Because its much cheaper (well it used to be), and does the job and has lower current noise than newer bipolar opamps? Why would you waste money on a more expensive part that's not actually needed? Yes you can put a jet engine on a moped, but why?
I love these sorts of discussions. I noted the question of "so what are your requirements" slip by the discussion with no answer... but that is probably for the best if this is to stay on a purely subjective basis.
Of course there are alternatives to the 5532/4 and at the level of performance that the 5532/4 provides the practical difference is negligible. But hey, of course if an alternative costs an order of magnitude more and is painted hot pink, it damn well has to be better, no?
Of course there are alternatives to the 5532/4 and at the level of performance that the 5532/4 provides the practical difference is negligible. But hey, of course if an alternative costs an order of magnitude more and is painted hot pink, it damn well has to be better, no?
How about two 5532's in parallel, with current sharing resistors. How about "N" 5532's in parallel? Douglas Self built an amplifier with N=32 of them in parallel; it drove 8 ohm loads.
How about configuring the op-amp - whatever it is - in a zero impedance configuration? I think that's the preferred method when driving a transformer?
@jlinkels Yes, I'm aware of the conversion factor when the secondary isn't loaded. The low end linearity suffers though and needs to be compensated for, as the DCR is still affecting the loading.
@egellings The 5532/34 doesn't have the best open loop gain, relying on NFB to linearize it, along with higher noise. I do agree that it is transparent when driving medium impedance loads, but it doesn't have the guts close to a 4580, which is often used for headphone applications: The OPA1612 and 1692 are good suggestions. I totally forgot about those. I built a headphone amp with the 1692 and liked what I heard. Its very sensitive to PCB layout, as are other higher performance op amps.
@cvanc The LM4562 doesn't sound as good as the NE5532 to my ears. Parallel 5532s do sound better and have lower noise. The thing I dont like about the 4562 is the midrange, especially when loaded heavily. It becomes very forward and nasal with higher perceived distortion, even though its supposed to measure cleaner than the 5532.
@newvirus2008 I'm familiar with the AD8066 which I believe is similar to the AD8056, but isn't rail to rail architecture. I may take a closer look at that one, although the noise is higher than the 5532/34.
@googlyone I was pretty sure my goals were clear in my first posts. I wanted something better than 5532/34 with lower distortion, noise and better current capability. When driving inductive loads, the current capacity matters more, especially with high inductance loads.
@cowanaudio The NE5532/34 does in fact sound good and I'm definitely a big advocate for it. The input current offset is high and the noise goes up alot with higher impedance sources. Its being fed by a 100k source and that's going to be an issue with the input offset. Thats much essier to accommodate with a fet input op amp.
@egellings The 5532/34 doesn't have the best open loop gain, relying on NFB to linearize it, along with higher noise. I do agree that it is transparent when driving medium impedance loads, but it doesn't have the guts close to a 4580, which is often used for headphone applications: The OPA1612 and 1692 are good suggestions. I totally forgot about those. I built a headphone amp with the 1692 and liked what I heard. Its very sensitive to PCB layout, as are other higher performance op amps.
@cvanc The LM4562 doesn't sound as good as the NE5532 to my ears. Parallel 5532s do sound better and have lower noise. The thing I dont like about the 4562 is the midrange, especially when loaded heavily. It becomes very forward and nasal with higher perceived distortion, even though its supposed to measure cleaner than the 5532.
@newvirus2008 I'm familiar with the AD8066 which I believe is similar to the AD8056, but isn't rail to rail architecture. I may take a closer look at that one, although the noise is higher than the 5532/34.
@googlyone I was pretty sure my goals were clear in my first posts. I wanted something better than 5532/34 with lower distortion, noise and better current capability. When driving inductive loads, the current capacity matters more, especially with high inductance loads.
@cowanaudio The NE5532/34 does in fact sound good and I'm definitely a big advocate for it. The input current offset is high and the noise goes up alot with higher impedance sources. Its being fed by a 100k source and that's going to be an issue with the input offset. Thats much essier to accommodate with a fet input op amp.
You " waste time " with 5532 5534 because literally nothing new outperforms it by any measurable margin. They are also dip8 and 0.30 each.
I wouldn't regard it a waste of time using an NE5532/34 for general audio purposes. If it was good enough for Philips in their higher end CDPs then it should be good enough for most audio applications. My issue with it is the lack of current output. Otherwise its more than adequate, especially in parallel configuration with 10 ohms at each output. I have a "few" ceramic package Signetics 5532s which could be put to good use here.
My question is though, is it better to apply local NFB at each section of 5532 or globally at the sum of all outputs to a buffer going in?
