Help to design a PCB for my PI design!

[Qiaozhi]

Yes! the noise is about 10KHz. BTW the before prototype PCB wasn't like this...
I will design a PCB... If I get free time.

Best regards,
1843

I doubt very much that the noise is due to the PCB layout.

Have you tried disconnecting the power supply circuit and hooking the PI up to a bench supply? Or you could simply disable the 7660, and feed the 7805s from a second 12V source. That will confirm if the noise is actually due to the 7660.
It the noise is still there, then it could be due to ground bounce caused by the ATMega8.

[1843]

OK porkluvr, I'm going to do some experiments...

Best regards,
1843

[porkluvr]

porkluvr,
Thanks a lot for the schematic.
There are some errors in the schmatic:
Best regards,
1843

I wasn't intending any components changes but It was late (and it's late now). No excuse, but I was tired.

About your voltage converter noise,
Try making a pi filter on the charge pump output by adding one L and one C, between charge pump and the 7805.
68uH and 220uF might work but the best combination would depend on your clock frequency and how much noise is too much. I can offer no mathematical analysis for selecting the best component values.

It's late again. (big yawn)

[1843]

Are you sure the noise is coming from the supply and is not from the micro? I had allot of problems with digital signal noise affecting the analog circuits in my design.

Yes! the noise is about 10KHz. BTW the before prototype PCB wasn't like this...
I will design a PCB... If I get free time.

Best regards,
1843

[hobbes_lives]

Are you sure the noise is coming from the supply and is not from the micro? I had allot of problems with digital signal noise affecting the analog circuits in my design.

[1843]

I tested the above circuit. Not real improvement I observed. The 7660 noise is still there...affecting the preamp.

Regards,
1843

[1843]

I found the best way! Using a totem-pole (same as the gate driver) DC converter instead of the backlight controller for supplying the preamp.
100% synced!


The backlight controller is not critical.

Best regards,
1843

[Aziz]

But the diode in series with the MOSFET compensate it. Doesn't it?

1843

No, this wouldn't help much.

Try replacing the 7660 with a discrete BJT DC/DC converter or mosfet driver IC DC/DC converter. This will definitely help.
Aziz

[1843]

But the diode in series with the MOSFET compensate it. Doesn't it?

1843

[Aziz]

Hi 1843,

One way is to use a P-MOSFET for driving the coil, and a synced 7660 to provide the negative rail.
Like the PD3200. What's your idea?

Regards,
1843

P-MOSFET's have a much lower avalanche breakdown voltage. We need as high flyback voltages as possible, to "kick the target stronger" and get more target signals.
Aziz

[1843]

One way is to use a P-MOSFET for driving the coil, and a synced 7660 to provide the negative rail.
Like the PD3200. What's your idea?

Regards,
1843

[Aziz]

Hi 1843,

Hi Aziz,

Do you think if the DC converter be synced, It can provide enough current for all the parts?

I tested it when Carl suggested, but the ATmega was restarting all the time.(because of voltage drop).

Also the working frequency of 7660 gets half if it is synced.


Best regards,
1843

the 7660 is specified only for low current draws. Synch'ing this would not help in this case. A discrete DC/DC converter is capable to deliver more current. Look to my proposal http://www.geotech1.com/forums/showp...&postcount=454 (post #454) for a discrete solution. It will deliver much more current. The higher the current source, the lower the ripple noises and voltage drops.

A simple high current DC/DC converter could be made with mosfet driver IC's (series TC426, containing two drivers). These chips can be configured for different DC/DC converters or paralleled for more current draw. See the data sheet for more info: http://ece-www.colorado.edu/~ecen451...p2/tc426_8.pdf


I personally wouldn't supply the micro on the DC/DC converter side. But it would require much design changes on your schematics. So another DC/DC converter could help here. The clock for the DC/DC converter must be sourced from the Timer/PWM output of the micro. A very simple synchronisation could be made, when on every cycle start (after the TX-on time), the PWM is reprogrammed again. This will compensate the uneven timings to the cycle period. A clock frequency of 10-50 kHz should be enough.

Aziz

[1843]

Hi all,

the reason for the bad performance of the two stage amplifier version is as follows:
The first stage is an inverting amplifier with its input resistance of 1k. This is too low for such tiny signals. The non-inverting version with NE5534 has roughly 100k input resistance (=input resistance of op-amp).

Another issue is the DC/DC converter. It will produce more ripple noises, when there is more current draw. As I mentioned, the DC/DC converter must be synchronized to the cycle period to have always same and repetitive ripple noises (correlation to cycle period).

A discrete DC/DC converter should perhaps be considered to have lower ripple noises. But there is no other way: it must be synchronized.

Aziz

Hi Aziz,

Do you think if the DC converter be synced, It can provide enough current for all the parts?

I tested it when Carl suggested, but the ATmega was restarting all the time.(because of voltage drop).

Also the working frequency of 7660 gets half if it is synced.


Best regards,
1843

[Aziz]

Hi all,

the reason for the bad performance of the two stage amplifier version is as follows:
The first stage is an inverting amplifier with its input resistance of 1k. This is too low for such tiny signals. The non-inverting version with NE5534 has roughly 100k input resistance (=input resistance of op-amp).

Another issue is the DC/DC converter. It will produce more ripple noises, when there is more current draw. As I mentioned, the DC/DC converter must be synchronized to the cycle period to have always same and repetitive ripple noises (correlation to cycle period).

A discrete DC/DC converter should perhaps be considered to have lower ripple noises. But there is no other way: it must be synchronized.

Aziz

It is unpolite and out of any cultural behavior what i done in my recent posts. I am sorry because of that. But sometimes we must use certain methods and rude "politics" to achieve some results. If i wasn't nerve and harras this man, God only knows when he will decide to share hex file here. Understand?