KG Balanced Dynahi build discussion thread

[Aive]

What's the reason for 5 pf? Just out of interest, keen to learn..

[G600]

I had them at home, and Mr Gilmore came with this : 

 

5pf is 3db down at 278Khz

3pf is 3db down at 500khz

 

So looks like the difference is way out of reach...

[G600]

Hello there.

I started biasing my two boards (onboard version), and need help...

PSUs are two 30V S22 with 160VA transformers, boards have lowered gain, red LEDs, ON MJE15030G and 15031G, every sand is from Mouser, no matching as I don't have tools for that.

 

The first one : no offset from start, easy as a pie, but the bias can't go above 0.4V with the 681 Ohm resistor.

I've changed for 500 Ohm but I can't go above 0.74V ! Still no offset by the way.

What can be wrong with it ?

 

The second one : nightmare.

The left branch goes well to 0.75V (681 Ohm resistor too), and becomes incredibly hot, while the right channel stops at roughly 0.5V (and stays cold).

Then I started tweaking the THAT340 pots, I could bring the bias close to correct at the right branch, and suddenly I heard something like an intermittent short, plus a bad smell from this right branch...

I guess I'm good for a BJT sanity check, but what can explain the odd behaviour at first start ?

 

Thanks.

[GrindingThud]

On the boards with jfets, do I want to match GR, BL, or V?

[kevin gilmore]

BL

[fishski13]

G600,

 

you need to match the transistors.  i would just rip out all the sand and make the effort to match them.

 

in my mancave, the on-board heat sinks will do 50mA/500mV per output device, measured across each paralleled 20R pairs, in free air... if you want a safe junction temp - thats around 3W dissipation per 2.5" tall heat sink with BJTs that have positive temp-co properties.  depending on chassis ventilation, you may have to go lower.  

Edited October 15, 2014 by fishski13

[G600]

Thank you.

Which ones are critical ? Can I match only small bipolars, or power transistor need to be matched too ?

Another one : matched sand goes in the same rail, or each rail has to be matched with the other ?

 

Regarding your experience, do you think that onboard Dynahi must be biased to only 50mA ?

My first board (the one that stops at 74mA with 500 Ohm) becomes hot, but the second one's left channel goes hotter. I think there's something wrong with that one.

[fishski13]

match NPN to NPN and PNP to PNP per board. i was able to basically get identical matches with the MPSW06/56 for each board. i bought 100 pieces each since they're cheap.  i think 1-2% variance in hfe would be fine.  for the MJE15030/31, within 5% is good.  i would recommend buying 2x more than you need.

 

if you don't match the sand, you could have significant differences in bias across each of the devices.  you need to measure across each and every one of the 20R resistor pairs to make sure all of the output devices are biased reasonably close to one another.  you don't want one biased at 50mA and another cooking at 90mA.

 

in my chassis, i was able to get the bias to 42-43mA per output device with a reasonable junction temp.

Edited October 15, 2014 by fishski13

[GrindingThud]

Did anything happen to the schematic repository? Links seem to be down.....

[kevin gilmore]

Its back up.

 

So...  critical updates forced the machine to reboot. At which time the machine determined

that the cmos battery was low, and stopped.  Have new battery, will install Monday.

[Pars]

Where is the schematic repository? Must have missed that...

[kevin gilmore]

one repository, some schematics there.

one of these years I will clean it up.

 

if people know of schematics that are current and need to

be in the public place, let me know.

Edited October 18, 2014 by kevin gilmore

[G600]

Are the plastic washers necessary for the MJEs ?

I'm sure not using them I made a mistake that can explain my problems... why not a short with a heatsink !?

At least I have aavid's thermal pads, but damn, I'm dumb 

[kevin gilmore]

the mje parts I use are fully isolated, so I just use thermal pad, stainless steel screws, stainless

steel washers.

 

if you have the non-isolated parts, then you have to use the special plastic rings

[Aive]

So I got my CRO finally

 

I'm seeing some weird behaviour on output of my SuSy Dynahi with square waves. Getting some serious ripple on the positive and negative sections of wave duty cycles at lower frequencies, and losing the 'squareness' completely at higher frequencies. Suspect low-pass filtering frequency is too low presently and I'm filtering out important harmonics... this is particularly evident with the 10 kHz wave photos below.

