About the layout constrains: convenience matters when building 16 channels, you're right to keep things as much as possible on pcb.
Both the opa1642 and the opa1652 are good options.
Using a jfet opamp, you can get rid of the servo (especially at low gain), which simplifies the design and cut down cost. Since those opamps exist in quad package (with very good crosstalk figures), you could also simplify the layout and power supply quite a bit.
Btw, I'd rather take "signal postgain" at the output of the buffer.
And, finally, you've got a useful calculator for noise in opamps circuits here:Opamp noise calculator
Both the opa1642 and the opa1652 are good options.
Using a jfet opamp, you can get rid of the servo (especially at low gain), which simplifies the design and cut down cost. Since those opamps exist in quad package (with very good crosstalk figures), you could also simplify the layout and power supply quite a bit.
Btw, I'd rather take "signal postgain" at the output of the buffer.
And, finally, you've got a useful calculator for noise in opamps circuits here:Opamp noise calculator
Alright, sounds like I could get some high input impedance here without changing the design (as all of the mentioned opamps use the standard pinout).
So I will implement the following changes:
- Use OPA1652 or ISL28210 (looks very promissing) instead of LME49720
- Use R1=R2=R3=R4=100kOhms 0.1%
- Change R12=R13 to not equipped
- Change R8=R9=1Meg
- Use SLF7055T-100M2R5-3PF as output inductor
PSU:
- Remove RF filter
- Insert SLF7055T-100M2R5-3PF between buffer caps
But after discussing all these details, maybe I could ask you, 00940 and abraxalito:
Could you try to rate the whole project in its latest state (and implemented the above listed changes) from the following three points of view:
- Which real world sound quality do I have to expect?
- Does it sound completely transparent (e.g. free from audible distortions) to a good but average ear (more precisely a vocalist)? [I know your ears are trained to find very small flaws, so I don't think, I could satisfy them with my design 😛]
- What are the biggest flaws of the design atm and are there simple methods to improve them?
And finally let me use this chance to just say a big thank you for all your advice and hints so far!!
So I will implement the following changes:
- Use OPA1652 or ISL28210 (looks very promissing) instead of LME49720
- Use R1=R2=R3=R4=100kOhms 0.1%
- Change R12=R13 to not equipped
- Change R8=R9=1Meg
- Use SLF7055T-100M2R5-3PF as output inductor
PSU:
- Remove RF filter
- Insert SLF7055T-100M2R5-3PF between buffer caps
But after discussing all these details, maybe I could ask you, 00940 and abraxalito:
Could you try to rate the whole project in its latest state (and implemented the above listed changes) from the following three points of view:
- Which real world sound quality do I have to expect?
- Does it sound completely transparent (e.g. free from audible distortions) to a good but average ear (more precisely a vocalist)? [I know your ears are trained to find very small flaws, so I don't think, I could satisfy them with my design 😛]
- What are the biggest flaws of the design atm and are there simple methods to improve them?
And finally let me use this chance to just say a big thank you for all your advice and hints so far!!
Last edited:
Bleh, there's no reason to think I've better ears than you.
I've heard quite a few amps based on buffer+opamps and you really have to mess things up to have flaws and distortion. Don't worry. Even if you spend a fortune on headphones, they will be your limiting factor.
I would add to your list of changes:
- get rid of the servo, it's really not needed with jfet opamps (even without an output cap).
- move postgain after the buffer.
- c19 doesn't do anything useful.
- I'd put c20 right at the jack pins. r27 can go there too (wherever you like).
A small question: why not make C18 a leaded cap ? There's more choice in bipolar caps in leaded packages.
I've heard quite a few amps based on buffer+opamps and you really have to mess things up to have flaws and distortion. Don't worry. Even if you spend a fortune on headphones, they will be your limiting factor.
I would add to your list of changes:
- get rid of the servo, it's really not needed with jfet opamps (even without an output cap).
- move postgain after the buffer.
- c19 doesn't do anything useful.
- I'd put c20 right at the jack pins. r27 can go there too (wherever you like).
A small question: why not make C18 a leaded cap ? There's more choice in bipolar caps in leaded packages.
The LT part looks fine except its only a single and its just a tad on the noisy side (15nV input voltage noise). Being only a single means it'll cost you board space which you might wish to optimize. ISL28210 is cheaper per channel (based on Digikey small qtys), is a dual and considerably quieter.
