EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Prehistoricman on October 01, 2017, 05:01:00 am
-
I have an audio transceiver system that I bought from aliexpress (https://www.aliexpress.com/item/1-PCS-2-4G-NRF24L01-Wireless-Digital-Audio-Transceiver-Module/32314185928.html). I'm using this to transfer audio directly from a guitar pickup. Even without the screaming interference, this is a problem in itself. The input impedance for this transmitter is absolutely abominable at 600 ohms. I've already tried the guitar directly into it (wishful thinking) and it's pretty quiet and has treble loss.
The interference showed up in this guitar -> transmitter setup. I connected the guitar ground to power ground and the signal ground. I clearly hadn't done my prayers to star grounding that day and I got my dueful punishment.
(https://www.eevblog.com/forum/projects/wireless-audio-transmitter-interference-and-grounding-issue/?action=dlattach;attach=356293;image)
(As you can see, L, R, and 5V each have their own ground. And the guy who wrote the silkscreen had a neurotic twitch around the rotate and mirror shortcut keys)
So in this situation, the solution to the interference was easy (kinda... there was still a small output noise). Now, to address the input impedance...
guitar -> unity gain buffer (LM324) -> transmitter
(https://www.eevblog.com/forum/projects/wireless-audio-transmitter-interference-and-grounding-issue/?action=dlattach;attach=356291;image)
This is one of a few arrangements I've tried regarding where to connect the grounds. This one has the opamp's negative supply run through the signal ground pin.
I also noticed here that the transmitter appeared to be interfering magnetically too. Upon moving it near the guitar's ground, the audio out squealed like a baby. Having the antenna pointing out vs lying on the board also appeared to make some difference.
I've also tried using an audio transformer between the opamp and the transmitter. No difference
So my question is:
Can I arrange the grounds in some way to stop this OR, how do I use the other 3 opamps in my LM324 to create a differential output that will play nicely with the signal ground?
-
Do you have a schematic of transmitter. "Normally" in split ground
designs you want them all to join at a specific common point. Here
you have one ground being modulated by the xmitter power and the
ADC on the xmitter. I would start there.
Also pay careful attention to bypassing, combination of polymer
tantalum and ceramic disk appropriate. Consider shielded cable
from mic to board as well.
Regards, Dana.
-
Thanks Dana
I looked up NRF24L01 and it just seems to be the TX chip.
I looked on the board and found the marking "XL-01m" and once looking this up, I found this:
http://www.lillyelectronics.com/download/2.4G-NRF24L01-XL-01M-V1.pdf (http://www.lillyelectronics.com/download/2.4G-NRF24L01-XL-01M-V1.pdf)
Gives useful advice such as:
"Ground near the power supply as far as possible" p. 8
I placed a 100nF ceramic right next to the transmitter and that didn't appear to help
I placed a 30nH huge inductor on the 5V to the transmitter, no difference
I somehow managed to swap the left/right channels once. I think my multimeter tip hit something. Very strange, and it reverted after a while
This board does have a 3.6V regulator and input decoupling so I don't really know what to think any more. Shouldn't those two things basically eliminate power supply noise?
-
I attempted to make a similar setup about 8 years ago.. using an Atmega and NRF24L01 to transmit 16-bit @48kHz (for guitar as well! but tested with mono audio).
Even though the AVR can read I2S with its USART in SPI mode, and the NRF24L01 can transmit at full 2Mbps, (= ~1Mpbs user through-put) there was just no way it could actually achieve the throughput. It dropped samples, and got plenty of errors. I strongly suspect that these boards either drop the sample rate down, or skip samples, both of which degrade the audio quality. Then of course there is the inevitable bit error rate of wireless links. Nordic semi released a chip dedicated to wireless stereo audio. Although it find its way into a few products it was quickly dropped because even the 4Mbps channel it used couldn't really deliver 'clean' low latency stereo audio without pops/clicks, and dropouts.
In order to rule some things out, a quick check list.
-"Clean" power supplies for both transmitter and receiver. Use Batteries.
- Scope the power supply lines on both boards, whilst they may have regulators, that doesn't mean to say they have the correct decoupling capacitance or ESR - for some LDO's this matters a great deal.
