The headphone people are quite right. Even in buffered applications, current loading a filter/buffer improves the sonics. I have just tried this on an Assemblage 3.1 signature, further modified. With AD825 output buffered by BUF634, I have 1 channel with buffered CRD at 1 mA, and resistor loading on another at about 1.2mA. They sound different and a lot less dry in the mid/hi than the unloaded units; definitely better. The CRD side sounds somewhat less forward than the resistor loaded side, and the effect is reversed when channel is reversed, validating my hearing.
My conclusion:- well worth shifting the operating point of even 'good' sounding opamps. The OPA 627, with its many supporters but not me, can be made to sound decent with about 1.6-2 mA loading, but still not fully open.
Does anyone have access or know source of an extender which allows me to fit a two fet current regulators on top of a DIL setup please. This will facilitate comparisons and tweaking???
My conclusion:- well worth shifting the operating point of even 'good' sounding opamps. The OPA 627, with its many supporters but not me, can be made to sound decent with about 1.6-2 mA loading, but still not fully open.
Does anyone have access or know source of an extender which allows me to fit a two fet current regulators on top of a DIL setup please. This will facilitate comparisons and tweaking???
cascode current source opamp adapter
Maybe I'll design one. The BrownDog opamp adapter seems to be doing fairly well, so I am feeling inspired. If I were to do this, I would lean towards a adapter that is basically a passthrough. It would make the IC taller, but not extend beyond the 400x400 mil boundary of the DIP socket. The opamp plugs into the cascode adapter which plugs into the socket on the board. The FETs lurk under the opamp on the adapter.
I want to do some experiments with a CRD/FET cascode instead of a dual FET cascode. It would take less space and might be better albeit more expensive than 2 FETs due to the tighter current regulation of the CRD coupled with the lower capacitance of the cascoded FET.
I am not sure if I can talk BrownDog into making this one, too. It is a bit more esoteric and would involve onboard components which is sort of of out of their scope, but I'll look into it.
It might be possible to get BrownDog or someone else to make a total solution module a la LC Audio, complete with opamps and cascode current source. The drawback is you are stuck with whatever opamp is chosen, so I am not as keen on this approach. I'll follow up after Memorial Day weekend.
Maybe I'll design one. The BrownDog opamp adapter seems to be doing fairly well, so I am feeling inspired. If I were to do this, I would lean towards a adapter that is basically a passthrough. It would make the IC taller, but not extend beyond the 400x400 mil boundary of the DIP socket. The opamp plugs into the cascode adapter which plugs into the socket on the board. The FETs lurk under the opamp on the adapter.
I want to do some experiments with a CRD/FET cascode instead of a dual FET cascode. It would take less space and might be better albeit more expensive than 2 FETs due to the tighter current regulation of the CRD coupled with the lower capacitance of the cascoded FET.
I am not sure if I can talk BrownDog into making this one, too. It is a bit more esoteric and would involve onboard components which is sort of of out of their scope, but I'll look into it.
It might be possible to get BrownDog or someone else to make a total solution module a la LC Audio, complete with opamps and cascode current source. The drawback is you are stuck with whatever opamp is chosen, so I am not as keen on this approach. I'll follow up after Memorial Day weekend.
Re: Re: cascode current source opamp adapter
If you really believe the extra socket contacts are going to ruin your circuit then air wire it yourself. Personally, I think that is nonsense. I got a lot of (groundless? muahaha!) complaints for not including bypass capacitors and a ground tether on the BrownDog opamp adapter design, yet the adapter works great, fits where no LC Audio module can, and accepts your choice of opamp. It is much better off as it is, and so will the cascode adapter be if I do it. Most people don't want to do microsurgery and air wiring. If it plugs in and it works it will be popular. Ironically, I will probably be able to fit a bypass cap on the cascode adapter, but it will go from +V to -V. Better than nothing. You could always balance the cascode adapter on top of your BrownDog adapter and solder it that way.
Take a second glance at my previous post. I'm not sure that it is better, I want to look into it. A dual FET cascode takes more space, actually. CRDs are tiny glass packages the size of small signal diodes.
If I do the cascode adapter I may use all surface mount components if I can find someone to make them. I'm off to the local surplus shop to see if I can find some FETs and CRDs to play with.
