Blue Hawaii BJT Build thread

[GeorgeP]

1 hour ago, mypasswordis said:

Uh oh, I see GeorgeP is crazy enough to want to set up another T2 run... are there going to be some CCS modifications to the board?  

Yeah, starting to have second thoughts as a couple interested people have already gone silent... but I am an optimist...

And if there are any changes, pleeeeease keep the tube and transistor spacing the same - that said, I am good with the status quo.

[JoaMat]

Can a DC/DC converter be used as it is or is some filtering needed?

[kevin gilmore]

yes that is the idea, something with at least 1kv isolation. 12v to 12v, series resistor, tantalum cap, and then a resistor to get to the right led current.

but you do need 4 of them per amplifier, but they are cheap

any changes to the board will keep all the spacings intact.

[JoaMat]

this one, 12V to 12V and 1/4W

[kevin gilmore]

probably a regulated version might be better

 

[JoaMat]

Lots of converters at Mouser – just have to find a suitable one in stock. Alternatively 9 volts cells....

[Kerry]

What a about a regulator after the 12V to bring it to 9V or so?  You can probably get lower noise on a separate regulator than a built in one.

[kevin gilmore]

that works too.

seems like the world stock of 2sc4686a are about to go extinct.

[JoaMat]

The squirrels sits on them.

[spritzer]

I guess that makes me a squirrel then...   

[JoaMat]

squirrels on Iceland? didn't know that...

[JimL]

On 4/24/2017 at 2:52 AM, sorenb said:

I am not aquinted with the work of Gary Pimm, and as you stated, his website is no longer up.

Not really, Jungs measurement shows deterioation above 1kHz - Jung even mention this in his text.

Still, the cascoded fets demonstrate higher impedance in comparison to their bjt counterparts. - but seems to lack in linearity.

You're correct, I was away from home when I wrote that.  Pimm's measurements also show some deterioration above 1 kHz, but so does his measurement of a 2G resistor.

In any case, Pimm did have some more complex current sources which exhibited even better test results than a cascoded MOSFET current source.  I never claimed that a cascode MOSFET was the ultimate, but it does give excellent performance for a minimum in parts cost.  The downside is the need to adjust individually, whereas a BJT current source that uses the on-voltage of a transistor or LED is pretty much set and forget.  Pros and cons.

Edited April 25, 2017 by JimL

[sorenb]

3 minutes ago, JimL said:

It's true that Jung's measurements appear to show deterioration above 1 kHz, however you will note that his test limit also deteriorates above 1 kHz so it's not clear whether that change is real or due to his coming up against the test limits.

I guess Jung is aware of the limitations of his test setup. Anyways, you brought up the results to support your claim - which it really doesn't.

4 minutes ago, JimL said:

In any case, Pimm did have some more complex current sources which exhibited even better test results than a cascoded MOSFET current source.  

Well, the Pimm web-site isn't up anymore, so it is kind of hard to discuss.

5 minutes ago, JimL said:

I never claimed that a cascode MOSFET was the ultimate, 

never said you did. 

6 minutes ago, JimL said:

it does give excellent performance for a minimum in parts cost

Agree.

7 minutes ago, JimL said:

The downside is the need to adjust individually, whereas a BJT current source that uses the on-voltage of a transistor or LED is pretty much set and forget.  Pros and cons.

Don't consider it a downside as such.

I guess it depends on what ever you consider "good", "best" or "ultimate" ... as far as I have experience the various CSS all deteriorates at some point, some earlier(lower frequency) than others. I guess that is caused by the output capacity of the device(s) used. In that area the fet's doesn't really excel.
 

[JimL]

Corrected comment above.  However, note that in Pimm's measurements as posted by laowei, the 2 gigohm resistor impedance also deteriorates above 1 kHz.  Actually, if you figure out the effective capacitance where the deterioration occurs, it is on the order of a few pf, which is more than acceptable, considering that the capacitance of the headphone is about 100 pf.  The added load on the active device is negligible.  The BJT may appear to be more linear because its impedance is significantly lower.  Remember, if the overall impedance is 10-fold lower, but the effective capacitance is the same, you won't see the deterioration until 10-fold higher frequency.

 

For electrostatic headphone amps, I prefer the higher impedance because the capacitative part of the headphone impedance is so high.  This pushes more of the signal current to the headphones.

Edited April 25, 2017 by JimL

[sorenb]

47 minutes ago, JimL said:

Corrected comment above.  However, note that in Pimm's measurements as posted by laowei, the 2 gigohm resistor impedance also deteriorates above 1 kHz.  Actually, if you figure out the effective capacitance where the deterioration occurs, it is on the order of a few pf, which is more than acceptable, considering that the capacitance of the headphone is about 100 pf.  The added load on the active device is negligible.  The BJT may appear to be more linear because its impedance is significantly lower.  Remember, if the overall impedance is 10-fold lower, but the effective capacitance is the same, you won't see the deterioration until 10-fold higher frequency.

Well, that is a good point - after reading you post, I did sim the T2 CSS against the 10m90/DN2540 depletion, and although the T2 comes out with a lower impedance in comparison, it maintains it well above 20kHz as oppose to the depletion cascode ...I assume it is not only a question about doing well at DC, but also how well across the entire range 

[JimL]

By the way, I agree with you that a single MOSFET current source doesn't have a very good input capacitance.  The real question is, is the BJT capacitance less than, equal to, or greater than the MOSFET capacitance.  As a single device I would bet the BJT is better, but when cascoded, I don't know.  The cascoding definitely minimizes the effective input capacitance.

Edited April 25, 2017 by JimL

[JoaMat]

I’ve been using the cascaded CCS in T2 since December 2014. It’s my favorite CCS, also used in Blue Hawaii and other amplifiers. Kevin’s idea using DC/DC converter with the T2 CCS is appealing. So I removed unnecessary components and put in 9 volts cells.

It works (why shouldn’t it?). Personally I prefer cascaded CCS but if there is a nice way to introduce DC/DC converters I might reconsider. Battery cells is not very convenient.

[joehpj]

10M90s has 900V max voltage. How to use it on T2?

[kevin gilmore]

you can't, that is the problem.  there are other depletion mode ixys parts with 1kv or more, but they are not sharp cutoff and have more capacitance so you have to cascade them.

[spritzer]

Well there is always the option of just running it at 1000V and crossing fingers and toes...

[joehpj]

Joamat used it since it since 2.5 years ago, so it seems stable. Or maybe he used some other parts?

Or be careful not to crank up the volume too much  

Edited April 30, 2017 by joehpj

[spritzer]

The 2SA1968's would most run just fine at 1000V (some even at 1050V) so there is always some margin built into these parts.  That said, I wouldn't bet on them all being able to handle it. 

[sorenb]

3 minutes ago, joehpj said:

Joamat used it since it since 2.5 years ago, so it seems stable. Or maybe he used some other parts?

probably because Joamat is not using his T2 at full blast, thus the 10m90 will see less than1kV

[JoaMat]

Deleted

Edited May 2, 2017 by JoaMat

[wink]

Not unless they want a magic smoke machine........