[dgardner]

I've been looking to test drive some new layout tools for PCB. I've built a Gilmore Dynalo using the toner transfer method with the Headwize project posting. Works great.

My goal is to complete a new PCB layout and share it with the group. I also plan to have the boards fabricated at a boardshop like PCBnet. Maybe organize a group buy. One of the goals is to create a single PCB that can be populated both bridged or non-bridged. This may have been done by someone else before, but I'm sure another version of the board won't hurt the DIY folks.

I'd like some feed back from the group to make sure I've incorporated the details eveyone might be intested in. I hope this is helpful, and doesn't open up a can of worms.

Schematic for review

EDIT:/fixed link.

 

[dgardner]

Some suggestions are:

1. Trim pots in parallel with the 500 resistors to dial in FET bias currents.

2. Trim pot across power rails with wiper to DC ADJ if Op-Amp servo not populated.

3. What lead spacing would be most universal for Power Rail Bulk Decoupling?
Would 0.300" be alright to support Electrolytics?

4. Are larger holes sizes for the output stage emitter resistors desireable?

 

[Pars]

How about the mods suggested by Sijosae or his fellow Korean DIYer, with the 1N4148 diodes thermally bonded to the output trannies? Not sure how many have tried this or not.

Also, if a PSU section is done on it, room for alternative diodes such as FREDs or Shottkys would be nice.

A nice big ground plane would probably be a plus

 

[dgardner]

Are the diodes you're referring to look like this?

Which junction temps are you trying to match with the bonding?

 

[dgardner]

One major question is whether to include some of the better PSU work that has already been done onto the same board. One advantage might be better regulation since there will be no long connecting wires. Lots of very small stand-alone layouts have already been done for just the amp board. I did not find a bridged output version that required no jumper wires. Those who want modularity have other choices.

Is there any interest in combining a Gilmore PSU and Gilmore Bridged on the same multilayer PCB? This would give first timers a much easier time with the PSU and the Amp as a single build project. Board cost might be higher, but what the heck.

What might make sense would be to make the PSU section include:
1) discrete layout for bridge rectifiers, you can pick your own diodes
2) option for very small caps across the rectifiers.
3) standard LM317 and LM337 circuits with trim pots.
4) REF02 + standard OPA548 output op-amps with all the stabilization fixes.
5) lots of pad options for output caps (if you dare load 'em up)

It would probably make sense to have inputs on the board from some external torroidal transformer, since no one will agree on that.

 

[dgardner]

Does anyone know *why* HeadAmp's KG reference amp uses four instances of what looks to be the single ended board? Am I missing something with the bridged output implementation? I thought only one front end circuit was needed to drive the two output sections. Maybe this is a taboo subject.

 

[Garbz]

In the spirit of modularity I sugest keep them separate. The whole it's easier arguement really goes down the drain when you consider there's only 3 wires to connect between the psu and the ampboard. In my case i've connected the two via some simple screw terminals, but make sure they are marked. I damged a good $40 worth of parts plugging it in backwards.

My sugestions are:
1. Keep the amp and psu separate.
2. Keep decoupling caps (large electrolytics and small films) on the amp board close to the output stage.
3. Discrete componetns are good for the PSU's rectifiers however they did take up considerable space. Although this might not go down to well consider mayb haveing a plug for rectifiers. There's not a lot of current going through them so plug choices are pretty large, and people could roll their own rectifiers like sijosae does.
4. PSU should have a groundplane
5. Option to add a ground loop breaker
6. Although it's popular to have the parts all over the shop in the psu with separate heatsinks on each, i find if the transformer is 50VA 2x25v then the VRs get HOT. My current PSU layout has all active components on one side allowing me to bolt a massive heatsink onto the side.

7. For the amp board I'll re-itterate your sugestions. It's imporntant to give poeple the choice of DC offset correction. Some poeple prefer to swap-out LEDs, I prefer to have a 1K 25turn pot in the 500ohm resistor's place.
Some people prefer to use a Servo, some prefer a large 20K pot from the supply pins to DC-adjust.

Umm that's all for me

/EDIT: consider larger capacitance next to the output stage. a 300mil electrolytic could easily be a 470uf 16v FK series Black Gate like in my case. While the total of your caps comes to the same as mine, I have space for 0.1uf bypass caps next to each of them.

 

[Pars]

For the Sijosae-posted mods, check this thread, post 18 I believe. The output transistors are the ones the diode is bonded to.

 

[dip16amp]

Quote:

Originally Posted by dgardner Does anyone know *why* HeadAmp's KG reference amp uses four instances of what looks to be the single ended board? Am I missing something with the bridged output implementation? I thought only one front end circuit was needed to drive the two output sections. Maybe this is a taboo subject.

