I've built the Rod Elliot 113 headphone amp and although it sounds pretty good you only just need to move the pot (20k) before it's very loud.
I'm using the standard circuit except for the 20k Pot
Headphone Amplifier
It looks like the gain is 23 (unless I've calculated wrongly), I'm thinking that reducing R4 to something like 4.7k to 6k would reduce the gain, I would appreciate any thoughts on if it's this simple or if anything else needs to change.
Setup is as follows:
Inputs: Pioneer 565 DVD, Sabre based DAC, PC (Xonar d2X)
Headphones: Beyerdynamic DT990 pro (250 ohm).
Cheers
Andy
I'm using the standard circuit except for the 20k Pot
An externally hosted image should be here but it was not working when we last tested it.
Headphone Amplifier
It looks like the gain is 23 (unless I've calculated wrongly), I'm thinking that reducing R4 to something like 4.7k to 6k would reduce the gain, I would appreciate any thoughts on if it's this simple or if anything else needs to change.
Setup is as follows:
Inputs: Pioneer 565 DVD, Sabre based DAC, PC (Xonar d2X)
Headphones: Beyerdynamic DT990 pro (250 ohm).
Cheers
Andy
A gain of 23 is way too much.
On my own discrete headphone designs I use a gain of 2.5, which is enough for 5VRMS output if you use a 2VRMS source. That gives me 100mW into my 250 Ohm DT880 PRO, which is enough for 116db peaks, more than sufficient, unless you like getting hearing damage.
The problem with low gain settings, are that unless the headphone amp has been desgined with that in mind from the beginning, you could end up with stability issues.
However, since this design is a composite amp, with a simple discrete buffer within the feedback loop of an Opamp, I cant really give you a definitive answer, you might just have to try and see if it works. 🙂
On my own discrete headphone designs I use a gain of 2.5, which is enough for 5VRMS output if you use a 2VRMS source. That gives me 100mW into my 250 Ohm DT880 PRO, which is enough for 116db peaks, more than sufficient, unless you like getting hearing damage.
The problem with low gain settings, are that unless the headphone amp has been desgined with that in mind from the beginning, you could end up with stability issues.
However, since this design is a composite amp, with a simple discrete buffer within the feedback loop of an Opamp, I cant really give you a definitive answer, you might just have to try and see if it works. 🙂
Last edited:
Hi Neutrality,
Thanks for the post. I recon my inputs will vary between 1.5 to 2 volts so your gain sounds good for me.
I see your point regarding stability at lower gains, I should have mentioned that I'm using an LM4562 op amp for the voltage gain, which I think should be good for low/no gain.
I don't know much about the transistor buffer stage though and how this is affected, I would be grateful for any opinions on this.
Cheers,
Andy
Thanks for the post. I recon my inputs will vary between 1.5 to 2 volts so your gain sounds good for me.
I see your point regarding stability at lower gains, I should have mentioned that I'm using an LM4562 op amp for the voltage gain, which I think should be good for low/no gain.
I don't know much about the transistor buffer stage though and how this is affected, I would be grateful for any opinions on this.
Cheers,
Andy
I have built the P113 using the LM4562 with R4 set to 3k9. If you want to go lower than this bear in mind that R5 & R6 also present a load to the opamp, so it might be better to increase R3 to 1k5 or 2k2 if you need less gain.
I used a 10k pot but changed R1 to 47k. Ideally, R1 should be substantially higher than the pot value. I also omitted R2 completely. This now results in "unbalanced" input currents so you should ideally use a FET op amp like the OPA2132 or OPA2107, but I found the LM4562 having low input bias current and offset voltages for a bipolar works fine. It also gives slightly more sparkle to the sound than the OPAs. However, the difference is only slight. Nowhere near the difference you would expect in a CMOY design.
A few other points that might help. Different opamps may not make a huge difference but the quality of the pot does. Don't use a cheap carbon pot - it sounds horrible. C2 should be a bipolar electrolytic. Bypass them with a film cap. I just had a few 1uf Wima polyesters but they gave an increase in treble definition.
