Hi
These look suitable: http://www.analog.com/media/en/technical-documentation/data-sheets/OP470.pdf
These look suitable: http://www.analog.com/media/en/technical-documentation/data-sheets/OP470.pdf
This is the 1998 version, right? The 2006 one uses all dual types, no quads. The quads in the older one are NJM2059, quad version of NJM4559. Better than their '58 cousins, but still not especially fast by today's standards.
The circuit uses highish impedances in order to keep the opamps from breaking too much of a sweat, and it runs 'em all in inverting mode, removing several common distortion mechanisms. Extremely low voltage noise is not required, but current noise shouldn't be too high, so that leaves either JFET or low-Ib bipolar parts.
I would try:
* OPA4132 or OPA4134 (FET input)
* OPA4227 (bipolar, but fairly low input current noise density)
Since both of these are faster than the original parts, some local rail decoupling capacitance seems advisable.
Even if you run master level as hot as possible without clipping and reduce LF/MF/HF level accordingly, output noise level on each channel will be 3.6 µV (20 kHz BW) at the very least, even with ideal noiseless opamps. That's the resistor noise of the inverting part in the output amp, contributing 12.8 nV/sqrt(Hz) at 295 K. Which is why I suspect any reduction in output noise will be fairly modest (under 2 dB when plugging some OP data into the calculator), though linearity especially at high frequencies should improve measurably.
For a substantial improvement in output noise, a major rework would be required.
The circuit uses highish impedances in order to keep the opamps from breaking too much of a sweat, and it runs 'em all in inverting mode, removing several common distortion mechanisms. Extremely low voltage noise is not required, but current noise shouldn't be too high, so that leaves either JFET or low-Ib bipolar parts.
I would try:
* OPA4132 or OPA4134 (FET input)
* OPA4227 (bipolar, but fairly low input current noise density)
Since both of these are faster than the original parts, some local rail decoupling capacitance seems advisable.
Even if you run master level as hot as possible without clipping and reduce LF/MF/HF level accordingly, output noise level on each channel will be 3.6 µV (20 kHz BW) at the very least, even with ideal noiseless opamps. That's the resistor noise of the inverting part in the output amp, contributing 12.8 nV/sqrt(Hz) at 295 K. Which is why I suspect any reduction in output noise will be fairly modest (under 2 dB when plugging some OP data into the calculator), though linearity especially at high frequencies should improve measurably.
For a substantial improvement in output noise, a major rework would be required.
Last edited:
Thanks for the analysis. The 4227 I already discovered, though its price is giving me some pause...lol The all dual amp circuitry was also done surface mount if I understand the history. The particular one I have is an '02 model, and is the thru-hole/DIP construction, which I believe rules out the TI/BB models you mention( for sure the 4134 anyway).
cheers,
Douglas
cheers,
Douglas
Indeed, OPA4134 is SOIC only for some reason. But another one available in DIP just came to mind, TLE2074. A bit higher in noise though.
Re: OP470, Samuel Groner's measurements of the related OP471 (similar except for a degenerated input stage promising better linearity at somewhat higher voltage noise) indicate generally good performance except for a really wimpy output stage in both current capability and bias levels. These may still be serviceable but they would be involved in output driving here as well, and I wouldn't feel comfortable having them do that.
There still is the old MC33079D. But again, not a great 600 ohm load driver.
The choice among quad opamps isn't exactly what I'd call breathtaking...
Re: OP470, Samuel Groner's measurements of the related OP471 (similar except for a degenerated input stage promising better linearity at somewhat higher voltage noise) indicate generally good performance except for a really wimpy output stage in both current capability and bias levels. These may still be serviceable but they would be involved in output driving here as well, and I wouldn't feel comfortable having them do that.
There still is the old MC33079D. But again, not a great 600 ohm load driver.
The choice among quad opamps isn't exactly what I'd call breathtaking...
Last edited:
That's a really weird mix, in fact. CH1 LF and MF outputs (with the highest levels) use 4580s, which have good output drive capability. CH2 LF and MF outputs use 2059 sections instead, as do all of the HF outputs. So one channel has better output drive capability than the other. Now most power amps won't be too heavy on their sources, but still...
