| Russ White |
I am a complete amateur, but I am excited to show what I think is a very good amplifier. It is designed as an operational amplifier, and can be scaled upward and downward for both current and voltage. I opened this thread because I thought it might be educational to look at the design along with its challenges. I have learned a lot from such design threads, and I want to start some conversation on the approach.
I have been debating with myself if this circuit is super symmetrical or not. To be honest, I am not really sure. I would ask Nelson to please clarify for me if it is or not. :) I do believe it is fully symmetrical, and that the summed balanced output has significantly less distortion than each individual output when referenced to GND (though even that is quite low).
Let me explain the motivation for the circuit. There are two major reasons I designed it.
1) I wanted a discrete I/V for my balanced PCM1794 DAC currently going by "COD".
2) I wanted a fully symmetrical op amp which could swing high enough voltage to directly drive power amps such as Nelson's F4.
Here are some key features:
1) The output common mode voltage can be nulled to zero via the error amp section. R34/R35 are a 1K pot used to set the null point. The input common mode voltage is not present on the output.
2) The balanced output distortion should be very low and noise as well if suitable devices are used.
3) It should work equally well with current source or voltage source inputs.
4) It actually works. :) I have tested it.
5) Its scalable. You could easily produce a Class A or Class A/B power amplifier with the same general scheme.
So please Nelson or anyone please give me some feedback on the design and its potential merits and pitfalls.
I hope someone gets something useful out of it.
Cheers!
Russ |
|
|
| AndrewT |
| quote: | Originally posted by Russ White
I am a complete amateur, but I |
show copyright on the schematic.
Just in case I want a feed a few lawyers to argue over whether someone has copied my creation? |
|
|
| Russ White |
| quote: | Originally posted by AndrewT
show copyright on the schematic.
Just in case I want a feed a few lawyers to argue over whether someone has copied my creation? |
I only put the notice to recognize my very real work. Nothing more. Honestly I have not invented anything to be sure. |
|
|
| Nelson Pass |
| quote: | Originally posted by Russ White
I have been debating with myself if this circuit is super symmetrical or not. |
If you apply symmetrical feedback to Q1 and Q2, then the
answer is yes.
:cool: |
|
|
| Russ White |
| quote: | Originally posted by Nelson Pass
If you apply symmetrical feedback to Q1 and Q2, then the
answer is yes.
:cool: |
Well that settles that. :) Thank you Nelson. Well in this respect then it clearly falls under your SuSy patent and I will give you due credit. :)
Now the most important question, does it seem to be any good? :cool:
Cheers!
Russ |
|
|
| Nelson Pass |
Why don't you build it and find out?
:cool: |
|
|
| Russ White |
| quote: | Originally posted by Nelson Pass
Why don't you build it and find out?
:cool: |
I have already built a prototype but with different devices(2N5550 and 2N5401) and a slightly different error amp section it was very similar. It worked well, but I only did one channel as a sanity check.
I will build this circuit and report my results.
Thanks for your encouragement. :) |
|
|
| regal |
I will be following this closely. There is a huge void in decent analog stages for these new damned differential out DAC's. IC opamps have been the only solution in the DIY world
Really only PS Audio as a discrete stage for the PCM179x and it isn't the best.
Thankyou for working on this. |
|
|
| fierce_freak |
| Looks nice, Russ :) |
|
|
| Nelson Pass |
Considering that you are looking at this for I/V, you might want
to consider JFETs for the gain devices. My experience is that
they are much more immune to digital noise.
:cool: |
|
|
| Russ White |
Thanks folks.
I have been playing with the circuit today. I have tried a few things:
1) CFPs on the buffers instead of emitter followers.
2) Discrete darlingtons using BD139/MJL4281A(x 2) to scale it up to 100W output or so. :)
Some nagging issues I still need answers to.
1) Will the diamond buffer type output stage stay thermally stable? Or do I risk thermal runaway using BJT output devices.
2) Is there a more effective way to compensate the amp than I have chosen?
3) The error amp section is a bit tricky, especially as you scale the amp up. I think I have a decent solution that.
I will post some schematics/simulations soon.
Cheers!
Russ |
|
|
| Russ White |
| quote: | Originally posted by Nelson Pass
Considering that you are looking at this for I/V, you might want
to consider JFETs for the gain devices. My experience is that
they are much more immune to digital noise.