- If you don't have a scope, a quick and dirty method, hook the power supply up to a guitar amp via a 100n cap. If the LDO is oscillating you will hear it.
- Your buffer circuit has a low input impedance for guitar pickups. Using the 10k||4.7k gives input impedance of ~3.2k. Make those 1Meg each.
- The boards you have include RF power amplifier and LNA, in order to boost range. Without a shield covering these, they can kick out quite a bit of EMI, and also pick it up.. in fact, without careful design, adding an RF front end like this can *reduce* range, by increasing the bit error rate. Keep the buffer near the guitar, and the transmitter a few metres away. With the receiver even further away (20m+).
If it has the RF front end I think it has, then the output power is 100mW, 20dB, which is LOUD. It could easily swamp the receiver if it is near.
I'm not trying to claim these boards are trash, they *may* be able to carry fair quality audio, but I strongly suspect that you will not get a quiet clean audio channel with them. I spent a long time investigating wireless lossless digital audio, and I'm sure mono (for guitar) can be done on these sorts of RF units (CC2500, NRF24L01 etc..) but requires a more complicated protocol than the AVR can achieve. They are probably just sending raw audio samples over the link, and not caring about lost packets.
If you manage to get them working reasonably well, I'll eat my hat, buy a pair, and test them myself.
tl:dr - don't think the boards can transmit audio at the quality they claim.
-
-"Clean" power supplies for both transmitter and receiver. Use Batteries.
I did this initially, then I get some small boost boards to get to about 5.2V. No apparent difference in noise at that point.
- Scope the power supply lines on both boards, whilst they may have regulators, that doesn't mean to say they have the correct decoupling capacitance or ESR - for some LDO's this matters a great deal.
- If you don't have a scope, a quick and dirty method, hook the power supply up to a guitar amp via a 100n cap. If the LDO is oscillating you will hear it.
Okay I did it quick and dirty and all the positive rail had this 1.6KHz noise on it, even after the regulator. Basically every place I probed (including grounds, to a degree) had the same noise on it.
I probed every pin of the ADC and the only pin without noise (including digital and analog supply and ground) was pin 12, which is a digital clock input.
- Your buffer circuit has a low input impedance for guitar pickups. Using the 10k||4.7k gives input impedance of ~3.2k. Make those 1Meg each.
I was simply hastily putting it together using parts that were on my desk. I should buy a resistor kit :-[ Also I can't set the DC offset to be mid-rail as the opamp has a 3V headroom at 5V supply.
- The boards you have include RF power amplifier and LNA, in order to boost range. Without a shield covering these, they can kick out quite a bit of EMI, and also pick it up.. in fact, without careful design, adding an RF front end like this can *reduce* range, by increasing the bit error rate. Keep the buffer near the guitar, and the transmitter a few metres away. With the receiver even further away (20m+).
If it has the RF front end I think it has, then the output power is 100mW, 20dB, which is LOUD. It could easily swamp the receiver if it is near.
Yeah, I already noticed that. It swamps out my wireless headphones on all channels when they're within 30cm.
Assuming EMI is the main issue, how would I go about shielding the board? I do have a copper shield or two hanging around
I'm not trying to claim these boards are trash, they *may* be able to carry fair quality audio, but I strongly suspect that you will not get a quiet clean audio channel with them. I spent a long time investigating wireless lossless digital audio, and I'm sure mono (for guitar) can be done on these sorts of RF units (CC2500, NRF24L01 etc..) but requires a more complicated protocol than the AVR can achieve. They are probably just sending raw audio samples over the link, and not caring about lost packets.
If you manage to get them working reasonably well, I'll eat my hat, buy a pair, and test them myself.
tl:dr - don't think the boards can transmit audio at the quality they claim.
I also suspected that. The listing claims 24-bit which is highly excessive. The WM8738 ADC (http://pdf1.alldatasheet.com/datasheet-pdf/view/88091/WOLFSON/WM8738/+57193UyLI.yPPdaSEECh+/datasheet.pdf) is definitely capable of that though. I might do a 17dB gain (ish) on the input and then a divider on the output to get a lower effective noise floor. It's worth noting that the receiver also has a similar, but far less pronounced, interference issue. I may also use both right and left channels and then subtract the difference between them at the receiver.