Rod recommends 120 ohm output resistor but this makes the output impedance quite high. I used a 4r7 resistor instead.
Hope this helps.
I used a 10k pot but changed R1 to 47k. Ideally, R1 should be substantially higher than the pot value. I also omitted R2 completely. This now results in "unbalanced" input currents so you should ideally use a FET op amp like the OPA2132 or OPA2107, but I found the LM4562 having low input bias current and offset voltages for a bipolar works fine. It also gives slightly more sparkle to the sound than the OPAs. However, the difference is only slight. Nowhere near the difference you would expect in a CMOY design.
A few other points that might help. Different opamps may not make a huge difference but the quality of the pot does. Don't use a cheap carbon pot - it sounds horrible. C2 should be a bipolar electrolytic. Bypass them with a film cap. I just had a few 1uf Wima polyesters but they gave an increase in treble definition.
Rod recommends 120 ohm output resistor but this makes the output impedance quite high. I used a 4r7 resistor instead.
Hope this helps.
Richard,
Thanks for all the info. Before I saw your post I changed r4 to 2.2k and it was a bit hairy with the PCB pads so I don't think I dare risk swapping again. I will get a better 10k pot as the one in using is an akps 20k from a goody bag. Re the bipolar cap in using a polarised electro and it seems ok. I have some 2.2uf wima film caps would these be ok for bypass? Do you know what this bi cap does and also how bypassing helps?
Thanks again for the info.
Andy
Thanks for all the info. Before I saw your post I changed r4 to 2.2k and it was a bit hairy with the PCB pads so I don't think I dare risk swapping again. I will get a better 10k pot as the one in using is an akps 20k from a goody bag. Re the bipolar cap in using a polarised electro and it seems ok. I have some 2.2uf wima film caps would these be ok for bypass? Do you know what this bi cap does and also how bypassing helps?
Thanks again for the info.
Andy
Andy
Polarised electrolytics work best with a reasonable level of DC voltage across them. In this circuit we are only talking about millivolts. Rod Elliott in his application note (you can access this if you bought the PCB from him - he will have sent you the password) presumably changed the cap to a bipolar because of this.
However, if it works fine don't change it - no damage will result. Bypassing an electrolytic with a film cap sometimes yields an improvement in audio quality - depending on the circuit. Electrolytics are technically the worst kind of capacitor but are unavoidable for large values.
In this case there was an improvement, not spectacular but nonetheless an improvement. This is with Sennheiser HD650 headphones. You should be able to solder the Wima caps you have to the underside of the PCB in parallel with the electros. Give it a try. No harm will result by leaving them in place even if you think they have not made things better.
Polarised electrolytics work best with a reasonable level of DC voltage across them. In this circuit we are only talking about millivolts. Rod Elliott in his application note (you can access this if you bought the PCB from him - he will have sent you the password) presumably changed the cap to a bipolar because of this.
However, if it works fine don't change it - no damage will result. Bypassing an electrolytic with a film cap sometimes yields an improvement in audio quality - depending on the circuit. Electrolytics are technically the worst kind of capacitor but are unavoidable for large values.
In this case there was an improvement, not spectacular but nonetheless an improvement. This is with Sennheiser HD650 headphones. You should be able to solder the Wima caps you have to the underside of the PCB in parallel with the electros. Give it a try. No harm will result by leaving them in place even if you think they have not made things better.
Just finished my build of p113, what a fantastic headphone amp. I used a 100k ALPS pot (like the op, all I had handy), and two opa627 on an adapter. Even without burning it in, the soundstage is excellent.
I'm using DT880 (600 ohms impedance) without any volume control problems.
What is the technical reason a 100k pot is inappropriate here?
I'm using DT880 (600 ohms impedance) without any volume control problems.
What is the technical reason a 100k pot is inappropriate here?
Noise.
Sorry for my non-technical response; I'll leave it to someone else to elaborate.
