Any interest in GC input buffer and power supply regulator pcb's?

[randytsuch]

If boards do not become available, I'll probably try etching some. I know nothing about it, but if you can etch one side, why not two? Aren't the only extra requirements be that one side is protected while the other etched and that the respective patterns be reasonably well mated? Can't be that hard for a small board? I like the idea of compact layouts (just because).

Sheldon

I have not etched before, but I would think you would just etch both sides at once.

The thing you can't do at home is make plated through holes (at least I don't know how) This means the pad on the top side won't be connected to the pad on the bottom, unless you run a wire through the hole.

For lining up top and bottom, there should be registration holes in the corners of the board (at least three of the corners). This should make it pretty easy to line up.

Randy

 

[Sheldon]

Re: It works!

I still have a little tuning to do. I'm getting about 90mv offset in one channel and some hum, about 75mv in the other but very low hum. Maybe a bad joint or grounding arrangement to chase down. Time for a break before I tackle those.

Sheldon

So time to tune. The setup as described earlier is one of Brian's non-inverted boards with Pedja's jfet buffer and his discreet reg. p.s. I'm getting a little more hum on one channel, though not bad (I have to be within a foot or so on 90 db speakers, as compared to a few inches for the quieter channel). And I'd like to reduce the DC offset, though I'm not sure it's essential. The max I see is 100mv which at 8 ohms is still only 1.25 mw. The table below is self explanatory, except for the "buffer grounded" column. Here I have the buffer output shorted to the signal ground on the amp board. Here's the data:

Normal Operation ... Buffer Grounded

....L ........ R ......L....... R
AC 0.4mv 1.1 mv 0.4mv 12 mv
DC -75mv -100mv -10mv +140mv

One other note: When I touch the buffer output in the right channel with the multimeter probe, it get's a little quieter in normal mode, but the hum get's louder in the "buffer grounded" mode. Anything obvious I should look for?


Sheldon

 

[moving_electron]

Aren't the only extra requirements be that one side is protected while the other etched and that the respective patterns be reasonably well mated?

The degree of difficulty is higher on a double sided board but not prohibitively so.

1) the masking for the two sides needs to be aligned so that the pads etc. will be in the right place on each side. For most boards this has to be done pretty accurately.

2) You have to mask, expose, develop and etch each side seperately and as you mentioned each side has to be protected as the other side is worked on.

3) you need double sided medium which is somewhat more expensive, especially if you are using presensitized boards.

4) your chance of success on a double sided board is lower - on the order of the square of the chance of success on a single sided board. So if you have a success rate of .75 on a single side, your success rate on a double sided board will be on the order of .75x.75= 56% If your rate is 50% on a single sided board then your success rate on a double sided board will be something like .5x.5= 25%. [Of course if your chance of success on single sided is .99999 then your chance on a double sided is very high .]

So it probably comes down to the goal:
If the goal is to have a board artwork that can enable the most people and facilitate the most success for first time etchers then single sided is the way to go.

If the goal is to have the most compact board with no jumpers then double sided has the advantage. Quite a few people could make this work but some will be left out or unsuccessful.

 

[Sheldon]

Re: Re: It works!

So time to tune. The setup as described earlier is one of Brian's non-inverted boards with Pedja's jfet buffer and his discreet reg. p.s. I'm getting a little more hum on one channel, though not bad (I have to be within a foot or so on 90 db speakers, as compared to a few inches for the quieter channel). And I'd like to reduce the DC offset, though I'm not sure it's essential. The max I see is 100mv which at 8 ohms is still only 1.25 mw. The table below is self explanatory, except for the "buffer grounded" column. Here I have the buffer output shorted to the signal ground on the amp board. Here's the data:

Normal Operation ... Buffer Grounded

....L ........ R ......L....... R
AC 0.4mv 1.1 mv 0.4mv 12 mv
DC -75mv -100mv -10mv +140mv

One other note: When I touch the buffer output in the right channel with the multimeter probe, it get's a little quieter in normal mode, but the hum get's louder in the "buffer grounded" mode. Anything obvious I should look for?


Sheldon

Fixed the hum problem. Discovered that I has miswired one of the LP filter caps. Hum from either side is now inaudible from 10-20 cm away. Last issue is the DC offset. Sorry for the wasted bandwidth.

