Group Build: Dynafet

[Pars]

Did you try it without the servo in? Usually I see how low I can get offset and then put the servo in.

By "works" do you mean it hasn't caught fire yet or it actually makes sound?

[luvdunhill]

Did you try it without the servo in? Usually I see how low I can get offset and then put the servo in.

2mV without the servo. More like +-1mV. even at the lowest bias, the buffers run pretty toasty. So fracking hard to get a probe in edgewise on these boards.

Anyone know how much current you can put through a sigma22, assuming unlimited heat sinking for the four MOSFETs? Kinda wish I was using four simga22s

Edited June 26, 2009 by luvdunhill

[DigiPete]

hope I don't need more than two KG Dynahi PSUs...for the balanced.

I thought the Fet board should draw less than a dynahi...

[luvdunhill]

pics. Note current, servo (or lack thereof), and offset. It was actually at 0.00mV for like a minute, so I wasn't trying too hard to get the money shot

[Pars]

Yeah, I always get them as good as I can, then pop the servo in and relax. Under 1mV or +/- 1mV without the servo is plenty good, at least on I/V boards with no feedback (those are harder to get stable than the dyna---- are).

[j4cbo]

What's that heatsink you're using?

[Smeggy]

Looking good!

Nice to see pics

[luvdunhill]

What's that heatsink you're using?

Conrad MF35-151.5:

Conrad Heatsinks - Products

[digger945]

1A is gonna be 30W or the limit of these devices. When I was playing with the one I built it got the toroid pretty hot in a hurry.

What do you have on the gates? I had difficulty getting the gate voltages down with the adjustments and values used. It would be good to change some values around to be able to get the gates down a little, and in a more comfortable adjustment zone. I'll do some tinkering later.

Sexy lookin' board.

[vvs_75]

Anyone know how much current you can put through a sigma22, assuming unlimited heat sinking for the four MOSFETs? Kinda wish I was using four simga22s

"Two paralleled high-current, highly reliable MOSFETs (rated at 17A each) serve as the "pass" transistor of each rail.

The high-current MOSFETs are not normally the limit of how much current the σ22 PSU could supply, as long as they are adequately heatsinked.

The maximum current limit is determined by the rating of the power transformer, the rectifier diodes (the specified MUR820 devices are rated at 8A), and the AC line fuse. There is otherwise no current-limiting circuit in the σ22, which allows it to supply peak currents of many amperes. High transient bursts of current are always available, which some amplifiers require to avoid clipping and distortion."

AMB

[luvdunhill]

1A is gonna be 30W or the limit of these devices. When I was playing with the one I built it got the toroid pretty hot in a hurry.

What do you have on the gates? I had difficulty getting the gate voltages down with the adjustments and values used. It would be good to change some values around to be able to get the gates down a little, and in a more comfortable adjustment zone. I'll do some tinkering later.

Sexy lookin' board.

I don't quite understand. The gate voltage is directly proportional to the current draw of the output stage. So, you are saying that the thing gets hot? Well, yes it does. How low do you want to go? I'd be willing to try to lower things, but it seems that raising the pot is the ticket, to say 50K? Let me know what you'd like to try and I can give it a go.

At this point I'd like to look closely at the buffers and try to get them to run a bit cooler. I need to somehow work on the back side of the board so I can measure some voltages here and there. I'll also get some music flowing for Naaman

[Beefy]

The minimum bias with the 20K pot is 750mA. Lamers may need to up this value. Accidentally had the bias at 3A for a bit... oops.

Running at 1A now. The transistors in the buffer are hot, I cannot touch them for more than a second.

Wow. Just wow!

[luvdunhill]

A few numbers for j4cbo and digger945:

FETs drawing 329mA, 322mA, 336mA, 315mA. Total draw on negative rail is 710mA. Toroid (160VA I think?) is cold to the touch.

Gate voltages 3.39V, 3.22V, 4.55V, 4.57V

2Ω/2W resistors dissipating 0.286W (definitely want to try down to 1Ω, devices are well-matched and we have plenty of dissipation available even at 0.5W.. could stack the resistors and go even lower..)

4.7Ω/1W resistors dissipating 0.001W

10Ω/1W resistors dissipating 0.0002W

221Ω (R27-R30) 1.11V, 1.11V, 1.74V, 1.73V across them

I agree with digger945 that we need to work out a solution for lower bias, but I'm pleased as punch so far.

I'm assuming we will want to run the same dissipation numbers with around half this current draw, as well as maybe a quarter?

I run these same FETs at 70mA in the Blowtorch... but it's a preamp after all. They sound better in this application at 100mA, but I've never had the heat sinking to go higher than that, so stuck with 70mA.

Edited June 26, 2009 by luvdunhill

[luvdunhill]

Wow. Just wow!

yeah, I guess the 500mA on the datasheet is a guideline. Needless to say, I lunged for the power switch pretty quick... more or less was a pulse test I have had them running at max current for a bit though, just for the heck of it and they survived.

[Beefy]

yeah, I guess the 500mA on the datasheet is a guideline. Needless to say, I lunged for the power switch pretty quick... more or less was a pulse test I have had them running at max current for a bit though, just for the heck of it and they survived.