Looking at this ADC, it appears to have an input resistance of 20k. That's much higher than the listing claimed, and seems reasonable given the fact that guitar direct into the board worked moderately well.
I agree that a custom solution here would probably be best. For guitar purposes, you really don't need a lot of sample frequency. I would settle for 22KHz sample rate at 16-bits, duplicated 6 times in packets for redundancy. That's within the limits of the NRF24L01.
-
Also tested with a scope. This is what the 5V rail looks like.
And the second picture is a point that I found on the bottom of the board that has the signal that appears to be causing this. Leaky cap maybe?
-
Yeah, that's less than ideal eh.
You mention a 5v rail. As the NRF24L01 requires 3.6V max, I'm guessing the regulator on board is 3.3V, powered by the 5v rail. As for the AVR.. I'm assuming that's powered by the 5v rail. Again, I will ask 'how are you powering this? I mentioned batteries in a previous post, but then you said 'boost'. Don't use a boost converter, start with a clean, 5V supply, that is capable of 1A. It looks like you're powering it with a boost converter, that is going into current limiting.
The bottom trace could be anything but I suspect its to do with tx/rx switching - VDDPA is powered when its transmitting, and is used to tell the RFX2401 RF front end to switch to tx, and turn on the PA. Could be wrong, that's a guess.
If the front end could be powered by the on-board voltage regulator, and will draw a fair bit of current. I would check the current draw from your 5V supply when its running.
-
So I've just finished testing with an HP 6554A, and using the power supply reduced interference (in the receiver output) compared to the battery depending on how I angled the antenna. However, the supply rails are just as noisy as before.
Draws about 100mA at 5V
-
It could well be interference in your setup form the noisy transmitter circuitry. What kind of decoupling do you have? The standard 10uF + 100nF cap would be useful, but as you had already tried the 100nF cap apparently no effect then I'm out of idea's. Having a power supply, with decoupling and then long thin wires to the board doesn't cut it, in fact I'd solder a large electrolytic right across the power input pins to be sure.
There is also, as you mentioned, ground loops, since the guitar/buffer is single ended, and shares ground with the power supply, there is the possibility of a ground loop. You said you tried using an audio transformer between the buffer and transmitter, but exactly how did you do this? The schematic for the buffer you posted has no output cap, so it carries a DC bias, which I suspect isn't helping. Also an LM324 isn't great for audio, but shouldn't be overly noisy as a buffer.
So, whilst this seems like a sod to track down, lets make a check list:
- 5v supply. should be capable of >200mA, and have plenty of decoupling. Short wires to board. Measure voltage at the input pins, not elsewhere on the board.
- the onboard 3.3V reg (could be 3v). I still think this is either taxed in terms of current output, or has a dodgy output cap. Whilst ceramic caps are always used fro decoupling, they can play havok with LDO's control loops, and also their capacitance can drop significantly with a DC bias near their limit. Often chinese manufacturers just use caps with what seems like the right value, say 4.7uF @6.3v for a 5V supply, not realizing its capacitance can drop to 15% of that, far below the bare minimum an LDO needs to hold regulation.
- The guitar buffer. You were spot on about the opamp's asymmetrical output (lm324 can go near GND, but only VCC-1.5), but I would still increase the value of those resistors a couple of orders of magnitude. And add an output cap to remove the DC bias - most Audio ADC's have built-in biasing anyway, so expect an AC input, and they filter out DC anyway. If you have noise on the 5v line, or whatever rail is powering the opamp, then this will be carried to the buffer input via the bias resistors, and possibly also through the ADC itself if its powered off the same rail.
Once the 5V, 3/3.3V and Audio in lines are relatively clean, if you still get the interference, then the 1.6kHz could be a clue. The scope looks like its a regular switching, like the RF part switching from tx to rx.
