An Improved Dual Class-A Output: The "DCAO2"

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:

An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

This thread is a preview for an updated Dual Class A Output, the DCAO2.

Update September 16,2016. The current pre-driver schematic using a dual op amp instead of the THAT1240 is here: https://www.proaudiodesignforum.com/for ... 9862#p9862

The original Dual Class-A Output began as an experiment in current-boosting a THAT1646 using "Common Mode" Drive: https://www.proaudiodesignforum.com/for ... ?f=6&t=140

Common mode drive uses the Sense inputs to drive both outputs together, rather than differentially. The "normal" differential input is grounded.
By combining the two outputs, which have internal ballasting resistors, the already high output current of the THAT1646 (or DRV134) can be doubled.
The thread also shows how to use the 1646 to drive transformers differentially.

In a later thread the THAT1646 is shown with external current-boosting transistors: https://www.proaudiodesignforum.com/for ... p?f=6&t=23

Common Mode signal drive and differential bias drive are combined and are the beginnings of the Dual Class-A Output Headphone and Line Amp.

The Dual Class-A Output PC board was developed. The documentation is located in the Build section of the forum: https://www.proaudiodesignforum.com/for ... ?f=7&t=509

The Dual Class-A PC boards found their way into transfer consoles and headphone amps.
Many of them are used to drive transformers.
When I need to use phones its my go-to gadget.

In a later thread it's turned into a 10W Class-A power amp: https://www.proaudiodesignforum.com/for ... ?f=6&t=679

After selling out of the first run of PC boards I decided to make some improvements.

The most obvious improvement is a gain stage. This is a no-brainer.

The original design using the THAT1646/DRV134 was primarily a line amplifier/transformer driver and unity gain.
Having a gain stage and trim is needed for both Headphone Amps and Transformer Drivers.
Transformer drivers need minor gain trims to offset insertion loss.

The second was to explore eliminating the THAT1646/DRV134 due to their internal build-out resistors.

Along the way I found how to reduce the open loop distortion of the output even further.
Those build-outs were not a problem after all.
As it turns out signal gain accuracy in the upper and lower drivers is very important.
The THAT1646/DRV134 have great gain accuracy; THAT1240s have even better.

By virtue of having a gain stage, there is the ability to provide overall global negative feedback.

Rod Elliot wrote:
[T]he open loop behaviour of a power amp should minimise crossover distortion before any feedback is added.

Source: http://sound.westhost.com/class-a.htm
I couldn't agree more.
Having a highly linear open loop stage - one that doesn't need feedback - is only going to get better with it.

The final result in the DCAO2 is a highly linear open loop output stage with the option to operate closed loop to extend output power.
The already low distortion open loop output, which does not need NFB at the 100 mW level, gets a power boost and THD reduction at levels exceeding 1 Watt.

Schematic

Image
Dual Class-A 2 "DCAO2" Schematic.

Circuit Description

The output stage is identical to the original Dual Class-A using two BD139 NPNs, with one BD139 as a bias reference, and a single BD140 PNP.

The THAT1646/DRV134 are changed to a pair of THAT1240s in an interesting yet functionally-identical configuration.

The output of IC3 is the sum of signal plus bias.
The output of IC4 is the sum of signal minus bias.

Thus, the outputs driving the transistor bases move in common mode in relation to input signal but the Vbe reference is mirrored differentially around signal.

After this was drawn I realized that it is the exact same circuit block as the MS matrix: https://www.proaudiodesignforum.com/for ... p?f=6&t=15

Note that although the outputs can be run open loop, the DC bias is closed loop (nested) with thermal feedback from the heatsink and local DC feedback around the THAT1240s.

The original motivation was to eliminate the internal build-out resistors in the THAT1646 or DRV134.

Op amps were tried first and what became obvious is that gain error in the common mode channel (the signal path) has a large effect on even-order distortion.
As it turned out the effect of the build-out resistors was minimal.

Although a trim could be employed, the THAT1240's gain error, 0.05 dB max, is so far superior to the THAT1646 or DRV134 or 1% resistors that using them eliminated the requirement to provide the trim.

Description

Input is to a THAT1240 or THAT1246 depending on the application.
J1 Allows the board's two inputs to be connected in parallel or cross-coupled.
J2 is for applications where a balanced input is not required and the TAHT124X is eliminated.

Output from IC1 (or the input connector) flows to an optional Level pot.
The level pot can either be an external control or a board-mounted trim pot.
J3 bypasses the connections to the pot if it is not used.

J4 allows the output stage to be driven directly from the balanced line receiver or the gain stage.
When J4 is in the A position, the gain stage is in circuit and global feedback can be used.
The B position of J4 takes signal from the balanced line receiver. In this mode the output stage runs open loop similar to the original Dual Class-A.

IC2A provides gain and permits optional global feedback to be introduced.
Input AC coupling is optional.
I'm showing a "Turbo NE5532" - an NJM2114 - in this location but a 5532 works just as well.
In fact a number of op amps do but I wouldn't suggest substituting them willy-nilly unless you have a 'scope to check stability.
Tuning involves changing the 10 pF locally around the NJM2114.
The op amp stage needs to be slower than the output when global feedback is used. See: viewtopic.php?f=12&t=13

J5 selects the feedback option.
When J5 is in the A position, global feedback is introduced.
If an open loop output stage is desired with the gain stage active, J5 should be jumpered in the local feedback B position.
This closes the loop locally around IC2A.

The original Dual Class-A typically had output offsets less than 10 mV, typically less than 5 mV.
Due to the introduction of a gain stage which can be DC-coupled, an offset trim was added.

IC2 is an active attenuator providing a zero Ohm output impedance DC trim voltage.
The offset trim can be injected into one of two points depending on whether the gain stage is used.
J6A should be jumpered to IC2A when it is used.
If the output is driven directly from IC1, then J6B should be connected to the offset trim stage.

A build-out resistor should always be used in Line Amp applications whether global feedback is used or not.
The build-out should be 47-50 Ohms.

For open loop outputs a build-out is not required for stability and isolation from capacitive loads.
The open loop output impedance is less than 2 Ohms and the maximum capacitive load is determined by output current and device dissipation.

Closed loop operation requires careful consideration of the build-out value to isolate the output from external load capacitance.
For headphone applications, where the anticipated cable and load capacitance is low, it can be 0 Ohms.
However there is no short circuit protection so the choice of 0R or 10-47 Ohms is up to the builder.

This is an FFT at 1.5 Watts into 15 Ohms.

Image
Dual Class-A 2 DCAO2, Rload = 15 Ohms, 1.5W Output, Red is Input Generator Monitor Offset by 12dB, Green is Output.

The generator output, in Red, is offset by 12 dB to compensate for the 12 dB gain of the output stage. The THD of the source is 0.0024%, the THD of the DCAO2 at 1.5W into 15 Ohms is 0.0025%.
You can see the output THD-3 in green barely "peeking" above the tip of the generator's 3rd harmonic.
Without null testing its hard to see but the THD contribution at 1.5W appears to be 0.0001%.

The output Iq is 90 mA to provide heavy Class-A current but overall NFB is applied to extend output power from the 100 mW level to >2.5W level.
Thus an already excellent 100 mW line and headphone amp can have its power extended up to 1-3 Watts and still operate in full Class-A at lower power levels.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by JR. »

That looks very respectable... A little weirdness in the top octave where some distortion in the source appears to cancel out?

I am tempted to build a proper sound system for my TV... Right now I'm using a glorified computer speaker system (has a sub) but sometimes I hear problems with the bass... sounds like control voltage feedthrough, perhaps the computer speaker has some compression in it so a crappy or misadjusted gain element? Or perhaps the audio coming from the satellite is corrupted.

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: An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

That looks very respectable... A little weirdness in the top octave where some distortion in the source appears to cancel out?
Thank you. I think some of that hash above 10 kHz may be the shielding (or lack therof) owing to the liberal use of clip leads.

There is some small THD "reduction" which is a cancellation. I need to do a null test.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by Gold »

Looks great Wayne. I'll have to take a closer look at that gain stage. There is a lot to take in.

I was holding off using this for a line amp until I could figure out a gain stage. This of course is far better than anything I could come up with.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

I had the unit hooked up for listening tests, enjoyed it a lot, then gave the ears a rest for a few days.

