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
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

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

Post by JR. »

Those press in prototype fixtures have a couple pF of capacitance between adjacent pins (due to the internal construction).

I got bit once prototyping a digital circuit (not me but a tech working for me did). The single digit pF between pins caused an undesirable gate delay and upset some ram select lines timing.

So you may have some accidental capacitance around the circuit.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5442
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 »

So you may have some accidental capacitance around the circuit.
So true.
The capacitance between pins is about 2 pF.
What I've found is that they tend to make unstable circuits more so.
If it's borderline it will usually show up.

In this circuit the impedances around the pre-driver are low. (1K which is a tad too low IMHO.)

One trick I use - not on ICs - is to separate nodes with a grounded strip as a guard ring.
I've done that on mic preamp input networks.
The one to watch out for is coupling between the inverting and non-inverting inputs.

Contact resistance is also something to watch out for as well as those protoboards that are not shielded with a metal, grounded baseplate.
Sometime in the 1990s Global Specialties stopped bonding the black binding post to the baseplate, went to cheaper pins and plastic.
I like the old ones.
I discovered the floating baseplate while troubleshooting an SPI interface which had crosstalk.
Once the baseplate was grounded it cleaned up.
The one in the pic is plastic.

If you know their limitations you can usually work around them on audio and low frequency digital circuits.
They're sure a time-saver and if a circuit winds up over-compensated due to the stray capacitance it can always be reduced - or eliminated - once it is laid out.
But, I usually find the protoboard translates pretty well to the layout.

If I can get 0.00X% THD on the protoboard with this circuit I'm pretty sure I can pull it off on a real PC board.
User avatar
mediatechnology
Posts: 5442
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 wanted to post an updated op amp-based pre-driver schematic.

Image
DCAO2 Op Amp-Based Pre-Driver Circuit Fragment.

The gain stage/error amplifier is IC1A.
Local HF AC feedback is provided by the 10 pF capacitor.
Global feedback is from the output.
If the gain stage is not used as an error amplifier for the output, feedback is taken from IC1A's output.

IC2A and IC2B are the pre-drivers which replace the THAT1240 in the original circuit.
The Vbe reference is added to signal at the non-inverting input of IC2A.
The output of IC2A drives the output NPN transistor.

IC2B subtracts the Vbe reference from the input signal.
It's output drives the PNP output transistor.

The 200 Ohm "AC Balance" trim pot adjusts the open-loop even-order distortion.

The circuit impedances are low and the resulting signal currents relatively high.
To improve linearity the 1K feedback resistors were increased to 2K.

IC1A Buffers the signal input; IC2B provides a "zero Ohm" source impedance to prevent the input signal from coupling back into the Vbe reference/Vbe multiplier (not shown).
The AC performance of IC2B must be good in order to maintain a low output impedance versus frequency.

The 1k feedback resistor in series with IC1B's inverting input should be used with NE5532/2114 family op amps because the inputs have back-to-back diodes.
The resistor limits fault current into the input and are not required for BIFETs or the OPA1612.

The 10R output resistors in series with the transistor bases may or may not be required.
These "base stopper" resistors seem to be required with the OPA1612 which has rail-to-rail common emitter outputs.
Near the clipping point, the OPA1612 will briefly oscillate as it begins to hit the wall.
Adding the resistors, cleans it up and do not seem to affect open or closed loop linearity.
When the PC board is laid out and the signal lengths shorten they may not be required.

IC2 should not be a 5532/2114 family part due to the aforementioned response/phase "kink."
The OPA2134, OPA2604 and the OPA1612 all work great.
The OPA1612 is excellent in both IC1 and IC2 locations.

IC1 can be an NJM2114, 5532 or OPA1612.
The bipolar input parts seem to work slightly better (lower THD) than the BIFETS.

I'll draw a complete schematic and post it soon.
User avatar
mediatechnology
Posts: 5442
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've revisited this project and have rebuilt the protoboard.
Right now I'm doing listening tests and hope to post a schematic in a few days.
User avatar
mediatechnology
Posts: 5442
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 »

In between outdoor projects I've been tweaking the design of the Dual Class-A V2.0.

As promised I'm posting the schematic and some test results.


Image
Dual Class-A "DCAO 2" Headphone and Line Amplifier Schematic

Circuit Description

Input is received by a THAT1240/1246 balanced line receiver.
Jumper J1 allows the second channel to be driven by the first channel either in or out of polarity.

J2 bypasses the THAT124X when balanced inputs are not required. When unbalanced, the line receiver IC is not inserted.
J3 selects either an on-board trimmer (T), fixed gain (F) and variable (V) using a front panel level control.

AC coupling is optional.
I'm considering a dual footprint for a 47 uF bipolar electrolytic or a 2u2 film with the bias resistor and op amp selected accordingly.

The output stage can run either open loop or closed depending on the jumper positions of J4 and J5.

For outputs up to about 65-100 mW, the heavy Class-A output current provides nearly identical performance open or closed.

Op amp B in the first group provides 12 dB gain and when feedback is global (output closed loop) serves as the error amplifier.

Jumper J4 bypasses op amp B to provide direct connection to the pre driver. This is ideal for unity gain line driver applications <100 mW.

When J5 is set in the local feedback position, op amp B provides gain.
When J5 is set in the global position feedback is taken from the output to extend the output power (in Class A/B) up to about 3W.
The 10 pF capacitor around op amp B assures that the error amp is slower than the output.