Exactly, and depending on opamp type and its common-mode distortion performance, linearity with increased source impedance doesn't exactly improve either. With a 100k, source impedance can go as high as 25k, which is a bit much. Incidentally, voltage noise on the OPA627 is equivalent to <2 kOhms, so noise would actually be limited by the pot at some settings even with a 10k type.
This amplifier pretty much is a plain vanilla opamp + AB buffer design (plus bias bypass). It should perform well. A few notes though:
1. As mentioned, the stock gain of 27 dB is excessive. With a weak source (~1 Vrms) and K240Ms or the like, I might see R4 values of up to 8k2 or 9k1 making sense (i.e. up to 20 dB), otherwise you'll probably be best off in the 2k2...3k9 range (10..14 dB). The buffer should have a bandwidth of 50+ MHz, so it would take a very fast opamp (by audio standards) to run into trouble even at unity gain.
2. R2 may still be good for keeping RF out, but as shown, the impedance seen by the +input is always greater than the one seen by the -input at any volume setting. Drop it below R3||R4, and you'll see them equal at some low volume at least.
3. According to simulation, you can replace C3-C4 with just one cap (connected from C3+ to C4-). Seems counterintuitive, but it appears to be totally equivalent.
This amplifier pretty much is a plain vanilla opamp + AB buffer design (plus bias bypass). It should perform well. A few notes though:
1. As mentioned, the stock gain of 27 dB is excessive. With a weak source (~1 Vrms) and K240Ms or the like, I might see R4 values of up to 8k2 or 9k1 making sense (i.e. up to 20 dB), otherwise you'll probably be best off in the 2k2...3k9 range (10..14 dB). The buffer should have a bandwidth of 50+ MHz, so it would take a very fast opamp (by audio standards) to run into trouble even at unity gain.
2. R2 may still be good for keeping RF out, but as shown, the impedance seen by the +input is always greater than the one seen by the -input at any volume setting. Drop it below R3||R4, and you'll see them equal at some low volume at least.
3. According to simulation, you can replace C3-C4 with just one cap (connected from C3+ to C4-). Seems counterintuitive, but it appears to be totally equivalent.
Last edited:
Thanks for raising these great points.
Generally my favourite opamps (on dual adapters in this case) are AD825 and OPA627, however I notice that OPA627 in this circuit gives the best bass clarity (thought the sound is noticeably 'darker'). I'm not familiar with voltage noise measurements - are you suggesting that the noise is limited by the pot, no matter what value pot, hence the 100k pot should be ok in this instance? I certainly don't hear any noise, quiet is a grave!
Generally my favourite opamps (on dual adapters in this case) are AD825 and OPA627, however I notice that OPA627 in this circuit gives the best bass clarity (thought the sound is noticeably 'darker'). I'm not familiar with voltage noise measurements - are you suggesting that the noise is limited by the pot, no matter what value pot, hence the 100k pot should be ok in this instance? I certainly don't hear any noise, quiet is a grave!
No, certainly not. Depending on volume setting, the 100k pot would give up to 10 dB more noise than a 10k, the effect being greatest around the -6 dB point. Actually it wouldn't become as bad as this here, as the parallel R2 would limit maximum source impedance to ~12.7k (and distort the pot characteristic near the upper range). So the noise penalty over a 10k would probably be more like 5 dB in this circuit. Normally R2 should be > Rpot.
Total input-referred noise is probably best calculated like this:
1. Determine impedance seen by noninverting input.
2. Ditto for inverting input.
3. Add the two.
4. Determine equivalent input noise density en = sqrt(4 k T R) ~= sqrt (1.63E-20 * R) @room temp
5. RMS sum resistor en and specified opamp en, i.e. en,tot = sqrt(en,R² + en,OP²)
6. Total noise Vn = en * sqrt(bandwidth) ~= en * 141 @20 kHz
Actually I created a little spreadsheet last year that does this sort of stuff and conveniently calculates output noise and SNR. There you go. It's an OpenDocument thingy (Open/LibreOffice, M$ should have converters nowadays). I don't have an R2 equivalent in there, but it should be useful as-is.
