A Lateral MOSFET Power Amplifier Design (second try)

[Sevy]

What is idle OS current in your sim? In mine Rv2 is set to 222 Ohm. Idle current is 270mA, most likely that's why my squares look different.

Good Morning,
I started again from my clean asc file and redo all the modifications you did in your. I revised all the values and unless I miss some, they are all identical. If I measured your MOSFET current (that's the Iddle current right?) I get:

Id(Q504): 0.274714 device_current​

Ig(Q504): -2.40841e-11 device_current​

Is(Q504): -0.274714 device_current​

Id(Q604): -0.278794 device_current​

Ig(Q604): -5.03482e-11 device_current​

Is(Q604): 0.278794 device_current​

Now I get similar on my simulation:

Id(Q504): 0.274714 device_current​

Ig(Q504): -2.40841e-11 device_current​

Is(Q504): -0.274714 device_current​

Id(Q604): -0.278794 device_current​

Ig(Q604): -5.03482e-11 device_current​

Is(Q604): 0.278794 device_current​

And the square waves are also identical! I guess I was tired yesterday after more than 10 hours on this... The mornings are much better for my concentration! lol.

Thanks again...

​

Here are my files.

 

[Sevy]

Interesting how the Square Pulses are affected when I add the D501/D601 protection circuits...

Without Protection:

With protection:

 

[Sevy]

I think I am ready to layout the PCB!

The data sheet of the Exicon ECX10P20 (ECX10N20 similar) say:

Vdss Drain-Source Voltage: -200V​

Id Continuous Drain Current: -8A​

Pd Allowable Power Dissipation Tcase=25C: 125W​

---

Under 8 Ohms load, Vin 1.2Vp 1kHz:

N-Period=8
Fourier components of V(out)
DC component:-0.032476

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+3 3.317e+1 1.000e+0 90.06° 0.00°
2 2.000e+3 2.715e-4 8.186e-6 67.90° -22.15°
3 3.000e+3 2.731e-4 8.233e-6 112.94° 22.88°
4 4.000e+3 9.971e-5 3.006e-6 106.15° 16.10°
5 5.000e+3 6.669e-6 2.011e-7 -145.20° -235.26°
6 6.000e+3 6.175e-5 1.862e-6 85.88° -4.18°
7 7.000e+3 5.287e-5 1.594e-6 3.60° -86.46°
8 8.000e+3 4.464e-5 1.346e-6 82.49° -7.57°
9 9.000e+3 7.105e-5 2.142e-6 1.52° -88.54°
10 1.000e+4 3.399e-5 1.025e-6 86.40° -3.66°
Partial Harmonic Distortion: 0.001254%
Total Harmonic Distortion: 0.001306%

---

Under 8 Ohms load, Vin 1.2Vp 20kHz:

N-Period=4
Fourier components of V(out)
DC component:-0.032472

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+4 2.764e+1 1.000e+0 91.19° 0.00°
2 4.000e+4 4.287e-3 1.551e-4 68.90° -22.29°
3 6.000e+4 1.474e-3 5.331e-5 166.95° 75.76°
4 8.000e+4 1.700e-3 6.150e-5 51.69° -39.50°
5 1.000e+5 5.244e-4 1.897e-5 8.88° -82.31°
6 1.200e+5 1.213e-3 4.390e-5 38.42° -52.77°
7 1.400e+5 1.299e-3 4.701e-5 -24.56° -115.74°
8 1.600e+5 7.950e-4 2.876e-5 36.24° -54.95°
9 1.800e+5 1.411e-3 5.103e-5 -38.77° -129.95°
10 2.000e+5 4.241e-4 1.534e-5 47.69° -43.50°
Partial Harmonic Distortion: 0.019708%
Total Harmonic Distortion: 0.021097%

---

Under 4 Ohms load, Vin 1.2Vp 1kHz:

