|
Post by dotnet on May 19, 2008 16:12:30 GMT
Hi all, I received three 4562 samples from NatSemi today and had to try them in my CMoy-like amp. Surprisingly, they work (almost) well as a drop-in replacement for OPA2132. I didn't expect that, as they are quite different from the FET op-amps...
Remaining problems are a rather high DC offset (best sample 50mV/120mV, worst 250mV/250mV), and a bit of a noise floor. I'm going to change the R's tomorrow (hopefully) and that should take care of both issues.
Even now, the amp sounds a lot better. In comparison, the 2132 is really stuffy, constricted and breathless.
I'll post an update when I've got the resistors changed.
Cheers Steffen.
|
|
Will
Been here a while!
Ribena abuser!
Member since 2008
Posts: 2,164
|
Post by Will on May 19, 2008 19:20:02 GMT
It'd be really interesting to see what you come up with. Reading about someone else's tatting is always good!
|
|
toad
Been here a while!
I am the Super Toad, the Original Toad, the Whole Toad and nothing BUT the toad.... don't forget it!
Posts: 1,223
|
Post by toad on May 19, 2008 20:51:26 GMT
The AD823 is well worth a try in a CMoy too. I have tried a few op amps in my CMoy and the AD823 always get swapped back in after a few days
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 19, 2008 21:15:22 GMT
Steffen A higher noise floor often means instability problems. SandyK
|
|
leo
Been here a while!
Team wtf is it?
Posts: 3,638
|
Post by leo on May 19, 2008 23:09:10 GMT
I personally would not use a LM4562 without any external help on its output One op-amp thats hard to beat at driving loads on its own is the AD826, don't think it would work well in a cmoy circuit though, probably go unstable, I'll have to try it Also Toads suggestion is a good choice, if you decide to try the AD823 my advice is to look closely at its datasheet
|
|
|
Post by dotnet on May 20, 2008 4:44:48 GMT
Steffen A higher noise floor often means instability problems. SandyK Hi Alex, I was most apprehensive about instability, and checked current draw and heat all the time. The noise I'm getting is not overly present, though, audible only during quiet passages. I did the numbers with the resistors I'm using right now, and the s/n ratio should be around 80dB. I reckon this is in the ballpark of what I'm hearing. I think it's mostly due to the very high R2. I'm going to drop R2 from 1M to 100k, and reduce the gain from 6 to 2 or 3. This should bring the s/n ratio close to 100db, and also reign in the DC offset to under 10mV. I hope it stays stable. I've just received the resistors here at work, and will be putting them in tonight... Cheers Steffen.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 20, 2008 6:16:36 GMT
Steffen You already seem to know that the LM4562 doesn't like high value resistors ! If it's already stable, reducing the gain should make it more so.BTW, they need a series resistor at their output,when driving a typical interconnect,or even a short shielded cable to a volume control (!) but I don't think it needs to be as high as 100 ohms like many people use. I usually shove in 68 ohms. Alex
|
|
FauDrei
Been here a while!
Posts: 489
|
Post by FauDrei on May 20, 2008 12:03:38 GMT
Ha... lately I use my cMoy exclusively for burning-in of the new opamps... I also experienced slightly higher noise floor when using LM4562 in a cMoy.
Yes, this LM4562... quite a (strange) beast: when I used it as a straightforward replacement for standard AD823 in green Solo - it did OK, but nothing special (preferred LM6172 and/or AD8397). But after Leo's Solo resistors tweak this LM4562 is, at least to my ears, a class above all other opamps (ones that I've tried).
Go figure.
|
|
|
Post by dotnet on May 20, 2008 13:33:47 GMT
Ok, the jury is back, and they have a verdict... I tried 3x gain first, with R3=1k and R4=2k. The noise floor was gone, there is black silence now without input signal. The output DC offset dropped to 4mV/11mV, which is fine and a sign that the theory works... I wasn't entirely happy with the sound though, I thought some loud notes of certain frequencies had something like a cymbal trail, if that makes sense. I don't have an oscilloscope so I can't be sure what's really going on there. Also annoying was the fact that "too loud" happened already at the 11 o'clock position of the volume pot. Out went the 2K and in the 1k, for a gain of 2. The sound is now really clean and transparent, the DC offset dropped further (1.9 and 4.6mV). I'm a happy camper now, and leave it like that. The 4562 in conjunction with much lowered gain and input load is definitely a worthwhile upgrade for the OPA2132, at least with my headphones. Oh yes, I should mention them, 60ohm Koss KSC-75. I also tried my low-ohm Ety's 6i and they sounded horribly tinny and wrong. I then discovered that their filters were clogged and need to be replaced... They did play very loud though, and there is no evidence that the 4562 would have any trouble driving them. For low-ohm phones it might be a good idea using an output resistor of a few dozen ohms, esp if the phones are plenty loud. This is supposed to improve the 4562's stability further, but I haven't tried this. I'd describe the sound now as spacious and transparent, quite forceful where required and with good voice reproduction, and with a good tight bass (to the extent as all that is possible with headphones ;D ) Especially the bass has improved. It is slightly on the lean side now, but that suits me well. It was far too thick and mushy previously with the KSC-75s. In hindsight, the only thing I'm not sure about is whether I should have gone for those expensive Welwyn resistors. I figured rather than buying 10 to get a matching pair I go for the 0.1% ones, at about $4 each all up... During all my testing the CMoy was driven direct from an X-DAC (not sure which version, it says "HDCD" at the front and has a tubular case). Cheers Steffen.