Although I've not listened to OPA1652 nor OPA1642 I wouldn't recommend them (SQ-wise, THD-wise of course they're impeccable) based on what I've noticed in the datasheets. In particular the PSRR for the '42' looks to be negative in practice for the +ve rail. Compare the OL gain plot with the +ve rail PSRR plot - the PSRR figure is lower. I'm also wary of the OL output impedance plot rising at LF. The '52' is a CMOS part (hence not JFET input) and I've not had decent sound out of one CMOS opamp that I've tried.
1- ah ok, that would simplify the design a lot
2- yeah I already added this to my TODO list, after you wrote it, just forgot to mention it last time
3- you're right with the current value - but I just copied it from C20 as a jumper for C18 (if not equipped). Sorry for this misleading. A resistor with 0R would be more appropriate. But most likely I will completly remove C18 and C19 (independent from the DC servo)
4- Oh, you're right. I added it to the schematics as a RF bypass for induction from the headphone wire (there will be some meters of wire prior to the headphone). It will be much more effective directly at the headphone jack.
As ESR doesn't matter for DC blocking I headed for a smaller volume (which usually is provided by SMD). But most likely I will remove it anyway.
Maybe, could you put in numbers what DC offset is ok on the headphone output? (some uV, below mV, ...?)
Thanks for your feedback!
Alright, so I think I'll head for the ISL28210.
Do you have some overall rating for my project, as asked for above?
Your expectations are your own responsibility, if you expect nothing then any performance won't lead to disappointment 🙂
To me 'completely transparent' goes beyond being free from audible distortions. Something can be less than transparent and still be not annoying in any sense. Just when you compare a bypass then you'll notice the slight lack of transparency - perhaps in terms of the 'impact' (some would say PRaT perhaps).
Just as an example last night I built up a new design classA SE amp for my headphones (nothing particularly special, AKG240). How it improved on the last one was in how punchy it sounded. Prior to listening to this new one I'd have said the previous amp (also classA SE) was transparent - it was only in comparison that I noticed the latest one was better at conveying the rhythmic impact of the original performance.
The difference in design between the two is fairly simple - the later design has better PSRR.
I suggest you're speculating rather than know 🙂 I'm a newbie in this game - have only been building headamps for under a year.
The design's not complete yet, but I'd say the biggest weakness if you're aiming for maximum transparency is that its not a SE classA design. That's not a flaw, just a practical constraint because of your requirement for so many channels in a small physical space. ClassA is woefully inefficient - in SE form never better than 25%. So probably totally impractical within the constraints you've set yourself.
This is just my personal opinion (practice) so no need to give it much weight but a few (i.e. single digit) mV doesn't worry me as regards offset. It depends on the impedance of the headphone - with mine (I measured about 60ohms DC) then 6mV is going to give 100uA which really is nothing even compared to the normal bias currents in a classAB output buffer. But then again my most recent designs have been capacitor coupled...
About offset: under 10mV is usually considered as no big deal, except maybe for the most sensitive IEM.
I'd still keep the output cap if you're building for production/recording. It's a cheap safety measure.
I'd still keep the output cap if you're building for production/recording. It's a cheap safety measure.
Btw, I wouldn't worry much about the opa1642's PSRR. From the datasheet's graph, all its positive PSRR is indeed coming from feedback (give or take 1 or 2db). But, since you're operating at low gain, you have plenty of loop gain, even at 20K (about 44db). And the declining gain wrt frequency is compensated by the RC filter in the opamps' supply, which gets more effective as frequency increases. You'll have at least about 110db of effective supply noise rejection over the whole audio band.
Thanks for your rich feedback.
It sounds to me, like it is worth giving it a try.
I hope I didn't bore you too much with all my questions. It's just that 400 to 500 Euro (the approximate over all costs of that project) mean a lot of money to me as a student with no regular income. So I cannot just try out, as I won't get a second try, if you know, what I mean 😛
I will finish the design including all changes and then, of course, post the schematics and layout again.
And if you don't mind, you can have a look at them again.
It sounds to me, like it is worth giving it a try.
I hope I didn't bore you too much with all my questions. It's just that 400 to 500 Euro (the approximate over all costs of that project) mean a lot of money to me as a student with no regular income. So I cannot just try out, as I won't get a second try, if you know, what I mean 😛
I will finish the design including all changes and then, of course, post the schematics and layout again.
And if you don't mind, you can have a look at them again.