N-Period=8
Fourier components of V(out)
DC component:-0.0324061

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+3 2.764e+0 1.000e+0 90.06° 0.00°
2 2.000e+3 2.318e-5 8.385e-6 102.49° 12.43°
3 3.000e+3 1.470e-5 5.320e-6 11.46° -78.60°
4 4.000e+3 3.051e-6 1.104e-6 -82.35° -172.42°
5 5.000e+3 1.234e-6 4.464e-7 -173.93° -263.99°
6 6.000e+3 1.925e-7 6.964e-8 94.53° 4.47°
7 7.000e+3 5.222e-8 1.889e-8 -172.12° -262.18°
8 8.000e+3 8.074e-8 2.921e-8 92.86° 2.79°
9 9.000e+3 6.203e-8 2.244e-8 5.44° -84.62°
10 1.000e+4 3.383e-8 1.224e-8 -86.73° -176.79°
Partial Harmonic Distortion: 0.001000%
Total Harmonic Distortion: 0.000994%

---

Under 4 Ohms load, Vin 1.2Vp 20kHz:

N-Period=4
Fourier components of V(out)
DC component:-0.0323762

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+4 2.764e+1 1.000e+0 91.25° 0.00°
2 4.000e+4 7.358e-3 2.662e-4 76.79° -14.46°
3 6.000e+4 6.592e-3 2.385e-4 174.41° 83.16°
4 8.000e+4 3.086e-3 1.117e-4 63.49° -27.76°
5 1.000e+5 3.282e-3 1.187e-4 158.13° 66.89°
6 1.200e+5 2.277e-3 8.236e-5 50.36° -40.88°
7 1.400e+5 1.322e-3 4.783e-5 139.04° 47.79°
8 1.600e+5 1.860e-3 6.729e-5 39.25° -52.00°
9 1.800e+5 5.580e-4 2.019e-5 77.74° -13.51°
10 2.000e+5 1.628e-3 5.889e-5 30.76° -60.49°
Partial Harmonic Distortion: 0.041445%
Total Harmonic Distortion: 0.046482%

---

Then I lowered the input voltage at 0.75 V for no distortion of the sine wave. I did a test at 1 kHz 2 Ohms load:

Under 2 Ohms load, Vin 0.75 Vp 1kHz:

N-Period=8
Fourier components of V(out)
DC component:-0.0321799

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+3 2.073e+1 1.000e+0 90.07° 0.00°
2 2.000e+3 5.081e-4 2.451e-5 102.55° 12.48°
3 3.000e+3 5.616e-4 2.710e-5 176.39° 86.32°
4 4.000e+3 2.172e-4 1.048e-5 103.37° 13.31°
5 5.000e+3 3.016e-4 1.455e-5 179.59° 89.52°
6 6.000e+3 1.477e-4 7.127e-6 99.16° 9.09°
7 7.000e+3 1.677e-4 8.090e-6 176.95° 86.88°
8 8.000e+3 1.128e-4 5.440e-6 96.18° 6.11°
9 9.000e+3 8.866e-5 4.277e-6 175.19° 85.12°
10 1.000e+4 9.044e-5 4.363e-6 93.48° 3.41°
Partial Harmonic Distortion: 0.004289%
Total Harmonic Distortion: 0.004495%

 

[Sevy]

And for the Power Output...

Under 8 Ohms load, Vin 1.2Vp 1kHz:

Under 8 Ohms load, Vin 1.2Vp 20kHz:

Under 4 Ohms load, Vin 1.2Vp 1kHz:

Under 4 Ohms load, Vin 1.2Vp 20kHz:

Under 2 Ohms load, Vin 0.75 Vp 1kHz:

 

[minek123]

I think I am ready to layout the PCB!

a) what idle current you will use ?
270mA idle, however it's definitely possible, seems little high for 1 pair. If you lower it, you need to verify square waves sim (at 1kHz and 20kHz)
something between 50mA and 100mA seems more reasonable. What the book says?

b) have you done stability simulations ? What is phase margin and gain margin of this amp? This is very important.

 

[Sevy]

something between 50mA and 100mA seems more reasonable. What the book says?

I remember that in the video, Bob Cordell explains that he likes to push the idle current to 200mA to allow the MOSFETs to already be at a good temperature at peak current demand. As the original circuit has 2 MOSFETs in parallel, I agree with you that 100mA should be the target.

b) have you done stability simulations ? What is phase margin and gain margin of this amp? This is very important.

I did it in the beginning when I was reading in parallel the chapter on LTSpice. But I should redo it since we modified some parts values.

