I am still considering building one of these to replace my dynalo, but had questions regarding the input section.

 

From what I understand, the THAT340 bipolar input can be replaced with individual JFETs simply by replacing the individual NPN or PNP in the THAT with N-ch or P-ch FETs?

 

C -> D

B -> G

E -> S

 

Advantages would seem to be higher input impedance and lower offset? Any other changes required to the circuit? Would you want to remove the input series resistors, shown as 5K on the schematic (R18/R33) and jumper these?

 

I still have enough 2SK109/2SK389 or single equivalents to do this. Or is the THAT340 recommended? Anyone listen to either version and have comments?

Edited February 4, 201511 yr by Pars

its supersymmetry, and the 5k resistors are part of the feedback.

if you go with fets you can multiply both Rf and Ri by 10 to increase

the input impedance.

really does not matter for this. definitely does matter for the input impedance

for the ubal to bal converter.

Thanks Kevin. This would be for the multiamp, not the ubal to bal converter (I guess these are not the same thing?).

 

Rf is R16 and R19? I assume by Ri you mean input, so the 5K resistors R18 / R33?

yep

Another stupid question time

 

When replacing the THAT340 with JFETS, would the NPNs be replaced with N-ch, and PNPs replaced with P-ch, or vice versa? For a NPN, the base needs to go positive by the diode voltage, or +0.7Vdc before it will turn on. Somehow it seems that a P-ch should be used here, but not sure. Like I said, stupid question

 

Might not be an issue, as in going thru parts inventory, I have a bunch of 2SK170s but can't find the 2SJ74s I thought I had. If anyone has around 8 (or 4 matched Idss) they would sell, shoot me a PM.

 

Also, this is applicable to anything using complementary pairs of 2SK170 / 2SJ74, but I ran across this last night on DIYaudio:

 

http://www.diyaudio.com/forums/swap-meet/232547-fs-2sk170-2sj74-matched-sets-10.html#post3966396

 

Has anyone run across this before, and what do you think of it? Not having a curve tracer, no way for me to verify this. Also see attached pdf.

 

 

Believe I figured it out looking at the original dynalo schematic. NPN -> N-ch, PNP -> P-ch.

Yup, N-channel is N-channel whether it is a fet or a bjt. 

 

I would just just the That340 though.  Even Kevin likes it a lot which has caused us to look into bipolar inputs for some other amps. 

Isn't the input impedance awfully low using the THAT340? I may use it anyhow, though I already have a 4-ch RK27 50K ohm pot. I have to do this cheap if I am going to do it right now. Can embellish later.

 

I could use some 2SJ74s if anyone has a few they could spare

It is on the low side so a 10K pot has to be used.  TKD makes some nice ones...    

 

Ask Justin if he has any J74's he can't match to n-channels. 

Isn't the input impedance awfully low using the THAT340? I may use it anyhow, though I already have a 4-ch RK27 50K ohm pot. I have to do this cheap if I am going to do it right now. Can embellish later.

 

I could use some 2SJ74s if anyone has a few they could spare

I have 2 matched pairs of 2SJ74BL bought from Jack at Tech DIY, that I can give to you. Bought them for an F5 , but used the 2SK170s in my KGST. I'm in Shenzhen China ATM, and won't get back home until Thursday. If you still need them then, PM me.

Kevin,

 

After looking at the datasheets for the BJTs, 16mA seems rather low for these devices. Even the 18mA you noted on the previous page seems low. These seem pretty linear all the way out to 70-100mA. Any reason not to go higher than 18mA on these?

 

I am currently measuring hFE on the PNPs (from Avnet). Still waiting for the NPNs from Mouser. At ~16mA, using this circuit, with 100K Rb and my Fluke in series on the base in uA mode, I am not experiencing the transistors even getting warm after 2-3 minutes.

 

 

Current parameters I am planning on using for matching:

 

V = 7.5Vdc

Rb = 100K

Ib = ~70uA

Ic = ~16-17mA

 

The original 2SA1015 / 2SC1815 in the Dynalo are considerably lower power handling. Looking at the datasheet, I can see where 16mA was a good operating point for these.

 

Thanks.

 

BTW, anyone still looking for thru hole devices: Avnet has the MPSW56s (around 1500 still left). Mouser still has the MPSW06.

Edited February 27, 201511 yr by Pars

18ma @ 20 volts power supply is 360mw.  about 1/3 of full power, and they definitely

get hot.  But you can crank it if you want.

Still have not finished installing the balanced input jacks, but the protector circuit fits nicely in there.

