AURA

[andy_c]

This one? It's still using VAS lead-lag compensation...

Yes, I was thinking more in terms of frequency-dependent gm.  But looking at that design more closely, I see there's only about 2:1 variation, unlike your technique which is in a completely different ballpark.

[bogdan_borko]

Hi syn08. What`s the advantage in using separate drivers for every output bjt and why colector resistors in drivers and predrivers of the putput stage?

[syn08]

Hi syn08. What`s the advantage in using separate drivers for every output bjt and why colector resistors in drivers and predrivers of the putput stage?

Share the current, same reason as using multiple output devices (not related to SOA). Killing the parasitic inductances in the PCB traces.

[syn08]

PCB ready to go for manufacturing, early next week . 10.7" x 4.7", uses mostly SMD, except for the power trannies and some large film caps and resistors.

[stinius]

Wow, looks beautiful. I’m impressed 

[stinius]

Hi O

Any update?

How does the bias spreader track the temp?

Stein

[syn08]

Hi O

Any update?

How does the bias spreader track the temp?

Stein

Nothing yet, it will last a couple of weeks until I'll get the boards.

[syn08]

Boards are here...

[ Guests cannot view attachments ]

[syn08]

Half of the hard work is done... Heatsink ready to receive the board.

[ Guests cannot view attachments ]

And the board installed on the heatsink. Heatsink is 0.35C/W and will also be forced cooled by temperature triggered fans.

[ Guests cannot view attachments ]

Now, on to the fun part 

[megajocke]

Very nice! It looks beautiful. What do you use for insulating the transistors from the heatsink? I have mica pads that are just large enough to use with TO-264, but I have some trouble with them slipping around out of under the device when screwing the devices down. I wonder if I should try spring clamps perhaps.

A board with the transistors underneath does seem like a very good idea to reduce parasitics, but I really hate repairing amplifiers where you have to unsolder or unscrew all the output transistors to access the bottom of the board. Especially if you have isolating pads that you need to align when reassembling. I guess this one isn't going to see much abuse though and it seems rather conservatively designed so hopefully no repairs will be needed. 

Also quite unusal to see SMD emitter resistors. What size and type are they? And while those are SMD types, the feedback resistor is that big honking TO-247 in the middle of the heatsink, isn't it?

It will be interesting to see how good it performs! 

[syn08]

Very nice! It looks beautiful. What do you use for insulating the transistors from the heatsink? I have mica pads that are just large enough to use with TO-264, but I have some trouble with them slipping around out of under the device when screwing the devices down. I wonder if I should try spring clamps perhaps.

A board with the transistors underneath does seem like a very good idea to reduce parasitics, but I really hate repairing amplifiers where you have to unsolder or unscrew all the output transistors to access the bottom of the board. Especially if you have isolating pads that you need to align when reassembling. I guess this one isn't going to see much abuse though and it seems rather conservatively designed so hopefully no repairs will be needed. 

Also quite unusal to see SMD emitter resistors. What size and type are they? And while those are SMD types, the feedback resistor is that big honking TO-247 in the middle of the heatsink, isn't it?

It will be interesting to see how good it performs! 

You are certainly a very perceptive person 

- I use TO-247 micas purchased from EBay quite some time ago. They are just fine in size.
- Except for the power devices, not much to do on the bottom of the board if you use SMD.
- Mouser has the 0.2ohm/2W SMD devices and they are dirt cheap.
- Yes, that's the 270ohm/70W thick film feedback resistor. Freaking expensive  but haven't found any cheaper alternatives for a noninductive device.

Her's a picture of the assembled board, before soldering the power devices, and a detail of the dense SMD area on the left. I will fire up the entire amp, from a monster 2x60Veff xformer and the attached monster power supply (built out of Lelon 3300uF/100V caps that I was able to crop for $1.50 a pop), over the weekend and be back with some measurements, I'm expecting THD20<50ppm @ 500W into 4 ohm and a minimum of 350V/uS slew rate before installing the input RFI filter.

[syn08]

I powered up amp from my lab supplies (Sorensen 60V @ 10A) and I got two preliminary conclusions:

- The good one: the amp is rock stable. Not a trace of instability for the rather conservative (for me) 3MHz ULG.
- Thermal tracking is crap. Although it simulates fine with Bob's models for the NJL and the MUR120 diode, in real life it's significantly undercompensated. At this point I'm not even sure if it can be fixed in this version of the PCB. With the values in the schematic per post #26, the current is way to high. Adjusting R84 and R85 fixes the current value, but also undercompensates the entire output stage.

[Pete Basel (PB2)]

I powered up amp from my lab supplies (Sorensen 60V @ 10A) and I got two preliminary conclusions:

- The good one: the amp is rock stable. Not a trace of instability for the rather conservative (for me) 3MHz ULG.
- Thermal tracking is crap. Although it simulates fine with Bob's models for the NJL and the MUR120 diode, in real life it's significantly undercompensated. At this point I'm not even sure if it can be fixed in this version of the PCB. With the values in the schematic per post #26, the current is way to high. Adjusting R84 and R85 fixes the current value, but also undercompensates the entire output stage.

It sure looks nice, almost as nice as Webers, need some chrome on it!

[syn08]

Each time I'm building something, it reminds me how crappy the models we are "working" with are.

The current through the VAS was about double of that simulated. After lowering the current through the VAS to 7.7mA (replaced the D15 and D16 zeners with 12V, lowered R24 and R86) now the thing stabilizes at about 140mA (28mV across the emitter resistors, which is about right). As expected this thing dissipates well over 100W in idle, the heatsink is at 60-70 degs, so forced cooling will be mandatory. I don't like the idea, but I should probably increase a little the Vbe multiplier ratio and further lower the VAS current to 5mA.

One to another, Bob's idea for biasing is well undercompensated (at least for 10mA through the VAS). Not sure if this thermal compensation was ever tried before in the real world (it was simulating ok) and not sure what the root cause is. Either the 2SC5171/2SA1930 drivers have an unusual high tempco, or the diode and NJL models are way off? I'll eventually try to find out. Anyway, I was much happier with the Sansui style bias compensation that I was using in the VSOP amp.

Quick test with signal into 4ohm load. At almost 400W in the load, just before clipping, THD20=130ppm before even rerouting the power and load cables for a THD20 minimum, this is encouraging, it will certainly go well under 100ppm after the cabling is optimized. The good news is that after a couple of minutes at full power the bias stays solid (I would prefer it a little overcompensated though). Clipping is nice, not as nice as an active clamp, but the sticking is minimum.

[ Guests cannot view attachments ]

Unfortunately I forgot and installed the input RFI filter (680ohm/150pF) right away, so I don't know the real SR. The step response is though excellent and (as expected) clearly dominated by the input filter.

[ Guests cannot view attachments ]

EDIT: Now, to put together some fans for cooling this beast (and the load) and then do some serious measurements.

[Pete Basel (PB2)]

Let me ask is the TO126 on the heatsink the Vbe Multiplier? 
If it is you are then multiplying the 2 fast response thermal trak
diodes plus the one slow response BE junction as I see it - would
you agree?

What gain factor are you using in the Vbe Multiplier?