How many transistors to get for Dynalo?

[robzy]

Hey guys, im about to order the transistors and FETs for my dynalo, the problem is that im having a very hard time figuring out how many i should get to factor in matching.

So please help me out - how many do i need of each?

Thanks,
Rob.

 

[Garbz]

replace one of the 500ohm bias resistors near the LED with a 1k pot and the cost of transistors decreases considerably as matching isn't all that important anymore.

Just for reference there are 24 transistors and I ordered 36 and got matching pairs mostly.

 

[dgardner]

You should try to keep the BJTs in the same Hfe (gain) class if possible. Likewise it helps to have the two JFETs in the same Idss (porportional to gain) class. This is denoted by the letter suffix at the end of the part. (Y, GR, BL) Check the datasheet. If you have trouble getting the BJTs in the same gain class, let me know via PM. I have several hundred of each in the GR gain class.

 

[robzy]

Quote:

Originally Posted by Garbz Just for reference there are 24 transistors and I ordered 36 and got matching pairs mostly.

Hrm... i think i might go for 36-40 just to be on the safe side. Wouldnt a pot in that part of the amp be in the signal path and thus be bad?

Quote:

Originally Posted by dgardner You should try to keep the BJTs in the same Hfe (gain) class if possible. Likewise it helps to have the two JFETs in the same Idss (porportional to gain) class. This is denoted by the letter suffix at the end of the part. (Y, GR, BL) Check the datasheet. If you have trouble getting the BJTs in the same gain class, let me know via PM. I have several hundred of each in the GR gain class.

How exactly does one keep BJTs in the same gain class when ordering? Or is it just a luck-of-the-draw-thing?

Rob.

 

[dgardner]

Quote:

Originally Posted by robzy How exactly does one keep BJTs in the same gain class when ordering? Or is it just a luck-of-the-draw-thing?

It turns out that www.bdent.com will sell them to you by gain class, if you insist on it. Unfortunately, www.mcminone.com will not, and has recently shipped some orders with differing gain classes despite asking for the same class.

 

[dgardner]

Quote:

Originally Posted by robzy Wouldnt a pot in that part of the amp be in the signal path and thus be bad?

The pots are used to set the value of the 2mA constant current sources that are formed by the 1.6VLED+BJT+500R circuits. Using the pots to make rough bias current adjustments helps to bring the DC offset on the Output down to a managable level, so the op amp servo's limited control range will work properly. You will likely start off with a fairly large DC offset if you don't match any transistors. (large is 150mV or more). It depends on whether you consider this "in the signal path". Most headfi'ers consider it acceptable.

 

[Garbz]

I'd be more worried about the LED than the pot. Afterall it's so far out of the actual signal path (which I would consider the output stage transistors and the resistors immediately connecting to the FETs) that it's negligable. Plus since it's not a stereo pot there's no channel missmatch.

It would be the same arguement as using different resistors, but once again this is soo far back that I seriously doubt even the most golden ears could tell the difference.

 

[robzy]

Okay, so we are talking R11, R15 & R41 R45 i assume.

As far as i can tell - these resistors already have pots (R10, R14 and R40 and R44) in parallel with them. Does that means i will have two pots running in parralel?

And a question i have wanted to ask for a while but never got around to it: What do the 4 pots in the schematic (R10 R14 R40 R44) do?

Rob.

 

[Garbz]

In the schematic you can see each resistor has a pot in parallel. These are designed in a mutually exclusive way. Use a resistor or a pot, but not both. Infact you only really need 2 pots as if one side is off it can be compensated using the other side. I've put pots into R10 and R44 close to the power input just for visual effect, and use 500ohm resistors in R15 and R41.

The pots adjust the current going into the LEDs. The LEDs form a constant current source which biases the input stage of the amp. Small changes here in the order of 1-50ohm will cause large changes (comparatively) in the offset of the channel you fiddle with. This allows you to be less stringent on the matching of parts as you can compensate with the pot. After you have an amp that's running at 0mV when warm, throw in the opamp and let it keep drift under control.