|
|
|
Post by PinkFloyd on May 20, 2008 21:46:53 GMT
The only real advantage the RC55Y have over 1% metal film resistors is the lower 15ppm temp coeff compared to 50ppm with 1% types ;D
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 20, 2008 22:04:02 GMT
I find it almost bordering on obscene to spend $4 a pop on resistors.In mil spec devices and special instrumentation devices they have a place, but not in general audio applications. It's far better to get good quality resistors in the first place, then if necessary, use a reasonable quality Digital Meter to match values used in differential pairs and feedback network resistors (gain setting) of both channels. Even then, the volume controls typically used are going to make a mockery of all that close matching, except in the input differential pair of a discrete circuit.
SandyK
|
|
rickcr42
Fully Modded
Rest in peace my good friend.
Posts: 4,514
|
Post by rickcr42 on May 20, 2008 23:20:16 GMT
Something is WAY wrong there bud.No way you should see that unless your power supply is way out of balance between the +/- sides OR you have such a high impedance differential at the input that your "balanced input" is anything but hence creating a higher offset than the max spec of the chip. I'm gong to assume the latter and even more so after reading your post further down where you used 1 Meg Ohm in the R2 position which is WAY WAY TOO HIGH under any circumstances (makes for a dandy AM radio receiver without input compensation ) and which unless R3 is ALSO 1M would mean you have a serious issue with input current balance hence your high DC offset at the output (input DC offset X Gain=Output offset) When using a bipolar input opamp You need to balance the INPUT or you increase the DC offset by the magnitude of the imbalance so R2 needs to shadow R3 (or you will have to take other measure to reduce the offset and none are easy or inaudible). But ! no free lunch here either ! you want R2 to have a high enough value that you input impedance provides an easy to drive full frequency response yet go to high and make the value of R3 that value and you have a situation where the feedback loop resistor,R4 will be HUGE in an amp with a x10 gain though less of a problem than having too small a resistor at R2 so my best guess here would be : P1-10K R2-47K R3-47K R4-470K Rout-(see blow) The noise issue is another problem and will have very little to do with R2 it being in a SHUNT position rather than a SERIES position.this means yes,different resistor types MAY sound varuable but noise factors will remain constant no matter the value if everything else is correct in the circuit layout. Next up-Series Output Resistance : Well since the output is already short circuit protected you have no need to use an inline (series) resistor to protect the output from shorts so any resistor here would be more for "loop" isolation. this is where any RF on the headphone cord has the potential to enter the output which the amp output itself being a very low resistance means would have no place to go BUT because the negative feedback loop connects to the output and is the fed back to the input and because the input IS a high resistance value (in ohms) means we now have a path for the RF signal to not only GO but to be amplified by a factor of the amp stage gain so for instance if your amp has a gain of x10 then the Rf signal will be amplified by x10 ! Worse ? because this signal is only "seen" by the inverting input unlike the input audio signal which is "seen" at both the non-inverting (signal) and the inverting (output fed back to the input) there will be ZERO common mode cancellation it not being a "common mode" signal. The fix ? Only allow any Rf to see a low impedance thus having no PATH to go elsewhere and this means you need to decouple the output JACK from the output CIRCUIT (the op-amp) with the simple method being a small value resistor. where this gets tricky is it just so happens that the best value for decoupling the output ;120-300 ohms,is also about the WORST value for driving headphones unless they A-were designed to be used with an amp complying with the IHF standard "120 Ohm" load or B-The actual headphone impedance is a minimum of X2 the value so for 120 ohms your cans need to be 240 ohms or bewtter unless designed to IHF loads for lower values you need to walk the "six of one,half dozen of the other" walk balancing dynamic current delivery under load with feedback loop RF rejection. Quick and dirty ? Use a 250 dual pot between the amp and the headphones set to "zero" resistance then with music playing increase the resistance until you can actually HEAR the music change (usually it will be the dynamics,the LIFE of the music .At that point back off a smidge,remove the pot from the circuit,measure the resistance value between the input terminal and the wiper then solder in a resistor of the closest stock value. The higher the better but NOT if the music suffers and it will using headphones that need more current drive than they do voltage drive to move the diaphragms ! good luck rick out
|
|
leo
Been here a while!
Team wtf is it?