Op amp A and Q2 are the Vbe bias reference.
Q2 is mounted on the heatsink to provide thermal feedback.
Op amp B buffers the Vbe reference due to the high signal currents in the pre driver.

Op amps A and B in the second group emulate the THAT1646/DRV134 in the original Dual Class-A.
Op amp A sums the input signal and Vbe to drive the NPN output.
Op amp B differences the input signal and Vbe to drive the PNP output.
With no signal the NPN sees +Vbe; the PNP sees -Vbe.

The 200R trim balances the input drive signal of Q3 to match Q1.
This minimizes even-order THD contributions.
To adjust the trimmer J5 is set to "open loop," the output is loaded at about 100 mW, and the even order distortion nulled.

The 33R base resistors isolate the transistor base capacitance from the op amp outputs.
It was found that the OPA1612 and LME49720 benefited from having them.
Currently I'm using BD139/140s which have unspecified, but high, Cob. (Estimates run 100 pF).
The KSA1381/KSC3503 (2SA1381/2SC3503) at <3 pF might allow the 33Rs to be reduced somewhat.

At levels below 100 mW feedback isn't doing much.

The quiescent current is 100 mA and the output, even when open loop, is highly linear class-A.

The following FFTs overlay the input (in green) over the output in red.
The vertical scale of the input has been adjusted to offset the +12 dB gain and attenuation in the analyzer input.

The first FFT is 65 mW at 10 kHz. The THD is estimated to be 0.0002%

Image
Dual Class-A 2 "DCAO2" 65mW 30R Load 10kHz Red is Output Green is Generator Monitor

The next FFTs are at 1W at both 1 and 10 kHz:

Image
Dual Class-A 2 "DCAO2" 1W 30R Load 1 kHz Red is Output Green is Generator Monitor

The THD signature in the above shows that the DCAO2 is cancelling a bit of second harmonic. The THD is estimated to be 0.0004%

Image
Dual Class-A 2 "DCAO2" 1W 30R Load 10 kHz Red is Output Green is Generator Monitor

The THD at 10 kHz at 1W is estimated to be about 0.0038%.

I've spent a lot of time making the circuit work with a wide variety of op amps.
I don't recommend the NE5532/NJM2114 for the second group of op amps due to the "kink" in its gain/phase response.
When two 5532/2214 are cascaded inside the feedback loop there is significant response peaking.
Bandwidth is king for low THD in the second stage.
Op amps not having this kink, which are most, FETs and many BJT inputs work fine.

The OPA2134, OPA2604, OPA1612, OPA1642 and LME49720 all work very well.
My favorite is the OPA1612 with the LME49720 a close second.
User avatar
mediatechnology
Posts: 5442
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 decided to re-check the DCAO2's closed loop stability with output load capacitance.
It can drive 1 uF without oscillation.
Headphone cables should be fine...

So I decided to try driving speakers with the DCAO2 output operating closed loop.
The Iq is about 90 mA.

By doing a null test between the input and output using the 'scope I can see where it begins to "break."
I'm using real music for test monitoring both output at the speaker terminal and the input then subtracting in the 'scope front end.

The DCAO2 is limited by power supply and 1R collector resistors but the null "breaks" somewhere around 20V p-p into 8R.
At 8 Ohms, with my limited 500 mA supply it looks like it can make 5-6 Watts easily.

For headphones and line amps I think we've got plenty of headroom. :lol:
User avatar
mediatechnology
Posts: 5442
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've been spending the last few days putting the finishing touches on the layout for the DCAO2 and submitted the prototype run on Thursday.
I should have boards by Tuesday to check out.
My eyes are tired from doing layout. :roll:
User avatar
mediatechnology
Posts: 5442
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 »

DCAO2 Dual Class-A CCIF (ITU-R) IMD 19/20 kHz 1:1 Distortion Measurements

Image
DCAO 2 Dual Class-A CCIF ITU-R 19/20 kHz Distortion

Green is the Left Channel loaded in 30R at 750 mW. 1 kHz side bands are about -90 dB.
Red is the Right Channel unloaded.

The DCAO2 was using global feedback. Rbuildout was 0R.
Gold
Posts: 677
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

That's great. I really need a good 'bump box' for -10dBv to +4dBu conversions. I think I may include a headphone amp. It could be a little transfer station front end for cassette transfers and the like at less than the studio rate.
User avatar
mediatechnology
Posts: 5442
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 »

Thanks Paul.

I've made it so that the output can run open loop like the original Dual Class-A or closed loop.
With the high idle current it uses to maintain Class-A and good open loop THD, feedback doesn't start providing distortion reduction until 65-100 mW.

The one advantage closed loop operation has at low power levels is very low output impedance.

The open loop output impedance measures about 1.9 Ohms 20-20 kHz.
Closed loop measures 38 milliOhms at 100 and 1 kHz rising to 77 milli Ohms at 10 kHz.

I added a jumper to shunt the Build-out resistor so I can make quick comparisons.
This lets people jumper it for "0" Ohm outputs, <2 Ohm in "open loop" or a fixed Ro >2R.

My last round of listening tests were done with closed loop "0" Ohm outputs (vs 33R).
I'm really pleased at the tighter low end I get with the Sony MDR-7506s.

Though I've always suggested using some build out to provide short-circuit protection the 1R collector resistors and current-limited power supply provide some short duration tolerance to dead shorts.
Post Reply