ThermalTrak+TMC amp

[dadod]

Hi Damir,

Thank you for this wonderful amplifier.
I finally managed to assemble two channels and had a first listen yesterday.
All I can say is that's a very good sounding amplifier that I can highly recommend.
It's very quiet, smooth, balanced, with firm bass and delicate highs.

In my build at least I had to deal with a couple of problems that I would like to mention:

1. I had a parasitic oscillation of aprox. 100Mhz which I tracked down to the input cascode pair.
Putting a small cap from Q3, Q4 base to ground solved this (I tried various values (3.3uF, 100nf, 1.5nF all of them working fine) and finally installed a small 4.7nF MKT in that position)

2. I installed a BAV21 diode from Q12 collector to Q15 base which slightly improved clipping behavior.
Now I only notice a very slight sticking in the negative side.

I also would like to mention that I deviated from the suggested list of parts
and used a pair of 2SK170-BL as input fets, KSA992 and KSC1845 instead of 2SA970 and 2SC2240 and BD139 instead of 2SC3503 for Q1(I would appreciate your opinion about the changes Damir, are they O.K. by you?).

Future builders should also in addition to suggestions of posts #453 and #463 install a wire link in R15 position and shouldn't forget C19 and C24 which I put in the underside of the PCB.

A big thanks also to Terry who was kind enough to order and send me two pcb's that made this build possible.

Best regards,
George

P.S. In my final build I installed C3, C4 and D11 as offset was too high for my liking without them.

Hi George,
Nice work, you installed TT transistors differently and that is also one good option.
Strange that you've had that 100MHz cascode oscillation, I had only oscillation when C19, C24 where badly soldered. I can't see the value of those caps, did you use C0G type?
KSA992 and KSC1845 instead of 2SA970 and 2SC2240 are good substitute, I have them but not used in cascade, so I am not sure if that was reason for the oscillation. C7 should prevent that oscillation. BD139 for Q1 should be OK.
I am against of use diode from Q12 collector to Q15 base as it increase distortion to much, and my amps clipping was quite good without it, and why to force this amp to clipping.
Yes, I also tried without C3//C4 but offset fluctuated to much, one would expect that jfets allow omitting those caps.
BR Damir

 

[AndrewT]

..................

1. I had a parasitic oscillation of aprox. 100Mhz which I tracked down to the input cascode pair.
Putting a small cap from Q3, Q4 base to ground solved this (I tried various values (3.3uF, 100nf, 1.5nF all of them working fine) and finally installed a small 4.7nF MKT in that position)

That location is the same as adding a Q9 collector to ground for the CCS. I wonder if the Ccb on the CCS is actually preventing the CCS from oscillating?

That makes me think "what if the Ccb was installed as collector to +ve supply rail?"
Would that have a similar effect?

 

[AndrewT]

............. I also tried without C3//C4 but offset fluctuated to much, one would expect that jfets allow omitting those caps.
BR Damir

Shorting across C3/4 changes the amplifier from AC coupled to mixed AC & Dc coupled.
It now has closed loop gain at DC and all tiny DC waverings get amplified and appear as wavering output offset. Whereas, with AC coupled the offset waverings are attenuated by ~30dB.

 

[dadod]

How high a rail voltage was this designed to operate at? I will spin the 2N5xxx transistors. I actually thought I had checked the pinouts when at first it didn't work. I guess I was mistaken. I will fix it in the morning and report back. This is the last thing on my work bench. Going to need a new design to try building after this.

Hi Terry,
Have you succeeded with that PS Regulator? If used with the TT amp it should put it to the next quality level.
BR Damir

 

[gagnou]

Hi Damir,

I think C19, C24 are old possibly Phillips made caps I found in a local store.
They are 220pF in value @ R.T. and falling to 208pF after blasting them with a heating gun at quite a high temperature.

Regarding the oscillation I had a clue when I read Andrew Russell's input on cascode transistors http://hifisonix.com/wordpress/wp-content/uploads/2010/10/Cascode-Oscillation-in-Audio-Amplifiers1.pdf
Could it be trace inductance to Q3, Q4 base or the elevated capacitances of the jfets causing the problem?
Before I tried :
1) to increase C19, C24
2) slightly bigger C5 and C6
3) changed the design to EF2 (removed Q2, Q19)
but none of the above worked.


Hi AndrewT,

gain of 1 @ DC makes the wavering very small (under 1mV in this case after some minutes)
Without the cap it started @ 160mV and gradually fell to +-20mV max wavering after 10 - 15 minutes.

Regarding your comment about Q9 I don't understand it.
Q9 seems to be connected as an emitter follower or am I wrong?

Best regards,
George

 

[dadod]

Hi Damir,

1. I had a parasitic oscillation of aprox. 100Mhz which I tracked down to the input cascode pair.
Putting a small cap from Q3, Q4 base to ground solved this (I tried various values (3.3uF, 100nf, 1.5nF all of them working fine) and finally installed a small 4.7nF MKT in that position)

I simulated with that cap 4.7nF connected between Q3 Q4 bases to the ground and it doubled THD20k. Try max 2.2nF here as it not increase 20 kHz distortion.
BR Damir

I noticed now that 1.5nF was enough to stop oscillation, use that.

 

[still4given]

Hi Damir,

I could not get it to work. I may try etching a new board and start fresh. I probably have some broken parts due to the reversed transistors earlier. It's too hard to find them. The amp sounds really good so it may be worth trying it with the matching PSU. I still have the board sitting off to the side on my bench so I won't forget to deal with it.

I'm really happy that George has his working. I still have a couple extra pair of boards if someone else is excited about building one. It would be a shame for them to just sit unused.

Blessings, Terry

 

[AndrewT]

...............Hi AndrewT,
....................

Regarding your comment about Q9 I don't understand it.
Q9 seems to be connected as an emitter follower or am I wrong?
.............

I opened the sch.pdf in post1.
The two transistor labels over-write. It appears to be Q6 and Q6.
I thought Q9 was the CCS transistor and Q6 was the mirror transistor. The CCS transistor's collector lead goes straight down to the two bases in the LTP and feeds the cascode ref voltage.

 

[dadod]

I opened the sch.pdf in post1.
The two transistor labels over-write. It appears to be Q6 and Q6.
I thought Q9 was the CCS transistor and Q6 was the mirror transistor. The CCS transistor's collector lead goes straight down to the two bases in the LTP and feeds the cascode ref voltage.

You vent to much in the past, schematic was changed few times after that.

 

[gagnou]

Hi AndrewT,

I re uploaded the schematic for your reference.

 

[AndrewT]

I had an old sch from post1.
Ignore my comment about Q9. it is a different transistor.

 

[dadod]

Hi AndrewT,

I re uploaded the schematic for your reference.

As George had problem with the cascode oscilation, here is updated schematic with added cap od 1.5nF between cascode bases and the ground (soldered on the PCB backside). This should solve possible problem if used different cascode transistors. R37 increase from 1k to 6.8k to prevent sticking at the clipping.
Damir

 

[obscurus]

Hi Damir,

I tried to make your PS regulator before but failed to complete ... Now I want to try again.

This time, I want to remove the protection part ans use mosfet instead of bjt so I modified your PS regulator a bit. But due to my lacking in skill, I'm not sure if I did thing right. Could you have a look at what I'm having now?

What I did was:

- Remove the big resistor.
- Remove protection part.
- Change CCS from R to jfet and a resistor because I have a lot of them.
- Use VR on both rail so I can change both rail volatge drop.

Thanks!

 

[dadod]

Hi Damir,

I tried to make your PS regulator before but failed to complete ... Now I want to try again.

This time, I want to remove the protection part ans use mosfet instead of bjt so I modified your PS regulator a bit. But due to my lacking in skill, I'm not sure if I did thing right. Could you have a look at what I'm having now?

What I did was:

- Remove the big resistor.
- Remove protection part.
- Change CCS from R to jfet and a resistor because I have a lot of them.
- Use VR on both rail so I can change both rail volatge drop.

Thanks!

You need a drivers for the big mosfets, the LTP current is not enough. Try to simulate it with different loads. Why would you use separate adjustment for the both rail voltage? If you drop to much voltage on the regulating mosfet, result is to much power dissipation on it. When I designed my regulator I wanted that voltage drop to be 5 to 6 V about, just enough to have good regulation at full output current.
There is a simple way to make capacitance multiplier if you don't want to have any protection incorporated.

 

[obscurus]

You need a drivers for the big mosfets, the LTP current is not enough. Try to simulate it with different loads. Why would you use separate adjustment for the both rail voltage? If you drop to much voltage on the regulating mosfet, result is to much power dissipation on it. When I designed my regulator I wanted that voltage drop to be 5 to 6 V about, just enough to have good regulation at full output current.
There is a simple way to make capacitance multiplier if you don't want to have any protection incorporated.

I had adjustments on both rail because I think I may need to have lower voltage drop with the use of low Rds mosfet. May be I just need it one rail since I can change value of R9 directly on board.

With the current schematic, if R load = 2 ohms -> I = 11A -> the ripple was about 70mV.

Could you give me more info about how to add the a driver for the mosfet? I don't have good electronic background to know how to do so?

Also, how to have apacitance multiplier the simple way? Some info, link will be a great help.

 

[dadod]

I had adjustments on both rail because I think I may need to have lower voltage drop with the use of low Rds mosfet. May be I just need it one rail since I can change value of R9 directly on board.

With the current schematic, if R load = 2 ohms -> I = 11A -> the ripple was about 70mV.

Could you give me more info about how to add the a driver for the mosfet? I don't have good electronic background to know how to do so?

Also, how to have apacitance multiplier the simple way? Some info, link will be a great help.

Elliott has many nice thing on his page, like cap multiplier: Capacitance Multiplier Power Supply Filter
This is mosfet PS regulator, not built yet.

 

[obscurus]

Elliott has many nice thing on his page, like cap multiplier: Capacitance Multiplier Power Supply Filter
This is mosfet PS regulator, not built yet.

What is the purpose of Q1 && R2? I didn't see them in BJT version.

And how to adjust the limit current of the PS Regulator?

 

[dadod]

What is the purpose of Q1 && R2? I didn't see them in BJT version.

And how to adjust the limit current of the PS Regulator?

Q1 is the driver for the regulating mosfet, R2 was on the BJT version.
R9 defines the current limit, but you said you don't want any protection, so just omit it and corresponding circuitry.

 

[obscurus]

I will build the mosfet version of PS Regulator, I managed to fit the whole schematic into a 100 x 100 mm, two side board based on your BJT layout BUT i have to leave D8 and D108 because I couldn't manage any space for them. Could I solder them direct to C7 and C107 leg?

I will order the board from local PCB company, I have all components for it.

 

[dadod]

I will build the mosfet version of PS Regulator, I managed to fit the whole schematic into a 100 x 100 mm, two side board based on your BJT layout BUT i have to leave D8 and D108 because I couldn't manage any space for them. Could I solder them direct to C7 and C107 leg?

I will order the board from local PCB company, I have all components for it.

Just don't use D8 and D108, not needed.
Bear in mind that I did not build that version with the mosfet transistors, just simulated, you are going to be the first one.