Maybe another topic (as part of this project):
Which pot would you suggest? As I need 16 of them, of course, it should be as cheap as possible, but I want to spent enough money to not degrade the system performance.
Is it worth, spending money for a conductive plastic pot (like Bourns ProAudio) or does it make no difference if I use f.e. an ALPS pot (for less that half of the price)?
Which pot would you suggest? As I need 16 of them, of course, it should be as cheap as possible, but I want to spent enough money to not degrade the system performance.
Is it worth, spending money for a conductive plastic pot (like Bourns ProAudio) or does it make no difference if I use f.e. an ALPS pot (for less that half of the price)?
I know what you mean about willing to do it right without too much wasted money. 😉
This said: the pcb costs are really a small part of the total cost (see for example smartprototyping or other similar chinese sellers). The case, the main board pcb, the power supply, all the active and passive parts will be the expensive part of it. I strongly suggest building one amplifier, see if it works as you wish and then only solder the rest. If you have to modify the pcb, all the rest is still useful and you've only lost 20€.
Bourns and ALPS make a lot of pots... any specific models in mind ? The ALPS RK097 is the bottom line at about 2€. Consider buying more than needed and keep only the ones with the best matching in between channels. The next step up is the 91 serie from Bourns. About 10€ but conductive plastic and a decent reputation.
Consider registering and asking for inputs at groupdiy.com. It's a forum dedicated to building pro audio equipement. They could be of serious help for the overall layout, etc. You'll probably get real life experience about the long term reliability of the various pots.
This said: the pcb costs are really a small part of the total cost (see for example smartprototyping or other similar chinese sellers). The case, the main board pcb, the power supply, all the active and passive parts will be the expensive part of it. I strongly suggest building one amplifier, see if it works as you wish and then only solder the rest. If you have to modify the pcb, all the rest is still useful and you've only lost 20€.
Bourns and ALPS make a lot of pots... any specific models in mind ? The ALPS RK097 is the bottom line at about 2€. Consider buying more than needed and keep only the ones with the best matching in between channels. The next step up is the 91 serie from Bourns. About 10€ but conductive plastic and a decent reputation.
Consider registering and asking for inputs at groupdiy.com. It's a forum dedicated to building pro audio equipement. They could be of serious help for the overall layout, etc. You'll probably get real life experience about the long term reliability of the various pots.
I would also try the OPA1652. It sounds really good(for a JFET op-amp) and is less than $2 at Mouser.
In my opinion, it's the best sounding of the OPA16x2 series including the bipolar input OPA1662.
Questions, when they're from curiosity are never boring as they stimulate creativity. Its claims that are boring and you've not been making those 🙂
Yes, these are the models I have in mind. ALPS R097 as "simple" pot or Bourns 91 as conductive plastic pot. But I'm amenable for alternative recommendations.
I'm not recommending them for your application as it would mean a lot of redesign work needed, but there are alternatives to traditional pots that are worth consideration. There are digitally controlled attenuators and there are DC controlled amplifiers - the latter needing no software. An example of the first is the LM1972, the second, AD603.
Getting an appropriate user interface is the biggest challenge with the digitally controlled pots - up/down buttons just don't have the ease of use of a pot. Voltage controlled amps don't suffer from this drawback but traditional VCAs have SQ issues which were discovered by an ex-colleague of mine when working at SSL.
Just a few random notes on the circuit:
1. While 1-OP balanced inputs generally get the job done, consider adding unity gain buffers for a better tradeoff between input impedance and noise. This is what line receiver ICs tend to do, too, plus laser-trimmed resistors for better CMRR. The THAT1200 series features a common-mode bootstrap circuit on top of that, for improved (transformer-like) CMRR performance in real life.
2. Any good headphone amplifier should provide some means of gain selection. Headphone sensitivity may vary by 30+ dB (from K240Ms to sensitive IEMs), and covering all of that without running into pot-related channel balance issues or limited output level can be rather tricky to say the least.
3. I suppose the power switch is placed there in order to be able to turn on the amplfier one module at a time as needed? I hope there's a mains power switch then as well. On the primary side I'd consider the odd protection element like a MOV. I don't see any fuses either?
4. Consider what your input protection diodes will do when power to the device is turned off but the source isn't. Oops.
5. Output ground return should be kept low impedance. Usually not too big a deal unless you were using long and thin wires to connect your output jack, but I'll mention it anyway. Crosstalk measurement on the finished thing with a load attached is recommended.
