Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

4W/2W Converter and 2175Hz Notch Filter

Status
Not open for further replies.

Bruce Glazier

New Member
upload_2015-7-28_9-6-3.png


upload_2015-7-28_9-6-47.png


upload_2015-7-28_9-10-43.png




Hello,
Looking for some advice on the forum again since I had really useful help the least time I posted. The scope picture (Picture 3) is at the output of Picture 1 on C36 with no input connected, inputs being C28 (Tx+) & C29 (Tx-) and also C30 (Rx+) & C31 (Rx-). This noise is present on the output and I can't seem to figure out what is causing it. Picture 2 is the power circuit, which takes +12VDC in and creates the +5V and -5V rails for the op-amps. The IC is a MAX660ESA (https://pdfserv.maximintegrated.com/en/ds/MAX660.pdf). I had managed to eliminate some noise on the rails by connected FC to VCC which switched the MAX660ESA to operate in 40KHz mode instead of 10KHz mode.
Also, FYI, the notch filter portion is not my design but it is a design taken from a working module so any ideas about what could be changed there would be appreciated because I honestly have a hard time wrapping my head around it.
Thank you for any input and I am open to answer any questions.
-Bruce
 
Step one:
Remove +12V, and stop U2 from working. Power from a known good +/-5V supply.
OR
If this is too much work; stop U2 and power from +12V and a good -5V supply.

I want to know that the noise is (is not) coming from U2.

Step 2:
Look at the output of each op-amp. What ones are quite/ not quite.

On your last picture I see two frequencies. What are they?
 
Opamps do not need a negative supply if they are biased at about half the positive supply voltage. The Mickey Mouse switching negative voltage generator is causing the modulation (about 70kHz) and the 757kHz is one of the opamps oscillating.
 
Thank you for your input ronsimpson and audioguru. I'm pretty new in-practice so I appreciate the help. I'm going to try your suggestion's and report back.
 
So I removed the U2 and attached a -5V source and the noise was gone, so thanks ronsimpson. audioguru, I unfortunately do not have the luxury of redesigning this for my current run, I have to use these boards so using +12VDC and GND and biasing the op-amps isn't really an option. I'm wondering if there is anything I can do to reduce the oscillation on the switching supply or if there is a more suitable replacement?
Thanks again!
-Bruce
 
"The disadvantages of a switcher are that the switching component adds quite a bit of noise to the output power. I’ve seen many different switching frequency signatures. The better switchers have all of their noise up in the megahertz range, so that there’s virtually no noise in the audio band. More commonly the switching frequency is up in the tens of kHz with some sub-harmonics extending into the audio band, plus possibly also some low-frequency ripple. Then there are the really cheap switchers, with a switching frequency right in the middle of the audio band." https://tangentsoft.net/audio/opamp-ps.html

I just found this explanation and I am wondering something, would I be better off over-driving the switching regulator with an external crystal in the MHz range? It seems that is an option but that might get kind of messy using jumper wire to add it to the board. I think the best solution would be to drop a better IC in it's place. I am having a tough time locating a good fit.

Thanks,
-Bruce
 
Hi Bruce,
Now that the source of the problem has been identified you could try more decoupling on the - 5 volt rail and possibly the +12 volt rail. First I would suggest that with the negative rail still powered from the external -5 volts that you re connect the +12 volt to the MAX660. If the noise comes back then noise is being injected into the +12 rail by the MAX660. You could try adding a decoupling capacitor of about 100 nF right between ground pin and the 12 volt pins of the MAX660 and increasing the value of C46 If you post the results of these test myself or others will advise on the next step. Posting a picture of the copper side of the PCB with the ground tracks highlighted would also be helpful in case there are any sections of the track that both carry current to the MAX660 and analogue signals.

Les.
 
Thanks Les,
I must clarify though that the input to U15, (the +5VDC voltage regulator) is the +12VDC and the input to the negative voltage switching regulator is the +5VDC. For my test I left the +12VDC and the +5VDC in circuit and replaced the -5VDC regulator only. So I will try your test to check if the MAX660 is injecting noise into the +5VDC rail (sorry I should have clarified this better).
-Bruce
 
Hi Bruce,
I was mistaken. I assumed U15 was 12 volts out. I have no excuse as you stated in the original post that the op amps ran from + and - 5V. It is a nice change to see concise information about the problem. There are too many posts where finding out what the question is the most difficult part. The comments I made still apply about first identifying if the MAX660 is injecting noise into the +5V supply and then trying to filter it out before dealing with noise on the -5V supply.

Les.
 
Hi Les,
So it seems that with the MAX660 still on the board, except with the Vout pin lifted, and the -5V powered externally, the noise is still not present. So that means that the MAX660 is not injecting noise back into the +5V rail but only onto the -5V rail. With that discovery, I think I should focus on how I can best patch the MAX660 (drive with external crystal/decouple better?) OR change to another -5V regulator with a much higher frequency built-in oscillator.

I think I am going to see if attaching a crystal is a viable option and report back.
Thanks again,
-Bruce
 
Hi Bruce,
That's progress. First try adding a 100nF capacitor between pin 3 and pin 5 (Ground and -5 out.) of the MAX660 and increasing the value of C48. If that does not fix the problem then add a resistor of about 100 ohms in series with the -5 volts to the op amps and add another capacitor between ground and the op amp side of the resistor. Try about a 10 uF capacitor (and maybe a 100nF in parallel with it.) This is trying to filter the output noise from the MAX660. Check that the 100 ohm resistor does not drop the -5 volts too much. If that does not work then I have run out of suggestions for the moment.