:cool: |
Thanks, I had thought about that too, but was wondering if a JFET front end would limit my ability to scale the amp voltage swing upwards more than I want to. I suppose I could cascode them?
Cheers!
Russ |
|
|
| MASantos |
Russ, not sure if you've seen this. Check out this hread at headwize.com and this suport site. It's a discrete diamond buffer design with JFET input stage and BJT output stage developed by Ti Kan (AMB) and several DIYers. Maybe you can get some ideas from it for your I/V opamp.
Regarding your design, when you do the circuit board. I would provide pads to bipass the VAS stage and use it only as a buffer. This would be good to buffer loop outs or phono stages. Think about it.
Manuel |
|
|
| regal |
The issue with the Jisbos or others is the balanced to unbalanced transfer that is required to get the most out of the PCM1794.
That is why this project is so important. |
|
|
| Russ White |
So what do you call a super symmetrical symmetrical amp?
SuperDuperSymmetrical?
Something like this:
:cool: |
|
|
| Russ White |
.step f=1000
Fourier components of V(out+,out-)
DC component:-3.27687e-010
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 4.000e+00 1.000e+00 -32.39° 0.00°
2 2.000e+03 1.342e-09 3.354e-10 95.94° 128.33°
3 3.000e+03 4.704e-07 1.176e-07 -99.82° -67.43°
4 4.000e+03 1.200e-09 3.000e-10 -112.95° -80.56°
5 5.000e+03 4.092e-09 1.023e-09 15.99° 48.38°
6 6.000e+03 1.012e-10 2.529e-11 125.34° 157.73°
7 7.000e+03 8.622e-09 2.156e-09 141.21° 173.60°
8 8.000e+03 7.298e-10 1.825e-10 -45.90° -13.51°
9 9.000e+03 7.516e-09 1.879e-09 79.39° 111.78°
Total Harmonic Distortion: 0.000012%
.step f=5000
Fourier components of V(out+,out-)
DC component:-6.20175e-010
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 5.000e+03 4.000e+00 1.000e+00 -161.95° 0.00°
2 1.000e+04 6.723e-10 1.681e-10 -129.45° 32.50°
3 1.500e+04 4.544e-07 1.136e-07 -139.38° 22.57°
4 2.000e+04 8.950e-10 2.238e-10 4.50° 166.45°
5 2.500e+04 1.143e-08 2.857e-09 -64.66° 97.29°
6 3.000e+04 1.257e-09 3.142e-10 -147.05° 14.90°
7 3.500e+04 1.071e-08 2.678e-09 12.83° 174.78°
8 4.000e+04 1.754e-09 4.385e-10 3.39° 165.34°
9 4.500e+04 3.851e-09 9.628e-10 39.78° 201.74°
Total Harmonic Distortion: 0.000011%
.step f=10000
Fourier components of V(out+,out-)
DC component:-9.36479e-011
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+04 4.000e+00 1.000e+00 36.10° 0.00°
2 2.000e+04 1.998e-09 4.996e-10 -41.23° -77.33°
3 3.000e+04 4.448e-07 1.112e-07 80.07° 43.97°
4 4.000e+04 1.483e-09 3.707e-10 132.02° 95.92°
5 5.000e+04 1.637e-08 4.094e-09 -163.38° -199.47°
6 6.000e+04 5.747e-09 1.437e-09 -119.47° -155.57°
7 7.000e+04 2.426e-08 6.064e-09 -39.62° -75.71°
8 8.000e+04 7.066e-09 1.767e-09 -80.97° -117.07°
9 9.000e+04 4.998e-09 1.250e-09 -4.83° -40.93°
Total Harmonic Distortion: 0.000011%
.step f=20000
Fourier components of V(out+,out-)
DC component:1.97193e-009
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 4.000e+00 1.000e+00 72.20° 0.00°
2 4.000e+04 1.220e-08 3.051e-09 139.37° 67.17°
3 6.000e+04 4.752e-07 1.188e-07 155.16° 82.97°
4 8.000e+04 1.762e-08 4.406e-09 -87.29° -159.49°
5 1.000e+05 4.120e-08 1.030e-08 13.55° -58.65°
6 1.200e+05 3.689e-08 9.223e-09 72.96° 0.76°
7 1.400e+05 4.863e-08 1.216e-08 -167.45° -239.64°
8 1.600e+05 4.430e-08 1.108e-08 -145.67° -217.86°
9 1.800e+05 4.560e-08 1.140e-08 -47.79° -119.98°
Total Harmonic Distortion: 0.000012% |
|
|
| Russ White |
| The is the FFT with 8vpp output 20khz |
|
|
| BrianDonegan |
| Funny. I have a bunch of the 970/2240s and was thinking of building one of the prototype with them... |
|
|
| Russ White |
Pretty cool stuff! :)
Still there are some design features that this amp delivers...
