I've just finished piecing together a preamp for my LM3886-based chipamp. It uses relays to switch inputs and the volume is controlled by an LM1972 followed by an OPA2134 at a gain of about 7db (I wasn't able to run it at unity gain reliably - kept spiking awfully on turn on/off).
It's been a while since I used this amp, but after adding the pre the sound is remarkably different, at least from what I can remember. It's also different from my other, LM1875-based amp with a passive pre. The sound's much.. softer. Highs, like snare drums and cymbals, which I usually find pretty harsh, are very smooth. Perhaps a bit smoother than they ought to? On the other hand, the sound overall is really, very pleasant. Strings and pads sound awesome, for example.
Now, the OPA2134 datasheet mentions impedance matching to minimize distortion in non-inverting applications (which the LM1972 requires, iirc). The LM1972 has a variable output impedance between 25k and 35k. I was aiming to keep the feedback network at about 30k but during testing I ended up at about 18k (33k and 39k paralleled). Could this mismatch increase the distortion enough to cause the sound I've described above, or is it simply "the Burr-Brown sound"?
Also, if anyone has any other comments on the importance of impedance matching of the OPA2134, I'd be interested in hearing about that as well. 🙂
It's been a while since I used this amp, but after adding the pre the sound is remarkably different, at least from what I can remember. It's also different from my other, LM1875-based amp with a passive pre. The sound's much.. softer. Highs, like snare drums and cymbals, which I usually find pretty harsh, are very smooth. Perhaps a bit smoother than they ought to? On the other hand, the sound overall is really, very pleasant. Strings and pads sound awesome, for example.
Now, the OPA2134 datasheet mentions impedance matching to minimize distortion in non-inverting applications (which the LM1972 requires, iirc). The LM1972 has a variable output impedance between 25k and 35k. I was aiming to keep the feedback network at about 30k but during testing I ended up at about 18k (33k and 39k paralleled). Could this mismatch increase the distortion enough to cause the sound I've described above, or is it simply "the Burr-Brown sound"?
Also, if anyone has any other comments on the importance of impedance matching of the OPA2134, I'd be interested in hearing about that as well. 🙂
Impedance-mismatch crossed my mind while I was still on the second passage... 🙂
I don't have much experience with particular devices but I can assure you that impedance (mis)matching can significantly influence frequency responce and alter the "sound". 😉
I don't have much experience with particular devices but I can assure you that impedance (mis)matching can significantly influence frequency responce and alter the "sound". 😉
Think you have something more than just the "Burr Brown" sound going on here.
The only real way to tell, is do some tests with a 'scope and look how the amp behaves.
Does the sound change if you change the opamp for something else. Try something like a TL072, an NE5532 and an old 4558 (really) and see if the same listening impressions are there.
The only real way to tell, is do some tests with a 'scope and look how the amp behaves.
Does the sound change if you change the opamp for something else. Try something like a TL072, an NE5532 and an old 4558 (really) and see if the same listening impressions are there.
The OPA134 does not like to be loaded hard. See the data sheet and also the work of Samuel Groner. He made an in debth comparison of a lot of popular op amps.
Do you realy think that Burr-Brown tries to make a "home sound" ?
I do not think so. I asume that they try to make their Opamps as good as they can in a technical sense.
P.S. D.Sef measured the OPA134 too and he found raised treble distortion when loaded down to 600 ohm. Also if you have a mismatch in impedance of the feedback resistor and the source impedance you end up with common mode distortion in this OP in the case of a non inverting series feedback arangement.
The OPA827 and ADA4627 are in improvement if you want to use Fet input.
Do you realy think that Burr-Brown tries to make a "home sound" ?
I do not think so. I asume that they try to make their Opamps as good as they can in a technical sense.
P.S. D.Sef measured the OPA134 too and he found raised treble distortion when loaded down to 600 ohm. Also if you have a mismatch in impedance of the feedback resistor and the source impedance you end up with common mode distortion in this OP in the case of a non inverting series feedback arangement.
The OPA827 and ADA4627 are in improvement if you want to use Fet input.
Great, more fiddling. 🙂 One of the vias on the board is getting a bit tired from me de-/soldering parts to it, so I wasn't too keen on more testing if nothing seemed out of order. Will now, and grab a hold of some higher precision resistances as well, while I'm at it.
The OPA should not be loaded down by much. The chipamp.com amplifier boards I'm using have an input impedance of about 23k (if I'm reading the schematics correctly), well above the 600 the OPA has problems with.
I'll try to get a hold of an oscilloscope and see if I can measure the output with the current (~18k) and more appropriate (~30k) impedance in the feedback network. Hopefully the latter will prove sufficient, given that the LM1972 has a variable output resistance. Would hate it if I should have to compensate this in the feedback network. (Though it would be interesting 🙂)
What would be useful tests? Square waves at a few different frequencies? Sines?
The OPA should not be loaded down by much. The chipamp.com amplifier boards I'm using have an input impedance of about 23k (if I'm reading the schematics correctly), well above the 600 the OPA has problems with.
I'll try to get a hold of an oscilloscope and see if I can measure the output with the current (~18k) and more appropriate (~30k) impedance in the feedback network. Hopefully the latter will prove sufficient, given that the LM1972 has a variable output resistance. Would hate it if I should have to compensate this in the feedback network. (Though it would be interesting 🙂)
What would be useful tests? Square waves at a few different frequencies? Sines?
Look round about post #1017
http://www.diyaudio.com/forums/chip-amps/154106-best-sounding-audio-integrated-opamps-51.html
You have to measure and test all the time. Squarewave testing is extremely useful and shows up many problems.