Sheldon

 

[Pedja]

Here we go again. Bypass caps are not on-board at all, though holes are provided if you choose to mount under the board, or you can mount straight to the pins or whatnot.

Maybe: R2 above R1; then the output bypass cap right of this new R2 position; the reg itself maybe moved a bit down, so R1 becomes closer to it; previous R2 position stays empty so this place may be used for the input bypass cap.

The 1M resistor is rear-mounted. There's room to slide it over and mount above the two resistors if that would be preferable.

What I was initially about was:

An externally hosted image should be here but it was not working when we last tested it.

Then I thought it can be done this way:

An externally hosted image should be here but it was not working when we last tested it.

Then I thought about the strayed capacitance from the output to the input:

An externally hosted image should be here but it was not working when we last tested it.

But I thought also the cascode can be made tighter, though the V- supply path pays the price…

An externally hosted image should be here but it was not working when we last tested it.

"Incoming and outgoing traces around the reg itself are narrowing" - if this refers to the traces getting narrow to clear the other reg pins, a question: Is it better to have smaller traces that are consistent in size vs. thicker and narrow right at the reg?

It is OK now, previously they were getting narrow too early. Btw, you are not supposed to use just one track width, you may combine a few tracks of less width to get better results.

Do not be too upset about anything of these notices, take them as a final thoughts…

Two more important things at the moment... Where the filter is supposed (if the filter is supposed) to be mounted? And, on the amps boards, the cap from the mute pin to ground is missing.

Pedja

 

[Sheldon]

Re: Re: Re: It works!

Last issue is the DC offset. Sorry for the wasted bandwidth.

Sheldon

Did a bit of searching, which I should have done before I asked about the offset. As I understand it, the DC offset, assuming it isn't a function of DC input (which shouldn't be the case here, as the buffers are coupled to the amp boards via a 4.7uf cap) is caused by bias current in the amp chip. So, as Carlos pointed out in another thread(http://www.diyaudio.com/forums/showthread.php?postid=438843#post438843), it's somewhat of an inherent problem with the NI versions. And its magnitude depends at least partly on manufacturing variation. So how about injecting a little DC offset as suggested here: http://www.diyaudio.com/forums/showthread.php?postid=385198#post385198

Anyone see problems with this approach? I could use the output from the regulated supply or from the buffer regulators.

Sheldon

 

[Pedja]

Sheldon,

You may put 10k instead of 22k input resistor on the non-inverting input (ignore filter’s frequency for the moment). You may go also somewhat below 10k. I’d expect this to solve the issue in certain degree. No idea if it will be sufficient (for you at least).

Pedja

 

[cjd]

Hmm...

why hadn't I considered moving the pins to allow more space?!

FWIW, I've only found any posts to be quite helpful. I have no illusions about being a great designer in this arena, as I know I'm missing too many rules and standard practices (all of which have solid reasons for existing).

Then again, it could just be that none of you have suffered through an art critique (which, I must say, are incredibly harsh and demanding).

I'll continue pondering these. Every few days I give them a fresh look over and invariably find things to tweak.

Pedja, your illustration sequence is amazingly helpful in ways beyond just providing layout suggestion. Thanks!

C

 

[Pedja]

Btw, you are not supposed to use just one track width, you may combine a few tracks of less width to get better results.

one track width should be one track of the given width

So how about injecting a little DC offset as suggested here: http://www.diyaudio.com/forums/showthread.php?postid=385198#post385198
Anyone see problems with this approach? I could use the output from the regulated supply or from the buffer regulators.

To answer the question, I think it won't work. Even in your case of defined DC resistance seen by the non-inverting input.

Pedja

 

[Sheldon]

Sheldon,

You may put 10k instead of 22k input resistor on the non-inverting input (ignore filter’s frequency for the moment). You may go also somewhat below 10k. I’d expect this to solve the issue in certain degree. No idea if it will be sufficient (for you at least).

Pedja

Thanks Pedja,

I tried 9.5K and 6.6K. It does indeed lower the offset, as would be expected if the offset is a result of bias current. I assume that would raise the LP filter cutoff some? As the offset is under 50mv, I can't imagine that it would be a problem. Even at 100mv it may not be a problem. Just the ordinary quest for perfection, I guess.

To answer the question, I think it won't work. Even in your case of defined DC resistance seen by the non-inverting input.

