200W MOSFET CFA amp

[dadod]

Some people are interested in more powerful CFA amp, and this is 200W//8ohm and 400W//4ohm.
It uses lateral MOSFET output transistors like Hitachi 2SK134/135 and 2SJ49/50 or similar from Toshiba 2SK1056 2SJ162 or from other producers.
Interestin part of this amp is use of TPC compensation, but unusual one with one capacitance branch connected to the output instead to the VAS collectors.
That brings 20kHz distortion down significantly.
This thread is open to discussion and suggestions as it’s tried in simulation only.
As CFA uses very low feedback resistance there is significant power dissipation on series resistor, and I show two schematic, one with lower FB resistors and other with double values. I case if Rf is 220 ohm at full power dissipation go up to 14 W and in case if Rf is 480 ohm it halved. Still this resistor should be parallel combination resistors to distribute power dissipation. first is low FB resistance, this gives lower distortion and higher slew rate.
Damir

Zener protection diodes moved from drivers emitter to drivers bases on suggestion from JohanB (not good idea) 200W MOSFET CFA amp - Page 87 - diyAudio

Argument about stability: What is important is the outer feedback loop
200W MOSFET CFA amp - Page 67 - diyAudio

Better protection suggested by Wahab: 200W MOSFET CFA amp - Page 88 - diyAudio

First listening 200W CFA, post #740: https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-74.html#post4278243

Index:
1. 200W CFA VMOFET, the last schematic and the BOM is here on post #770 200W MOSFET CFA amp and the .asc zip file is here on page #775 200W MOSFET CFA amp

2. 100 W CFA assembling instruction on post #911 200W MOSFET CFA amp

3. PS regulator BJT version assembling and testing instruction post #912 200W MOSFET CFA amp
PS regulator layout errors post #919 200W MOSFET CFA amp
PS Regulator BOM post #988 200W MOSFET CFA amp

4. Corrected 100 W CFA schematic and BOM post #947 200W MOSFET CFA amp

5. astx real amp 200W measurements. post #851 200W MOSFET CFA amp

6. some RNMarsh real amp 200W measurements. post #787 200W MOSFET CFA amp

7. 200W CFA assembling instruction with schematic, balanced add schematic and corresponding BOMs.post #1230 www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-123.html#post5311565

8. 200W CFA VMOSFET printed board rev. 1.3.2 has some silk screen errors:

200W CFA VMOSFET-dado-1_3_2
Silk scree errors:

D21 rename to D8 and revers
D20 rename to D5 and revers
D23 rename to D20 and revers
D5 rename to D23
D22 rename to D21 and revers
D8 rename to D22
Q9 rename to Q5
Q7 (on the left) rename to Q9 2SC3503

explanation on page #1289 link http://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-129.html#post5539928

9. 200W CFA board rev 1.3.2, corresponding balanced add board and PS regulator MOSFET board with BOMS. page #1405 https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-141.html#post5725721

10. 200W CFA some soldering explanation, unbalanced input and balanced inputs, post #1415.
There is silkscreen error on 200W boards rev.1.3.2, input connectors wrongly marked, IN SIG GND should be GND SIG https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-142.html#post5740471
post# 1506 https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-151.html#post6031569

11. MOSFET PS regulator testing, post #1416. https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-142.html#post5740493

12. MOSFET PS regulator oscillating in some cases, remedy on post #1456. https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-146.html#post5904223

12. 100W CFA board rev 1.4 and PS regulator (BJT) rev 1.3 post #1491 https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-150.html#post6000469

13. BJT PS regulator short connection in rev 1.3, look how to rectify it #post 1535 https://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-154.html#post6158297

14. Shared gerbers for 100W version #1,588 https://www.diyaudio.com/community/threads/200w-mosfet-cfa-amp.243481/page-80

 

[dadod]

Here is schematic with higher FB resistance.
Damir

 

[mcd99]

Maybe add diode clamps across the VAS/TIS to clean up the clipping. Looks like saturation.

Do you need a Vbe multiplier with L-FETs? Maybe simple resistor biasing would work.

 

[dadod]

Maybe add diode clamps across the VAS/TIS to clean up the clipping. Looks like saturation.

Do you need a Vbe multiplier with L-FETs? Maybe simple resistor biasing would work.

This is typical TPC clipping, but here is no sticking, and the amp is going fast out of clipping.
Still I will try to find good way to make clean clipping, not sure what is the best way.
The Vbe multiplier is here to simplify Bias adjustement no for thermal control, as simple trimer can lose conntact and could provoke the bias go to high and burn mosfet. In the Vbe the resistor between base and collector is a trim pot and if loses conntact the bias is going down(but you know that).
I am not yet sure if some thermal compesation is needed for the drivers.

 

[mcd99]

The Vbe multiplier is here to simplify Bias adjustement no for thermal control, as simple trimer can lose conntact and could provoke the bias go to high and burn mosfet. In the Vbe the resistor between base and collector is a trim pot and if loses conntact the bias is going down(but you know that).