If you really believe the extra socket contacts are going to ruin your circuit then air wire it yourself. Personally, I think that is nonsense. I got a lot of (groundless? muahaha!) complaints for not including bypass capacitors and a ground tether on the BrownDog opamp adapter design, yet the adapter works great, fits where no LC Audio module can, and accepts your choice of opamp. It is much better off as it is, and so will the cascode adapter be if I do it. Most people don't want to do microsurgery and air wiring. If it plugs in and it works it will be popular. Ironically, I will probably be able to fit a bypass cap on the cascode adapter, but it will go from +V to -V. Better than nothing. You could always balance the cascode adapter on top of your BrownDog adapter and solder it that way.
Take a second glance at my previous post. I'm not sure that it is better, I want to look into it. A dual FET cascode takes more space, actually. CRDs are tiny glass packages the size of small signal diodes.
If I do the cascode adapter I may use all surface mount components if I can find someone to make them. I'm off to the local surplus shop to see if I can find some FETs and CRDs to play with.
Re: Re: Re: cascode current source opamp adapter
[QUOTEIf you really believe the extra socket contacts are going to ruin your circuit then air wire it yourself. Personally, I think that is nonsense. IIf I do the cascode adapter I may use all surface mount components if I can find someone to make them. I'm off to the local surplus shop to see if I can find some FETs and CRDs to play with. [/QUOTE]
----------------------------------------------------------
Why have more contacts than one needs to. They DO deteriorate with time.
2N crds are no longer available in UK; only J50x and CRx which are larger.
[QUOTEIf you really believe the extra socket contacts are going to ruin your circuit then air wire it yourself. Personally, I think that is nonsense. IIf I do the cascode adapter I may use all surface mount components if I can find someone to make them. I'm off to the local surplus shop to see if I can find some FETs and CRDs to play with. [/QUOTE]
----------------------------------------------------------
Why have more contacts than one needs to. They DO deteriorate with time.
2N crds are no longer available in UK; only J50x and CRx which are larger.
While waiting for FETs, etc. I decided to start looking at possible designs for a FET cascode current source adapter.
First, silly as it may sound, I evaluated the feasibility of using discrete non SMT components. 4 TO-92 FETS will barely fit between the rows of socket pins. If capacitance buffering resistors were to be included they would somehow have to be stashed on the underside. With all the holes needed for the FET leads, it would be hard to find room for the resistor mounting holes. Leaving out the resistors means the opamp will see more of the capacitance of the FETs, but it would still work, and no SMT would be involved.
Onward to the more sensible SMT option. One current source goes on each side of the adapter. Assuming 2N5484 FETS, Idss is 1-5mA with 4 being typical. Desired current is about 1mA, but without picking FETs will typically be much higher. This means that the 100 Ohm capacitance buffering resistors will need to dissipate more power than is possible for an 0805 SMT resistor, which is the largest that will comfortably fit on the board. Thus we will also need a current limiting resistor on the source of Q1, or find some SMT CRDs to use instead. Crystalonics makes such a device, but I do not know how available or economical they would be. It will be tight squeezing two resistors and two FETs on each side. The resistors would have to be 0603 to fit.
First, silly as it may sound, I evaluated the feasibility of using discrete non SMT components. 4 TO-92 FETS will barely fit between the rows of socket pins. If capacitance buffering resistors were to be included they would somehow have to be stashed on the underside. With all the holes needed for the FET leads, it would be hard to find room for the resistor mounting holes. Leaving out the resistors means the opamp will see more of the capacitance of the FETs, but it would still work, and no SMT would be involved.
Onward to the more sensible SMT option. One current source goes on each side of the adapter. Assuming 2N5484 FETS, Idss is 1-5mA with 4 being typical. Desired current is about 1mA, but without picking FETs will typically be much higher. This means that the 100 Ohm capacitance buffering resistors will need to dissipate more power than is possible for an 0805 SMT resistor, which is the largest that will comfortably fit on the board. Thus we will also need a current limiting resistor on the source of Q1, or find some SMT CRDs to use instead. Crystalonics makes such a device, but I do not know how available or economical they would be. It will be tight squeezing two resistors and two FETs on each side. The resistors would have to be 0603 to fit.