That's not terribly surprising. DT880-600s aren't all that sensitive, around 96 dB SPL / 1 V. They'd probably still keep noise generators like a TDA2822 in check, or your average integrated amp for that matter. Designs like P113 were basically created for headphones like these.
Your amp now has a worst-case noise level of around 50 µV @20 kHz BW. That's like 10 dB SPL with these cans. At normal volumes it would probably be even less, about 4-5 dB SPL. The threshold of audibility is in the 20-30 dB SPL vicinity, depending on headphone type. IOW, this combo should be as dead silent in theory as you found it to be in practice.
You should also be around the -30 dB point in terms of volume setting, which is a totally unproblematic range for most any pot.
Now enter some headphones or in-ears which are 30 dB more sensitive (they're not uncommon), and it's a totally different world. Noise becomes distinctly audible, and operating a volume pot around -60 dB usually is not much fun even if it does keep decent channel tracking. Hence, this thread.
(Though I'm a little surprised that the OP would have run into trouble with DT990s, which aren't that much more sensitive.)
Last edited:
Sgrossklass,
I just wanted to confirm the trouble I had with the DT990's wasn't to do with unwanted noise, hiss etc. It was to do with the standard gain of the circuit design which was huge so even with not sensitive beyers I only had to move the pot (20k ohm) by a few degrees from minimum before my ears were blown off 😱.
Cheers,
Andy
I just wanted to confirm the trouble I had with the DT990's wasn't to do with unwanted noise, hiss etc. It was to do with the standard gain of the circuit design which was huge so even with not sensitive beyers I only had to move the pot (20k ohm) by a few degrees from minimum before my ears were blown off 😱.
Cheers,
Andy
I find the stock gain perfect for my non-over sampling ad1865 dac with passive I/V......though the output is about 125mVrms so I can certainly see how the p113 would have too much gain for a source with 2vrms.
That's 125 mVrms fullscale? That would be 24 dB under CD player level. So with a standard CDP you'd be getting along just fine with a whopping 3 dB gain in the headphone amp, in spite of the rather insensitive DT880-600s. That would seem to fit; even with rather insensitive 600 ohm cans (HD420SL) my home listening levels tend to be at a few dozen mV at best, and I tend to be setting my Clip+ for about 80 mVrms FS with typical ReplayGained (89 dB SPL / -14 dBFS avg) non-Classical recordings.
That would already put us at -55 dB with a CD source and stock P113. Which happens to be about the range where people start complaining about touchy volume setting. And that's with insensitive cans, albeit at lowish levels. Oops. (I'm not sure how I got to -30 dB earlier.)
It's certainly interesting. Stock P113 gain is sized for an input sensitivity of about 400 mVrms (minimum level required for maximum output). That already is 6 to 8.5 dB higher than on many integrated amps, yet still makes for a touchy volume pot. (Complaints about touchy volume or bad tracking are not uncommon on integrated amps, as these are dimensioned like they were 20+ years ago but average levels on CDs have gone up by 10+ dB.) Then again, stock P113 could drive DT880-600s to about 115 dB, while your average 100wpc speaker amp will get maybe 105 dB out of average sensitivity speakers. The difference in input sensitivity is not enough to offset this, the headphone combo still remains somewhat more sensitive. With more sensitive cans, it gets even worse.
Still, it might not hurt to double check the volume pot for goofs - like miswiring or wrong type (e.g. linear instead of log). Murphy never sleeps. 😉
That would already put us at -55 dB with a CD source and stock P113. Which happens to be about the range where people start complaining about touchy volume setting. And that's with insensitive cans, albeit at lowish levels. Oops. (I'm not sure how I got to -30 dB earlier.)
