings.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by JR. »

mediatechnology wrote:
Then subtracting two constant current biased Vbe is not remotely dividing it by 2, or taking a square root.
Are you talking about the role of Q3 and Q4?
They neither divide it by two or take the square root.

Image

The output of the detector is the mean of the sum of the squares at 3mV/dB scaling. (At Q3 emitter or Q4 emitter.)
Q4 provides a -Vbe level shift.

If the scale factor is defined as 6 mV/dB then the 3mV/dB mean of the sum of the squares is divided by two which is a log-domain square root of the mean of the sum of the squares.
Nothing electrically divides by two in the detector, certainly not Q3 or Q4.
A 3/6 change in scaling ratio divides it by two.

Q4 and associated components simply drop the final output by one Vbe. (And buffer the timing network.)

An actual THAT2252 has its log converter biased up by one Vbe at the noninverting input to offset itself from the absolute value current sinks.
(To give the rectifier current output compliance.)
The 2252's level shift then has a two Vbe offset to subtract: One from the log converter and one from the rectifier/log domain filter.
In the 2252 V2 = Vbe; V3 = 2Vbe.
See: http://www.proaudiodesignforum.com/imag ... ematic.JPG

The 2252 clone shown above has op amp A's non inverting input at ground.
It doesn't need to be biased up by one Vbe because the current rectifier ahead of it can work into a virtual ground.
Q4 (V3) only needs to offset one Vbe (Q3).
As a consequence, and to make the 2252 clone compatible with the THAT2252, only a single Vbe offset is required.
In the 2252 clone V2 = 0; V3 = Vbe.

There is no square root operation performed by Q3 or Q4.
If I look at the output with 3mV/dB scaling it's not giving me RMS.
If I use a 6mV/dB scale factor after the detector's output (either electrically or by definition) then a square root is performed on a 3mV/dB "mean of sum of squares" measurement and I get linear dB/log-scaled RMS readings.
Sorry to be so pedantic about this, over the decades I have performed several RMS conversions to drive meters (using both analog and digital technology). For analog, the integration of the x^2 current is generally done by driving an op amp virtual earth input with CT across the feedback path, the square root is then extracted by pulling that integral of X^2 current through the second diode connected junction that is in series with the X^2 junction. Sorry again examples of this topology are all over the literature.

What dbx/THAT did is elegant, sounds good, and works...(in combination with their VCA, which is their business).

I learned long ago not to argue with people about what "they" hear... if it sounds good it is good.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

If it makes you feel better John I can put a gain of 1/2 on the output and make the scale factor 3mV/dB in order for it to pass "RMS certification."

The bottom line is that it needs to be functionally equivalent to a THAT2252.
The TS-1's current rectifier is a keeper as an absolute value.

The first circuit I posted http://www.proaudiodesignforum.com/imag ... 800_BW.jpg isn't electrically the same as a 2252 because it has no level shift.
It really needs a fifth transistor matched to the other four.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by JR. »

mediatechnology wrote:If it makes you feel better John I can put a gain of 1/2 on the output and make the scale factor 3mV/dB in order for it to pass "RMS certification."

The bottom line is that it needs to be functionally equivalent to a THAT2252.
The TS-1's current rectifier is a keeper as an absolute value.

The first circuit I posted http://www.proaudiodesignforum.com/imag ... 800_BW.jpg isn't electrically the same as a 2252 because it has no level shift.
It really needs a fifth transistor matched to the other four.
No that won't but you do not need to make me feel better... :D

My ideas would make it less of a 2252 mimic, and not the design goal. (I still like the idea of adding degeneration resistors to current mirror emitters if you use that rectifier topology and the emitters are separately accesable.)

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

(I still like the idea of adding degeneration resistors to current mirror emitters if you use that rectifier topology and the emitters are separately accessible.)
I do want to try that.
It's in the 80s here today in February. Very hard to stay inside for long.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT2252 RMS Detector Replacement Using A THAT300 Array

Post by mediatechnology »

Gary Hebert of THAT sent me a variation of the current rectifier which uses Q2 to pre-bias the rectifier in class A-B similar to the THAT2252.
Gary liked JR's common base stage, Q9.


Image
THAT2252 replacement using THAT300 and current rectification with class A-B bias optimized for True Power Summing.

Current rectification is performed by op amp A, Q1-Q4 and Q9. (See also: viewtopic.php?f=6&t=856&start=16)
Q3 rectifies positive inputs. Q4 mirrors the current in diode-connected Q3.
Q2 provides class A-B bias for transistor Q1.
Q1 rectifies negative inputs.
Q1's collector current is a nearly-identical copy of the negative input current.
The collectors of Q1 and Q4 are combined to provide an output current equal to the absolute value of the input current.

The absolute value current output feeds the log converter at the inverting input of op amp B.
This particular log stage level-shifts the input to the converter, using Q5, so that the emitter of Q8 sets around 0V.
The THAT4305, 4315 and 4320 RMS detectors have a similar level shift.
This reduces temperature sensitivity when the detector output is combined with another channel for True Power Summing.

Since the timing capacitor (Ct) can now see either polarity I added op amp C to bias the -Ct terminal to below ground.
For polar electrolytic capacitors the -300 mV may not be significant and Ct- could be grounded.
Film timing caps could also be grounded.

Tantalum capacitors, which have tighter tolerance and lower DA excel as a Ct.
Tantalums prefer that the Ct- terminal be either biased negative by op amp C or tied to the -15V rail to avoid any reverse-polarity.
(A non-polarized Ct could also be made from two back-to-back capacitors. In the case of tantalums this is more expensive than op amp C.)
If Ct- is tied to the negative supply it is subject to supply noise and should also be rated at >16V.

The return current through Ct- is quite large and limited by op amp B's output current.
When Ct is grounded the return trace should be large.

Q8's dynamic impedance isolates op amp B's output from the large capacitive load of Ct.
When Ct- is referenced to op amp C's output it is also isolated from Ct's capacitive load by Q8.

When Ct- is connected to the -1.3V reference Ct's charging current is isolated from ground and returns to the supply rails.
Another advantage is that low-voltage tantalums can be used.

In this configuration a stereo True Power Sum detector uses 4 THAT300, and 3 LME49720 vs. 3 THAT300 and 3 LME49720 in the previous example.
Post Reply