LM4562 alternatives for LF SDR applications

[ledtester]

This Elektor SDR clock receiver utilizes a LM4562 to amplify signals in the 60 kHz - 200 kHz range (like MSF, WWVB, DCF77, etc.):

https://www.elektormagazine.com/magazine/elektor-288/61435

(PDF is available from the above link)

AI says that the LM4562 is frequently used in SDR projects.

Currently they run $8.80 a piece in unit quantities. I was wondering if there are cheaper alternatives or what operational specs are important in searching for an alternative.

For instance, is the OPA1656 ($2.90/piece) a suitable replacement?

https://www.ti.com/lit/ds/symlink/opa1655.pdf

The attachment shows how the LM4562 is used in the project.

[Benta]

At a glance, I'd say that the LM4562 is total overkill for that application.
Major consideration is low noise, but 55 MHz BW is unnecessary.

[exe]

I'm not sure what are opamp requirements, but LM4652 is strange in this circuit. I'm also not sure it needs to be low-noise. Anyway, OPA1656 looks like it will fit it. As, probably, many other opamps.

The circuit is.. strange. They have 100 ohm resistor. Presumably it should filter power supply noise. But with quiscent current of 10mA alone that will consume 1 volt, leaving opamp 4volts.

Is there any evidence that the circuit in the article even works?

[MariuszD]

Radio noise in the LF band is high, so I don't think low-noise requirements are critical. It is unknown if there are further gain stages. It is unknown what gain they were trying to achieve in IC1B. High gain requires a high GBW. The LM4562 can achieve 60dB without feedback at 70kHz.
That's not too much.

[magic]

NE5534 has just a bit more noise and similarly high GBW up to 200kHz, but it's single and not unity gain stable.

[edavid]

NE5534 has just a bit more noise and similarly high GBW up to 200kHz, but it's single and not unity gain stable.

The problem with NE5534 (or NE5532) for the circuit as given is that it doesn't work very well from a 5V (or 4.2V if you use the RC filter) supply voltage.

[exe]

it's single and not unity gain stable.

It has external compensation)

[magic]

It has external compensation)

which slows it down

However, I must say that I have tried running 5534 from different vendors (Signetics, TI, maybe NJM too) at unity gain with no compensation and without much capacitive load on the output they didn't oscillate. YMMV.

The problem with NE5534 (or NE5532) for the circuit as given is that it doesn't work very well from a 5V (or 4.2V if you use the RC filter) supply voltage.

Good point, but neither does LM4562 to be honest (rated 5V minimum). Both will have a somewhat limited output range (not R2R).

And don't be misled by TI's datasheet laziness: 5532/5534 were originally rated down to 6V total supply and the supply current vs supply voltage plot shows that they reach operational bias at a bit less than 5V. So for a one-off build it may work, if you can live with 2Vpp output.

[edavid]

And don't be misled by TI's datasheet laziness: 5532/5534 were originally rated down to 6V total supply and the supply current vs supply voltage plot shows that they reach operational bias at a bit less than 5V. So for a one-off build it may work, if you can live with 2Vpp output.

It appears that the ADC reference is 3.3V, so you are only losing 1 bit (if you manage to set your gain just so).

I wonder if the reviled TL07xH would be a good choice here, at least for the second stage. It might have to be divided into two stages to get enough bandwidth.

[floobydust]

Just a note, TI changed the "circuit design" PCN20231114002.1 Feb. 2024 on a bunch of other op-amps like NE5532, LM833, MC34078 etc. so their "new" performance is not known.
The NE5534 design was changed in the 80's when TI bought the Signetics stuff, but to date... I don't see that IC getting yet another circuit change/die-shrink cost improvement.

For VLF circuit ideas, maybe look at the Blitzortung Lightning Detection VLF pre-amp. But constant revisions, changes over the years and complicated. I don't know all of the circuits. At least they includes mains hum filtering, which that Elektor circuit is missing I think.
They used NE5534's then settled on the THS4531.
I think they are at Rev. 20 now... https://docs.lightningmaps.org/hardware/lightning-detectors/hardware_reference/#pcb-20-complete-boards

[Vovk_Z]

Interestng than OPA1656 has a really wide supply range. It starts from +-2.25 VDC (or 4.5 VDC single supply) and works up to +-18 (36) VDC.

[David Hess]

The high gain-bandwidth product keeps distortion low at higher frequencies, especially with gain.

The OPA1656 should work fine.

[bsdphk]

If you will be using a loop-antenna, the AD797 is probably the best there is, but it can take some experimentation to avoid feedback/singing.

Also: Lots of interesting ideas at vlf.it

[Benta]

If you will be using a loop-antenna, the AD797 is probably the best there is, but it can take some experimentation to avoid feedback/singing.

DCF77, WWVB etc. receivers usually have a ferrite rod antenna. I don't see a loop antenna as a viable solution for 60 kHz.

[bsdphk]

The reason single-frequency "clock" receivers use a tuned-to-near-resonance ferrite-rod antenna, is that they only need 5Hz-ish bandwidth, and you get a LOT of amplification for free that way, making it possible to run forever on a battery.

(non-resonant) Loop antennas are perfectly reasonable for VLF work:  Here is some work I did two decades ago on loran-C and other time signals using a loop made from a paint-bucket lid:

https://phk.freebsd.dk/loran-c/

Here is my second generation work using an ARM with internal 1MSPS ADC:

https://phk.freebsd.dk/AducLoran/

Enjoy the gif's:  Nowhere else will you see the fabled "The Dance of the Nightwave"

[edavid]

If you will be using a loop-antenna, the AD797 is probably the best there is, but it can take some experimentation to avoid feedback/singing.

Except that OP was looking for something cheaper than the LM4562, not something twice as expensive 

[profdc9]

Another alternative is using a low noise BJT like BC547 or 2N4401 with a cheap op-amp like LM358.  I attached a document that describes how they work from Vojtech Janasek.  I attached a Qucs-s noise simulation of it.  This solution would avoid the need for any specialized op-amps.

[PCB.Wiz]

Another alternative is using a low noise BJT like BC547 or 2N4401 with a cheap op-amp like LM358. 

The idea is good, tho perhaps with a slightly better opamp, as the slew rate and cross over of a LM358 is poor.

That circuit seems to have a strange bias setup: the DC bias (base) has more gain on the positive path than the negative path (emitter), so it will never have a stable operating point.

[profdc9]

Yeah, a better op-amp would work too.  I put a 1k resistor on the output to prevent crossover distortion by always sinking current, and even 1V at 100 kHz is only 0.1 V/us, well within the slew rate of LM358.  I tend to design for the lowest common denominator crap if I want circuits to be reproduced.  Nevertheless, something like MCP6022 can be used too.

Maybe the components could use some tweaking, but it does seem to have a stable bias point.  Note the transient simulation is stable.  Also, the nodes are labeled with the DC voltages/currents.

[PCB.Wiz]

Maybe the components could use some tweaking, but it does seem to have a stable bias point.  Note the transient simulation is stable.  Also, the nodes are labeled with the DC voltages/currents.

Strange, LTSpice says no stable DC point, and I need to reduce R8 to 42.98k to get close, HFE of BC847A is 180 - or 29k with a lower HFE part.
It simple shoots thru the bias point, as the path is regenerative.

I can get a stable DC operating point, if I add a lossy inverter to the positive feedback path, so it becomes properly negative.