Pedja

Would the injected offset affect the LP, even if a fairly high value resistor is used? Or does it create a feedback issue? I assume that when you say it won't work that it would have an adverse affect on sonics? National looks like they do something similar when using positive power supply.

This continues to be an excellent learning experience, thanks to you and all the experienced users that are willing to share.

Sheldon

 

[Pedja]

I tried 9.5K and 6.6K. It does indeed lower the offset, as would be expected if the offset is a result of bias current. I assume that would raise the LP filter cutoff some? As the offset is under 50mv, I can't imagine that it would be a problem. Even at 100mv it may not be a problem. Just the ordinary quest for perfection, I guess.

Yes, I personally wouldn't care much about that. With 9k5 you may use filter values of 33nF/1kOhm/4n7, which will still may be generally fine with 6k6. But hey, you might like it better as it is at this moment? I’ve learnt from Joe “Just Listen To it” Rasmussen that the input LP filters should be tuned by ear…

Would the injected offset affect the LP, even if a fairly high value resistor is used? Or does it create a feedback issue? I assume that when you say it won't work that it would have an adverse affect on sonics? National looks like they do something similar when using positive power supply.

In the given schematic the non-inverting’s input DC path varies with a pot’s position and DC will change with a pot’s position. (Btw the schematic in fact comprises the cap inside the feedback loop while the guess is it doesn’t.)

If we assume this path is constant (i.e. AC coupled source) it may be set to some stable value, but you will need megamultiturn pot for this. The one I normally use is 20 turns unit. The offset on the output to be corrected is, say, 100mV. The gain is, say, 25. The voltage we need at the input is 4mV. We are starting from, say, +/-25V, so 50V at the pot’s ends. Every turn is 2.5V. To tune it for 4mV, we must operate within 1000th part of one circle. Now, we may start with 5V and get more relaxed demand to tune it inside 100th part…

Pedja

 

[MalichiConstant]

Think small

"Now, we may start with 5V and get more relaxed demand to tune it inside 100th part…...."

Gee... most the applications I have seen use resistors to set the voltage drop across the trim pot to a few tens of millivolts swing. If you are having to null out more than a few millivolts you have other problems like mismatched resistances to ground for the positive and negative inputs to the amplifier IC. To large vaules for the feedback and input resistors will cause problems also. The offset errors are mainly from the input transistor base currents and the larger the value of the resistors the larger the offset voltage. Do some RC filtering on the resistor feeds to the trimpot or you can wind up with a small portion supply noise fed into the input of the amp which you don't want. Often a couple of diodes to ground with the pot across them are used. The low drop with schottky diodes is even better. for reducing the voltage range and decreasing the pots sensitivity. When using just resistors Pick the right values so the pot wiper can only swing a few 10 of millivolts and you can get by with a single turn pot. Its done all the time and a single turn trim pot gives you some visual feedback or where it is set.

 

[Sheldon]

Dc Offset

Thanks Guys,

Once again; ask a question and get thoughtful and intelligible answers - and from as far away as halfway around the world. Not just solutions, but at least two possibilities and the explanations to go with. That's very cool. Thanks again for taking the time to enlighten a beginner. By the way, the amp sounds very good.

Sheldon

 

[Pedja]

Welcome back, Malichi, I think we’ve been missing you here, including those who claim they have not. They need you in fact. For whatever the purpose is. That is a life.

(Yet I must say that we did not at all suspect that you’ll be back. Otherwise the fact you’ve been gone couldn’t have any sense.)

Gee... most the applications I have seen use resistors to set the voltage drop across the trim pot to a few tens of millivolts swing.

Can be done with resistors, but…

When using just resistors Pick the right values so the pot wiper can only swing a few 10 of millivolts and you can get by with a single turn pot. Its done all the time and a single turn trim pot gives you some visual feedback or where it is set.

A few tens (say 50) of mV means a 1000th of the starting 50V. Now you’ll need to know the used voltages very precisely and you will need probably 0.1% resistors, maybe weird values if voltages are (and probably are) mismatched. Single turn pot? Maybe, but you will need 0.01%.

I thought if you really need small offset, you approach it generally different way.

If you are having to null out more than a few millivolts you have other problems like mismatched resistances to ground for the positive and negative inputs to the amplifier IC.

I thought this was supposed to solve something like that. But I may be wrong.