I forgot an almost natural upgrade to the NJM2059, btw - the NJM2060 (NJM4560 quad). Faster, better output drive (and going by higher quiescent current, more output stage bias as well), hence much lower HF distortion, noise about the same. Looks like they never released any quad versions of 4565 or 4580.
It doesn't look like this should be a problem here, but it's always a good idea to keep power supply capabilities and cooling in mind when swapping opamps. Swapping a bunch of TL07x for NE5532 or similar could result in a nasty surprise.
I forgot an almost natural upgrade to the NJM2059, btw - the NJM2060 (NJM4560 quad). Faster, better output drive (and going by higher quiescent current, more output stage bias as well), hence much lower HF distortion, noise about the same. Looks like they never released any quad versions of 4565 or 4580.
It doesn't look like this should be a problem here, but it's always a good idea to keep power supply capabilities and cooling in mind when swapping opamps. Swapping a bunch of TL07x for NE5532 or similar could result in a nasty surprise.
Last edited:
Working! Had a pretty busy week so I apologize if I've fallen behind on any threads on here that I need to respond to.
My latest BIG project goes into the semiconductor fab soon, so it'll be a little while before the world gets to see it
As for other projects, I have a few ideas in the works. I've been taking a lot of data for a presentation I'm giving on distortion in op amp circuits (both from the op amp and from other components) and that always gets my brain going. I do have an article coming out sometime in May on a different way to stabilize difference amplifiers when driving capacitive loads. The article uses headphones as an example cap load and includes an impedance sweep of my Sennheiser HD555s up to 10MHz.
John, thanks for jogging my memory of that particular model. With this project I am looking at needing some 2-amplifier variants as well. It looks like 49740 only comes in quads...with my stone knives and bearskin tools I found LM4562 looks very like two of the 49740 amps. Am I on the right track, or is there actually a 2-amp chip package of the 49740?
cheers,
Douglas
You're tools led you down the right path
LME49720 is the dual version of the LME49740. But as far as I know the LM4562 is pretty much the same part. I work on the TI/BB op amps so I can't say for sure what those National guys did with part numbers
That's a good point, I'm actually not aware of what the circuit schematic in question looks like. Even on bipolar input devices with bias current cancellation, there can still be a noise benefit to matching source impedances. I was playing with this the other day with an OPA1612, and got about a 25% reduction in noise from matching the impedances at the two inputs.
The problem with swapping in low noise FET input devices is that their input capacitance can create a second pole in the loop gain curve with high impedance feedback resistors, potentially creating stability issues.
The schematic of the unit in question can be obtained online (it's the AC 23B (1998 version)). Lots of 20k-47k resistors and 50k-100k pots.
Isn't this rather a phenomenon that only occurs with types employing input bias current cancellation? (Bias cancellation circuitry adding common-mode noise - I think Samuel Groner discussed this somewhere, or was it Douglas Self?)
Which possibly was why the original design uses bipolars, and why you commonly find a small capacitor across the feedback resistor in inverting amps.
Matching source resistance to reduce noise is most effective in op amps which DON'T use bias current cancellation. This is because the input current noise in these op amps is mostly correlated. I think it is mainly produced by the shot noise in the tail current divided by the input transistor beta. Matching source resistances converts the correlated current noise into a common mode voltage which is then removed by the CMRR of the op amp.
However, in op amps that use bias current cancellation, this introduces 2 uncorrelated noise sources into the input bias current, and matching source impedances will only reduce the correlated noise.
Theoretically, the broadband current noise spectral density of an op amp should be In=sqrt(2*q*Ib) where Ib is the input bias current, and q is the charge of an electron. However, in bipolar input op amps which use Ib cancellation the current noise does not follow this rule due to the additional noise of the bias cancellation circuit. Worse yet, this noise is mostly uncorrelated, so it can't all be cancelled, but some of it can.
