ThermalTrak+TMC amp

[keantoken]

Note: I just replaced the attachment with a new version, if the old one didn't work, maybe it will work now...

[mightydub]

Originally Posted by dadod

Thanks mightydub,
You are stepping a temperature for whole circuit and I do it for separted transistor groups, so I can simulate predriver/driver independetly of output transistors and its TT diodes.
If you look this thread a bit back you can see that I use similar Vbe multiplier.
Here, as this triple predrivers are connected in diamond connection, Vbe multiplier need to be different and that is what I tried to simulate.
I use Cordel's models too.
dadp

I think you missed something - the basic process was to hold the temp of the outputs constant and vary temp for the pres and drivers to determine the necessary slope, then hold the temp of the pres and drivers constant and vary the temp of the outputs. Then at the last step put it all together. (see post #2.)

Anyway - I'm following your progress with interest!

[LineSource]

Originally Posted by dadod

Here is a new TT amp with diamond triple OPS. All CCS are with deplation mode mosfet DN2540 and this simplified a quite this amp and 20k THD is still low as 3ppm at full power.
dado

From my experience, JFETs and MOSFETs have a wide production variation. I have always found it necessary to hand match JFETs in order to get consistent performance. I would not select a depletion MOSFET for a low noise current source where an fixed/consistent current is required in production. You may want to sample several DN2540 parts from differnt vendors to feel comfortable that a multi-national DIY design will match your personal prototypes.

I use high hfe bipolar transistors and low noise LEDs for discrete current sources because of their well documented performance and proven production consistency. Your earlier design, although more complicated, would probably have better performance and more consistent worldwide production.


The datasheet on the DN2540 shows the wide VGS variation I have experienced, and the datasheet only gives the typical GFS transconductance spec with no min/max numbers.


VGS(OFF) Gate-to-source off voltage min= -1.5V max= -3.5V
GFS Forward transconductance only typical = 325 mmho spec'ed.
----No min or max spec is provided in the datasheet.

[dadod]

Originally Posted by LineSource

From my experience, JFETs and MOSFETs have a wide production variation. I have always found it necessary to hand match JFETs in order to get consistent performance. I would not select a depletion MOSFET for a low noise current source where an fixed/consistent current is required in production. You may want to sample several DN2540 parts from differnt vendors to feel comfortable that a multi-national DIY design will match your personal prototypes.

I use high hfe bipolar transistors and low noise LEDs for discrete current sources because of their well documented performance and proven production consistency. Your earlier design, although more complicated, would probably have better performance and more consistent worldwide production.


The datasheet on the DN2540 shows the wide VGS variation I have experienced, and the datasheet only gives the typical GFS transconductance spec with no min/max numbers.


VGS(OFF) Gate-to-source off voltage min= -1.5V max= -3.5V
GFS Forward transconductance only typical = 325 mmho spec'ed.
----No min or max spec is provided in the datasheet.

Thank you for your comments. I've got a bunch of DN2540 in TO220 case and was thinking how to use them.
The main question in this design is diamond OPS predriver and how to bias it correctly.
If I put predrivers and drivers on separate small heatsink, thermal drift supposedly should cancel and what is left to be thermaly compensated are TT output transistors. Problem I expected is low Vbe bias voltage(1.2V) but in simulation the main problem I found was predriver/driver thermal drift and not Vbe multiplier with it's task to compensate output only. with different predriver transistors I get very different result, so I suppose the models are not good regarding thermal simulation. What BJT model parameter defines tempco?

mightydub, I did the simulation exacly how you suggested, keeping predrivers/drivers at fix temperature and test Vbe multiplier with TT diodes, and then keeping TT transistors and diodes on fix temperature and test predrivers/drivers slope.
dado

[greenm01]

After much troubleshooting, and some transistor supply problems, I finally got around to completing the amplifier. This is my first DIY amp from the ground up. Many thanks to Dado for all his help and guidance. I've learned quite a bit along the way.

The TT-Triple sounds superlative. The ThermalTrak BJTs are excellent performers, and the amp sings wonderfully. Beautiful dynamic range. Punchy bass. Wonderful highs. This amp will take center stage in my Linkwitz Pluto active speaker clone.

Here's a YouTube video of the amp playing "Regular Pleasures" by Patricia Barber on Audio Engine 2 speakers.

