What to replace an LMV358 opamp with in my EHX Holy Grail ?

[j_flanders]

After doing a ton of tests and experiments I finally managed to 'accidentally' damage an opamp (LMV258M) in one of my EHX Holy Grails.
This could be an excellent opportunity to see if a better opamp could bring down the (idle) noise (hiss) which these, otherwise great sounding pedals, suffer from.

I've googled for hours, and can't seem to find what I'm looking for.
Each time I think I found a substitute, I see the supply voltage is wrong, CMRR isn't better, noise is the same, package is wrong, too expensive, etc.
When I give up I come here.
So, what do people suggest I buy and try?

- Supply Voltage: 5V
- Rail to rail
- Excellent CMRR and noise figures
- Dual
- Package? SOIC, SOP, SO8 (are these all the same?)
Here's a picture of the pcb in case that can help to determine the package/size:

Hi-res original: https://i.postimg.cc/pWxtWNzP/DSC07724.jpg

Here is a link to the LMV358 datasheet: https://tinyurl.com/5n77fwx8

Here is a link to the Holy Grail schematic:
https://www.synthxl.com/wp-content/uploads/2020/01/Electro-Harmonix-EC-0060-Holy-Grail-1.pdf

There are two LMV358 in there and they're being used for:
-input buffer
-output buffer
-bias voltage buffer
-differential to single ended converter (I assume CMRR is important here)

I'm willing to order 10pcs if needed (and when not as expensive as an OPA1612 for example)
I'm in Europe so preferably an online store near me.
Since this is SMT (luckily not too small) I'd like to get the right one the first time.
Should I get the LMV358A or is there something better that fits?

[Rob Strand]

Yes, it's amazing how many devices are out there when you speak generally and how few end-up on the list when you start pinning down requirements.

What about the TLV342?  Available in SOIC ( 8 ) package.  The SOIC pic show 10 pin in the datasheet - looks like a mistake as the part has 8 pins, and the top of the datasheet says SOIC ( 8 ).

https://www.ti.com/lit/ds/symlink/tlv342.pdf

For PSRR and CMRR it's good to check the PSRR and CMRR frequency response plots as many opamps have crap PSRR and CMRR at higher frequencies.   The LMV358 datasheet only has tabulated data so it's difficult to compare details.   The LMV358 has a typical CMRR of 65dB whereas the TLV342 is up around 80dB.

The thermal noise from the 2x270k resistors alone on the first stage are going to contribute a lot of noise.   With the LMV358 it's not just the noise voltage but also the noise current.  The noise current with the 270k resistors further adds to the noise from the resistors.   The TLV342 has a lower noise voltage and a much lower noise current which helps reduce noise on the first stage.

I didn't do an exhaustive search and I don't have experience with these parts.

[PRR]

Well, the designer of this Holy Grail didn't care, so why should you?

Right at the input {as Rob says}  we have 270k dead resistance, modestly loading guitar and heavily hissing. More than ANY opamp. Power thriftiness is moot when the CS4811 sucks 200mA. Yes, the DAC filter needs some CMRR but not a lot and it needs it at bitclock rate, not audio. Q1 also passes audio, actually precious clean audio, but is not R2R; in any case not-R2R would only lose a dB or two headroom. Max chip level is 1Vrms (2Vrms in diff mode), so only 2.8V peak to peak needed, not 5V R2R. At the end, R22 adds a last dash of hiss.

It sure could be improved, but not with opamp swaps; and it would be significant re-engineering.

[antonis]

At the end, R22 adds a last dash of hiss.

It surely does, but lowering its value should result into loading Q1 (at least, for Blend pot CCW) leading on to proportionally lowering R31 value resulting into more current draw..

P.S.
The above isn't the designer's sole "carelesness" but, IMHO, he might thought that too many compromises are OK for a paltry guitar pedal..
(after all, it's not faced as a hi-end design..)

[j_flanders]

What about the TLV342?  Available in SOIC ( 8 ) package.  The SOIC pic show 10 pin in the datasheet - looks like a mistake as the part has 8 pins, and the top of the datasheet says SOIC ( 8 ).

The thermal noise from the 2x270k resistors alone on the first stage are going to contribute a lot of noise.

