OPA627's
Absolutely agree. I built a circuit on Vero board and referenced the OPA627/637 datasheet to select decoupling. Excellent amplifier for critical audio applications although very expensive. 😱
I just looked on ebay for buffer pcb or preamp pcb there are many to choose from like this dual one for a couple of bucks.
Dual OP AMP Preamp DC Amplification Empty Board FOR NE5532 OPA2134 OPA2604 AD826 | eBay
Cheers George
still shrill 627
well tried a lot but same result
OPA627: OK have the DIP to fit straight into socket. Let the CDP on over night, for an IC like that it should be plenty to settle in. All recaping I did some months a go and should also be settled.
SUpply after recap: 1 set of 100uF silmic + 680pF PP for each rail for IV - and another set for buffer/filt (though shared L+R). See diagram. Also have 100nF ceramics straight under pins for rails and in between rails (+ to -). Now this morning put further 18uF panasonic FR between rails and 4,7uF FC on rails straight under pins.
Gain: changed TZ from 1,5k->600 and buff/filt Av 1½->2. I.e. a total of 0.5 gain compared to what it was before. And it works so I can turn my volume knob a bit more
Sound: tried this setup now.. but sound is same as yesterday
OPA627: very open/resolving..involving .. but sharp/shrill in up mids =impressive but tiring long term
ADEL2020: also rather open.. and some sharpness but not as much, but is also somewhat uninvolving
the other opamps I have can have better control of damping the
'shrillness' but not as open as the 627 and ADEL.
Test material used for evaluating mods:
1) Clapton, Reptile, track 1 & 2.
Sharp guitars w snap and grunt..delicate percussion and keyboard in background..mix between open and more crowded instrumentation+ there is voice
He normally has more muddy mixings, but this one is ok..and there is possibility to test if the details can be resolved from all the grunt in the forground and can be placed in space
2) (the torture test) : Richard Galliano, French touch, track 1
Dominant accordion in foreground, very strong voiced and sharp. Gentle jazz drum/perc. + double bass in back ground. The accordion has some transitions that are very sharp .. but does they cut like a razor or a chain saw? (i.e very very easily get very shrill!)
+ few other albums with mix of electric and acoustic instruments
This evening Further Mod A-class Biasing:
read 2-4mA should do it and OPA627 particularly would 'gentle up' in the highs
so 3mA (4,7k from out to - rail)
Actually not really hear any change? (mayby very slightly less 'sss' on voice and more grain in voice and the metallic part of percussion (the resonant 'poing' part))
then 6mA (2x 4,7k ||)
still no real change. OPA627 is still shrill to an extent I doubt I will have long term pleasure listening to some of my favourite materials (and the graininess might have picked up little w 6mA bias )
SO what can be done to tame the devil?? ..any other tricks up someone sleeves?
(reintroduce output coupling caps and use silmics instead of the PP I had first = i.e. sound shaping..ok w me, am not after neutral anyway)
thanks
/Patrick
well tried a lot but same result
OPA627: OK have the DIP to fit straight into socket. Let the CDP on over night, for an IC like that it should be plenty to settle in. All recaping I did some months a go and should also be settled.
SUpply after recap: 1 set of 100uF silmic + 680pF PP for each rail for IV - and another set for buffer/filt (though shared L+R). See diagram. Also have 100nF ceramics straight under pins for rails and in between rails (+ to -). Now this morning put further 18uF panasonic FR between rails and 4,7uF FC on rails straight under pins.
Gain: changed TZ from 1,5k->600 and buff/filt Av 1½->2. I.e. a total of 0.5 gain compared to what it was before. And it works so I can turn my volume knob a bit more
Sound: tried this setup now.. but sound is same as yesterday
OPA627: very open/resolving..involving .. but sharp/shrill in up mids =impressive but tiring long term
ADEL2020: also rather open.. and some sharpness but not as much, but is also somewhat uninvolving
the other opamps I have can have better control of damping the
'shrillness' but not as open as the 627 and ADEL.