Posts: 3,638
|
Post by leo on May 21, 2008 1:49:06 GMT
I find it almost bordering on obscene to spend $4 a pop on resistors.In mil spec devices and special instrumentation devices they have a place, but not in general audio applications. It's far better to get good quality resistors in the first place, then if necessary, use a reasonable quality Digital Meter to match values used in differential pairs and feedback network resistors (gain setting) of both channels. Even then, the volume controls typically used are going to make a mockery of all that close matching, except in the input differential pair of a discrete circuit. SandyK They are also magnetic and sound shit! Try the RC55Y's in feedback positions, they make the highs squeaky
|
|
|
Post by dotnet on May 21, 2008 4:11:14 GMT
I find it almost bordering on obscene to spend $4 a pop on resistors. I agree. However, they were the ones I could get overnight, the 1% ones would ship in 6-8 days... The non-inductive 1% types weren't available from any local source as far as I could see. For a one-off it doesn't really matter much. Cheers Steffen.
|
|
|
Post by dotnet on May 21, 2008 4:30:46 GMT
Something is WAY wrong there bud.No way you should see that unless your power supply is way out of balance between the +/- sides OR you have such a high impedance differential at the input that your "balanced input" is anything but hence creating a higher offset than the max spec of the chip. The power supply is well balanced. There is quite some sample variation among the three chips I've got. One sample has a high but identical (left vs right) DC offset. You'd imagine they'd send their best measuring samples out, wouldn't you...? Are you saying R2 and R3 should be similar? Right now, R2 is 100k and R3 is 1k, yet the DC offset is very small. R2 as 47k is something worth trying, I'm sure the X-DAC will have problem driving that However, a x10 gain is far too much for my phones. Right now I'm using x2, and even that could be lower. I don't know how the 4562 will behave when approaching x1, though. I punched the resistor values and 4562 specs from the data sheet into the noise calculator at tangentsoft.net, and R2 did indeed have the biggest impact on noise. Dropping it to 100k make the noise almost disappear, as per the calculator result. Lowering the gain reduced it further. I decided to go without one, since there doesn't appear to be a need for it. I got that idea from a forum post elsewhere, as a cure to instability with low-ohm loads. Thanks for you detailed and informative reply! Cheers Steffen.
|
|
|
Post by dotnet on May 21, 2008 4:39:03 GMT
Try the RC55Y's in feedback positions, they make the highs squeaky I have, and I can't detect any squeakiness. Can you elaborate? Are you saying their inductance is too high and inhibits feedback for high frequencies? I'm running a low gain, maybe it doesn't matter as much then. I shall try different resistor types, and will look out for the non-inductive ones. Cheers Steffen.
|
|
leo
Been here a while!
Team wtf is it?
Posts: 3,638
|
Post by leo on May 21, 2008 8:45:35 GMT
They was tried in a active linestage, several other people reported the same results complaining they made the sound glassy when used in the feedback
Normal 1% metal films sounded better to me, they are cheap and supplied in packs of 50 so matching wasn't a problem when needed.
All I know is that I don't like them and won't be using them again, probably something like a Cmoy the type doesn't make much difference
|
|
|
Post by dotnet on May 21, 2008 14:58:15 GMT
All I know is that I don't like them and won't be using them again, probably something like a Cmoy the type doesn't make much difference And I appreciate your input. I will definitely try other metal film types, I'm just about to order a bunch from RS. It's just that I needed that head amp back in service quickly as I've pulled my MF A120 apart to replace the scratchy volume pot (and all of the electrolytics while I'm at it). Listening to headphones at home isn't my first choice but better than no music at all Re the CMoy, I think it has a lot of limitations but if operated in its sweet spot it should perform pretty well. I believe simplicity is a virtue, and while you can fix certain problems with added circuitry you often buy more problems in return. That said, I think you can never spend enough effort on a decent power supply... I'll report back when I've tried other resistors. Cheers Steffen.
|
|
|
Post by dotnet on May 27, 2008 7:17:12 GMT
Ok, just so you don't think it's all honky-dory...
I'm back to using the 2132. The sound of the 4562 was airier and wider but after a day or two I admitted to myself that something odd was being added. I can't describe what it is, it just doesn't sound smooth and clean in comparison. I'd love to stick an oscilloscope probe to it, until then I stick to the meaty if somewhat constricted sound of the BB's. They are more honest sounding.
Oh, and this is still with the R55Y resistors, I haven't tried other MF types yet. Gotta put that Musical Fidelity amp back together first...
Cheers Steffen.
|
|
|
Post by nelsonvandal on Oct 24, 2008 16:56:34 GMT
LM4562 is impressive in many ways - clean and transparent but just not musical. No X-factor. I've tried it a lot of times (and LME49720/10) and never clung to it.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 24, 2008 20:13:12 GMT
nelsonvandal. As I have posted elsewhere,the LM4562 is a fabulous chip if correctly implemented. It is is how ever very demanding of a power supply, unlike devices like the OPA2134 which are less fussy with their PSU requirements. In the Jaycar thread, many members are getting fabulous results from the LM4562 after fitting the low impedance JLH addon.
SandyK
|
|