I've no equipment to do any signal testing, apart from a simple multimeter. I did some simple loopback tests with my soundcard but the results were utter garbage. I'm looking at picking up a Tektronix 2215 second-hand (looks like around $100). Should be more than sufficient for my needs, I presume?
if you feel that the sound leaves something to be desired, I would put my money on that ancient LM1972. The opamp you mentions is much more transparent than that chip.
There are much better level control chips these days.
jd
Well, the sound I'm hearing is not radically different. It's more of a nuance in high frequency noises (like percussion). So I'm not sure I dislike it, but I'm trying to find out if it's "wrong" in a sense (ie. ringing or such) or if it's just "different", which I can live with.
I think I'll do a few side-by-side tests of my two amps, as well as changing the values in the resistor network. Measuring would be interesting but I need to check my budget for this month first. Want to get a hold of some ply for a new pair of speakers as well.
I think I'll do a few side-by-side tests of my two amps, as well as changing the values in the resistor network. Measuring would be interesting but I need to check my budget for this month first. Want to get a hold of some ply for a new pair of speakers as well.
When planning this preamp I looked around and the only digital pot ICs I found that seemed any good were the LM1972 and the PGA2310. Seeing as the latter is about three times as expensive, I went for the LM1972.
What chips other than these do you suggest I look into?
Hi Mosfet,
For what it's worth I find the OPA134 and it's multipart companions quite bland subjectively also. The treble is smooth if a tad tinny, mids somewhat dry and bass weak and without substance. The thing that most stops me using them is that tonally they are not correct - eg the brass section in a large orchestra isn't, well, brassy enough, and male vocals sound wrong. This is compared with the 5534 in a line level preamp where they can safely be swapped (ie moderate loading and a compensation cap fitted for the 5534). The 5534 has some top end grain, but overall much more air in the mids, proper bass and correct tonality. I can only guess these differences are down to the harmonic structure of the distortion produced by each. It's amazing to me that the difference is audible as I've measured the THD using each opamp in the preamp and both are below -100dB (the limit of the analyser I used) over the whole 20Hz to 10kHz range.
PS None of the other expensive audio opamps sound quite right to me either - generally I only use 5534 or discrete ccts.
For what it's worth I find the OPA134 and it's multipart companions quite bland subjectively also. The treble is smooth if a tad tinny, mids somewhat dry and bass weak and without substance. The thing that most stops me using them is that tonally they are not correct - eg the brass section in a large orchestra isn't, well, brassy enough, and male vocals sound wrong. This is compared with the 5534 in a line level preamp where they can safely be swapped (ie moderate loading and a compensation cap fitted for the 5534). The 5534 has some top end grain, but overall much more air in the mids, proper bass and correct tonality. I can only guess these differences are down to the harmonic structure of the distortion produced by each. It's amazing to me that the difference is audible as I've measured the THD using each opamp in the preamp and both are below -100dB (the limit of the analyser I used) over the whole 20Hz to 10kHz range.
PS None of the other expensive audio opamps sound quite right to me either - generally I only use 5534 or discrete ccts.
Did some A/B comparisons with my other amp today. Hooked one amp up to each speaker, running only a pair of Tangband full-range drivers in sealed boxes, no filters. Most listening tests were inconclusive, though the LM1875 amp did produce som extra noise when playing high frequency sines (~14k).
The most marked difference was sines at ~2k. It seems the input signal is getting a bit too strong for the LM1972 attenuator. I'm only running it off a soundcard through quite a few feet of crappy wire, so the levels should not be out of the ordinary. Turning down the volume of my soundcard and instead increasing the volume on the LM1972 got the two amps much, much closer in sound. Gone was the harsheness (well, probably some form of clipping), and now just a smooth sine.
I'm running the line input directly into the attenuator (after a relay and coupling capacitors). Would buffering before the LM1972, possibly dividing down the signal to a few hundred millivolts, help? What would that do to my SNR?
I found someone online was having trouble with this chip and DC. Measuring the speaker outputs of the amp, I'm getting a low 2mV with the signal attached (compared to about 100mV when inputs are left floating), so I imagine that's not the issue.
The most marked difference was sines at ~2k. It seems the input signal is getting a bit too strong for the LM1972 attenuator. I'm only running it off a soundcard through quite a few feet of crappy wire, so the levels should not be out of the ordinary. Turning down the volume of my soundcard and instead increasing the volume on the LM1972 got the two amps much, much closer in sound. Gone was the harsheness (well, probably some form of clipping), and now just a smooth sine.
I'm running the line input directly into the attenuator (after a relay and coupling capacitors). Would buffering before the LM1972, possibly dividing down the signal to a few hundred millivolts, help? What would that do to my SNR?
I found someone online was having trouble with this chip and DC. Measuring the speaker outputs of the amp, I'm getting a low 2mV with the signal attached (compared to about 100mV when inputs are left floating), so I imagine that's not the issue.
Just found another 1972 project where the inputs are divided by a pair of 120 ohm resistors (one in series, the other to ground), which apparently solved clipping issues in that case. Going to try this tomorrow. 🙂
If I understand you correctly then 240 ohms loading is way to low for a line level output.
Using an opamp as high impedance buffer, with a low value ( say 1 to 2 k) divider on the output is best all round.
If your DC offset varies with the inputs open and floating and connected then you have something unaccounted for in the design... it shouldn't happen. The input bias currents should be the same under both conditions.
Make sure it really is an offset variation and not due to stary pickup from a floating input.
- Status
- Not open for further replies.