Hello ... fellow DIY audio enthusiast here. I am intrigued by the IC AMP pdf found at the Blackgate website. Here is the link:
http://www.blackgate.jp/ebg7.htm
The design is reminiscent of the Denon UHC-MOS circuit used in the POA series. Of course the capacitor regime is a big ridiculous and obvious advertisement for the BG N-Series. I priced the monoblock out at about $581 for capacitors alone!
The design seems very nice! I am thinking of replacing the OPA627 with a discrete or maybe a tube voltage gain stage. Any comments/suggestions? I like the simplicity. I'd drop the volume pot and run a 47K to ground in lieu of the pot/100K to ground. I am also not too keen on the complex BJT regulation. I'd rather just run a pair of LM317's for the low current supplies. Any P/S gurus know why they use this scheme?
BTW ... I happen to have 1 NOS 2SJ217 and 3 SK1665's handy to build something. And NO ... they AREN'T FOR SALE! Now I just need one more 2SJ217 ...
http://www.blackgate.jp/ebg7.htm
The design is reminiscent of the Denon UHC-MOS circuit used in the POA series. Of course the capacitor regime is a big ridiculous and obvious advertisement for the BG N-Series. I priced the monoblock out at about $581 for capacitors alone!
The design seems very nice! I am thinking of replacing the OPA627 with a discrete or maybe a tube voltage gain stage. Any comments/suggestions? I like the simplicity. I'd drop the volume pot and run a 47K to ground in lieu of the pot/100K to ground. I am also not too keen on the complex BJT regulation. I'd rather just run a pair of LM317's for the low current supplies. Any P/S gurus know why they use this scheme?
BTW ... I happen to have 1 NOS 2SJ217 and 3 SK1665's handy to build something. And NO ... they AREN'T FOR SALE! Now I just need one more 2SJ217 ...
analog_sa said:If you take out the BGs there is precious little left in that design...
And a tube voltage gain will be either very difficult or completely impossible
How do you figure? Just throw in a simple 12AT7/6DJ8 gain stage and a coupling capacitor in lieu of the OPA627. DONE! The P/S would add a bit of complexity, but would add some tube warmth. Very similar (AKA scaled down version) of the Peter Millett Hybrid Headphone amp using a MOSFET to sync the current after a tube linestage.
mlloyd1 said:
indeed, a girly-man op amp driving power mosfet gates unaided?
mlloyd1
K.I.S.S.! That is my theory when building DIY audio ... see my "Dream System" in my sig to understand where I am coming from.
A few questions for the SS gurus:
1. The OPA627 has a gain of 40 getting feedback from the speaker terminal. So, if I have a 2Vp-p input, wouldn't that allow the op-amp to clip the output since the FET's only have a 25V rail? 2*40=80V!
2. I want to replace the OPA627 with a discrete section. Does anyone have any direct experience with the Nelson Pass discrete? With so many out there, it would take eons to evaluate them. Just looking for a clean simple design.
My take on the circuit is this ... someone please confirm or correct me.
The signal goes into the op-amp and gets an impossible gain (still confused here)? The signal travels through 2 parallel capacitors of an ungodly value (could I use a smaller cap ... say 47uF?). The UHC-MOS design is solid and I assume the CRD E-102 section is there to bias the OPA627 to Class-A (I can't understand why else it is there). Seems pretty simple to me. If that is the purpose of the CCS section, I see no reason to simplify the schematic as follows:
1. Change the op-amp supply to a LM317 voltage regulator with +/-15V rails
2. Change the CCS supply to a LM317 current regulator at 1ma
That reduces the BG cap count considerably saving some cash!
craigtone said:
Tube warmth is great but first the circuit has to work.
There is no easy way to DC couple a tube to the mosfets. It is the dc connection which keeps the output offset low thanks to dc nfb. And even if you somehow managed this how would you keep reasonable distortion in a class AB output stage? The opamp has oodles of open loop gain for nfb correction; how much open loop gain does a tube have?
The gain of the amp is indeed x40. This means higher than usual sensitivity. So 2v p-p is a bit out of the question - max output is less than opamp supply voltage.
Wrong assumption. It's there to bias the MOSFETs.
The only workable suggestion you have is to replace the opamp with a discrete. Whether this would sound better is far from clear. What is certain is that dc offset will suffer.
analog_sa said:
So then this whole article is rubbish?
http://www.tubecad.com/april_may2001/2001_04-05.PDF
analog_sa said:
Last I checked, a FET was biased with voltage not current. The op-amp output could do that on it's own without the need for the CRD. The CRD allows for low power signals to run in Class-A until the output current exceeds the bias current. I am assuming that is why the 2.2K needs to be adjusted ... to reduce crossover distortion.