It's certainly interesting. Stock P113 gain is sized for an input sensitivity of about 400 mVrms (minimum level required for maximum output). That already is 6 to 8.5 dB higher than on many integrated amps, yet still makes for a touchy volume pot. (Complaints about touchy volume or bad tracking are not uncommon on integrated amps, as these are dimensioned like they were 20+ years ago but average levels on CDs have gone up by 10+ dB.) Then again, stock P113 could drive DT880-600s to about 115 dB, while your average 100wpc speaker amp will get maybe 105 dB out of average sensitivity speakers. The difference in input sensitivity is not enough to offset this, the headphone combo still remains somewhat more sensitive. With more sensitive cans, it gets even worse.
Still, it might not hurt to double check the volume pot for goofs - like miswiring or wrong type (e.g. linear instead of log). Murphy never sleeps. 😉
Last edited:
Exactly (almost) right – in fact, the AD1865 used with 200R i/v resistors (unbalanced) gives 127mVrms! I’m using 91R MK132 resistors so the output is probably less than 100mVrms. The P113 is almost acting like a higher current i/v gain stage, and gives listening volumes around 12PM on the volume pot. Works very well indeed!
I once had an NAD integrated amplifier, with a cd player as the source the volume couldn’t go past 9PM without being too loud, but the channel tracking was absolutely terrible!
I once had an NAD integrated amplifier, with a cd player as the source the volume couldn’t go past 9PM without being too loud, but the channel tracking was absolutely terrible!
Thought I'd hook my p113 up to a standard 2vrms cd player to see what other users are experiencing and oh boy, the gain is FAR too high, even for my 600 ohm headphones. The pot also gets a lot noisier, to the point of being annoying during silent passages in music.
I also read on rod elliott's page that this amp could be used to power efficient speakers. So, I took the volume pot out & attached it to some 90dB floor standing speakers I have and lo and behold, with a low gain preamp and standard 2vrms source, we have a beautiful 5W power amp! In fact, the sound with speakers easily rivals my 6V6 push pull tube amp AND uses 5% of the energy!
Presumably the 120r resistors at the output are not needed if we are using this circuit with 8 ohm speakers?
I also read on rod elliott's page that this amp could be used to power efficient speakers. So, I took the volume pot out & attached it to some 90dB floor standing speakers I have and lo and behold, with a low gain preamp and standard 2vrms source, we have a beautiful 5W power amp! In fact, the sound with speakers easily rivals my 6V6 push pull tube amp AND uses 5% of the energy!
Presumably the 120r resistors at the output are not needed if we are using this circuit with 8 ohm speakers?
Actually, he himself writes,
Which seems about right for medium power transistors like those used. (That's about 600 mA peak at full power into 8 ohms.) Earle Eaton used beefier types in his amp.
Ultimately there's only so much an amp with a single-stage buffer can drive though. A 5532 can supply up to 25 mArms or so, and you get a beta of maybe 50 with good low-droop transistors. That would still be good for ~10 W into 8 ohms (voltage limited at +/-15 V) or 6 W into 4 ohms, but your average LM317/337 or 7815/7915 supply would collapse first and require some beefing up.
Ultimately you'd be looking at a two-stage (e.g. EF2) buffer and a matching power supply, possibly some Zobel network at the output as well. Still, an opamp-based concept makes a fair bit of sense for a low-power amp. You could get maybe 20-25 watts into 4 ohms with common opamps, which is a workable amount for sure.
This amp uses an opamp with CFP output, though I must say I don't like the biasing with only one diode. it's either underbiased or thermally unstable.
No, unless you're happy with a 40 mW / 8 Ohm, current output amp that is. 😉
Are you sure you're not using a linear taper potentiometer instead of a logarithmic taper? Rod has an article on his site about how to fake a log taper from a linear pot.
If you're not sure, set the pot at its halfway position and measure resistances from the wiper to the other two points. They should vary by maybe 10:1 for a log pot.
I am a newbie in this sector but i want to build this Rod Elliott's headphone amplifier. I have collected NE 5532 chip and all other components for this project. Do anyone have the pcb layout for this project? If so, then please let me have it. Hope you people will help. Thanks.
- Status
- Not open for further replies.