1) It nullifies(very close to zero) common mode DC offset. This can actually be tweaked to bring to zero.
2) It features symmetrical folded cascodes which improves the open loop gain a lot.
3) I designed it myself. :)
Cheers!
Russ |
|
|
| regal |
| quote: | Originally posted by Russ White
Pretty cool stuff! :)
Still there are some design features that this amp delivers...
1) It nullifies(very close to zero) common mode DC offset. This can actually be tweaked to bring to zero.
2) It features symmetrical folded cascodes which improves the open loop gain a lot.
3) I designed it myself. :)
Cheers!
Russ |
#1 is the key. The key to finally reaping the benefits of the PCM179x Dac's. A discrete solution has not been publicly shared in the DIY world.
This is a big deal.
How do I get on the waiting list? |
|
|
| ejaouen |
I am also on the waiting list :
discrete totally balanced (differential ins and outs) I/V stage for the COD dac (pcm1794) will be awesome I am sure !
Also is there a way (resistor value choice) in this circuit to select output voltage range ? I need 1 to 2 Vrms max balanced output range depending on the power of the following power amp (active 3 way system).
waiting for that I/V stage to happen. |
|
|
| Russ White |
It is in the works. :) Will have something I am hoping to have it laid out this week.
I am laying out the first of two versions, both I have high hopes for (at least they simulate well even with very nasty loads). They are both folded cascode circuits, but one is quite a lot more complex.
Cheers!
Russ |
|
|
| ejaouen |
Any news on this side of discrete active balanced I/V stages ?
Also do they include anti-image low pass filter solutions ?
thanks |
|
|
| Russ White |
| quote: | Originally posted by ejaouen
Any news on this side of discrete active balanced I/V stages ?
Also do they include anti-image low pass filter solutions ?
thanks |
It definitely coming along.
Yes, the PCB will support optional low pass filter caps. |
|
|
| ejaouen |
Thanks for the reply
Did you already hear some of your discrete active I/V prototypes (with the cod) ?
If so do you think this discrete way is going to be worth it over the ths4131 I/V solution ? |
|
|
| PigletsDad |
My take on a supersymmetric discrete op-amp has been up at http://www.battletonphoenix.co.uk/linebal.pdf for some months.
Pros and cons:
1) My circuit does not have complementary symmetry. I reasoned that once you introduce one sort of symmetry, and cancel all the even harmonics, additional forms of symmetry offer minor benefits, but put the parts count up.
2) Limited output drive in my version - it doesn't have emitter follower stages.
I have some simulation results somewhere but not to hand. I will look them out to compare with Russ; I expect his circuit will have much lower distortion, but mine is low enough.
I suspect they are at different points on the complexity versus simulated performance trade-off curve. |
|
|
| Russ White |
| quote: | Originally posted by ejaouen
Thanks for the reply
Did you already hear some of your discrete active I/V prototypes (with the cod) ?
If so do you think this discrete way is going to be worth it over the ths4131 I/V solution ? |
I have tested it on protoboard, it does work very well, it just a matter of finalizing the PCB.
It is very comparable to the THS4131(it is a very similar circuit) in terms of performance, and actually a bit better at driving low impedance loads. It actually makes a wonderful headphone amplifier even for 16-25ohm cans. |
|
|
| Russ White |
| quote: | Originally posted by PigletsDad
My take on a supersymmetric discrete op-amp has been up at http://www.battletonphoenix.co.uk/linebal.pdf for some months.
Pros and cons:
1) My circuit does not have complementary symmetry. I reasoned that once you introduce one sort of symmetry, and cancel all the even harmonics, additional forms of symmetry offer minor benefits, but put the parts count up.
2) Limited output drive in my version - it doesn't have emitter follower stages.
I have some simulation results somewhere but not to hand. I will look them out to compare with Russ; I expect his circuit will have much lower distortion, but mine is low enough.
I suspect they are at different points on the complexity versus simulated performance trade-off curve. |
That is a simple elegant circuit for sure. :) Nice work.
Mine is quite a lot different though in many respects.
In practice I personally have found my circuit not very useful without the output stages. But i have thought about several options here, including ICs such as NS new buffer chip, or one of TIs buffer chips. Wrap those in the feedback and you should have a great amp.
BTW, I probably will not use complementary symmetry, that was just a whim. In reality it does not yield enough benefit for the added complexity/cost.
I will post some details of the amp soon.
Cheers!
Russ |
|
|
| PigletsDad |
I suppose I should add that I assume that Nelson's patent applies to my circuit as well, in those territories where the patent has been granted.
It was certainly designed with Nelson's ideas very firmly in mind, as the simplest discrete circuit I could think of that gave DC accuracy and the supersymmetry principle. |
|
|
| Russ White |
Here is the circuit as I am laying it out now.
It has short circuit and SOA protection.
It is thermally stable thanks to the CFP output.
The distortion should be very very very low. :)
The PDF is of the simulation, reference designators and such are not finalized.
It is a simulation of the circuit in I/V mode with filter caps for a Fc of about 200khz.
The circuit as shown would work with PCm1794A and output 2VRMS peak output.
Cheers!
Russ |
|
|
| Russ White |
here are the simulated THD number into the 300R/300pf load:
The first is the balanced THD, The second is the THD of a single output to GND.
Heightened Def Con from 0.000256019 to 0.000256019
Fourier components of V(out+,out-)
DC component:8.29446e-010
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.770e+00 1.000e+00 -5.68° 0.00°
2 4.000e+04 7.463e-10 2.694e-10 -20.80° -15.12°
3 6.000e+04 5.798e-08 2.093e-08 69.54° 75.23°
4 8.000e+04 7.130e-10 2.574e-10 -126.70° -121.01°
5 1.000e+05 1.230e-08 4.440e-09 149.26° 154.95°
6 1.200e+05 1.214e-09 4.382e-10 -63.77° -58.09°
7 1.400e+05 8.478e-09 3.061e-09 -50.92° -45.24°
8 1.600e+05 1.409e-09 5.087e-10 50.66° 56.34°
9 1.800e+05 3.645e-09 1.316e-09 -29.03° -23.35°
Total Harmonic Distortion: 0.000002%
Fourier components of V(out+)
DC component:-0.000220695
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 1.385e+00 1.000e+00 -5.68° 0.00°
2 4.000e+04 6.406e-07 4.626e-07 -15.68° -9.99°
3 6.000e+04 2.908e-08 2.100e-08 70.81° 76.49°
4 8.000e+04 1.410e-09 1.018e-09 139.62° 145.31°
5 1.000e+05 5.837e-09 4.215e-09 152.53° 158.21°
6 1.200e+05 1.175e-09 8.487e-10 -79.13° -73.44°
7 1.400e+05 4.244e-09 3.064e-09 -43.23° -37.55°
8 1.600e+05 1.138e-09 8.217e-10 14.40° 20.09°
9 1.800e+05 1.647e-09 1.190e-09 -40.27° -34.58°
Total Harmonic Distortion: 0.000046% |
|
|
| Russ White |
20 khz, same load
Green trace is balanced output.
Blue is SE. |
|
|
| ejaouen |
Are these discrete balanced I/V (and LPF) stages going to be available soon ?
From your trials of these : are they worth it over THS4131 opamps IVY stages ?
I will place an order of COD modules and wonder if discrete balanced I/V stage is coming.
thanks in advance |
|
|
| Russ White |
| quote: | Originally posted by ejaouen
Are these discrete balanced I/V (and LPF) stages going to be available soon ?
From your trials of these : are they worth it over THS4131 opamps IVY stages ?
I will place an order of COD modules and wonder if discrete balanced I/V stage is coming.
thanks in advance |
Hello,
Yes they are coming at some point. We just have a lot to do right now.
Measurement wise I doubt it will be much if any better than the THS4131, it may actually not be quite as good, but it will have more load current capacity, and it is discrete which will make some people happy. :)
I will have one measured and report when the project is finished.
Cheers!
Russ |
|
|
| Blitz |
| Hi Russ, very nice indeed...any progress on this one ? |
|
|
| Russ White |
Thanks Blitz,
I am testing a proof of concept output stage which I plan on applying to this circuit as well. If that goes well then I will proceed with this amplifier.
Thanks for your interest. :)
Cheers!
Russ |
|
|
|