I decided to take a couple of measurements using squarewaves...

I changed the local Cfb around the NJM2114 to 100 pF to reduce the squarewave overshoot.
I wasn't sure from my original build if that was a 10 pF or a 100 pF.
Turned out to have been 100 pF.

I then checked the Slew Rate and capacitive load stability with 0 Ohm build-out and feedback applied.

The capacitive load stability is more important in headphone applications than in line output applications.
In line amp outputs the build-out should be about 50 Ohms so load capacitance is not a stability factor.
In headphone amps some people do not want build-out resistors.
I prefer some just to provide short circuit protection but with the potential of no build-out resistor and feedback applied just how tolerant of capacitive load is it?

Slew Rate

The slew rate is 10 V/µs measured from the 1240 balanced input to the load including the gain stage.

Maximum Load Capacitance

The capacitive load stability with 0R build-out looks like it is tolerant of loads in excess of 10 nF.
1 nF is probably enough for 30 foot headphone cables; at 10 nF there was a little waveform abberation but no sign of oscillation.
At 100 nF there was the onset of damped ringing.
I think we're OK in the maximum capacitive load department.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by JR. »

I don't like to think in terms of slew rate, since that is an overload condition. Circuits can be gamed to exhibit very fast slewing during overload, and that has value for sample and hold circuits to get from here to there faster as long as recovery time is brief. For linear audio, slewing means you failed to properly LPF the front end or the circuit is indeed too slow. :oops:

Rise time, and/or power bandwidth are linear operational metrics. When you said square waves, I expected rise time.

JR

PS I've been kicking the tires on Bruno Putzeys class D modules but 70EU per channel has me in sticker shock (for 6 channels of audio). They also sell a switching PS module so it should be like lincoln logs or lego blocks, just not my idea of cheap.
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: An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

Almost sorry I brought it up.
The point is, at 10V/µs, it's fast enough for line amp and headphone applications.

Edit: Looks like the power bandwidth is about 127 kHz based on a Vpeak (max) of 12.5V.
As a point of comparison an NE5532 has a power bandwidth of about 110 kHz.

For more information on Slew Rate and how it relates to Power Bandwidth see: viewtopic.php?f=12&t=800
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

I tuned the local Cfb to 22 pF from 100 pF.
22 pF makes the overshoot just about critically-damped; 100 pF is overdamped.
(Measured with an NJM2114 gain stage from balanced input to power output, closed loop, gain of 12 dB.)

The rise time is about 250 ns.
The small signal -3dB point is around 1.4 MHz.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by mediatechnology »

The overshoot that I was seeing turns out to be peaking in the THAT1240 line receiver at the input and in the pre-drivers.

Not much can be done about that.
Limiting the rise-time with an RC network at the NJM2114 input could be used, but that's complicated by the potential of a pot or trim between the line receiver and dual op amp .

Local feedback around the gain stage reduces rise time, makes sure that the 2114 error amp is slower than the output, and provides stability for load capacitance.

Some empirical values for the local Cfb:

10 pF is enough to maintain overall stability.
For capacitive loads with 0R build-out it should be larger.
22 pF around the NJM2114 provides stability for Cloads > 1 nF.
100 pF Cfb reduces the overshoot and is stable with a Cload of 100 nF.
The 100 pF value filters the slight overshoot of the input and limits the rise time to the pre-driver.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

Re: An Improved Dual Class-A Output: The "DCAO2"

Post by Gold »

JR. wrote: PS I've been kicking the tires on Bruno Putzeys class D modules but 70EU per channel has me in sticker shock (for 6 channels of audio). They also sell a switching PS module so it should be like lincoln logs or lego blocks, just not my idea of cheap.
As I have recently found out prefab is much more expensive than DIY. It's a strong dollar. Now is the time to buy if you are. They do sound great. I guess it depends how busy you are and how much your time is worth to you. All you have to do is the casework which might not be your idea of fun.

OT: After getting a quote on a nice prefab cabin I've started researching house construction. Geez, framing doesn't seem too difficult or time consuming for a one story not too big structure. As I've been telling everyone, I seem to be incapable of doing things the easy way.
Post Reply