This week-end, I concentrated myself on the real parts list (Mouser) and the PCB because I wanted to do something else. In my first circuit project it was important to put the OPS driver close to the MOSFET on the heat-sink. But I didn't read any advise yet in the book concerning the best placing of the pre-driver and drivers in regards to the heat dissipation. And from what I read so far, MOSFETs don't dissipate heat as the BJT do. So I am starting to think that the Dissipante 3U casing Heat-sink will be a bit to much for the requirement of heat dissipation.

 

[Sevy]

Hi,
Here's a little update...
I have finished the PCB design. I can replace parts easily if needed. It has a full ground plane on the components side but only one pad is connect to it. The ground pad of the output connector, that serve also as the High Quality Ground (HQG).

I fixed the idle current of the MOSFETs to 100 mA. I did the DC Analysis, AC Analysis, Pulse Analysis, Transient Simulation at 1.2 Vpeak and the THD at 1 kHz and 20 kHz analysis, also at 1.2 Vpeak.

DC Analysis results:

RV1 = 225 Ohms​

I(R3): 0.00400462​

​

RV2 = 358 Ohms​

I(R502): 0.0231564​

I(R602): 0.0231564​

Is(Q502): -0.100738​

Is(Q602): 0.104821​

​

V(vin) V(out)​

--------------- ---------------​

848.53 mVrms 23.455 Vrms​

1.2 Vp 33.13257 Vp​

AC Analysis :

Transient Simulation :

Is this the right calculus to get the Slew Rate.

THD at 1 kHz :

N-Period=8​

Fourier components of V(out)​

DC component:-0.0326915​

​

Harmonic Frequency Fourier Normalized Phase Normalized​

Number [Hz] Component Component [degree] Phase [deg]​

1 1.000e+3 3.317e+1 1.000e+0 90.06° 0.00°​

2 2.000e+3 2.713e-4 8.178e-6 67.33° -22.73°​

3 3.000e+3 3.824e-4 1.153e-5 157.82° 67.76°​

4 4.000e+3 1.046e-4 3.154e-6 107.60° 17.54°​

5 5.000e+3 2.281e-4 6.877e-6 -177.46° -267.52°​

6 6.000e+3 6.945e-5 2.094e-6 95.59° 5.53°​

7 7.000e+3 1.427e-4 4.303e-6 -179.47° -269.53°​

8 8.000e+3 5.072e-5 1.529e-6 88.92° -1.14°​

9 9.000e+3 9.170e-5 2.765e-6 -179.55° -269.60°​

10 1.000e+4 4.088e-5 1.232e-6 83.12° -6.94°​

Partial Harmonic Distortion: 0.001707%​

Total Harmonic Distortion: 0.001747%​

THD at 20 kHz :

N-Period=4​

Fourier components of V(out)​

DC component:-0.0330803​

​

Harmonic Frequency Fourier Normalized Phase Normalized​

Number [Hz] Component Component [degree] Phase [deg]​

1 2.000e+4 3.313e+1 1.000e+0 91.19° 0.00°​

2 4.000e+4 4.707e-3 1.421e-4 72.73° -18.46°​

3 6.000e+4 7.019e-3 2.118e-4 158.30° 67.11°​

4 8.000e+4 1.322e-3 3.990e-5 32.82° -58.37°​

5 1.000e+5 5.755e-3 1.737e-4 143.13° 51.93°​

6 1.200e+5 2.022e-3 6.102e-5 -4.04° -95.23°​

7 1.400e+5 4.062e-3 1.226e-4 135.93° 44.73°​

8 1.600e+5 2.454e-3 7.406e-5 -0.62° -91.82°​

9 1.800e+5 3.178e-3 9.591e-5 129.68° 38.49°​

10 2.000e+5 2.086e-3 6.295e-5 6.38° -84.81°​

Partial Harmonic Distortion: 0.036637%​

Total Harmonic Distortion: 0.015875%​

I shared the asc file and some captures of the PCB design.

Ground Traces Highlighted :

Feedback Traces Highlighted:

+55 VDC:

+15 VDC:

-55 VDC:

-15 VDC:

PCB Layout:

When I ordered the Dissipante 3U casing, I thought that the holes were already made on the heat-sinks based on the standard holes layout found on their website. So my first design was based on those holes dimensions.
But this option is only available with the 4U and 5U. Consequently I had to drill and tap the holes myself. After a couple of iteration of the PCB layout, I realized that I could have much more flexibility to forget those holes dimensions. But still I have one Heat-sink already drilled. So this design does use the same holes dimensions except for the two holes for the MOSFETs that will need to be drill and tapped if the PCB is used with the 4U and 5U.
And finally, the position of the Drivers and the MOSFETs are again under the PCB, but totally covered by it. Their screws will still be easily accessible via the four holes made in the PCB to let the screwdriver reach the screw.

Assembled in Hires 3D Plan view:

And some Hires 3D views :

 

[brian92fs]

It has a full ground plane on the components side but only one pad is connect to it.

A ground plane across the entire board? I'd recommend against this. For audio, I'd recommend against ground planes and use traces to better control the currents. If you do use a plane, isolate it to the small signal areas of board.

Just my $0.02.

 

[pbilous]

Looks neat.

Kind of high THD figures??? Can try swapping drivers to 2SC3503/2SA1381, maybe those will allow for less distortion.

Another suggestion is adding one more pair of output MOSFETs. You have it in your posts at the beginning of this conversation. Should help reducing distortion at 20kHz.

Have you looked into two pole compensation?

 

[jwilhelm]

Your next learning curve should be about loop areas in your layout design. + and - rails should be on top of each other so they cancel their noise. Sources should follow returns very closely. Input ground should route very closely to the feedback trace. Otherwise the traces will emit a lot of RF for the sensitive input stage to pick up. As Brian92fs suggests, ground planes are a bad idea. Better to use properly routed traces. Planes are better for high speed (digital) circuits. A single output pair amp should be tiny to minimize loop area.

 

[Sevy]

For audio, I'd recommend against ground planes and use traces to better control the currents

As mentioned , all ground traces are routed individually to the HQG. The ground plane have only one electric connection to the ground, at the output ground pad. But I am totally open to any comments.

My goal with this version is to keep all traces symmetrical with the most possible same length for the same residual resistance.

 

[Sevy]

Another suggestion is adding one more pair of output MOSFETs. You have it in your posts at the beginning of this conversation. Should help reducing distortion at 20kHz.

Have you looked into two pole compensation?

Based from others posts suggestions, I choose to use only one pair to eliminate oscillations.
I am surprised that you find distortion to be high ! Those are similar to the book original design no?

My next step is to experiment further with TPC. Also, the DC servo isn't yet on the board, but I still have the place to put it between the two +/-15 VDC supply.

 

[Sevy]

Better to use properly routed traces.

As noted above, they are! I can remove the ground plane with one click. All traces will still be connected. The ground plane is on the components side but isn't connected to any ground pad. They are all isolated from it and do route to their own trace to the ground pad HQG.
Any EMI should flow into the ground plane until they reach the HQG at the ground pad of the output connector, no?

 

[jwilhelm]

Look at you input ground trace compared to your feedback trace. Can they be moved closer together (looks like they could very easily be placed one on top of the other)? This is what I'm referring to as properly routed. Anywhere there is space between the two traces will be radiating noise.

It looks like you took the feedback source from the wrong location. It should come directly from the output inductor, not the opposite end of the output rail.

Mechanical design is as important and as tricky as the circuit design. You can easily ruin the best designed circuit by laying it out incorrectly. Thought also needs to be put into how you are going to put it into a chassis as well. A perfect board layout can be compromised easily if you need to run too much wire to hook it up. Input stage should be located away from the main transformer. Output transistors should be mounted on the heatsink slightly below center. With dual output transistors, they should be located close together to keep their temperature equivalent (not as important with lat-fets). The fun never ends!

 

[pbilous]

Based from others posts suggestions, I choose to use only one pair to eliminate oscillations.

Base stopper resistors and bypass caps at each MOSFET should be fine to stop those. You can add ~10R resistors to separate output MOSFETs power rails from the rest of the amp too.

…

Dug out my book and that is exactly what Cordell suggests.

C18,C19 bypass caps. Can safely increase those to 1uF and use film type. PCB layout needs to have the shortest ground loop length between V+ , bypass cap, transistor, GND then another bypass cap, transistor, V-, place caps as close as possible to transistor legs basically.

R45 isolation resistor.

