The "hot spot" for the RF susceptibility is with the loop directly over the LME49720 IC package. FWIW my OPA1612s are on DIP adapters to permit easy swapping of TH and SMT parts.
The nearby Wima polypropylene caps do pickup hum and are easily shielded but they are not a source of ingress to RF.
I haven't yet found a way to prevent RF rectification with the LME49720 other than a thorough shield.
The nearby Wima polypropylene caps do pickup hum and are easily shielded but they are not a source of ingress to RF.
I haven't yet found a way to prevent RF rectification with the LME49720 other than a thorough shield.
This is an FFT comparison of an LME49720 and OPA1612. The output of the notch filter has 40 dB gain. The oscillator output level is approximately +6 dBu. The OPA1612's 2nd and 3rd THD is in the range of -135 to -140 dBc.
My cell phone was on the other side of the room but a WiFi access point was about 3' from the prototype which has a bottom metal shield and a grounded top foil shield sandwiched between cardboard. Other than the 60 Hz and related components in the OPA1612 example there is not much benefit to the shield with that op amp. The LME49720 benefits greatly from the shield but not enough.
LME49720 FFT of oscillator output after notch filter. +40 dB gain.
OPA1612 FFT of oscillator output after notch filter. +40 dB gain.
The bursts seem to begin around 20 Hz and at first appear line-related. Triggering the 'scope from the 60 Hz powerlinline shows that the burst noise is asynchronous to the line frequency.
The NE5523, NJM2114, NJM2068, NJM2043, OPA2134, OPA2604, OPA1642 and MC33078 will all run in that circuit with various THD signatures higher than the OPA1612 but otherwise they look clean.
The LME49720 has lots of spurs.
Dimitri I'm not sure those links are to the RF problems. You might have had something different in your clipboard when you pasted them.
My cell phone was on the other side of the room but a WiFi access point was about 3' from the prototype which has a bottom metal shield and a grounded top foil shield sandwiched between cardboard. Other than the 60 Hz and related components in the OPA1612 example there is not much benefit to the shield with that op amp. The LME49720 benefits greatly from the shield but not enough.
LME49720 FFT of oscillator output after notch filter. +40 dB gain.
OPA1612 FFT of oscillator output after notch filter. +40 dB gain.
The bursts seem to begin around 20 Hz and at first appear line-related. Triggering the 'scope from the 60 Hz powerlinline shows that the burst noise is asynchronous to the line frequency.
The NE5523, NJM2114, NJM2068, NJM2043, OPA2134, OPA2604, OPA1642 and MC33078 will all run in that circuit with various THD signatures higher than the OPA1612 but otherwise they look clean.
The LME49720 has lots of spurs.
Dimitri I'm not sure those links are to the RF problems. You might have had something different in your clipboard when you pasted them.
In this oscillator the OPA1612 is my second choice to a good (and shielded) LME49720. On a DIP adapter it is very expensive.
The 1612 has almost rail-to-rail output into 2K, decent 600R drive, low input voltage noise, OK current noise for an Ib compensated part. There's a lot to like about it. Its a Ge-Si part. The FET version is the 1642.
Take a look at the OPA1612 open loop output impedance vs. frequency curve. Looks odd to me.
I don't know about the OPA1612 THD with high impedance sources. Its not a good choice for MM preamps due to its current noise but is an excellent choice for a MC or mic preamp with external transistors.
The 100 Hz rep rate looks like something polling to me. I shut down two access points in the shop but it was still there and wasn't able to move all the cellular devices out of range. I also have a Zigbee-based smart meter and energy monitor bonded to it that I can't shut down and RF weather stations. Could be Ethernet-related. Most USB spurs I see are at 1 kHz.
The LME49720 seems to rectify ambient RF envelopes. It's response is relatively broadband.
The 1612 has almost rail-to-rail output into 2K, decent 600R drive, low input voltage noise, OK current noise for an Ib compensated part. There's a lot to like about it. Its a Ge-Si part. The FET version is the 1642.
Take a look at the OPA1612 open loop output impedance vs. frequency curve. Looks odd to me.
I don't know about the OPA1612 THD with high impedance sources. Its not a good choice for MM preamps due to its current noise but is an excellent choice for a MC or mic preamp with external transistors.
The 100 Hz rep rate looks like something polling to me. I shut down two access points in the shop but it was still there and wasn't able to move all the cellular devices out of range. I also have a Zigbee-based smart meter and energy monitor bonded to it that I can't shut down and RF weather stations. Could be Ethernet-related. Most USB spurs I see are at 1 kHz.
The LME49720 seems to rectify ambient RF envelopes. It's response is relatively broadband.
To tame the LME's EMI addiction wind ten turns of fine wire around a 1,000 ohm 1/4 watt resistor and put that right at the chip's input. Some folks us 10 ohms others 100 ohms. Doesn't really affect the audio noise level as the plain resistance should be well under an ohm.
Demian this was in the BT thread years ago.
Demian this was in the BT thread years ago.
Yes. I tried the DECT phone near the DUT this afternoon and it went crazy. Normally its about 8 feet away. I only see it when the remote phone is taken off-hook near the DUT and the base station link is active however. When the DECT link is idle there is no proximity effect.
We have so many IOT devices on our wireless network finding all of them is difficult. Our neighborhood's Smart Meter mesh network for data backhaul is in the ISM band.