Test material used for evaluating mods:
1) Clapton, Reptile, track 1 & 2.
Sharp guitars w snap and grunt..delicate percussion and keyboard in background..mix between open and more crowded instrumentation+ there is voice
He normally has more muddy mixings, but this one is ok..and there is possibility to test if the details can be resolved from all the grunt in the forground and can be placed in space
2) (the torture test) : Richard Galliano, French touch, track 1
Dominant accordion in foreground, very strong voiced and sharp. Gentle jazz drum/perc. + double bass in back ground. The accordion has some transitions that are very sharp .. but does they cut like a razor or a chain saw? (i.e very very easily get very shrill!)
+ few other albums with mix of electric and acoustic instruments
This evening Further Mod A-class Biasing:
read 2-4mA should do it and OPA627 particularly would 'gentle up' in the highs
so 3mA (4,7k from out to - rail)
Actually not really hear any change? (mayby very slightly less 'sss' on voice and more grain in voice and the metallic part of percussion (the resonant 'poing' part))
then 6mA (2x 4,7k ||)
still no real change. OPA627 is still shrill to an extent I doubt I will have long term pleasure listening to some of my favourite materials (and the graininess might have picked up little w 6mA bias )
SO what can be done to tame the devil?? ..any other tricks up someone sleeves?
(reintroduce output coupling caps and use silmics instead of the PP I had first = i.e. sound shaping..ok w me, am not after neutral anyway)
thanks
/Patrick
Attachments
OPA627's
The OPA627 is not known to be sharp sounding. That suggest that the filter is messing with it. In my application I use it as a simple non-inverting buffer i.e. no capacitor filter components in it's feedback path. Just voltage buffer. I have never liked the sound of a Sallen key filter. The Arcam appears to use one. To many phase problems. I have even built 7th order GIC Bessel gyrator based filters. Really a task to build something like that. Complicated and in the end I got better filter results with a line transformer based filter. Super simple and no active devices in the signal path. ADEL2020... Best applied in place of the triple stack with standard feedback. As an I/V. It is a fairly fast video amplifier. Filter position.... You could try a OPA604 family part. That is what was commonly used there in the old days. Sound quality... It is possible to tune your CD player to sound anyway you like it. Will take a lot of experimenting. Testing... I have been able to verify my results on my cousin's stereo system. Even though we share not one component between our systems. It kind of comes down to taste. I am shooting for accuracy and resolution. I also found dynamics and excellent bass. Considering my loud speakers are handicapped somewhat in the bass being planar bass and large ribbon tweeters, I get shockingly low bass with excellent drive. Maybe it is having 300 WPC of power. 😀
The OPA627 is not known to be sharp sounding. That suggest that the filter is messing with it. In my application I use it as a simple non-inverting buffer i.e. no capacitor filter components in it's feedback path. Just voltage buffer. I have never liked the sound of a Sallen key filter. The Arcam appears to use one. To many phase problems. I have even built 7th order GIC Bessel gyrator based filters. Really a task to build something like that. Complicated and in the end I got better filter results with a line transformer based filter. Super simple and no active devices in the signal path. ADEL2020... Best applied in place of the triple stack with standard feedback. As an I/V. It is a fairly fast video amplifier. Filter position.... You could try a OPA604 family part. That is what was commonly used there in the old days. Sound quality... It is possible to tune your CD player to sound anyway you like it. Will take a lot of experimenting. Testing... I have been able to verify my results on my cousin's stereo system. Even though we share not one component between our systems. It kind of comes down to taste. I am shooting for accuracy and resolution. I also found dynamics and excellent bass. Considering my loud speakers are handicapped somewhat in the bass being planar bass and large ribbon tweeters, I get shockingly low bass with excellent drive. Maybe it is having 300 WPC of power. 😀
Patrick - from looking at your attached schematic you may well find an improvement in SQ from scaling up your resistor values to reduce loading on the OPA627 opamp's outputs. Firstly I'd make the 1k0 feedback resistor into 22k and the 1k8 shunt resistor into 39k. The amount of classA bias you need is directly related to the load impedance an opamp sees, it should be the minimum to keep the OPS in classA.