6. I'd agree with leaving the output coupling caps out, since those + a DC servo is just about the definition of redundancy. For good long-term reliability they'd have to be a low-leakage bipolar type anyway, unless you were willing to get fancy with two back-to-back and bias voltage.
1. While 1-OP balanced inputs generally get the job done, consider adding unity gain buffers for a better tradeoff between input impedance and noise. This is what line receiver ICs tend to do, too, plus laser-trimmed resistors for better CMRR. The THAT1200 series features a common-mode bootstrap circuit on top of that, for improved (transformer-like) CMRR performance in real life.
2. Any good headphone amplifier should provide some means of gain selection. Headphone sensitivity may vary by 30+ dB (from K240Ms to sensitive IEMs), and covering all of that without running into pot-related channel balance issues or limited output level can be rather tricky to say the least.
3. I suppose the power switch is placed there in order to be able to turn on the amplfier one module at a time as needed? I hope there's a mains power switch then as well. On the primary side I'd consider the odd protection element like a MOV. I don't see any fuses either?
4. Consider what your input protection diodes will do when power to the device is turned off but the source isn't. Oops.
5. Output ground return should be kept low impedance. Usually not too big a deal unless you were using long and thin wires to connect your output jack, but I'll mention it anyway. Crosstalk measurement on the finished thing with a load attached is recommended.
6. I'd agree with leaving the output coupling caps out, since those + a DC servo is just about the definition of redundancy. For good long-term reliability they'd have to be a low-leakage bipolar type anyway, unless you were willing to get fancy with two back-to-back and bias voltage.
Yes, thanks! That's what I am asking for.
Although, I have some questions/comments on your hints, I would be glad if you could answer.
Yes, that's the ultimate solution: Buffers and then low-impedance sourcing an bipolar Opamp (f.e. LME49720).
I didn't want to invest in the THAT1200 as I can easily achieve the same results with this solution. But I have a little space problem. I inteded to make the output cap optional if I am not contented with the work of the DC servo - like an assembly variant. (This also refers to comment 6.)
If I can rely on the work of the DC servo, I can remove the output cap and maybe have enough space for the input buffer opamp. But at this point I tried to avoid this by using a high impendance input (100k) and a FET input opamp (to avoid excessive noise).
But I'm currently neither satisfied with the choice of input opamp (OP1652 seems most promising, but has a significant current noise, ISL28210 looks very good, but is much expensive and a bit rare.), nor with the choice of DC servo opamp (TL072 is cheap and fits with medium audio performance, but has a high offset voltage (some mV), other FET-opamps are more expensive, bipolar opamps have too high offset currents for the 1Meg resistor).
Do you have some good proposals for these two opamps?
Or would you definitely suggest to insert an input buffer?
Do you speak about increasing gain or lowering gain? My input signal is at line level, so I don't think increasing the gain makes any sense (which only causes clipping). Thus I assume you are talking about lowering the gain for very sensitive headphones.
As implementation I would simply add a series resistor prior to the pot (f.e. 91k to a 10k pot) and add a switch to bypass it. Or would you suggest a more smart solution?
Yes, they are intended to switch off particular channels.
The schematics of the PSU only include the secondary side of the transformer.
The primary side consists of a mains jack, common mode choke, X and Y caps, a varistor, fuses for L and N, a mains power switch.
On the secondary side I didn't insert fuses atm. Short curcuit after the PSU should be handled by the LM317 and LM337.
Does this overall make sense or would you suggest to add anything?
Yes, depending on the source impendace this could cause some short curcuit as seen by the signal source. But don't have the opamps internal input clamps, too? Wouldn't this damage my input opamps, if I omit the external clamps, turn the power off and give some noticeably input signal to the inputs of the opamps?
I could add some series resistance prior to the clamps, but I will have to match it, to not additionally degrade CMRR performance.
What would you suggest?
I tried to optimize all return currents and impedances. Do you have some concerns to some place in the PCB?
If you refer to the ground trace to the output connector: I tried to feed it back to where the currents are sourced. Impedance should be no problem, as this is a 1mm track (5 mOhms/cm), so about 0.025 ohms - in the range of a wire.
I was adviced not to use a bipolar type but a FET type as input offset currents in combination with the 1 Meg input resistor would cause a significant additional input offset voltage (>10mV).
Can you suggest an opamp suited for this?
Thanks for your help!
- Status
- Not open for further replies.