Edited March 14, 201511 yr by GrindingThud

18ma @ 20 volts power supply is 360mw.  about 1/3 of full power, and they definitely

get hot.  But you can crank it if you want.

You make a good point as always

 

I was planning on running @ 16V or so. I'll see what happens, but will probably stay around 16-18mA. I don't really want to use OPA445s for the servos, so need to stay below 18V.

Finally got around to measuring 200 transistors: MPSW-06 NPNs from Mouser and MPSW-56 PNPs from Avnet. FIrst off, these seem to be tighter in spec than the Toshibas from the original Dynalo. The NPNs overall were a bit higher than PNPs. I used a base resistance of 221K ohms with power at 16V for these. I had a DMM in the base and collector legs set for uA and mA, respectively. Ic ranged from 15mA to 22mA or so; most fell within 19-21mA. I waited until the DMM measuring Ib hit a particular value and then recorded the Ic. Subjective, but it seemed like a good way to do it. I already had a home etched board with SIP sockets and posts on it for test leads, resistor, etc. I put resistor leads in the DUT SIP socket and tested the devices while still on the tape; didn't seem to affect the result.

Should be enough close matches for a multiamp I think.

Edited September 1, 20187 yr by Pars
replaced graphic

its supersymmetry, and the 5k resistors are part of the feedback.

if you go with fets you can multiply both Rf and Ri by 10 to increase

the input impedance.

really does not matter for this. definitely does matter for the input impedance

for the ubal to bal converter.

If I am using FET inputs instead of the THAT340, is there a reason I would want to increase the input impedance even more by going 50K for Ri and 250K for Rf? I wasn't sure whether your statement really does not matter for this applied to the ubal to bal, or to the multi amp.

 

Also, a couple of part questions (trying to do this with what I have around):

 

1) The 1uf integrator caps... any reason to get ceramic X7Rs if I have some polyester film laying around?

2) The 4.7uf film caps... I presume these are rail bypass caps? So value isn't particularly important?

3) If I do raise the impedance, does the compensation 5pf cap need to be changed?

 

Thanks!

Edited March 6, 201511 yr by Pars

Increasing it would be useful if you wanted to decrease a cap in proceeding gear (you asked if there was any reason, this is kinda obvious, perhaps I am not answering the question).

Good point Marc. I hadn't thought of that, though I am not sure it is an issue. From what I gather, using the FET input instead of bipolar already increases the input impedance considerably just by itself? Looking at the original Dynalo schematic, with the SJ109/SK389 input, there is no series resistance; my stuff has no issue with driving that.

Poly caps always beat ceramic in my book so go for it. 

 

As for the 4.7uf caps, I can't find the layput files here for this board but I'm pretty sure they are the bypass caps I wanted Kevin to put in. Any value that fits is fine there but moar is moar bettah!!! 

Thanks. I just looked at the gerbers and the 4.7uf are indeed bypass caps. With 100uf electros on the board, any reason to choose 4.7uf as a bypass? I'm not really up on the latest thoughts regarding bypassing and whether it is well thought of or not. I have in the past seen some pretty good reasons not to do it.

I just want some large film caps right on the rails. 

If I am using FET inputs instead of the THAT340, is there a reason I would want to increase the input impedance even more by going 50K for Ri and 250K for Rf? I wasn't sure whether your statement really does not matter for this applied to the ubal to bal, or to the multi amp.

 

Also, a couple of part questions (trying to do this with what I have around):

 

1) The 1uf integrator caps... any reason to get ceramic X7Rs if I have some polyester film laying around?

2) The 4.7uf film caps... I presume these are rail bypass caps? So value isn't particularly important?

3) If I do raise the impedance, does the compensation 5pf cap need to be changed?

 

Thanks!

Here I go quoting myself again

 

From Kevin's post:

 

if you go with fets you can multiply both Rf and Ri by 10 to increase the input impedance.

 

1)  Is there a reason I would want to increase the input impedance even more by going 50K for Ri and 250K for Rf? I wasn't sure whether your statement really does not matter for this applied to the ubal to bal, or to the multi amp.

Anyone have a transformer recommendation for a multi-amp? The sigma22 I had built up has a Triad FD7-36 56VA 2x18V secondary on it currently, which might be enough. The original Dynalo (headamp) used an Amveco 35VA transformer in it.

 

I think Kevin mentioned that each board draws ~300mA, so 30V * 600mA (2 boards) would be ~18VA?

Edited March 14, 201511 yr by Pars

Does anyone have a top board view with the part designations (not the values) per the schematic? If not, I'll figure it out.

 

Thanks!

Like this:

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

Edited March 16, 201511 yr by GrindingThud