@ Dave
yes seems a 3rd order Sallen-Key. Pluggin in R's and C's in online calculator ( sim.okawa-denshi.jp ) the 3 equivalent pole frequencies are 29KHz, 160Khz and 50Khz. Pulling feed back C11 out makes sense -> pushes 160Khz pole to inf. And the remaining 2order is at 50Khz w an 0.8 damping factor. Also the cap at TZ provides an extra pole at 100kHz already, so nothing lost really pulling feedback cap out = done!
Have a set OPA604 (inside my prime systems OPera ref. 2.2.. will only pull out as last attempt at things w the Arcam)
@abraxalito
OK the output signal load is a R14=33K internal in the Arcam and my pre input is 10K = ca 7.7k, in || bias load (4,7k or 2,2k) || 1k+1k feedback loop = 1,2k-922. Is that low? (I could change R14 to 100K+ and pre input to 33K or something). Anyhow w 3Vpeak swing over the 7.7k out it is ca 0.2mA signal load current. So 3mA bias =15x and 6mA bias =30x over = should be plenty?
Now for feedback loop was 1k+1k -> changed both to R11=R12=22k now -> Av=2 (and my TZ resistance i 600Ohm now -> overall lowering of gain by factor ½.) 22k+22k >> bias and signal load now.
so started testing where left off yesterday (but now feedback cap c14 = 0/cut, feed back resistors R11_R22=22k)
6mA bias (Rpull 4,7k -> 30x signal)
hmm better wrt shrillness reduced at least not 'ear piercing' on accordion. But still some grain-> loosing some detail and air/space -> sound stage compressed 1D nearly
3mA bias (Rpull 2,2k ->15x signal)
now get sharper again (not so much as w/o bias) ..too much shrillness, little more air and space back
hmm the 2sk170 I got (and not used) for AD844 bias: The GR measured genuine w Idss = ca 3mA. The BL fakes w Idss= ca 12mA but low gfs about 14 mS..-> so try JFET bias w a 1k series form output as suggest by tangentsoft.net
3.4(L)-3.6(R) mA bias (Rpull=1k + JFET ->15x signal)
so now I have a lot of space/air and detail back..but also sharpness/shrill
No bias
pull the bias netw. open. Now back close to where I started yesterday.. lot of space and air and detail and involvement..but too sharp/shrill
Conclusion
pulling feed back cap as such not the big change it seems. The bias and balance wrt feedback resistors and signal load impedance..does impact
and could also be the AD 'nursing' and new IV stage is giving lots more mid/top detail out than I am used too (narj not only, as the accordion IS too shrill compared to what it can be)
What to try next
seem the JFET did change things compared to just a pull resistors for same bias current -> try a Cascode with the fake 12mA JFETS..will it work w such low gfs of 14mS? (and then some resistor or trimmer -> trying go into the 6-8mA domain and see if 'gentle' comes back but does not kill air and space?)
As last resort put in output caps..MKP I don't believe will change a lot -> something more drastic -> lytics or PIO etc
and then finally try putting OPA604s in
else?
Thanks for continuing battling through my longish posts and giving valuable feedback
/Patrick
Last edited:
TZ node resistor (I/V)
Hi Patrick, When you reduced the TZ resistor you need to scale up the capacitors value. So halving TZ resistor requires doubling the capacitor from 1.8 nF to 3.6 nF. Worth a try anyway. 😉
Hi Patrick, When you reduced the TZ resistor you need to scale up the capacitors value. So halving TZ resistor requires doubling the capacitor from 1.8 nF to 3.6 nF. Worth a try anyway. 😉
Sort of ok now
yes sorry didn't mention. For simplicity just parralled both an extra resistor and cap across the original 1,5k + 1nF. So time constant remained same
Further trials on OPA627
8mA JFET bias
inspired by old thread by carlosfm: http://www.diyaudio.com/forums/chip-amps/38984-opa627-really-sings.html
Well ok, slight rounding of sharper edges, while remaining w most detail and air. But the full space/ depth not remained (possibly slight 'masking' of micro detail, damping 'sharps' but hiding fine space resolution?)