I also very much appreciate Dado's power supply regulator, which has built in speaker protection.

Pictures below:

[dadod]

Finally TT amp was tested and it is presented here as a complete project - I call it AudioCube. It contains Power Supply module, PS Regulator and output amplifier.
1. Power Supply module contains diodes bridge, capacitors bank and a fuses. It could be used without PS Regulator. Between first caps bank and second ones there are 0.22 10 W smutting resistors.

2. PS Regulator contains kind of a capacitor multipliers, overload/error protection and a loudspeaker protection, all in electronic way without any relay contacts.
The capacitance multiplier will drop input DC for 5 V abouth, needed to function properly. Overload protection will sense a current flowing from the + and – output and if triggered will shut down the PS regulator. There is provision for 6 A or 12 A of trigger current. The trigger is delayed and it depends of the current level an duration. If the trigger level was set for 12 A then it will start at 12.2 A and 250msec of a delay and will go up to 30A and 1.5msec of delay.
I hope this is good enough to protect all electronic connected down stream. Loudspeaker protection is copy of similar one used for JLH 80W MOSFET amp.
If one side of the power supply turn off other side is turned off instantaneously.

3. TT amp was gone prolong tests. First I tested thermal bias stability. One channel kept 18mV on one emitter resistor from moment of turn on up to hot heatsink.
I set the bias after small heatsink reached steady temperature and then I heated output heatsink with the heat gun and the temperature reached 50 degree.
Other channel behaved differently. It was undercompensated even with the same components value. I had to change R44 from 1k to 1.2K to get the same behavior as first channel. It could be different reason for it, TT diodes tempco is not consistent between TT transistors or the bias spreader transistor hfe influence the bias spreader tempco.
In attachments there are TT schematic(last and final version), PS regulator schematic, Power Supply module schematic and the grounding principle schematic.
If somebody interested to build this amp I will provide PCB layouts.
Attached some photos of built AudioCube, wooden cover is missing.

The sound – I like it very, very much, better the my modified JLH 80W MOSFET amp, or simple TMC amp with EF2 output stage. I could not compare it with commrcial amps as I have simple low power NAD integrated amp only.
dado

[dadod]

Sorry, Power Supply wrong schematic, this one is correct.
dado

[metal]

how did you make the case?

[dadod]

Originally Posted by metal

how did you make the case?

Some more photografs, I hope it is clear how from it.
dado

[greenm01]

Originally Posted by dadod

First I tested thermal bias stability. One channel kept 18mV on one emitter resistor from moment of turn on up to hot heatsink. I set the bias after small heatsink reached steady temperature and then I heated output heatsink with the heat gun and the temperature reached 50 degree. Other channel behaved differently. It was undercompensated even with the same components value. I had to change R44 from 1k to 1.2K to get the same behavior as first channel.

My recent experience:

1) I had the exact same problem with setting the bias correctly. One channel was stable at 18mV, but the other channel was undercompensated as you describe above. Unfortunately I burned out a power transistor by letting the emitter resistor voltage drift too high. I replaced the faulty output transistor and replaced R44 with 1.2k. The bias seems to be stable ~18mV on both channels. I also tested with my Wife's hair dryer on the external heatsink and it seems to be stable.

2) There was a very small ground loop (barely audible) that was mostly eliminated by switching the PS Regulator ground from star ground to the PSU capacitor connection point. Of course, the only bullet proof way to completely eliminate ground loops is to use balanced connections. In the case of TT-Triple (unbalanced input) I have ground loop noise 90% eliminated. I can still hear a little hum when I press my ear to the speaker and eliminate all background noise (HVAC, refrigerator, etc...) in my living space with -70dB volume. For practical purposes it's completely inaudible with any realistic input signal.

3) Occasionally the PS Regulator speaker protection will kick in for no reason and shut down the amplifier. I don't know if it's because of transients or what, but the frequency of occurrence is not consistent. Sometimes the amp stays on for 10 minutes, and other times go for an hour without fault. I need to do more testing to confirm this behavior.

4) Sound quality is phenomenal. The amp is completely transparent, and power dynamics are beautiful. If I close my eyes, my Linkwitz inspired Pluto speakers transport me directly into the concert hall. I love this amp.

Attached are photos of PS Regulator ground connection and latest internal photo.