Here is a rough draft using a current limiting resistor on the source of Q1 as per the Siliconix paper figure 2 with a capacitance buffering resistor between the cascode and the opamp:
http://www.vishay.com/document/70596/70596.pdf
I find myself wondering if the circuit would be better with the gate of Q2 tied to -V instead of Q1 source. Any thoughts? I have no FETs to play with atm.
http://www.vishay.com/document/70596/70596.pdf
I find myself wondering if the circuit would be better with the gate of Q2 tied to -V instead of Q1 source. Any thoughts? I have no FETs to play with atm.
Attachments
Dyslexics untie! I think I scrambled some numbers yesterday, in fact there is no power problem with 100 Ohm resistors, the power would be 100-1000 microWatts, not milliWatts. Nevertheless, it would be nice to reduce the current to somewhere between .5 and 1mA. It would also be nice if the resistors were the same value, so there is no issue with mixing them up during assembly.
Nevertheless, it would be nice to reduce the current to somewhere between .5 and 1mA. It would also be nice if the resistors were the same value, so there is no issue with mixing them up during assembly.
feasibility and pricing
I just talked to BrownDog about feasibility and pricing. The estimated preliminary pricing for a fully populated FET cascode current source opamp adapter is ~$10-20 and an unpopulated version with press fit socketed pins is ~$5-10, depending on quantity.
The primary application is biasing opamps into class A in legacy circuits such as CD players and headphone amps. Let me know if you are interested in this adapter, and whether you want it unpopulated or fully assembled.
I just talked to BrownDog about feasibility and pricing. The estimated preliminary pricing for a fully populated FET cascode current source opamp adapter is ~$10-20 and an unpopulated version with press fit socketed pins is ~$5-10, depending on quantity.
The primary application is biasing opamps into class A in legacy circuits such as CD players and headphone amps. Let me know if you are interested in this adapter, and whether you want it unpopulated or fully assembled.
CRD versus cascode
I am intrigued by the considerable sound quality difference between a buffered CRD 1k - Vishay J505 and a cascode system 2N5457-2N5457- 100R in three applications with moderate (10k) to high Mohm input impedances. The systems are; analogue out from a tuner; A/D buffer;
and D/A filter. in all cases the CRD sound coarser and with a different sound stage (more forward) than the fet CS. The latter is much more refined and the currents drawn are nearly identical at around 1 mA.
Why is this??
I am intrigued by the considerable sound quality difference between a buffered CRD 1k - Vishay J505 and a cascode system 2N5457-2N5457- 100R in three applications with moderate (10k) to high Mohm input impedances. The systems are; analogue out from a tuner; A/D buffer;
and D/A filter. in all cases the CRD sound coarser and with a different sound stage (more forward) than the fet CS. The latter is much more refined and the currents drawn are nearly identical at around 1 mA.
Why is this??
CDRs vs cascoded fets vs cascode/crd
I dont know if you have read this section, but they talk about the advantages of each option and why one is better than the other.
http://tangentsoft.net/audio/meta42/tweaks.html
cheers
I dont know if you have read this section, but they talk about the advantages of each option and why one is better than the other.
http://tangentsoft.net/audio/meta42/tweaks.html
cheers
Casscode current dose sound better to me
I have used Current sources to load Opamp's for quite a while and current loading improves the sound of almost any opamp. Going to a cascode current source is evean futher Refinement and as fmak states the differences are dramitic enough to warrent the effort. BTW i now like High Value Buffer resistors evean for cascode current sources.
I have used Current sources to load Opamp's for quite a while and current loading improves the sound of almost any opamp. Going to a cascode current source is evean futher Refinement and as fmak states the differences are dramitic enough to warrent the effort. BTW i now like High Value Buffer resistors evean for cascode current sources.
opamp current loading
Hi,
I don't know how you decided that 2mA load was OK, but there is an easy way to determine that.
Assuming that the object is to run the opamp output stage in class A (we DO agree on that, don't we?), determine the max output voltage you expect and divide that by the lowest loading impedance you expect.
Example: 5VRMs, across 2kOhm load gives max load current of 2.5mA RMS which is about 3,5mA pk-to-pk (for a sine wave). Taking some reserve, in this case 4mA would be OK.
You could of course do a series of listening tests at 1, 1.1, 1.2 etc mA, but that seems rather primitive and tedious, and I wouldn't put money on the reliability of the outcome of such a process.