7mA Cascode JFET bias
hm seem temperature sensitive? runs from 7,5->7mA in a few minutes. Sound wise well, little grainy, not as enjoyable as the 8mA single JFET
0 bias
is definately to shrill for me .. but has best air and deep sound stage
8mA JFET bias back
spent rest evening listening to this one. Is fine.. lots of detail and enjoyable.. though still little on sharp side (but in a level probably I can get used to w/o fatiguing)
So seem there is a trade off: either all micro detail and space /depth - but at the cost of edges being too sharp/shrill
Now wrt AD844 IV. Could it be this stage (on pin5 TZ) is the one where the detail vs sharpness 'knife edge balancing' originates from? (and buf/filt just fighting to convey it)
-> would stacking up from 3x to 4x AD844 have en effect in the 'gentling' department?
or is there a sweet spot wrt the load impedance on pin5 TZ?
thanks
/Patrick
HI Dave
yes sorry didn't mention. For simplicity just parralled both an extra resistor and cap across the original 1,5k + 1nF. So time constant remained same
Further trials on OPA627
8mA JFET bias
inspired by old thread by carlosfm: http://www.diyaudio.com/forums/chip-amps/38984-opa627-really-sings.html
Well ok, slight rounding of sharper edges, while remaining w most detail and air. But the full space/ depth not remained (possibly slight 'masking' of micro detail, damping 'sharps' but hiding fine space resolution?)
7mA Cascode JFET bias
hm seem temperature sensitive? runs from 7,5->7mA in a few minutes. Sound wise well, little grainy, not as enjoyable as the 8mA single JFET
0 bias
is definately to shrill for me .. but has best air and deep sound stage
8mA JFET bias back
spent rest evening listening to this one. Is fine.. lots of detail and enjoyable.. though still little on sharp side (but in a level probably I can get used to w/o fatiguing)
So seem there is a trade off: either all micro detail and space /depth - but at the cost of edges being too sharp/shrill
Now wrt AD844 IV. Could it be this stage (on pin5 TZ) is the one where the detail vs sharpness 'knife edge balancing' originates from? (and buf/filt just fighting to convey it)
-> would stacking up from 3x to 4x AD844 have en effect in the 'gentling' department?
or is there a sweet spot wrt the load impedance on pin5 TZ?
thanks
/Patrick
Last edited:
OPA627's
Hi Patrick, Well your results are not typical. Means for some reason your getting that shrill sound somewhere along the line. So question is what loud speakers are you using? Maybe it just more of the same and sums to shrill sound. At some point I'd like to see you take a fresh start at this thing. I went modular a long time ago. Downside is you need a lot of space to fit the new circuit boards. Using the triple stack as an I/V off the TZ point is well and good. The issue as I see it is no one has placed a 3rd order Sallen Key filter off the TZ as you have. Could be doing something odd. In my old CDB650 CD player I added an I/V board in front of the Sallen key filter. Since the original circuit was both I/V and filter I ended up having a phase reversal. Had to swap speaker wire polarity to correct the phase. Depends which input the circuit is using at the filter position. If it is + input you should be fine. If it is - in. You have phase problems. 😕 That all of course is going back to about 1990... A blast from the past.
Hi Patrick, Well your results are not typical. Means for some reason your getting that shrill sound somewhere along the line. So question is what loud speakers are you using? Maybe it just more of the same and sums to shrill sound. At some point I'd like to see you take a fresh start at this thing. I went modular a long time ago. Downside is you need a lot of space to fit the new circuit boards. Using the triple stack as an I/V off the TZ point is well and good. The issue as I see it is no one has placed a 3rd order Sallen Key filter off the TZ as you have. Could be doing something odd. In my old CDB650 CD player I added an I/V board in front of the Sallen key filter. Since the original circuit was both I/V and filter I ended up having a phase reversal. Had to swap speaker wire polarity to correct the phase. Depends which input the circuit is using at the filter position. If it is + input you should be fine. If it is - in. You have phase problems. 😕 That all of course is going back to about 1990... A blast from the past.