Thanks for the suggested opamp.
Given the rest of the circuit and those 270k resistors I guess there's no point in searching for an opamp with super excellent specs as other sources of noise will trump it.
Regardless, one is broken, so I need a replacement anyway and might as well pick a good one.

Well, the designer of this Holy Grail didn't care, so why should you?

I've e-mailed the designer and this was his reply: (I didn't ask him whether he cared )

[...]
As for the CS4811 noise, we had played with some different noise reduction techniques at the time of developing the Holy Grail, including companding, but ultimately decided against them and to live with the noise.
In hindsight, the U3 input stage is probably contributing a lot of noise. The LMV358 is a very noisy op-amp and it uses 270K resistors for its input and feedback impedance, these large resistor values don't help.

Right at the input {as Rob says}  we have 270k dead resistance, modestly loading guitar and heavily hissing. More than ANY opamp.

I've tried putting 1k resistors parallel to those 270k resistors and the hiss was only marginally less. But I don't know if 270k||1k is the same as 1k when it comes to hiss/noise. They were also clipped on, not soldered.
Muting the output of U3 (big cap to ground) removes most of the hiss at the pedal's output. I can only listen at the wiper of the blend pot though, as the part of U5 that forms the output buffer is currently broken.
So U3 is the source of most of the (idle) noise.
The "270k modestly loading guitar" is part of what I like about this pedal though.
I have several other Holy Grails (Holier Grail, Holy Grail nano, Holy Grail plus). They either have much higher or much lower input impedance making them too bright or too dull.
How would I go about changing the input buffer so it still has 270k loading but smaller resistors/much less hiss?

Power thriftiness is moot when the CS4811 sucks 200mA.  Q1 also passes audio, actually precious clean audio, but is not R2R; in any case not-R2R would only lose a dB or two headroom. Max chip level is 1Vrms (2Vrms in diff mode), so only 2.8V peak to peak needed, not 5V R2R. At the end, R22 adds a last dash of hiss.

Good to know, and thusly not specs I have to care about when selecting opamps.

Yes, the DAC filter needs some CMRR but not a lot and it needs it at bitclock rate, not audio.

Both AOUT+/- (DAC) outputs have a TON of hiss/white (broadband) noise. It's LOUD!, deafening, beyond reason.
At the output of the differential amp there's a LOT less of it.
The cutoff is at 12.5 kHz so my assumption was that that could not be the primary reason for the massive broadband (hiss) noise reduction.
I was guessing the differential amp sums the 180° out of phase reverb signals from AOUT+ and AOUT- and greatly nullifies/cancels the broadband noise.
Hence my concern about good CMRR specs there.
Please correct me if I'm wrong in my reasoning.

In his reply the ehx Holy Grail designer has also suggested two replacements for the LMV358 opamp:

Op-amps like the OPA1678 or TLC2272 would be much better choices. 

[Rob Strand]

Given the rest of the circuit and those 270k resistors I guess there's no point in searching for an opamp with super excellent specs as other sources of noise will trump it.

The 100k's on the output stage aren't great either.  You could probably reduce the value of those a bit, however, reducing them too much will mean you need to increase the caps and it will start to load down the blend pot..   You can end-up making a lot of changes for small gains.   Shaving the values in key areas here and there can reduce noise a bit.

There doesn't seem much point to R25.   It just adds noise.

The Vrefs have fairly small caps.  Not much filtering.

I've tried putting 1k resistors parallel to those 270k resistors and the hiss was only marginally less. But I don't know if 270k||1k is the same as 1k when it comes to hiss/noise. They were also clipped on, not soldered.

The noise of the parallel combination should be representative.  As a design it's not right to reduce them as it will load down the guitar.  Even with the 270k's (but also with the 1k's) you can get different results in these three cases:
- open input
- guitar connected
- input connected to ground

The inverting opamp configuration is inherently more noisy than the non-inverting configuration.

[j_flanders]

The 100k's on the output stage aren't great either.  You could probably reduce the value of those a bit, however, reducing them too much will mean you need to increase the caps and it will start to load down the blend pot.. 