Your sacrifices are everybody's gains

[Pars]

A few numbers for j4cbo and digger945:

FETs drawing 329mA, 322mA, 336mA, 315mA. Total draw on negative rail is 710mA. Toroid (160VA I think?) is cold to the touch.

Measured using the drop across the 2 ohm resistors? I assume this is something like Q27, Q31, Q28 and Q32 (half populated)?

Gate voltages 3.39V, 3.22V, 4.55V, 4.57V

Why the big discrepancy between the + and - rail gates?

2Ω/2W resistors dissipating 0.286W (definitely want to try down to 1Ω, devices are well-matched and we have plenty of dissipation available even at 0.5W.. could stack the resistors and go even lower..)

4.7Ω/1W resistors dissipating 0.001W

10Ω/1W resistors dissipating 0.0002W

221Ω (R27-R30) 1.11V, 1.11V, 1.74V, 1.73V across them

Once again, why the discrepancy between + and - rails (I assume these figures are for 1.11V for R27, R28, and 1.74V for R29 and R30)?

I agree with digger945 that we need to work out a solution for lower bias, but I'm pleased as punch so far.

I'm assuming we will want to run the same dissipation numbers with around half this current draw, as well as maybe a quarter?

I run these same FETs at 70mA in the Blowtorch... but it's a preamp after all. They sound better in this application at 100mA, but I've never had the heat sinking to go higher than that, so stuck with 70mA.

You discerned enough of the Blowtorch from that epic thread to build one?

[luvdunhill]

Measured using the drop across the 2 ohm resistors? I assume this is something like Q27, Q31, Q28 and Q32 (half populated)?

yes

Why the big discrepancy between the + and - rail gates?

Welcome to the world of MOSFETs. This is the main reason I've been hoarding the old Toshiba devices, as they are by far the most complimentary devices I've found, maybe they'll make their way into a headphone amp. If you want a good read, look at Charles Hansens comments with regard to lateral versus vertical FETs.

Anyways, even without empirical data we can predict this. Check the second graph on the datasheets here:

http://ezphysics.nchu.edu.tw/prophys/ael/File/Datasheet/sk214e2.pdf

http://ezphysics.nchu.edu.tw/prophys/ael/File/Datasheet/sj77e2.pdf

You can take the parameters from the circuit, but for discussions sake, let's arbitrarily choose Id = 200mA and Vds = -4V. For the 2SJ this corresponds to Vgs -3.0V for the 2SK this corresponds to Vgs 2.0V. That's a 1.0V difference right there.

If you want to run the actual numbers from the DynaFET, don't forget to figure in graph number four, which shows transfer characteristic as a function of Tc, you'll need to figure this in to get the measured numbers to add up, which they do.

But remember, the important thing is that all of the devices draw the same amount of current through the 2 ohm ballast resistors, and this is why the devices need to be matched. The 21mA difference in this case was from the absolute worse match in the pool. The other side had a 7mA difference, which was a better matched pair. The other pairs and quads should be even better. Also, I didn't match the 2 ohm resistors (I think they are 5% tolerance), and I'd recommend doing that as well.

You discerned enough of the Blowtorch from that epic thread to build one?

That thread and picking Kevin's brain. The kicker was keeping things stable, with respect to temperature, and the servo design. That and the literal 1% yeild of appropriate devices in the matching process. There are 8 JFETs (four if you have access to V grade devices), and 4 four MOSFETS and the rest of the devices are resistors. It's all about matching. I need to draw up some boards, but the distribution would have to be rather clandestine...

[n_maher]

Aren't you going to need a 4-wire meter to match the 2ohm resistors?

[luvdunhill]

Aren't you going to need a 4-wire meter to match the 2ohm resistors?

I was thinking about a Wheatstone bridge, or driving 100 mi to my dad's lab

[Pars]

Aren't you going to need a 4-wire meter to match the 2ohm resistors?

Got one. Including home-made leads. Bridge works too. My dad doesn't have a lab (and wouldn't have a clue as to what we are talking about, but I digress).

[luvdunhill]

heh, my dad is a EE professor and they have a commercial resistor bridge I could match the resistors on.

ok, I ran a few more numbers, you can check my math given the numbers above.

The output of the buffers are running at 15.5mA quiescent, which I'd assume means these transistors are dissipating 500mW and are rated to 800mW. The input is running at ~5mA. The two CCSes for the buffers are running at 5mA and 8mA, and are dissipating 150mW-250mW.

No wonder these are toasty.

[n_maher]

I know some of the build team have the resources and knowledge to do it, I was just pointing it out to those that might not. Carry on.

[Pars]

Good point Nate. And your post was not taken in a bad light whatsoever.

[digger945]

I mainly meant that I couldn't get the gate voltages down low enough to put the outputs into a somewhat decent bias, like you had originally targeted, somewhere in the 150mA range.

Think I'll wait until you do some more testing and probing with a load on the output to comment further. My little proto was not quite as organized.

[DigiPete]

Maybe a post later on to outline the parts matching and adustments needed to optimize the circuit would be appropriate, to guidethe technically less savvy builders.