3 VS 4 AD844's
George did this experiment a long time ago. As I recall he sad it was tough to hear any difference between a 3 stack and a 4 stack. Some guys with the TDA1541 did like the 4 stack. I don't think it will fix your "shrill" problem.
George did this experiment a long time ago. As I recall he sad it was tough to hear any difference between a 3 stack and a 4 stack. Some guys with the TDA1541 did like the 4 stack. I don't think it will fix your "shrill" problem.
Hi Dave
wrt speakers Avance Dana/Danish 2way dynamic speaker design w an Dynaudio Esotec treble and scanspeak 6½" woof/mid ->yes have modded the xover ½-1y ago.. went PIO and Mundorf SIO instead of MKP in the caps .. and from normal core wound coils to core free ribbons = yes has opened both ends. BUT not shrill on its own / smooth fine but present top .. and mid very involving mid. But see your point, more open in top in several parts of the chain = maybe to much of the good.
(saw you used ribbon speakers, well my prime system, which is packed down now, uses MG1.6QR, also modded w xover = w fantastic result..a friend said he could here the drums sticks where made of wood)
OK pulling all out / separate boards, think too much work now wrt this is 'just' my 2ndary system - but is really challenging my prime now (though bit different sound profiles)
One other thing might impact the IV..not so much the 3rd order Sallen key (now 2nd order after pulling feed back cap out)..the deemp network in between TZ load of 600Ohm and the filter. The deemp is a BJT switch driven thing..in off it loads w 4,7M||BJT-off.. would that have any impact..just thought about 4,7M vs 600 should not be able to disturb the TZ load, but since it is micro detail level the shrill problem seems to be..hmm maybe
/Patrick
AD844 Stacking...
Hi Patrick, Well it sounds like you have a nice set of speakers. Dynaudio drivers are among the best in a conventional 2 way system. Fine on the Maggies that are put away. I have a pair of MG2.5R's which sport a 3rd over crossover design that my cousin designed. The crossover makes all the difference. No maggie slam, which is a crossover point issue and slope problem. Superb resolution with decent bass although the lowest of the low are missing in action. Some output down to 30 Hz. Sallen key filters have phase shifts in the pass band. Don't know if yours is an inverting type or not. Since you have a separate I/V stage it is likely a non-inverting type. Deemphasis... Wasn't used much or on many CD's. Likely you could get rid of that. It may be affecting things even when off. Hard to predict that. Understand you don't want to redesign as such. Just get better quality out of the Arcam. I modded my old CDB650 to the maximum that I could. The first 1541 based standalone dac outperformed it without breaking a sweat. Interestingly the Mark II dac that followed that one took me years to push it to out perform that first standalone dac. Your on an interesting path. Hope it takes you less time to get satisfying sound. It can be a long journey of discovery. 😀
Think it were russians diyers !
I like a lot Lundhal on a cd player I have, but don't know if the cause is a less subjective climby treble : it's very hard to benchmark a traffo VS a very good cap ! So Today I use no caps at all with the OPA861 and no buff !
I like a lot Lundhal on a cd player I have, but don't know if the cause is a less subjective climby treble : it's very hard to benchmark a traffo VS a very good cap ! So Today I use no caps at all with the OPA861 and no buff !
AD844 Stacking...