Given the 'weird' 87.5k value I thought the reason behind those 100k resistances was:
25k (blend pot)/2 = 12.5k and (12.5k + 87.5k) / 100k is unity gain for the mixer/output buffer.

There doesn't seem much point to R25.   It just adds noise.

I've tried bypassing it in the past and it didn't help for curing the whining noise nor the white noise. Didn't hurt either.
Maybe it's a compensation resistor? Though should be 50k then, not?
Maybe it's something left over from rev A? It was not true bypass and the dry signal was routed there when in bypass.

The Vrefs have fairly small caps.  Not much filtering.

I've tried adding additional ceramic and electrolytic caps but couldn't hear a difference with regard to the whining noise or the hiss.

The noise of the parallel combination should be representative.  As a design it's not right to reduce them as it will load down the guitar.  Even with the 270k's (but also with the 1k's) you can get different results in these three cases:
- open input
- guitar connected
- input connected to ground

The inverting opamp configuration is inherently more noisy than the non-inverting configuration.

I must admit I only tried with open input.

Who knows, perhaps the way the whine is getting in is also letting in other more random junk from the power supply.

I often disconnect input to stages and connect them to Vref  or ground (assuming DC is OK).  Sometime you need to add an input cap.  If the noise goes away it's a sign something before that point is causing the noise.   Can be messy to do on a build board.  Not to mention an SMD build.

In one of my tests, trying to figure out where most of the white noise comes from, I tried shunting signal or inputs to ground.
Why would I get a lot more white noise when putting a big 100µF cap to ground at the input, parallel to R2

[antonis]

Why would I get a lot more white noise when putting a big 100µF cap to ground at the input, parallel to R2

'Cause you are dominating Signal to Noise ratio..
(too little signal into a noisy amp..)

[Rob Strand]

Sorry for the delayed response.  The router fried (again) this week.  New unit on the way.  The only reason I got back online is because I repaired one of my very old routers.

Given the 'weird' 87.5k value I thought the reason behind those 100k resistances was:
25k (blend pot)/2 = 12.5k and (12.5k + 87.5k) / 100k is unity gain for the mixer/output buffer.

Yes, it looks like they are trying to get unity gain for the mid position.   However, at the extremes the idea fall apart since the gain is above unity.  In mid position you have two signal paths so perceive "unity" gain for the summed signal could be slightly lower than the extremes.  There will be a "correct" input impedance to the inverting output stage to achieve that.

Another angle is to make the input impedance high and achieve closer to unity gain at all positions, but that won't necessarily be perceived unity gain.   What I'm getting at here is the opposite case: if you make the resistors small on this stage for noise reasons it could cause a large signal loss in the mid position.

I've tried bypassing it in the past and it didn't help for curing the whining noise nor the white noise. Didn't hurt either.
Maybe it's a compensation resistor? Though should be 50k then, not?
Maybe it's something left over from rev A? It was not true bypass and the dry signal was routed there when in bypass.

The resistor R25 is not part of the above argument at all.  It simply adds noise for no reason!  I would replace it with a 1k.

And yes, when you have a number of more dominant noise sources its contribution may be small, so you can't perceive a change.   If you could reduce those dominant noise source (unlikely in this design) then you would then start to see the noise of that resistor contributing.

In one of my tests, trying to figure out where most of the white noise comes from, I tried shunting signal or inputs to ground.
Why would I get a lot more white noise when putting a big 100µF cap to ground at the input, parallel to R2

When you have an inverting amplifier it's normal to get different results with the input open.  When the input is open the gain is unit from a noise perspective.   The noise comes from the feedback 270k and the opamp.   When you ground on of the input the noise comes from both 2x270k's and the opamp.   The noise with the input grounded is amplified by two because the non-inverting input now sees a gain of 2 amplifier.  The opamp noise roughly doubles.  The resistor noise of the 2x270k's is less by 1/sqrt(2) of the single 270k, because for noise the resistor appear in parallel, but the gain of 2 makes the noise effectively sqrt(2) higher overall.  So all the noise contributions go up for the inverting amplifier.

[j_flanders]

Thanks Rob. You are a wealth of knowledge!