My question is though, is it better to apply local NFB at each section of 5532 or globally at the sum of all outputs to a buffer going in?
I did think its funny to pull the 741 into discussion, as its literally an AWFUL op amp, only good for static measurements (if that). They sure thought it was hot stuff when it came out in the early 70s (ish). I remember pulling apart an old signal tester with a bunch of metal can 741s. We used them in low frequency sine wave generator circuits, where they still had alot of distortion. It was better to use a bunch of piggy backed chains of nand gates.
@profiguy... Have you ever tried to build and actually measure a device at the levels of noise and distortion a 5532/4 (let alone a few in parallel) make? It is really, really non trivial. There is a world of difference between what you see on a datasheet and a robust design on the desk in front of you, and another again between that and the thing integrated into a system.
Noise, distortion and current into 600R is available from a swag of devices at 0.000X percent level. So given other parameters which might be less obvious, like insane bandwidths and slew rates as one earlier post may have light heartedly suggested, they are all in that ballpark.
Now connect a transformer (and seriously...even a series resistor to a real world load) and measure the distortion at the op-amp output and THEN the distortion at your real world load. At that point you will probably want a strong cup of coffee, and start asking yourself what all the fuss is about with naval gazing over spec sheets for subsystem components.
I don't know your background but maybe I am picking at my own failings being an engineer.... but at the system level, if I was going to throw another 10 or 20 or 50 bucks at what sounds like an op amp driving a transformer, I would throw it at the transformer every day of the week.
Then I would throw a bunch of time at layouts, grounding etc, as I promise you even "reasonable go at this" will struggle in a real world application to get numbers anything like what you see on the datasheets at your output socket.
Or I would paint it all hot pink.
PS: Way off topic - but have any of you ever measured a power amplifier output into a properly gnarly load before and after the Zobel network...
Noise, distortion and current into 600R is available from a swag of devices at 0.000X percent level. So given other parameters which might be less obvious, like insane bandwidths and slew rates as one earlier post may have light heartedly suggested, they are all in that ballpark.
Now connect a transformer (and seriously...even a series resistor to a real world load) and measure the distortion at the op-amp output and THEN the distortion at your real world load. At that point you will probably want a strong cup of coffee, and start asking yourself what all the fuss is about with naval gazing over spec sheets for subsystem components.
I don't know your background but maybe I am picking at my own failings being an engineer.... but at the system level, if I was going to throw another 10 or 20 or 50 bucks at what sounds like an op amp driving a transformer, I would throw it at the transformer every day of the week.
Then I would throw a bunch of time at layouts, grounding etc, as I promise you even "reasonable go at this" will struggle in a real world application to get numbers anything like what you see on the datasheets at your output socket.
Or I would paint it all hot pink.
PS: Way off topic - but have any of you ever measured a power amplifier output into a properly gnarly load before and after the Zobel network...
profiguy asked for a device with higher output current as a 5532. We are talking about DIY here - so very low quantity.
So he should really tinker around with an old OPA and try a workaround (parallel etc) instead of using the right part - cause a few bucks?
I don't know if you ever designed a very low noise and THD circuit but I would not choose a 5532 to do so. There are better parts for a reason.
(e.g. I had to design a very low noise headphone amp for audiometric measurements. 136dB(A) S/N, THD hard to measure with my Audio Precision. Not every design needs to be just ok and 100dB S/N. It's DIY - where is the fun in building "just OK", you can buy that for cheap.)
So he should really tinker around with an old OPA and try a workaround (parallel etc) instead of using the right part - cause a few bucks?
I don't know if you ever designed a very low noise and THD circuit but I would not choose a 5532 to do so. There are better parts for a reason.
(e.g. I had to design a very low noise headphone amp for audiometric measurements. 136dB(A) S/N, THD hard to measure with my Audio Precision. Not every design needs to be just ok and 100dB S/N. It's DIY - where is the fun in building "just OK", you can buy that for cheap.)
I have to agree with both of your views on this @IamJF @googlyone
I'm looking for more current capability, which I could achieve using parallel 5532s but the issue would be open loop gain VS feedback required to make it behave. The layout is still very critical if you want the best possible performance but in all sincerety, I'd prefer a rail to rail device with fet inputs. The 5532s are excellent for low source resistance and have relatively high input offset. I'd need to design a servo circuit to eliminate the need for coupling caps. So with all these other requirements, I'd prefer something better than a bipolar 5532. I have a few signetics 5532ANs (lower noise version) to tinker with and maybe try in parallel. I can barely measure the little bit of higher order distortion of a well executed 5532 implementation with my Motu interface but I can unfortunately still hear the op amp in the circuit vs a BUF03 or OPA1692. Having worked on a truck load of consoles based on the 5532/34 series chips, I know their potential. The transformers will have a much larger influence on the sound than the op amps on paper but there are audible differences which favor the OPA1692 or AD8066. The later is a video device, so its in a whole different league of performance.