Installed the triple stack board into the Mark 1 dac. My first build that has been gathering dust for some time. The BF245A drift is in a word zero, so I see little issue with the 2 mA current source for dac offset null. Measuring at the output of the BUF03 output buffer was interesting. Before the series 604 Ohm load resistor I did see some drift. Maybe 2 mV's. After the resistor and effectively at the Edcor passive filter primary I measured zero. So direct coupling to the filter primary was no issue. Quality wise the test results are the same as before. Mark 1 (triple stack off TZ + BUF03) was 90% of the quality of the Mark 2 (DDNF + BUF03. The volume levels are exactly the same in this test. The PCM1704 version with DDNF + OPA627 + Lundahl LL1690 is still better on all accounts. Still trying to get an enclosure from Par-Metal to finish the PCM1704 version.
Installed the triple stack board into the Mark 1 dac. My first build that has been gathering dust for some time. The BF245A drift is in a word zero, so I see little issue with the 2 mA current source for dac offset null. Measuring at the output of the BUF03 output buffer was interesting. Before the series 604 Ohm load resistor I did see some drift. Maybe 2 mV's. After the resistor and effectively at the Edcor passive filter primary I measured zero. So direct coupling to the filter primary was no issue. Quality wise the test results are the same as before. Mark 1 (triple stack off TZ + BUF03) was 90% of the quality of the Mark 2 (DDNF + BUF03. The volume levels are exactly the same in this test. The PCM1704 version with DDNF + OPA627 + Lundahl LL1690 is still better on all accounts. Still trying to get an enclosure from Par-Metal to finish the PCM1704 version.
Attachments
Torchwood421
I see that you have quite a lot of interconnecting wire running around. I often face challenge with the wiring between boards and from the board to RCA sockets. The wire impose signature on sound even more than the parts used. What kind of wire are you using?
it's not easy to find some neutral sounding wire.
I see that you have quite a lot of interconnecting wire running around. I often face challenge with the wiring between boards and from the board to RCA sockets. The wire impose signature on sound even more than the parts used. What kind of wire are you using?
it's not easy to find some neutral sounding wire.
As for internal non shielded hookup wire, I always found single core (approx .3mm to .5mm) Teflon coated copper to be the most neutral sounding of all.
Cheers George
Wire
I stick with aerospace types. The bigger stuff is Silver with Teflon jacket. My take on that is that it is stranded Copper with Silver plate and Teflon insulating jacket. For the smaller stuff it is Tin with Tefzel insulation. It is supposed to tolerate 90 degree bends. In listening test I hear no issues with either wire. I tend to avoid PVC covered types. Wish I knew of a source for the smaller 24 and 26 AWG Silver Teflon types. I'd use that in a heartbeat. 😉
Hi Torchwood, thank you for your impressive tests and nice sharing.
I have a question please, when you talk about the 627 in unity gain, it is not also the warm known sound of this oap you prefer instead an other one more neutral with as good spec ? So also a matter of taste, or has it simply not this sound warm presentation in the circuit you test ?
I'm asking myself what is using Pedja Rogic in his flagship Model S dac as buffer, I know it is not the oap 861 in reverse like the AYA used to ?
What are the IC chip with diamond buffer design ? Could we use after a first I/V a numeric digital Pot which has no attenuation but gain (My poor understanging is some gives also some gain as the use is volume pot swaping in preamp units !) ????
Sorry if it has no sense, I have not a clue about electronic ! just the basics and empirics works that's all !
I have a question please, when you talk about the 627 in unity gain, it is not also the warm known sound of this oap you prefer instead an other one more neutral with as good spec ? So also a matter of taste, or has it simply not this sound warm presentation in the circuit you test ?
I'm asking myself what is using Pedja Rogic in his flagship Model S dac as buffer, I know it is not the oap 861 in reverse like the AYA used to ?
What are the IC chip with diamond buffer design ? Could we use after a first I/V a numeric digital Pot which has no attenuation but gain (My poor understanging is some gives also some gain as the use is volume pot swaping in preamp units !) ????
Sorry if it has no sense, I have not a clue about electronic ! just the basics and empirics works that's all !
Last edited:
Try a jewellery wire supplier for 99.99% soft silver wire and wrap it in PTFE tape or some PTFE tubing if you can find it. You could use some silk or pure cotton.
How is the final design for the DDNF boards going? Any chance of a group buy?