Good point but a couple of resistors in series could be used instead of a trimmer in a simple resistive bias. L-Fets appear to be very easy to bias as you know. My preference is not to use a Vbe multiplier, I worry about the potential for oscillation.

Have you got a good source for the a1381/c3503 transistors where you can get the same Hfe Rank? Wonder if mismatched ranks would cause a problem when used as drivers?

 

[dadod]

Good point but a couple of resistors in series could be used instead of a trimmer in a simple resistive bias. L-Fets appear to be very easy to bias as you know. My preference is not to use a Vbe multiplier, I worry about the potential for oscillation.

Have you got a good source for the a1381/c3503 transistors where you can get the same Hfe Rank? Wonder if mismatched ranks would cause a problem when used as drivers?

There is big problem with fakes, probably if you order from big sellers, but I can't recomand any, had bad experience.

 

[mlloyd1]

just to eliminate any confusion (especially these days trying to stay ahead of the counterfeiters), Toshiba didn't (doesn't) make any lateral mosfets. that was hitachi, then that division became renesas who currently still provides the 2sk105x/2sj16x laterals. others (like profusion) provide lateral mosfets with similar dies (i.e. similar performance). i think someone even provides double dies in a single device for higher power (but also more capacitance).

the nicely complimentary mosfets toshiba provided a while back (2sk15xx/2sj20x) are "verticals" with low (almost lateral-like) threshhold voltage.

not interchangeable without suitable circuit modifications

ok, here's some links since it's never good to trust my memory:

for "hitachi-like" laterals:

MOSFETs for amplifier | Renesas Electronics America

Magnatec. ALFET Lateral MOSFETs

Lateral Mosfet

for the toshibas now EOL'd:
2SK1530 | Products | TOSHIBA Semiconductor & Storage Products Company
and
2SJ201 | Products | TOSHIBA Semiconductor & Storage Products Company

mlloyd1

 

[Krisfr]

Are LTSpice models available for all of these from Renesas or ?

 

[mcd99]

There is big problem with fakes, probably if you order from big sellers, but I can't recomand any, had bad experience.

The only major supplier of these transistors I have used is Mouser and they have the current product KSA/KSC but they are of different hfe ranks.

Sometimes I wonder whether we should design these transistors out by using the dreaded 3 transistor VAS. Then with the cascoding we can use normal commonly available matched rank transistors. Regarding drivers I used bootstrapping and BC3x7-40s in my CFA which meant that the KSC/KSA could be avoided there. For drivers (if I am correct) the current gain of the transistors is more important.

Sorry for going slightly off topic.

 

[dadod]

just to eliminate any confusion (especially these days trying to stay ahead of the counterfeiters), Toshiba didn't (doesn't) make any lateral mosfets. that was hitachi, then that division became renesas who currently still provides the 2sk105x/2sj16x laterals. others (like profusion) provide lateral mosfets with similar dies (i.e. similar performance). i think someone even provides double dies in a single device for higher power (but also more capacitance).

the nicely complimentary mosfets toshiba provided a while back (2sk15xx/2sj20x) are "verticals" with low (almost lateral-like) threshhold voltage.

not interchangeable without suitable circuit modifications

ok, here's some links since it's never good to trust my memory:

for "hitachi-like" laterals:

MOSFETs for amplifier | Renesas Electronics America

Magnatec. ALFET Lateral MOSFETs

Lateral Mosfet

for the toshibas now EOL'd:
2SK1530 | Products | TOSHIBA Semiconductor & Storage Products Company
and
2SJ201 | Products | TOSHIBA Semiconductor & Storage Products Company

mlloyd1

Sorry, I was thinking that 2SK1056 2SJ162 where produced by Thoshiba too.

 

[dadod]

Are LTSpice models available for all of these from Renesas or ?

I think they are very similar and could be used in place of Hitachi, so simulation with Hitachi is good enough.

 

[s2udio]

Why not use Exicon 1020 power fets in the output stage.
And looks far over complicated...most will disagree I am sure .

 

[ilimzn]