Pedja

 

[Sheldon]

Dc Offset

If I understand Malichi's post correctly, the circuit could go something like this:

10k resistor and diode to ground on each side of the p.s. (+/-25V). With a schottky diode, the voltage between the two leads, taken between resistorand diode, would be about 800mv. To get resolution to about 5 mv would require 0.5% precision. If that's not practical with a 1 turn pot, could an additional divider be used between the pot and diode? Would the resistor/diode approach sufficiently eliminate the noise issues so that an RC filter would not be required for the offset circuit? If the pot is a high enough value (say 50k) could pots for both channels use one set of resistor/diodes?

Sheldon

 

[MalichiConstant]

Absolutlely. I think trimpots down in the 500 ohm and less region might be more typical and would be a small value in relation to the 10K and greater values seen for typical input resistors. The resistor feeds to the diodes can be two resistors with a capacitor to ground. Make the resistor vaules equal which give the highest resitance for feed to the capactior to ground.

Also, the point of using a diode to ground is minimize changes in voltage to the trimpots for changes in the raw supply voltage. The attenuation of supply voltage changes will be about the dynamic impedance of the diode divided by the resistance biasing the diode from the supply. For a few millamps bias I think the diode has an impedance of a few ohms, so it is pretty insensitive to the supply voltage changes

 

[Sheldon]

disreet reg

Here's the amp section I'm using: http://www.diyaudio.com/forums/attachment.php?s=&postid=331888

The input resistor is 220 ohms. I assume that the offset would be injected at the signal input of the amp, not the buffer. So if both channels were fed with one resistor/diode pair and 500 ohm pots (which would be set near the middle), I'd get about 500 ohms from one channel to the other. To eliminate crosstalk, I'd guess would require either higher value pots or completely separate circuits for each channel (best, I'm sure, but hey it's worth a try to use fewer parts)? The RC filter would be a low pass filter set to pass anything higher than 10hz or so to ground?

This might be worth a shot, just for the exercise. And that's a primary motivation for this whole project anyway.


Sheldon

 

[sek]

Hi,

may I ask a quick question regarding the turn on delay by muting the LM3886?

As I'm developing a different kind of regulated GC using LM3875 or LM3876 here , I want to make sure that no turn-on noise occurs with a simple 3-pin regulator PSU.

So, does a turn-on noise only occur due to the fact that the input buffer powers up slowly and experiences offset "noise" at turn-on? In other words, can I use the LM3875 (without muting circuit) safely with a regulator supply?

My circuit doesn't contain a buffer, one can assume that the "source" circuit is turned on and stable once the GC is powered up.

I appreciate your opinion and apologise.

Thanks,
Sebastian.

 

[Sheldon]

Hi,

may I ask a quick question regarding the turn on delay by muting the LM3886?

As I'm developing a different kind of regulated GC using LM3875 or LM3876 here , I want to make sure that no turn-on noise occurs with a simple 3-pin regulator PSU.

So, does a turn-on noise only occur due to the fact that the input buffer powers up slowly and experiences offset "noise" at turn-on? In other words, can I use the LM3875 (without muting circuit) safely with a regulator supply?

My circuit doesn't contain a buffer, one can assume that the "source" circuit is turned on and stable once the GC is powered up.

I appreciate your opinion and apologise.

Thanks,
Sebastian.

Check Pedja's site. He addressed that difference. Apparently, there is a slight "click" with the reg. p.s., and a significantly louder "boom" with the buffer. With the reg. alone, you should be ok. I have a position on my selector switch which shunts the buffer output to ground, which I use during turn-on. The unit generates very little heat at idle, so I may just leave it on and switch the selector to the safety position when I'm not using it.

Sheldon

 

[sek]

Yep I'm actually inspired by Pedja's site.

Apparently, there is a slight "click" with the reg. p.s.

That's the question...

The click might result from the two regulators starting up asynchronously.
How does the LM3875 start up when connected to an unregulated supply?

I'm currently thinking about implementing a shutdown circuitry into the regulator part (a BJT and a voltage divider as suggested in the LT108x et al datasheet) in order to "control" the turn-on after the transformer and pre-caps are stable... but then I could either use the LM3876/3886's shutdown...

Any experience about the difference in turn-on behaviour of the LM3875 between regulated and unregulated PSU anyone?

Thanks,
Sebastian.