In the mean time I went to my local electronics store and they ordered some OPA1678's for me.
It could take two to three weeks for them to arrive.

I'll get back when I have installed them. We'll see how much it helps or doesn't and then we'll continue with mods for smaller resistors where possible.

I'll keep you updated.

[Rob Strand]

Thanks Rob. You are a wealth of knowledge!

In the mean time I went to my local electronics store and they ordered some OPA1678's for me.
It could take two to three weeks for them to arrive.

I'll get back when I have installed them. We'll see how much it helps or doesn't and then we'll continue with mods for smaller resistors where possible.

I'll keep you updated.

Good luck with it all.   I admire your perseverance.  It's really what it takes to weed out these types of issues.  There's no short cuts.   In effect you have done what Electroharmonix have not!

[j_flanders]

The resistor R25 is not part of the above argument at all.  It simply adds noise for no reason!  I would replace it with a 1k.

It's left over from rev A.
In rev A the pedal had a DPDT switch and was not true bypass:

https://i.postimg.cc/RCzP8rwd/rev-A-bypass.jpg

In bypass, the input was switched to the non-inverting pin of the output buffer.
But that meant it had a gain of 2:
1+ R23/(R22 + part of the 25k blend pot) = 1 + 100k/100k = 2
So they added a voltage divider before the non-inverting input with R24 and R25, 100k/100k, cutting the input signal in half before the 2x gain output buffer.

In rev B and C they used a 3PDT switch and made the pedal true bypass (switching input, output and LED). They left out R34, C29 and R24 ("omit on Bom") but left R25 in there for no good reason apparently.

At least that's my take on it.

[Rob Strand]

The resistor R25 is not part of the above argument at all.  It simply adds noise for no reason!  I would replace it with a 1k.

It's left over from rev A.
In rev A the pedal had a DPDT switch and was not true bypass:
...
At least that's my take on it.

Very interesting.  R25 is definitely a left over.

[j_flanders]

In the mean time I went to my local electronics store and they ordered some OPA1678's for me.
It could take two to three weeks for them to arrive.

I'll get back when I have installed them. We'll see how much it helps or doesn't and then we'll continue with mods for smaller resistors where possible.

I'll keep you updated.

Here I'm back.
In all the possible online stores and sites (except Ali express) the SOIC OPA1678's were out of stock.
So, I went to my local electronics store, the guy asks me: did you check Mouser?
I sure did, not in stock there.
He checks on his computer: they are in stock!
Hmm, are you sure? I checked a few days ago.
Yes, we can order them for you.
Great, order me 10!
After a month they finally arrive, I take them home, peel back the strip from the reel, take one out and... see they are the SON version, not the SOP, which looks like this:

Soldering smt opamps is already a challenge for my old shaky hands, but this is simply impossible to solder for a DIY person.
Back to the store.
Apparently Mouser used the wrong photo for that opamp and the guy in the store did not look any further...
It also explains why he said they were in stock, while I said they weren't.
Here's Mouser using the photo of the SOP version (P stands for pins) for the SON version (N stands for No pins):
https://hu.mouser.com/c/semiconductors/amplifier-ics/operational-amplifiers-op-amps/?m=Texas%20Instruments&series=OPA1678
First photo should show an opamp with NO leads!

Conclusion: he ordered OPA1678IDRGR while I needed OPA1678IDR.
My lesson learned: don't ask OPA1678 SOIC version. Tell the exact number: OPA1678IDR
His lesson learned: read text, don't trust the photo.

Back to square one.
Next he suggested he could contact some brokers that could hunt down some for me but he warned that they could be prohibitively expensive.
A couple of days later he came back with a number: 5 euros a piece. (vs 0.86 euros at Mouser)
I bit the bullet and ordered 5.

Almost another month passed and they finally arrived.

I removed the old LMV358 opamps and soldered in the new OPA1678's.
The Holy Grail which I broke was fixed again!

But a big part of the reason why I specifically wanted those OPA1678's was to see how much they would bring down the white noise these pedals suffer from.
I took many measurements comparing the white noise in all the different Holy Grails I have.

I'll create a new specific thread for this.

General conclusion for this thread: OPA1678 is a good replacement for the LMV358 in a Holy Grail.