Renesas = new name of the semiconductor division of Hitachi.
J162/K1056 are actually J50/K135 in TO-3P like cases.
Exicon/Magnetec/Alfet are apparently all made in the same fab and are based on Hitachi J56/K176. Exicon parts are direct drop-in replacements, however Exicon offers selection by treshold (color dot marking) which is useful when paralleling the MOSFETs. In fact, there is some advantage in omitting source resistors in favor of selecting MOSFETs to parallel. Among other things, source is on the case so you can simply put all of them on the same heatsink without isolation, and use it as the output terminal, isolating the heatsink itself, which is actually best done by using a mounting profile (usually an aluminium L or T profile) for the MOSFETs and isolating this profile from the heatsink. The total thermal resistance ends up being lower on account of the larger mating surface between profile and heatsink, compared to the sum of mating surfaces between the transistors and the profile.
It should be noted that it's fairly simple to find marching N-ch laterals, not so simple with P-ch (one needs about 3-4 times as many P-ch MOSFETs to chose from to get the same number of close matches, as for N-ch).
Source resistors will improve matching at lower currents but the voltage drop across them is far to low to make a difference at high currents due to the relatively low Gm of the laterals. The models do not show that in the real world they can vary quite a bit. It's gotten better with the newer production parts (J162/K1056), but not drastically so, especially concerning the P-ch part. When laterals were first introduced a number of manufacturers used them in their designs, strating with Hitachi themselves, but there were also Kyocera, Sanyo, and a ton of PA amp manufacturers. In the HiFi realm, it's quite conspicuous that most amps were of fairly low power in order to use a single pair. Amps that used multiple pairs invariably had matched parts (at least as far as i can tell from experience with Hitachi, Kyocera, Sanyo, Strasser, and a numer of other manufacturer's amps).

 

[DUG]

Just a few questions/comments

R46/R47 not symmetrical?

No equivalent for R63 on negative side?

FET gate resistors...220R in positive side and 150R in negative side.

Output quiescent current not adjustable?

 

[Walkalone]

2SC/KSC3503 & 2SA/KSA1381 with ''C'' rank Hfe are not available.
Mouser & Digikey current have ''E'' for 2SA/KSA and ''D'' for 2SC/KSC.

 

[Bonsai]

Good distortion results and slew rate on your design dadod. I think this clearly demonstrates the role in a CFA that the feedback resistor plays in setting the amplifier gm - you are getting these good results because there is plenty of feedback at HF, but you are able to curtail it at HF with TPC and avoid running into phase shift problems due to the output stage - FET's no doubt allowing this as well.

 

[dadod]

Renesas = new name of the semiconductor division of Hitachi.
J162/K1056 are actually J50/K135 in TO-3P like cases.
Exicon/Magnetec/Alfet are apparently all made in the same fab and are based on Hitachi J56/K176. Exicon parts are direct drop-in replacements, however Exicon offers selection by treshold (color dot marking) which is useful when paralleling the MOSFETs. In fact, there is some advantage in omitting source resistors in favor of selecting MOSFETs to parallel. Among other things, source is on the case so you can simply put all of them on the same heatsink without isolation, and use it as the output terminal, isolating the heatsink itself, which is actually best done by using a mounting profile (usually an aluminium L or T profile) for the MOSFETs and isolating this profile from the heatsink. The total thermal resistance ends up being lower on account of the larger mating surface between profile and heatsink, compared to the sum of mating surfaces between the transistors and the profile.
It should be noted that it's fairly simple to find marching N-ch laterals, not so simple with P-ch (one needs about 3-4 times as many P-ch MOSFETs to chose from to get the same number of close matches, as for N-ch).
Source resistors will improve matching at lower currents but the voltage drop across them is far to low to make a difference at high currents due to the relatively low Gm of the laterals. The models do not show that in the real world they can vary quite a bit. It's gotten better with the newer production parts (J162/K1056), but not drastically so, especially concerning the P-ch part. When laterals were first introduced a number of manufacturers used them in their designs, strating with Hitachi themselves, but there were also Kyocera, Sanyo, and a ton of PA amp manufacturers. In the HiFi realm, it's quite conspicuous that most amps were of fairly low power in order to use a single pair. Amps that used multiple pairs invariably had matched parts (at least as far as i can tell from experience with Hitachi, Kyocera, Sanyo, Strasser, and a numer of other manufacturer's amps).

Thank you for participation here. Your knowledge about latfet is very wellcome and can help to deign better amp here. I know that it's posible to lower didtortion by omitting source resistors if someone can get good matching.

 

[dadod]

Just a few questions/comments

R46/R47 not symmetrical?

No equivalent for R63 on negative side?

FET gate resistors...220R in positive side and 150R in negative side.

Output quiescent current not adjustable?

You got good eye.
R46/R47 are different to set DC offset at output to zero. Actually it will contain two 12 ohm resistors and 100 ohm trimer.
About need for R63(to compensate for bias current fluctuation) I am not sure do I need the same on negative side, I have to simulate more.
MOSFET gate resistors are of different value as N and P MOSFER are with different input capacitance.
R64 is a trimpot actually (2k) used to set the quiescent current.

 

[dadod]

Good distortion results and slew rate on your design dadod. I think this clearly demonstrates the role in a CFA that the feedback resistor plays in setting the amplifier gm - you are getting these good results because there is plenty of feedback at HF, but you are able to curtail it at HF with TPC and avoid running into phase shift problems due to the output stage - FET's no doubt allowing this as well.

Yes you've got it right Bonsai, this twist in TPC(connection to the output instead to the VAS collector) improve HF distortion a lot and with BJT's it did not work(for now).

 

[dadod]

THD20k drops from 14.5 ppm to 2.52 ppm (50W) when C8 C10 moved from Q6 Q8 collectors to the output, this is 21 dB.