In my experience tracking down noise in systems like this *should* be easy, but never is. Only way is to split it into conductive noise (requires decoupling, galvanic isolation), and radiated noise (needs shielding). Going through each part of the circuit narrowing it down. :/
-
Decoupling:
100nF on LM324
47uF on pins of the transmitter
22uF + small ceramic on regulator output
The transformer was connected with ground and LM324 output on one side, and the transmitter's signal ground and r/l channel on the other side.
Potential divider on the input is now 820K and 1.5M. It's simply an accident in making the schematic that there's no output cap. I've had a 2.2uF in circuit the whole time.
I'm pretty sure now that it's radiated noise. I wrapped a shield in tape, and put it up against the antenna amplifier and the interference is almost all gone. Whether it was always radiated, or it was partly due to supply noise - I don't know.
While soldering the electrolytic across the regulator, I hit one of the 0402 resistors next to it. Those things are bastards! I did my best with a soldering iron and microscope, but still one side of it was disconnected. I only discovered this once I got home, where I only have a soldering gun. After much frustration, I got it on! With an iron tip that's 2x wider than the resistor is long.
-
Yeah 0402 are sods. well, specks of sods.. but it looks like they're still in place.
So, it sounds like you've cracked it. I vaguely remember a youtube video where a guy tested NRF24L01 boards with PA/LNA (the rxf2401 RF front end chip), one with shield and one without, and the shielded one gave huge range, whilst the shielded one was no better than a board without an RF amp. I'm not entirely sure why there would be such a dramatic difference.
a link: http://blog.blackoise.de/2016/02/fixing-your-cheap-nrf24l01-palna-module/ (http://blog.blackoise.de/2016/02/fixing-your-cheap-nrf24l01-palna-module/) (I don't completely trust this source, seems like he was using a terrible power supply before, which isn't a problem with the board, just the user).
I have been meaning to run a test, for a few years now, of these modules, to get an idea of the range, and bit-error-ratio, packet loss, etc... and I have the boards, so maybe I'll run some tests myself, and look for suitable shielding cans.
Example of shielded module: http://www.ebay.com/itm/2-4G-22dBm-100mW-nRF24L01P-PA-LNA-Wireless-Transmission-Module-Shielding-Case-/372009158817 (http://www.ebay.com/itm/2-4G-22dBm-100mW-nRF24L01P-PA-LNA-Wireless-Transmission-Module-Shielding-Case-/372009158817)
-
I don't think the saga is over yet...
I attached a shield, and while the interference is reduced, it's still very prominent. At least as big as the guitar signal itself.
Should I also shield the rest of everything, wires included?
-
What is your power supply for this? If there isn't much decoupling on the board then that will will allow more noise on the power lines, but I find it hard to believe it would be as significant as you've shown on your scope captures (100's of milivolts). When you say "the noise is as loud as the guitar signal" I don't see much point in testing it with a guitar signal and hooked up to an amplifier whilst the noise is present, unless it is only present when you plug in your guitar?
-
It's present at all times. I was using a battery and boost converter this time. I really don't think that the supply has much to do with anything here. With the transmitter disconnected, there is a very slightly hiss on the power lines (back when I was using my amp as a scope), and that's all. Now that I've added extra decoupling, it should matter even less.
After putting the shield on, I was simply testing the severity of the issue. This is why I have a signal vs interference comparison.
Oh I forgot to mention last time. If you have any news from your tests of a wireless module, send me a link to your thread/blog/video. Those kinds of tests greatly interest me, and I wish I had the equipment to find out myself.
-
Welcome to the nightmare of most mixed design engineers. Especially with audio and digital stuff.
It's all about proper grounding, PCB design, decoupling and proper power supplies.
(and you already discovered that shielding doesn't always work)
In general you separate signal ground, power grounds and (dirty) digital grounds.
You always use with start grounds, were the powersupply is (mostly) the middle of this star.
In general the grounding point for the signal ground is right in the beginning, so where the signal comes in.
For a signal output that means it's right at the DAC (and NOT at the output pins).
Make sure your power ground also has a separate grounding path to the start ground.
I have seen/measured cases were the difference was around 4-6dB with certain amplifiers!
In guitar amps it can be pretty bad because of the high gain.
Also be very aware of not making any ground loops.
So sometimes it's better to use only one connection point with shielding, instead of connecting it to many ground points.