R38, R40 base stoppers. Will need to tweak the value based on the different MOSFETs being used. LTSpice can help with that.

 

[Sevy]

Base stopper resistors and bypass caps at each MOSFET should be fine to stop those. You can add ~10R resistors to separate output MOSFETs power rails from the rest of the amp too.

So if I got it right, most of the suggestions above doesn't follow the original circuit. Is there a post where all these have already been addressed, instead of me repeating the bad choice, (lol) ? But I am not talking about a ready PCB. I do want to make it myself.

Don't get me wrong. I really appreciate every comments. But having follow the book and after having view the video, of course I understand that the later is dated 2015, I was hoping that the circuit was already at it's best. I understand also that the goal of it is to rebuild the dh-220.

I guess what I am looking for is one circuit, based on MOSFETs, that is already very well designed so I can simply reproduce it. I am not an engineer, I am a simple technician, and although I find them very interesting, all the design process aren't my priority for the moment. Maybe I just don't have the patience ;-)

 

[Sevy]

Can try swapping drivers to 2SC3503/2SA1381, maybe those will allow for less distortion.

Well here the choice is based on the TTC004B and TTA004B that I have in stock already. With the bad experience with the 2SB649A/2SD669A purchased from eBay, I bought a hundred of each TTx004B to replace the above. In the simulation though, those weren't available so I used 2SB649A/2SD669A.

 

[pbilous]

Yeah, I’d confirm everything in the simulation before doing PCB. Simulations can take significant time - that is normal practice at the companies. Book is just a general guide for the idea or the concept. Simulator is what allows you to fine tune the design with good precision for specific transistors and parts you will be using.

You probably underestimated the importance and time it takes for the simulation stage in the design.

Your circuit is pretty much ready, just needs few little tweaks to refine it. With such amount of components you already have a great foundation for excellent results.

 

[pbilous]

Well here the choice is based on the TTC004B and TTA004B that I have in stock already. With the bad experience with the 2SB649A/2SD669A purchased from eBay, I bought a hundred of each TTx004B to replace the above. In the simulation though, those weren't available so I used 2SB649A/2SD669A.

Spice models available from the Toshiba website if I remember correctly

Here what I used:

.MODEL TTA004B PNP
+ LEVEL = 1
+ TNOM = 25
+ IS = 7.5e-014
+ BF = 190
+ IKF = 0.47
+ ISE = 5e-011
+ NE = 2.4
+ NK = 0.63
+ XTB = 1
+ XTI = 2
+ TRC1 = 0.003
+ NF = 1
+ VAF = 6.8
+ VAR = 50
+ BR = 6
+ IKR = 5
+ ISC = 1.0e-21
+ NR = 1.015
+ NC = 1
+ RB = 4
+ RC = 0.125
+ RE = 0.05
+ CJC = 4.09E-011
+ MJC = 0.33
+ VJC = 0.75
+ CJE = 1e-11
+ MJE = 0.33
+ VJE = 0.75
+ EG = 1.11
+ TR = 1E-009
+ TF = 1.59E-009

.MODEL TTC004B NPN
+ LEVEL = 1
+ TNOM = 25
+ IS = 1.374e-013
+ BF = 137.2
+ NF = 1
+ VAF = 11.95
+ IKF = 0.5057
+ ISE = 8.098e-013
+ NE = 1.8
+ BR = 35.11
+ NR = 1
+ VAR = 1000
+ IKR = 10
+ ISC = 1.916e-011
+ NC = 1.5
+ NK = 0.55
+ RE = 0.115
+ RB = 0.885
+ RC = 0.0709
+ CJE = 5.78E-011
+ VJE = 0.75
+ MJE = 0.33
+ CJC = 2.89E-011
+ VJC = 0.75
+ MJC = 0.33
+ FC = 0.5
+ TF = 1.545E-009
+ XTF = 1
+ VTF = 1
+ ITF = 1
+ PTF = 0
+ TR = 0
+ EG = 1.11
+ XTB = 1.4
+ XTI = 6.467

 

[Sevy]

Based from others posts suggestions, I choose to use only one pair to eliminate oscillations.

Also, every reading I have done so far suggest to stay with one pair for an amp of 70 W and less. I don't expect to go above that for now...