Cheers,
Jan Didden
Hi,
I don't know how you decided that 2mA load was OK, but there is an easy way to determine that.
Assuming that the object is to run the opamp output stage in class A (we DO agree on that, don't we?), determine the max output voltage you expect and divide that by the lowest loading impedance you expect.
Example: 5VRMs, across 2kOhm load gives max load current of 2.5mA RMS which is about 3,5mA pk-to-pk (for a sine wave). Taking some reserve, in this case 4mA would be OK.
You could of course do a series of listening tests at 1, 1.1, 1.2 etc mA, but that seems rather primitive and tedious, and I wouldn't put money on the reliability of the outcome of such a process.
Cheers,
Jan Didden
Re: opamp current loading
You could of course do a series of listening tests at 1, 1.1, 1.2 etc mA, but that seems rather primitive and tedious, and I wouldn't put money on the reliability of the outcome of such a process.
---------------------------------------------------------------------
Simple impedance and level matching considerations don't always count in audio. If this were the case, then there should be no difference in sound quality amongst resistor loading, crd/fet-resistor loadong and cascode loading, given suitably high input impedance. This is not the case by common agreement.
My point is that the 1 mA criterion usually adopted by the headphone guys does not necedssarily apply. In my cases, I found 2mA to further improve the sound for a variety of opamps. The basis is very simple. You are operating in different portion of the transfer curve. I would speculate that the middle portion ie. 0.25 of maximum current, could be even better.However, this would require heat sinking.
Why do you mistrust hearing to the point of calling it tedious and primitive, which it is???!!!
You could of course do a series of listening tests at 1, 1.1, 1.2 etc mA, but that seems rather primitive and tedious, and I wouldn't put money on the reliability of the outcome of such a process.
---------------------------------------------------------------------
Simple impedance and level matching considerations don't always count in audio. If this were the case, then there should be no difference in sound quality amongst resistor loading, crd/fet-resistor loadong and cascode loading, given suitably high input impedance. This is not the case by common agreement.
My point is that the 1 mA criterion usually adopted by the headphone guys does not necedssarily apply. In my cases, I found 2mA to further improve the sound for a variety of opamps. The basis is very simple. You are operating in different portion of the transfer curve. I would speculate that the middle portion ie. 0.25 of maximum current, could be even better.However, this would require heat sinking.
Why do you mistrust hearing to the point of calling it tedious and primitive, which it is???!!!
The Ear Brain is far more Accurate than Given credit
The ear can locate a sound to within 1.0% accuracy, the Ear-Brain and nural network is able to compute faster than large mainframe Computors. So i always trust my hearing and moswt times it is in trying to remove a sonic fault that i often find a technicaly Elagent solution to a Problem of the circuit that measurments and simulation failed to Show. Measurements are more Dependable than simulations, whitch only give a starting point.
Anyway regarding my View of the amount of current to Bias an opamp at then no more than 1/10 Th the opamps nominal Output Current. Alot less would Probaly be good in most high Impedance situations, However capacitive loading by long interconnect cables may require Opamps who's outputs will see the Outside world then 3-5 Ma of Constant current is a safe bet on covering any possible Load situations Excepting Low impedance loads within the Audio range like Headphones But It is my posistion that Opamps should be Buffered by a Dedicated Buffer IC like the BUF-634 or a Discreet component Alternative, However since the gain stage is Monolithic then why not Keep the concept similar in the output stage.
The ear can locate a sound to within 1.0% accuracy, the Ear-Brain and nural network is able to compute faster than large mainframe Computors. So i always trust my hearing and moswt times it is in trying to remove a sonic fault that i often find a technicaly Elagent solution to a Problem of the circuit that measurments and simulation failed to Show. Measurements are more Dependable than simulations, whitch only give a starting point.
Anyway regarding my View of the amount of current to Bias an opamp at then no more than 1/10 Th the opamps nominal Output Current. Alot less would Probaly be good in most high Impedance situations, However capacitive loading by long interconnect cables may require Opamps who's outputs will see the Outside world then 3-5 Ma of Constant current is a safe bet on covering any possible Load situations Excepting Low impedance loads within the Audio range like Headphones But It is my posistion that Opamps should be Buffered by a Dedicated Buffer IC like the BUF-634 or a Discreet component Alternative, However since the gain stage is Monolithic then why not Keep the concept similar in the output stage.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.