I haven't tried input series inductors with the LME49710 but I did try adding a series resistors in the grounded non-inverting inputs which had no effect. Adding input inductance right at the input pins makes sense.
The LME49720 response to a grid dip meter sweep seems to show a broad peak between 50 and 100 MHz. The GDM maxes out at 250 MHz.
EDIT: I unplugged the DECT base station and the 100 Hz spurs disappeared in the LME49720 oscillator. Wow. Put a tape measure to it. The base is 8' 3" away from the LME49720. It looks like the DECT base station is constantly pinging the phones but the phones are only listening in low-power standby. That explains why there isn't a proximity effect near an idle phone. From a power conservation perspective it makes sense.
Odd that only the LME49720 responds to this.
Last edited:
LME49720 Oscillator Notched Output DECT 6 On
LME49720 Oscillator Notched Output DECT 6 Off
LME49720 Oscillator Notched Output DECT 6 Off With Top Shielded
The LME49720's EMI susceptibility extends all the way to 1.9 GHz. The average transmit power of a DECT 6.0 station is 4 mW maximum. The peak power is enough to cause 100 Hz ingress in an LME49720 from 8 feet away.
With a top shield for hum and the DECT phone off the LME49720 looks pretty good.
Thanks Jens for the heads-up.
Last edited:
Most DECT base stations will transmit what is called a "dummy bearer" every 10 ms when there is no connection. This dummy bearer allows the handset to lock to the base station, listen for incoming call messages or set up a call when making an outgoing call.
The dumme bearer is shorter than a normal "traffic bearer". The dummy bearer will be around 100 us and the traffic bearer will be around 400 us, or longer if it is a wide bandwidth audio connection. When a connection is active the base stattion will transmit a 400 us burst every 10 ms and the handset will also transmit a 400 us burst, 5 ms after the transmission from the base station. There are variations to this, but this is the normal use case.
When the handset is locked to a base station, but idle, it will not transmit and therefore not create RF noise.
Yes, it is a bit odd. I have seen something similar with a family of mixed signal devices though (I won't name the vendor). The previous generation (and later generations) were not extremely sensitive to RF radiation, but the this particular generation was very difficult to get quiet. It turned out that it was related to the I/O pads used for this generation. Some of the ESD protection stuff in the I/O pad created a resonance at around the RF frequency used in the system.
Perhaps something similar happened with the LME49720? And perhaps it could also be the case with some of the other National Semiconductor designs of the time?
Thanks Jens. That explains a lot.
My "shielding" - because it's still a prototype - is completely ineffective at RF frequencies due it not being completely enclosed and the large amount of series inductance my makeshift screen has in the drain wire. Works great for hum but an open door to 1.9 GHz.
I need an RF Cone of Silence:
I'm going to scrounge for a really big gift tin to sit the prototype in. Once I get a board made or build a Vectorboard it won't be so hard to shield it.
I think I want to document this problem and alert TI. It needs to be fixed.
I doubt that you can convince TI to change an old National Semiconductor part, which was close to becoming obsolete when moving to a different fab.
If you can, please also make them reinstate the LME49990 to active status
I think the EMI problem has prevented me from screening for burst noise not related to EMI. I haven't tested the latest LME49720 I just received but the EMI issue seems so far to be universal.
Now that I know to turn off the DECT base station during test I can re-check for process-related burst noise.
simon7000's suggestion to add input inductors seems to help EMI.
I put ferrite beads on the lines entering the inverting nodes.
Its difficult to see the difference with inductors due to proximity effects (DECT-me-DUT) when the prototype is unscreened. There are just too many variables.
When I put my electrostatic shield on top of the proto with the inductors in place it is as quiet with the LME49720 and DECT on as it is with any other op amp or the LME49720 with DECT off. With inductors the "cone of silence" shield worked.
When the oscillator is on a proper circuit board in a metal box it should be OK. EMI is definitely one more thing to watch out for with the LME49720. LME49720 start-up behavior is another: LM4562, LME49710, LME49720 Start-up Behavior - Pro Audio Design Forum
I agree TI may not want to fix this after bringing the part back once already. But they should at least be made aware of it. I did see one LME497X0 EMI post on the TI e2e forum asking about the specification but I'm not sure it went anywhere.
Now that I know to turn off the DECT base station during test I can re-check for process-related burst noise.
simon7000's suggestion to add input inductors seems to help EMI.
I put ferrite beads on the lines entering the inverting nodes.
Its difficult to see the difference with inductors due to proximity effects (DECT-me-DUT) when the prototype is unscreened. There are just too many variables.
When I put my electrostatic shield on top of the proto with the inductors in place it is as quiet with the LME49720 and DECT on as it is with any other op amp or the LME49720 with DECT off. With inductors the "cone of silence" shield worked.
When the oscillator is on a proper circuit board in a metal box it should be OK. EMI is definitely one more thing to watch out for with the LME49720. LME49720 start-up behavior is another: LM4562, LME49710, LME49720 Start-up Behavior - Pro Audio Design Forum
I agree TI may not want to fix this after bringing the part back once already. But they should at least be made aware of it. I did see one LME497X0 EMI post on the TI e2e forum asking about the specification but I'm not sure it went anywhere.
I doubt that you can convince TI to change an old National Semiconductor part, which was close to becoming obsolete when moving to a different fab.
If you can, please also make them reinstate the LME49990 to active status
I think you will have to dangle a carrot in front of them.
How many million did you want?
Last edited: