I'm looking for a low cost opamp for the crossover in an active speaker I'm building and I'm having trouble finding a good cheap opamp. I currently using the TL074 and there's more noise than I would like. Any ideas?
I'm looking for:
-Available in quad TSSOP package
-Low cost, say less than $1 but lower the better
-Noise in the range of 4-5nV/√Hz
-PSRR of ~100db
-+/- 5v supply
-Any other features that would help?
The NE5532 would be good but does not come in a quad package. Motorola's MC33079 also looks very nice but only comes in quad SOIC. TL974 has good noise but only 70db PSRR.
Thanks!
I'm looking for:
-Available in quad TSSOP package
-Low cost, say less than $1 but lower the better
-Noise in the range of 4-5nV/√Hz
-PSRR of ~100db
-+/- 5v supply
-Any other features that would help?
The NE5532 would be good but does not come in a quad package. Motorola's MC33079 also looks very nice but only comes in quad SOIC. TL974 has good noise but only 70db PSRR.
Thanks!
Last edited:
If you are actually hearing noise (hiss) then you might have other problems (such as where/how the opamp circuit relates gain wise to the rest of the amplification chain).
The TL074 is not the quietest opamp in absolute terms, that's true, but its all relative and for line level there should be no real problem.
The TL074 is not the quietest opamp in absolute terms, that's true, but its all relative and for line level there should be no real problem.
That would work, but is more expensive than I was hoping for.
A friend of mine suggested the NJM2745 in SSOP which is great and only costs $0.64, but unfortunately has a minimum order quantity of 100 from verical.
The hiss is pretty subtle and can only be heard within a few feet of the speaker. It's not offensive by any means. If I'm not mistaken switching from the TL072 (18nV/√Hz) to a part with 5nV/√Hz would drop the noise floor by 12 db, right?
My board is just has the line inputs running straight into a lm1036, then through the crossover. If I pull the coupling caps between the crossover and amp and short the amp inputs I get almost imperceptible noise. If I pull the coupling caps between the lm1036 and tone control and short the crossover inputs, I get the same amount of hiss as with the whole circuit operational.
The noise floor will drop this much only if your the TL074s are the dominant noise contributor which would suggest that the rest of your input chain is pretty darned quiet indeed.
However, it seems we have a similar approach to this sort of thing and for exactly the same purpose I bought the ON semiconductors MC33079PG.
These are through hole, 16MHz, 4.5nV/√Hz and distortion with lots of zeros in.
They are under GBP1.0 from RS in the UK so the usual pricing factor should make them fit your $1 limit.
You probably don't need anything this quiet but then, nor did I 🙂
However, it seems we have a similar approach to this sort of thing and for exactly the same purpose I bought the ON semiconductors MC33079PG.
These are through hole, 16MHz, 4.5nV/√Hz and distortion with lots of zeros in.
They are under GBP1.0 from RS in the UK so the usual pricing factor should make them fit your $1 limit.
You probably don't need anything this quiet but then, nor did I 🙂
The noise from the crossover section is dominant with the gain turned down. With the gain cranked up it's probably 1/2 or less.
The MC33079 would be great, but it's not TSSOP/SSOP 🙁
The MC33179 comes in TSSOP but has higher noise at 8nV/√Hz. Distortion is similar, CMRR/PSRR are higher. Newark list it at $0.64 but no stock.
Seems like all the good cheap stuff is in dual packages or quad SOIC. If I had to do it over again I would definitely use an SOIC footprint. Even if I revised the board finding the extra 4mm for an SOIC package would be tough.
I think my best bet at the moment is to use the TL974 and hope that the 70db PSRR is not an issue.
The OP1654 is certainly better, but the price is almost $3 in small quantities...
You have to consider the circuit impedances around the opamps, FET devices such as the [old] TL074 type were generally better in highish impedance circuitry, bipolar better with low impedance. Having to specify a quad device in a particular package is very limiting on your choices.
It's probable that my implementation is sub-optimal for noise as I'm not an electrical engineer and have limited experience with line level audio. And yes, the quad package is limiting, live and learn.
I've attached a schematic of the filter section for critique. The input labeled "RIGHT" connects directly to the output of an LM1036. The outputs on the right drive an amplifier with an input impedance of 70k. There are no gain stages before the lm1036. The output labeled RLINE goes to a line out jack for a sub, with a 150pf cap to ground near the jack (not shown).
The four opamps in the schematic are part of the same TSSOP package.
An externally hosted image should be here but it was not working when we last tested it.
Lol... its kind of in the middle impedance wise.
I think here a bipolar is probably best but there are things you can do as it stands. The input stage. Do you need as much gain as you have ? If you do then the 66.5k and 13.3k could be reduced by a factor of 10. The other "input" 66.5k is fine and will be shunted by whatever the source impedance feeding the unit is. Filters at the top. Same applies. Reduce the resistors by a factor of 10 and increase the caps by the same amount. The networks on the opamp outputs are fine value wise.
Another possible issue... the TL074 has pretty limited output drive ability and that is made even worse running at low supply voltages. This is something I have looked at and tested for my own curiosity and they are less than ideal tbh... well at least compared to other devices. For the filters and reducing impedances though I think you will be OK. 2k is an often quoted minimum sort of impedance for these opamps, in practice I find they are a bit better than that but don't come close to devices such as the NE5532 or LM4562 or FET devices like the OPA2134... are they available in a quad ? just a thought.
I think here a bipolar is probably best but there are things you can do as it stands. The input stage. Do you need as much gain as you have ? If you do then the 66.5k and 13.3k could be reduced by a factor of 10. The other "input" 66.5k is fine and will be shunted by whatever the source impedance feeding the unit is. Filters at the top. Same applies. Reduce the resistors by a factor of 10 and increase the caps by the same amount. The networks on the opamp outputs are fine value wise.
Another possible issue... the TL074 has pretty limited output drive ability and that is made even worse running at low supply voltages. This is something I have looked at and tested for my own curiosity and they are less than ideal tbh... well at least compared to other devices. For the filters and reducing impedances though I think you will be OK. 2k is an often quoted minimum sort of impedance for these opamps, in practice I find they are a bit better than that but don't come close to devices such as the NE5532 or LM4562 or FET devices like the OPA2134... are they available in a quad ? just a thought.
Yes, the gain is about right.
So the components highlighted in green are the ones you propose changing? Resistors divided by 10, caps multiplied by ten.
An externally hosted image should be here but it was not working when we last tested it.
That's it 🙂 That will keep the filter characteristics the same and it will certainly make some difference... but whether its as much as you hope... I can't answer. The lower impedance in conjunction with a more suitable opamp certainly would though.
Hi there!!
Have you consider the OPA1664 or the Opa1604 from TI as they can be had in a TSSOP package and they are Bipolar type designs as well,
OPA1664 - SoundPlusTM Low-Noise Audio Op Amp - TI.com
The LME49740 is another excellent Quad Opamp as well but it is only available in either a PDIP or a SOIC packages.
http://www.ti.com.cn/cn/lit/ds/symlink/lme49740.pdf
FWIW
jer 🙂
Have you consider the OPA1664 or the Opa1604 from TI as they can be had in a TSSOP package and they are Bipolar type designs as well,
OPA1664 - SoundPlusTM Low-Noise Audio Op Amp - TI.com
The LME49740 is another excellent Quad Opamp as well but it is only available in either a PDIP or a SOIC packages.
http://www.ti.com.cn/cn/lit/ds/symlink/lme49740.pdf
FWIW
jer 🙂
Last edited:
there is little point in changing the Z of the parts after the gain of +6 stage if you keep with fet input op amp
noise R contribution after the gain stage is scaled by the gain squared - it only adds to the input referred noise as R of 1/36th the value
the opa1654 or 1644 look as cheap as you can expect to find for the low noise performance in jfet
with low Vnoise bjt you do have to scale the Z of everything down to obtain the noise advantage given their much higher Inoise - so much so that you probably can't get the needed C in any acceptable quality dielectric in a size that fits
noise R contribution after the gain stage is scaled by the gain squared - it only adds to the input referred noise as R of 1/36th the value
the opa1654 or 1644 look as cheap as you can expect to find for the low noise performance in jfet
with low Vnoise bjt you do have to scale the Z of everything down to obtain the noise advantage given their much higher Inoise - so much so that you probably can't get the needed C in any acceptable quality dielectric in a size that fits
I'm using LME49740 in DIP14, For an active crossover. Now they aren't 'cheap' but not extortionate. Even better if you can acquire samples...
Ugh
Ok, thanks for all the suggestions. I ordered some TL974s, OPA1654s and OPA4377s and will swap them in tomorrow.
Unfortunately I'm realizing after a bit of testing that the lm1036 tone control is a bigger offender than the crossover section. I think I got it all wrong before because I was evaluating the noise subjectively, more on that later.
I made a little chamber to place a db meter on top of my tweeter so I could measure the noise before and after changes (see below). Not the most precise way of measuring but better than subjective assessment like I was doing earlier.
First I changed the resistors on the buffer (divided by 6.666) and I got a ~0.5 and~0.8 db reduction in noise at the tweeters. Changing the components in the latter stages did not seem to make much if any difference.
Then I realized I was measuring the noise from multiple sources and that I should pull the capacitors between stages and measure the noise. Here's what I got:
Power off: 35 db
Amp stage only: 49.1 db
Amp + crossover: 51.0 db
Amp + crossover + LM1036: 54 db
So the approx contributions are:
Amp: 49 db
Crossover: 47db (with lower value resistors)
Tone control: 51 db (!)
That makes the actual reduction in noise from changing the buffer resistors more like 2+db because I was measuring multiple noise sources. Because the tone control generates so much noise I should have disconnected it before measuring. Doh!
It seems even if I swap in ultra low noise opamps that reduce the noise of the crossover section by 10 db, I'll only get another db down overall with the tone control as is. 🙁
So the best way to reduce noise is to get it out of the tone control, followed by the xover, then the amp. Fortunately, looking at the data sheet it looks like I can get substantial reductions in noise by lowering the supply voltage of the LM1036. The datasheet shows an 8 db reduction at low gain when going from 12v to 9v and a 1.5 db reduction at high gain (see below). I'm at 15v so there's a lot of room for improvement.
Tomorrow I will lower the supply voltage of the lm1036 to 9v, swap in some lower noise opamps and report back.
Ok, thanks for all the suggestions. I ordered some TL974s, OPA1654s and OPA4377s and will swap them in tomorrow.
Unfortunately I'm realizing after a bit of testing that the lm1036 tone control is a bigger offender than the crossover section. I think I got it all wrong before because I was evaluating the noise subjectively, more on that later.
I made a little chamber to place a db meter on top of my tweeter so I could measure the noise before and after changes (see below). Not the most precise way of measuring but better than subjective assessment like I was doing earlier.
First I changed the resistors on the buffer (divided by 6.666) and I got a ~0.5 and~0.8 db reduction in noise at the tweeters. Changing the components in the latter stages did not seem to make much if any difference.
Then I realized I was measuring the noise from multiple sources and that I should pull the capacitors between stages and measure the noise. Here's what I got:
Power off: 35 db
Amp stage only: 49.1 db
Amp + crossover: 51.0 db
Amp + crossover + LM1036: 54 db
So the approx contributions are:
Amp: 49 db
Crossover: 47db (with lower value resistors)
Tone control: 51 db (!)
That makes the actual reduction in noise from changing the buffer resistors more like 2+db because I was measuring multiple noise sources. Because the tone control generates so much noise I should have disconnected it before measuring. Doh!
It seems even if I swap in ultra low noise opamps that reduce the noise of the crossover section by 10 db, I'll only get another db down overall with the tone control as is. 🙁
So the best way to reduce noise is to get it out of the tone control, followed by the xover, then the amp. Fortunately, looking at the data sheet it looks like I can get substantial reductions in noise by lowering the supply voltage of the LM1036. The datasheet shows an 8 db reduction at low gain when going from 12v to 9v and a 1.5 db reduction at high gain (see below). I'm at 15v so there's a lot of room for improvement.
Tomorrow I will lower the supply voltage of the lm1036 to 9v, swap in some lower noise opamps and report back.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
If you want cheap, use NJM2058. It's available in TSSOP and under a dollar in Mouser. Personally, I use OPA1644 but it isn't cheap 🙂
I see it in ssop, but no one has stock.
A friend of mine suggested the NJM2745, also ssop and nice specs:
low noise (5nV),
low distortion (0.0005%),
really high CMRR (110dB)
It's just $0.64 at Verical, but you have to buy 100. No one sells them in smaller quantities for a reasonable price.
I was going to get some LM1036's but I decided against it as the specs were 1000 times worse (+30db) for noise and THD then the rest of the opamps I plan on using!!
Digital Pots or some PGA2320's or something along that line would work better,except that you would have to employ an mcu to control them.
Or maybe a Digital Pot that has a built in Up-Down function for the tone controls.
ST has some too and they are a little bit better but not by much. 😉
Although they do have a better noise rating of -100db to -106db and .01% THD rating vs -80db noise and.06% THD for the TI part.
http://www.st.com/web/en/catalog/sense_power/SC913
I wasn't able to get any since they cut me off after only 5 sample orders and they were on the very next order that they denied me!! 🙁
FWIW
jer 🙂
Digital Pots or some PGA2320's or something along that line would work better,except that you would have to employ an mcu to control them.
Or maybe a Digital Pot that has a built in Up-Down function for the tone controls.
ST has some too and they are a little bit better but not by much. 😉
Although they do have a better noise rating of -100db to -106db and .01% THD rating vs -80db noise and.06% THD for the TI part.
http://www.st.com/web/en/catalog/sense_power/SC913
I wasn't able to get any since they cut me off after only 5 sample orders and they were on the very next order that they denied me!! 🙁
FWIW
jer 🙂
Last edited:
- Status
- Not open for further replies.