The problem with the bridged output design is that there are two outputs per channel and only one opamp servo to adjust the two DC offsets. The DC offset problem mainly comes from the dual FET 2SJ109 and 2SK389 packages which can change the DC offset as much as a 100 mV between each FET within the same package.
I straighten the pins of the J109 and K389 and rebend them to fit inline on a 16 pin socket which allow me to swap between the two FETs in the dual package. Out of five J109 and five K389, I only found one of each that would cause less than 10 mV of DC offset difference without the opamp. When the opamp servo is installed, it can only drive the DC offset somewhere inbetween the two different DC offsets. Most of the DC offsets ended up over 100 mV on the other FETs.
It is much cheaper and easier to have an opamp servo for each output than to try getting good FETs that match themselves. This way, I can get each DC offset less than 10 mV without the opamp by swapping ends of the FETs and changing LEDs. Then when I put in the opamp, it will drive the DC offset to less than 1 mV.
For balanced inputs and balanced outputs, it's just four seperate channels wired the same. For unbalanced inputs, the negative outputs would probably have the second stage connected to the other 5K ohm resistors while the positive outputs would be wired normally. The servo would probably also be wired differently for the negative output.

 

[dgardner]

Thanks for inputs and the explanation! Makes sense to me now. Does anyone want to take a shot at what the opamp server wiring would look like for the bridged output - negative output amp - with unbalanced input? (See previous post)

 

[dip16amp]

Quote:

Originally Posted by dgardner Thanks for inputs and the explanation! Makes sense to me now. Does anyone want to take a shot at what the opamp server wiring would look like for the bridged output - negative output amp - with unbalanced input? (See previous post)

You may have to connect the negative output to the other input of the opamp if it doesn't work like you have it.

 

[dgardner]

Here is an update to the schematic. It incorporates circuitry for the suggestions listed above.

Schematic - Page 1
Schematic - Page 2
Schematic - Page 3
Schematic - Page 4


For *bridged mode*, some questions still remain.
1) Exactly how do the servo(s) need to be connected? There are two now.
2) What is the wiring look like for a balanced input?
3) What is the wiring look like for an unbalanced input?
4) How is the negative output leg going to be driven?
5) How would an XLR input connector get wired?
6) Will dual mono headphone wiring be required? I assume - yes.

Any help clearing up these last questions would be much appreciated.

The goal is *no* long & nasty jumper wires going across the board, especially in bridged mode. Tiny bare wire jumps between some 0.100" holes seems ok to me if that's what it will take to get non-bridged and bridged out of the same layout nicely.

EDIT:/fixed links

 

[amb]

dgardner,

With your addition of D7, D8, D9 and D10, you need to reduce the value of R34 and R43 from 510 ohms to around 200 ohms. Without this change you will overbias the output stages, and cause the output transistors to run too hot and possibly damaging them.

Now for your questions:

• If you are using two completely independent amplifiers for the non-inverted and the inverted sides of the balanced amp (for a total of four amplifiers for the stereo pair), the servos for each should be connected exactly the same as in the unbalanced setup.
• For balanced inputs, then you will have two XLR jacks, one for the left channel and one for the right. The + signal on the XLR connector should go to amp-A's input, the - signal to amp-B, and the "ground" pin to the signal ground. See the image below. Note I used the same resistor number designations as in your pdf schematic. The triangles denote each amplifier block, with the amp-A's "+" input being the same as your diagram's "Input A", and amp-B's "+" input being your diagram's "Input B".
• For unbalanced inputs, you will need to modify amp-B to actually be inverting. This could be most easily done in the manner shown in another image below. Basically, you need to add Ra and Rb. Ra should be the same value as R23/R57 (in your case 1K ohms), whereas Rb should be the same value as R22/R55 (10K ohms).
• Amp-A's output goes to the "+" terminal of your headphone transducer and amp-B's output goes to the "-". Note that you will need to rewire your headphones so that the left and right channels' "-" wires are separated from each other, so that you'll have a 4-wire setup rather than a 3-wire.
• See above.
• Yes, that's what "balanced" is all about.

Hope this helps.

 

[dgardner]

Here is the latest update to the schematics I intend to goto layout with. If anyone would like to provide any more input, now is the time.

Schematic - Page 1
Schematic - Page 2

Here are a couple of supplementary pages that incorporate amb's information and attempt to show some the basic system level connectivity of a dynalo and a dynamid. This might be helpful to newbies that are studying the Gilmore Amplifier project pages for the first time.

Schematic - Page 3
Schematic - Page 4

Next up, I hope to post the component place views and the ground planes sections. Right now, I'm pretty sure that a four layer board is the way I'll go. The top will be a ground fill, the two inner layers will be for routing, and the bottom layer will be split planes for the power supply rails & maybe more ground plane under the FETs.

If anyone has ground plane ideas for a four layer stackup, now is the time to share them.

EDIT:/fixed links

 

[amb]

Looks good dgardner. One suggestion: In your schematic #4, you might want to wire the amp's signal ground lines to the XLR output connectors' pin 1 anyway, even though the headphones won't use it. This is so that if you connect the balanced output of the headphone amp to another device's balanced input using standard XLR-XLR interconnect cables, it would also work.