" The design seems very nice! I am thinking of replacing the OPA627 with a discrete or maybe a tube voltage gain stage. Any comments/suggestions? I like the simplicity. I'd drop the volume pot and run a 47K to ground in lieu of the pot/100K to ground. I am also not too keen on the complex BJT regulation. I'd rather just run a pair of LM317's for the low current supplies. Any P/S gurus know why they use this scheme? "
You could replace the OPA with a discrete, but as the OPA is so good replacing it is very highly likely to degrade the objective and subjective performance. According to the manufacturer's application notes the OPA627 is sensitive to layout (presumably owing to stray capacitance) so refer to their recommendations prior to commencing board layout.
(I have no experience with DC coupling tube stages to FET output buffers so can not comment on the technicalities of that proposal).
The gain of the stage as it stands is quite high as you noted, however common sense dictates that you would "turn it down" way before applying 2V to the op-amp input.
The 100K resistor does not just shunt the pot, but also provides a dc path to the op-amp input in the event of pot failure.
The CRDs provide current through, and hence develop a voltage across, the 2.2K resistors and the diodes, in order to set the output stage bias.
The 100 Ohm resistors counter the gate input capacitance of the FETs.
The two 100 micro Capacitors on the supplies to the op-amp pins seem very excessive, as do the four 2200 micro Capacitors coupling to the output stage.
The TL431 in the power supply is a precision part, and I refer you to it's datasheet for information. That said, you are right that the + / - 15 V supply could easily come from a set of 3-terminal regulators.
I am unsure how much load the output stage will draw so can not comment upon the applicability of the + / - 30 V regulator.
Other suggestions to consider:
Add a (suitably qualified) cap across the mains switch
Add an RFI filter to the mains input
Add a transient suppressor across Live and Neutral, after the switch and before the fuse
Good luck,
G.
You could replace the OPA with a discrete, but as the OPA is so good replacing it is very highly likely to degrade the objective and subjective performance. According to the manufacturer's application notes the OPA627 is sensitive to layout (presumably owing to stray capacitance) so refer to their recommendations prior to commencing board layout.
(I have no experience with DC coupling tube stages to FET output buffers so can not comment on the technicalities of that proposal).
The gain of the stage as it stands is quite high as you noted, however common sense dictates that you would "turn it down" way before applying 2V to the op-amp input.
The 100K resistor does not just shunt the pot, but also provides a dc path to the op-amp input in the event of pot failure.
The CRDs provide current through, and hence develop a voltage across, the 2.2K resistors and the diodes, in order to set the output stage bias.
The 100 Ohm resistors counter the gate input capacitance of the FETs.
The two 100 micro Capacitors on the supplies to the op-amp pins seem very excessive, as do the four 2200 micro Capacitors coupling to the output stage.
The TL431 in the power supply is a precision part, and I refer you to it's datasheet for information. That said, you are right that the + / - 15 V supply could easily come from a set of 3-terminal regulators.
I am unsure how much load the output stage will draw so can not comment upon the applicability of the + / - 30 V regulator.
Other suggestions to consider:
Add a (suitably qualified) cap across the mains switch
Add an RFI filter to the mains input
Add a transient suppressor across Live and Neutral, after the switch and before the fuse
Good luck,
G.
Gordy ... THANK YOU!
I showed the schema to an EE @ work who is a mentor to almost everyone and he said pretty much what you did. He suggested pots for the 2.2K's to have the ability to adjust the offset of the FETs and ensure symmetry of the output waveform by biasing each the same amount.
He also concurred that the CRD is to bias the diodes that set the voltage on the gate so the FET always conducts reducing crossover distortion.
Given the fact that the CRD is 1ma and that a FET input impedance is extremely high, would it not also bias the op-amp in Class-A mode?
I was looking for a response like yours to understand what each component does to determine which components need attention, higher quality, etc. I think I'll leave the series-pass regulators just because. The only other question is the cap across FET rails. I assume to balance ripple to promote symmetry? If that is the case, I am thinking a Solen film.
Thanks for the OPA627 layout tip ... I'll read that now. I think I'll just use a DIP socket so I can op-amp roll and maybe try a Burson discrete for grins and giggles.
One more thing ... what is the 300mA on the common source of the FETS? Quiescent current?