 

Measurement parameters:

- Square waves generated via Matlab. Also downloaded some square wave .wav files from the net for comparison.

- Outputted via my MBP -> DAC (AGD M7, USB) -> balanced output into SuSy Dynahi

- Amp is set at default gain as per original silkscreen, haven't changed anything since I built it.

 

Photos below show left channel, + (yellow) and - (blue) signals. Sorry for the bad quality of the photos - I really should hook the CRO up to my PC...

 

1khz square wave:

 

100hz square:

 

100 hz square (zoomed):

 

10khz square:

 

 

10khz square (zoomed ):

 

Any ideas for how to improve/fix this? 

 

Will be a fun problem to solve, gives me something to do again

 

Edit: Should also mention, measurements taken with no load on output of amp board.

Edited November 9, 2014 by Aive

[Aive]

Update: I may be an idiot. I presumed that what was being output from the DAC would be correct. But that's not the case - the issue is with the DAC.

 

CH1 = Amp output, CH2 = DAC output. DAC output is the same ripple BS with it disconnected from any load.

 

 

What a fool I am to accuse the KG amp! 

 

Now to fuck around with the DAC to see wtf is going on.......

 

 

[kevin gilmore]

garbage in, garbage out.

but the 10khz dac output is really messed up

Edited November 9, 2014 by kevin gilmore

[luvdunhill]

http://en.m.wikipedia.org/wiki/Gibbs_phenomenon

[GrindingThud]

Ok, finally got boards now to go with the sand.

Looks like they match this: http://gilmore.chem.northwestern.edu/dynahibal1.pdf

Have a couple/few questions:

The amp is 2011 Version 1.0 boards (offboard sinks)....anything weird or incorrect on the silkscreen? I've read through the various threads and think everything should be correct, but just checking. Are there advised mods?

The PS is version .1 and I think I need to either cut a trace and add a resistor for the current limit or just leave it out....which is recommended?

For headphone use, can I get away with one PS board? I have two but trying to make this as small as possible.

What is a suggested part number for the little pins I see people mounting the 2sc3381/sa1349 in?

Thanks in advance!

gt

[GrindingThud]

Ok, this one has thrown me for a loop. The board I have is different from the schematic and the layout...so I'm thinking I just don't have the right version drawing. There is a spot on my board for a resistor not in the BOM, 40KDT, which I assume is the 20K dual tracking one mentioned below. The is also one labeled 10KLO (R15?) that does not match location in the schematic which is what is confusing me (C13 is bypassing 10K on the schematic vs 20K on the board). Do I leave the 10K one out? Looking at the board, it seems the 40KDT (R16?) is correct, R15 is 20 and the 10 comes out....

Rained all day....been matching transistors...

http://gilmore.chem.northwestern.edu/dynahipowerschem.pdf

http://gilmore.chem.northwestern.edu/dynapower2.jpg

silkscreen error on the dynahi power supply board.

The resistor labeled as 40k for dual tracking should be 20k.

schematic is correct.

Edited November 27, 2014 by GrindingThud

[kevin gilmore]

you install either the LO resistor, or the DT resistor.

one is for dual tracking, the other is for local feedback

 

don't install the current limit transistors.

[lkong]

transistor matching questions:

 

Are 2sc2705 and 2sc1145 the only pair I have to match between NPN and PNP?

And I only need to match NPN to NPN and PNP to PNP for others.

 

Also, how close should i match the Hfe of 2sc3421 and 2sa1358?

 

Thanks.

[GrindingThud]

Try this:

transistor matching questions:

 

Are 2sc2705 and 2sc1145 the only pair I have to match between NPN and PNP?

And I only need to match NPN to NPN and PNP to PNP for others.

 

Also, how close should i match the Hfe of 2sc3421 and 2sa1358?

 

Thanks.

[GrindingThud]

46 M3 taps later.....I begin to wonder why I wanted offboard sinks (yes it will be on the other side when assembled):

[kevin gilmore]

not supposed to tap the heatsink.

 

drill it the right size. transistors and screws and washer from the top,

transistor, thermal washer,heatsink,circuit board,washer,nut