I've personally used this method in the last 3 amps, and I've never matched the FETs, I simply couldn't be stuffed.

 

[saab]

Do you think you could take a picture or two of what you have done Garbz? I am sure that it would be greatly appreciated.

Thanks
-John

 

[dgardner]

Quote:

Originally Posted by robzy As far as i can tell - these resistors already have pots (R10, R14 and R40 and R44) in parallel with them. Does that means i will have two pots running in parralel?

Quote:

Originally Posted by Garbz In the schematic you can see each resistor has a pot in parallel. These are designed in a mutually exclusive way. Use a resistor or a pot, but not both. Infact you only really need 2 pots as if one side is off it can be compensated using the other side. I've put pots into R10 and R44 close to the power input just for visual effect, and use 500ohm resistors in R15 and R41.

As the PCB designer for this board, I can say that the intent was to have 10K pots in parallel with the 500R resistors when building it. Garbz is correct, you could also use a 1K pot and omit the 500R.

The benefit in my mind about using the 10K pot in parallel with the 500R is that you get a more fool proof starting point for the adjustability. The downside is that you can only lower the effective resistance, which means you can only increase the value of the current source you are tuning. A way to compensate for this is to start with 550R to 600R and then tune down to the proper current value and DC offset. Try to avoid over biasing the amp. In order to check this, ground the input and try to get about 5.0V drop across the 5K resistors feeding the JFETs. If you drop more than 5.0V, you will overbias the amp and it may run too hot.

The benefit of using the 1K pot is that you can actually increase OR decrease each of the current sources. The downside is you must preset the pot to around 500R-550R to be in the proper range. Otherwise you could be wildly off on the current value and cause overheating if its too high, and cause performance problems if it's too low. Again, check the 5.0V drop across the 5K resistors feeding the JFETs.

 

[Garbz]

Lol as the suggester of the idea this is what i meant

I kind of like the notion of using the 10k in parallel. But then in the end we get a one sided option with increased parts cost (i'm just being picky

)

Here's what i'm using. Please excuse the fact that R39, R62 and the bypass caps are unpopulated, I'm waiting on parts.

In this case you can see that R44 (topleft pot) has a 1k pot in it, R45 is left unpopulated. On the otherside conversely R41 has a 500ohm resistor in it and R40 is left unpopluated.

Providing a bit of matching is done there won't be too much offbiasing (is that a word?) done.

 

[robzy]

A huge thankyou to both Garbz and dgardner! Not only do i now know how many trans to order but also have a better understanding of the circuit

Now i just gotta make sure that all the trans come from the same gain class and im set! (Im gonna get a couple extra just incase one of them is wwaayy off i think).

Ohh... one thing i just realised - would unmatched JFETs also be fixable with the current bias? And if not how many would i need for a decent match...

Rob.

 

[Garbz]

like I said i didn't even bother matching jfets. Infact unmatched jfets can be compensated by unmatching transistors or leds slightly which is what i did in my very first dynalo incarnation.

The current bias changes the jfet, this change is carried through all the way down through to the output stage. It doesn't matter which part isn't matched including LEDs too, this will fix it. But like Dan said don't overbias, so a bit of matching is necessary.

And be careful where you buy from. From wez components about 95% of the transistors were Y rated. out of thoes 70-80% matched reasonably well, but there have been horror stories on this forum from someone who ordered them and got Y for the PNPs and GR for the NPNs, not a single one matched.

 

[dgardner]

Quote:

Originally Posted by robzy would unmatched JFETs also be fixable with the current bias? And if not how many would i need for a decent match

Unmatched JFETs in the same Idss class are acceptable. In fact, I usually socket the JFETs and try different combinations of several 389's and 109's until the DC offset is the lowest. Buying 2 or 3 extras of each usually means you can have a little room to play. Since these parts are not exactly cheap, you can get away without extras. It's up to you.