I'm looking for more current capability, which I could achieve using parallel 5532s but the issue would be open loop gain VS feedback required to make it behave. The layout is still very critical if you want the best possible performance but in all sincerety, I'd prefer a rail to rail device with fet inputs. The 5532s are excellent for low source resistance and have relatively high input offset. I'd need to design a servo circuit to eliminate the need for coupling caps. So with all these other requirements, I'd prefer something better than a bipolar 5532. I have a few signetics 5532ANs (lower noise version) to tinker with and maybe try in parallel. I can barely measure the little bit of higher order distortion of a well executed 5532 implementation with my Motu interface but I can unfortunately still hear the op amp in the circuit vs a BUF03 or OPA1692. Having worked on a truck load of consoles based on the 5532/34 series chips, I know their potential. The transformers will have a much larger influence on the sound than the op amps on paper but there are audible differences which favor the OPA1692 or AD8066. The later is a video device, so its in a whole different league of performance.
Last edited:
By the way, despite what op-amp manufacturers think, there is no such thing as a 600 ohm load. Even when 600 was used (forty years ago), it was a line terminated in 600 ohm at each end. That meant that the op-amp (zero output resistance) drove the line via a series 600 ohm resistor into the 600 line, so the op-amp saw 1200 ohm, not 600 ohm.
The transformer will introduce far more distortion, although you could get clever and drive it from a negative output resistance to reduce its distortion. Patented by Bruce Metzler of AP, the Lundahl site shows how to do this.
The transformer will introduce far more distortion, although you could get clever and drive it from a negative output resistance to reduce its distortion. Patented by Bruce Metzler of AP, the Lundahl site shows how to do this.
Has anyone ever measured the difference between the 5532N and AN versions? I believe the AN has about half the current noise of the N (which is about 4nV/Hz). I have seen better slew rate from the signetics 5532s than Ti or JRC versions. The unity gain stability of all 5532s is excellent without needing excessive measures in the peripheral circuitry. Compared to the 4558/59/60 the 5532 is superior but not to the 4556 regarding output current. 4580 is pretty good in that regard and can come close to the 5532 in other areas, so it has an advantage in some applications when current is important ie headphone amps, line drivers. It still has a lot to do with the circuit layout and PS bypassing. I tried the 4580 with the outout biased externally and liked what I heard. Its an underrated chip which deserves more attention, but it still doesn't outdo an authentic old school signetics 5532 regarding overall sonics.
@EC8010 I agree that the actual load presented by theoretical 600 ohms is higher in most cases. The resistor termination is IMO there to linearize the load. Alot of times I use a zobel on the transformer's secondary to smooth out the load and optimize the square wave performance. That rarely effects the in band FR linearity in the audible range, but it reduces overshoot and improves stability with capacitive and inductance load components. I usually shoot for a 5k load on a 1:1 coupling with an advertised 600 ohm spec. It also depends on the winding DCR. In the case of the Jensen JT11-DMCF, it likes to be run at about a 3k - 7k load. I've tried many other transformers and always come back to this Jensen model. The low end linearity is great and doesn't sound heavy or bloated compared to other transformers I've tried, especially for the purpose of galvanic isolation.
Can you tell more about your usecase? It's an impressive transformer.
I hope we agree that the price of the OPA is not the main thing when using a $90,- transformer? In the datasheet they use AD797 + buffer. Which would need at least an OPA1611 when you can step down with the output current (which should be ok for normal use in my opinion). And get rid of the 1k source resistor ... doesn't make sense with these components.
But depending on your over all system you don't need noise that low and prefer a higher current reserve?
p.s.: The measurements I showed are from OPA1611 + LME49610 as Buffer and a symmetrical input stage with some switchable gain. Feedback resistors get in the 1-2k area to achieve that noise level so a high output current of the OPAs is a must.
IME OPA2156 is very good and robust driver for low-impedance x-formers:
Observations:
WIth these precautions in place, I haven't been able to break an OPA2156 even in heavy abuse mode.
Driving with negative source resistance to compensate DC winding resistance is not worth the hassle, IMO, at least with quality x-formers. If you do, then never ever overdrive the x-former at LF, it sort of latches into a nasty fold-over distortion. With cap-less zero-impedance drive this happens as well, but with less severe latching effect.