Les.
 
At U2 pin5 there is a cap to ground. Good.
You need to cut this and make a CRC filter. (so U2-5 and Cap is connected to VEE through a low pass filter)
I would add a 10 ohm resistor or something.
upload_2015-7-28_12-32-21.png

Sorry I can not draw a good resistor. LOL
The RC should be set to reduce the signal at the switching frequency.
A CLC filter would be better.
The value of R can be set so only 0.1 volts are lost across it.
 

Attachments

  • upload_2015-7-28_12-31-46.png
    upload_2015-7-28_12-31-46.png
    11.5 KB · Views: 246
Which opamps does the circuit use? Maybe their supply noise rejection is poor at the fairly high frequency of the charge pump.
Are you using one of the recommended brands of low ESR output capacitor?
 
Remind us the acceptance criteria for this design?
and how is this related? 4W/2W Converter and 2175Hz Notch Filter

Vin=12Vdc +/-?, two differential AC coupled signals combined with gain =? and filter = ?
  • V+ = 5V @ ?mA , Pout= ? noise max. = ??
  • V- = -5V @ ?mA Pout=? noise max = ??
Keep in mind the output impedance of +5V must be much lower than the dynamic load for good regulation. The dynamic load is limited by the ESR of the 10uF switched caps and internal resistance.
 
ronsimpson thanks very much, I will see if putting in a simple filter does the trick tomorrow.

audioguru Here is the op-amp I used: https://www.ti.com/lit/gpn/lmc6482 Admittedly, I did the design in a high pressure situation so I sort of rushed it and I did not notice that section about the low ESR caps to reduce voltage ripple before. I may have to try that to see if it makes a big difference.

Tony Stewart Good point Tony, I've failed to describe the basic parameters of the circuit. Vin is a regulated +12VDC from a AC/DC power pack. Ideally there shouldn't be any gain or loss through at the output, i.e. the output should match a single input OR the sum of both inputs. I can say that after testing using the -5V external supply, the noise range became acceptable (signal looked very clean) when using a 1.004KHz Sine Wave of -10dbm but I don't really have a defined threshold. The circuit should do two things, 1) Mix the differential AC inputs and 2) Filter out frequencies around 2175Hz. I'm not entirely sure what the current draw is on each rail at this time but I can say there are four identical circuits to Picture 1 on the board all using the same rails. And to address your last comment, thanks, I get the gist of what your saying but I'm not entirely sure if I can calculate what those values are and what they should be with my limited knowledge of impedance's. I've calculated the input impedance of the amplifier portion using **broken link removed** but I'm not sure how to apply that here. I do recall some theories from circuit design but I haven't touched that stuff since I graduated 3 years ago.

Thanks again all!
Bruce
 
Last edited:
The opamps are fairly slow but still have a power supply rejection of more than 100 times at the frequency of the negative voltage generator.
I think the output cap of the negative voltage generator does not filter its noises properly.
The very high frequency might be caused by positive feedback caused by adjacent coupling on your pcb.
 
Thanks audioguru, this is similar to your first comment about the two frequencies observes in the scope picture. From what I've understood from yours and ronsimpson's suggestions, I need to filter the switching frequency from the negative voltage convertor. I will be back at it tomorrow morning and will post with results. I maybe close to a feasible solution to my issue....thanks again.
-Bruce
 
Imagine what happens with two different capacitors of different values and ESR and charge voltages are shorted together.

THis is basically what happens when the pump regulator drives a square wave on the Cap with a positive diode clamp to each stage. The initial transient voltage will be the ESR ratio of source/(load+source) followed by a voltage sag determined by the capacitance ratio, load/(load+source)

So you want a lower source impedance ( much larger C and lower ESR) for this. THe Charge pump has internal ESR of 10 Ohms max. and limited to 100mA at max voltage.

Attenuating Noise has several general considerations;
Conducted noise on shared grounds and supply.... isolation
Radiated noise from coupling .... separation or shielding
%Load regulation noise from %ratio of source/load impedance (above),
Feedback instability at unity gain. ... stray positive feedback.. or insufficent phase margin.... test with small feedback caps to isolate cause.

Dynamically switching currents by driving capacitive charge pumps.
 
ronsimpson I tried putting in just a 1Ω resistor for an RC to make a cutoff frequency of ~14KHz but I did not change the two frequencies present at the output, which are 40KHz and 500KHz waveforms.

Tony Stewart I suppose I will try to change out the capacitor with a larger value and a lower ESR.
 
Since internal ESR is 10 Ohms max , adding 1 Ohm is insignificant.

I hope this is a tight layout, not long wires on a protoboard. CHeck noise on each output stage. If so, then add RC snubber to output or CM filter on input.


I would expect a twin-T active notch to perform with a deeper notch and flatter passband.

I thought the switch rate option was 80kHz.

Considering all the trouble with noise on this AC coupled low gain amplifier, I think you'd be far better off without the negative Vee and bias at Vin/2 using 12V if you need large signals or just an LDO such as 8V and AC couple the output.

Use a CM torroidal choke around all long input wire connections.
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top