I showed the schema to an EE @ work who is a mentor to almost everyone and he said pretty much what you did. He suggested pots for the 2.2K's to have the ability to adjust the offset of the FETs and ensure symmetry of the output waveform by biasing each the same amount.
He also concurred that the CRD is to bias the diodes that set the voltage on the gate so the FET always conducts reducing crossover distortion.
Given the fact that the CRD is 1ma and that a FET input impedance is extremely high, would it not also bias the op-amp in Class-A mode?
I was looking for a response like yours to understand what each component does to determine which components need attention, higher quality, etc. I think I'll leave the series-pass regulators just because. The only other question is the cap across FET rails. I assume to balance ripple to promote symmetry? If that is the case, I am thinking a Solen film.
Thanks for the OPA627 layout tip ... I'll read that now. I think I'll just use a DIP socket so I can op-amp roll and maybe try a Burson discrete for grins and giggles.
One more thing ... what is the 300mA on the common source of the FETS? Quiescent current?
"Given the fact that the CRD is 1ma and that a FET input impedance is extremely high, would it not also bias the op-amp in Class-A mode?"
The CRDs, resistors and diodes are equal in both halves around the centre node connected to op-amp pin 6, and as such it is likely that the centre node is at a neutral state and hence not biasing the op-amp into Class A.
"The only other question is the cap across FET rails. I assume to balance ripple to promote symmetry? If that is the case, I am thinking a Solen film."
That value of 0.01 micro is comparatively low, and I would assume that it works in conjunction with the inherent induction of the supply line wiring to zap any very high frequency noise pulses. It's too small to effectively contribute to ripple rejection. A film cap would be appropriate, but as it does not pass audio signal I would not spend much on it (pennies, cents...)
"Thanks for the OPA627 layout tip ... I'll read that now. I think I'll just use a DIP socket so I can op-amp roll and maybe try a Burson discrete for grins and giggles."
Ironically the OPA627 data that I read warns against sockets owing to their extra stray capacitance and the op-amps susceptibility to such!
Also, while I have an audience, please be a little wary of IC-rolling. It is often touted by those who do not understand the subject too well: I admire their enthusiasm, however it is not as clear-cut as they would have you believe.
What I mean is that all op-amps are not created equal and some have very specific requirements for optimisation. Hence blindly substituting one for another, and then judging the subjective result, does not necessarily tell you how the new op-amp performs, but instead only indicates how the new op-amp performs in a potentially un-optimised installation.
Layout, bypassing, supply voltage, gain level, bias current path resistance, input and feedback impedances, etc., all have to be optimised before true comparisons can be made.
By all means have a go, and good luck, just understand that your results may not be definitive or repeatable in other installations.
"One more thing ... what is the 300mA on the common source of the FETS? Quiescent current?"
Yes, I think it is quiescent.
The CRDs, resistors and diodes are equal in both halves around the centre node connected to op-amp pin 6, and as such it is likely that the centre node is at a neutral state and hence not biasing the op-amp into Class A.
"The only other question is the cap across FET rails. I assume to balance ripple to promote symmetry? If that is the case, I am thinking a Solen film."
That value of 0.01 micro is comparatively low, and I would assume that it works in conjunction with the inherent induction of the supply line wiring to zap any very high frequency noise pulses. It's too small to effectively contribute to ripple rejection. A film cap would be appropriate, but as it does not pass audio signal I would not spend much on it (pennies, cents...)
"Thanks for the OPA627 layout tip ... I'll read that now. I think I'll just use a DIP socket so I can op-amp roll and maybe try a Burson discrete for grins and giggles."
Ironically the OPA627 data that I read warns against sockets owing to their extra stray capacitance and the op-amps susceptibility to such!
Also, while I have an audience, please be a little wary of IC-rolling. It is often touted by those who do not understand the subject too well: I admire their enthusiasm, however it is not as clear-cut as they would have you believe.
What I mean is that all op-amps are not created equal and some have very specific requirements for optimisation. Hence blindly substituting one for another, and then judging the subjective result, does not necessarily tell you how the new op-amp performs, but instead only indicates how the new op-amp performs in a potentially un-optimised installation.
Layout, bypassing, supply voltage, gain level, bias current path resistance, input and feedback impedances, etc., all have to be optimised before true comparisons can be made.
By all means have a go, and good luck, just understand that your results may not be definitive or repeatable in other installations.
"One more thing ... what is the 300mA on the common source of the FETS? Quiescent current?"
Yes, I think it is quiescent.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.