- low offset, x-former can be DC-coupled for best LF response and distortion, and its high-Z input allows for easy AC-coupling there
- enough drive current even for demanding loads
- low enough noise for line-level applications
- 250 R safety resistors at the inputs help limiting differential input current when overdriven etc.
Observations:
- R//L breakout required at the opamp to isolate any x-former capacitance at high frequencies
- catch diodes to the supplies at the opamp output recommended to divert away back-EMF, ESD and potential back-power surges
- series Rs (2x 50R or so) recommend after x-former to limit short-circuit and back-power currents
- snubber on x-former secondary required as per mfgr's recommendation.
WIth these precautions in place, I haven't been able to break an OPA2156 even in heavy abuse mode.
Driving with negative source resistance to compensate DC winding resistance is not worth the hassle, IMO, at least with quality x-formers. If you do, then never ever overdrive the x-former at LF, it sort of latches into a nasty fold-over distortion. With cap-less zero-impedance drive this happens as well, but with less severe latching effect.
@IamJF No, I'm not tight for cash in terms of op amp choices. I've learned early on to use the best transformer I could afford. Even if I had more money, I'd still purchase the same ones. After all, the design will have M74A and T34B for mids and highs, so the driving electronics need to be up to that level of performance. With cabinetry, the running total so far is more than $6k. Yes, I know I'm crazy.
I agree with the noise vs headroom tradeoff. That's why I chose the transformer as it has one of the highest saturation points I've seen for its physical size, especially considering the low frequency performance.
This trans will be on the output side of an active analog crossover for a 3 way design. My source and amplification is entirely balanced, so I decided to use transformers for conversion to single ended (filtering) and then back. I envisioned buffers at each output to convert high impedance source from filter loading to drive the trans and covert back to floating balanced. I have experience with the OPA1692 and AD8066 for fet input devices (aside from the lesser OPA2604, LF353, TL0xx, etc).
I dont have a rough draft schematic yet. I wanted to start with the buffers, as they were the hinge point of rhe entire design. Each output will have a transformer on it as well as the input, using Jensen JT11P-1 on the input into an input buffer.
I have just under a grand into the iron alone, so I'm not looking to use cookie cutter buffers. I stll have a bunch of AD8066 but I think OPA1692 and 1612 are more suitable devices.
I agree with the noise vs headroom tradeoff. That's why I chose the transformer as it has one of the highest saturation points I've seen for its physical size, especially considering the low frequency performance.
This trans will be on the output side of an active analog crossover for a 3 way design. My source and amplification is entirely balanced, so I decided to use transformers for conversion to single ended (filtering) and then back. I envisioned buffers at each output to convert high impedance source from filter loading to drive the trans and covert back to floating balanced. I have experience with the OPA1692 and AD8066 for fet input devices (aside from the lesser OPA2604, LF353, TL0xx, etc).
I dont have a rough draft schematic yet. I wanted to start with the buffers, as they were the hinge point of rhe entire design. Each output will have a transformer on it as well as the input, using Jensen JT11P-1 on the input into an input buffer.
I have just under a grand into the iron alone, so I'm not looking to use cookie cutter buffers. I stll have a bunch of AD8066 but I think OPA1692 and 1612 are more suitable devices.
Last edited:
@KSTR That is the type of info I need. The challenges you mention here are the type of issues I want to have ironed out. I've seen buffers latch up trying to drive heavy primaries too hard. It gets much worse with DC offset and its known to cause oscillation onset with clipping the LF. This is why I want to be careful using alot of NFB to lower THD at lower gain.
I built a buffer with BD139/140 driven by 2 parallel OPA2134. It was a good sounding circuit, but needed alot of layout tweaks. I ditched that for a single ended circuit with 2SK150 at the output. I also tried an LM12 which surprisingly worked well, but those are obsolete and expensive.
I built a buffer with BD139/140 driven by 2 parallel OPA2134. It was a good sounding circuit, but needed alot of layout tweaks. I ditched that for a single ended circuit with 2SK150 at the output. I also tried an LM12 which surprisingly worked well, but those are obsolete and expensive.
I've had good success driving 600 ohm transformers using a composite amplifier. Two ICs contained within a single global negative feedback loop. The AD744 is wonderful for this job because it has JFET inputs AND dedicated I/O pins for external compensation, which are ultra handy in a composite amplifier. The HA-5002 is a video buffer with plenty of grunt; 600 ohm loads are child's play to this beast. Octopart price tables are attached.
_
_
Attachments
Thanks @Mark Johnson I'll give that buffer design a closer look. I'm not that thrilled about the coupling cap but its at least a Nichicon UES which I do have a bunch of and its sized for a low cutoff.