Alright, one further question, arising from the original warning about power consumption. The 2059 lists I_cc as 7 mA. If that is for one amp, the idle consumption would then be 28 mA...which is less than say...the 49740 at 18.5 mA per set of four. Now if that 7 mA is for the whole quad, then the newer, faster stuff is a significant jump up in power supply demand.
cheers,
Douglas
cheers,
Douglas
Looking at the NJM datasheet I believe 7mA is for the entire quad, so yes the LME49740 would be a significant jump in power consumption. All other things being equal, bandwidth and slew rate generally come at a cost.
Indeed. I like to regard any amplifier as a tradeoff between circuit complexity, power consumption and performance. (Individual semiconductors' performance figures in, too, e.g. when looking at pnp vs. npn performance depending on manufacturing process.)
The instant you can afford higher output stage quiescent current, for example, performance into lower-impedance loads tends to improve a fair bit (as distortion is a function of output impedance variation, which for a bipolar output stage hinges on 1/gm + Re, with gm = Ic/vT ~ Ic). And since the output stage tends to be the slowest part in practice, it may be possible to relax compensation and hence make the amp faster as well, further reducing distortion this way. That's what I would expect to be happening when going from NJM2059 (6 MHz GBW, 7 mA) to NJM2060 (10 MHz GBW, 9 mA).
Samuel Groner, in his set of opamp measurements, looked at two single FET input opamps of 1970s vintage - TL071 (1.4 mA, 3 MHz GBW) and LF356 (5 mA, 5 MHz GBW). As you may imagine, the difference in the 600 ohm load driving test was profound!
Higher power consumption is not necessarily a problem, but a look at regulator type and cooling, regulator input voltage, transformer size and rectifier diodes is advised in order to avoid nasty surprises. I remember a case where one guy installed several fancy discrete opamps at 25 mA per channel into his integrated amp which had come with standard audio types, and promptly complained about hum creeping in when his not-particularly-constant line voltage was running low - transformer secondary voltage sagged, and the regulator hit minimum dropout voltage. Oops.
Stock current consumption would be expected to be about 15 mA for the 4 duals + 70 mA for 10 quads = 85 mA. Simulating the voltage doubler supply with a 85 mA (current source) load, a transformer secondary emulation of 15 Vp (10.6 Vrms) @ 60 Hz and 5 ohms of series resistance gives good results, with about 21 V of DC output - but double the load current, and minimum voltage drops to 16.7 V. Standard linear regulators need to drop 1.8-2.1 V depending on current, so this may already mean hum.
I am guessing the actual transformer is a 12-0-12 secondary, maybe a 9-0-9. It could in theory be as small as 3-5 VA, but hopefully the designers were a bit more generous than that. Problem is, you don't want to oversize your xfmr too much either, or else you'll run into trouble with the rectifier diodes. Which in turn you don't want to oversize too much because bigger Si diodes are slower and generate more RF buzz which you may then need to thwart with 22-100 nF of parallel capacitance for each of D1-4. (Standard measure on any PSU you want to use with a shortwave / mediumwave / longwave receiver, particularly the portable variety.)
Some measurements are recommended. I wouldn't expect a total of 105 mA with NJM2060s to be much of a problem, but 200 with LME49740s may well be.
Anyway, assuming you do have some milliamps to spare, there is another way of tweaking distortion performance - output Class A biasing with either a resistor or a current source to one of the rails (often the negative one, which means you'll be using the npn half in an emitter follower push-pull output). This usually works best for types with generous current capability but low quiescent current. There's a parallel thread in the Chip Amps section discussing this right now. Due to the implications on current draw (SE class A is rather inefficient), it's something I'd use sparingly, only in critical positions.
Last edited:
In the PS, Rane did some silly stuff. The power trans is off board. A lot of their stuff ran this power adapter. The rectifiers, filter and regulator is on-board and appears to be a doubler( the 9-0-9 into a +/- 15 V reg'd supply would need one). Rane sez it is good for 750 mA...no quotes on what sort of a +/-15vdc load though.
Input to the rectifier/filter is through something like an Ethernet plug. Stepping up to a full voltage AC supply and drop the doubler might not be so bad an idea.
cheers,
Douglas
Input to the rectifier/filter is through something like an Ethernet plug. Stepping up to a full voltage AC supply and drop the doubler might not be so bad an idea.
cheers,
Douglas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.