An externally hosted image should be here but it was not working when we last tested it.
This is my "real" gainclone, e.g. noninverted but with (to?) big bypass caps:
2 x 18000uF (out of a Fostex PA-amp)
4 x 10000uF (Jamicon)
4 x 2200uF (Philips)
4 x 1 uF foil Siemens
4 x 0,1uF foil Wima
The first impression is impressive
First, I used a trafo out of an Kenwood KA-1500, result: Smoke. With a 225VA torroid, 2x22V, it works fine and cool.
But give me some time. I will compare soon with the regulated amp and the tube buffered versions, together with some friends. And then write about.
Franz
P.S.
One result is clear: for subwoofers use BIG caps!
Sorry Folks, I puted the last reply in this thread: but actually I switched back to my regulated, OPA627 Class-A buffered amp:
This is music, not the NIGC with BIG caps!
So, i will reduce the caps and listen to it.
In theory, the best buffer must be no buffer (not true for the inverted version, I know).
Franz
This is music, not the NIGC with BIG caps!
So, i will reduce the caps and listen to it.
In theory, the best buffer must be no buffer (not true for the inverted version, I know).
Franz
carlosfm said:Hi Joe, welcome!
It makes sense, very nice.
What caps do you use there, for instance, on the OPA627?
I suppose it would not be the same for the 637?
Do you notice any improvement in sound by doing this?
Thanks.
That depends on what the previous implementation was like (and I wouldn't be able to count the different possibles). If it was pour to start with and THEN this idea was used - just add two film caps, bring the signal return back to the junction of the two before it sees ground - THEN it could make a significant improvement in sound.
Re 627 vs 637, yes, the same applies. But the 627 has a more aggressive internal com cap. This means less open loop gain at HF. This means less PSR at HF and keeping voltage changes/modulations away from the comp cap that just bit more difficult. That would be my logic.
Need something that has reactance going down with increasing frequency i.e. a cap. The value of the cap? It ought to be larger if the load Z is low value and vice versa. It must be film as it needs MHz plus performance.
A 0.1uF is 16 Ohm @ 100KHz. So that should suffice for signal ICs, but a little higher value wouldn't go astray...
But this will not do with a gainclone (power) chip and 8 Ohm Load. That would require 4.7uF film or similar. This will return HF currents which now are much higher. These should be on the chips' pins and NOT across the larger remoter reservoir caps.
Or...
Now let us revise the above idea to a gainclone chip and Low Z Speaker Load:
An externally hosted image should be here but it was not working when we last tested it.
Since most of us are likely using Panasonic 1000uF/50V electros, which have good HF performance, low ESR etc. So they do a good job and 1000uF is enough to deal with 4-8 Ohm Loads.
But the bottom line is the same. Where do you return the now much higher power return currents? The problem is the same even if the scale is different.
Now we can also see WHY the 1000uF caps needs to be close to plus and neg on the chip. In fact right on the pins.
The speaker return? Bring it right back to the chip and those caps.
Joe R.
Re: joe's tip
It's gets complicated when we get beyond just one chip - all I can say is you need to analyse what's happening and be aware of the return HF currents getting back to the chip via the rail it uses as its signal ground for those HF currents.
Like I said, this is not an easy subject to explain but when dealing with it you must try to analyse and find ways to deal with it. Even very experienced techs and designers have found themselves scratching their heads on this, especially in more extended multiple IC topologies. And they may not even be dealing with sonic qualities, just trying to keep things stable. "Where is that weird oscillation pattern coming from?"
The concept of 'ground' is a fleeting subject.
There is no universal ground.
For example: Grounds can lie.
NEVER TRUST A 'GROUND' !!! TREAT ALL WITH SUPREME SUSPICION.
Really, when you think of it, I ain't got any dirt in my amps, so there are NO grounds in any of my amps.
Grounds are common connections to which signals and power are referred or referenced. Nothing more than that. That is textbook.
Grounds do not LOCK DOWN signals or currents. Grounds can move.
As for a 'layout tip' - you are right. I can say that for single chip situations, you can consider it thus. And that was my idea as a 'trick' to improve the 627, which I like (if it has to be an IC). Apply it to the gainclone chip too. I bet that the original Gaincard has the speaker return coming back to the electro cap x 2 junction.
Joe R.
matjans said:
It's gets complicated when we get beyond just one chip - all I can say is you need to analyse what's happening and be aware of the return HF currents getting back to the chip via the rail it uses as its signal ground for those HF currents.
Like I said, this is not an easy subject to explain but when dealing with it you must try to analyse and find ways to deal with it. Even very experienced techs and designers have found themselves scratching their heads on this, especially in more extended multiple IC topologies. And they may not even be dealing with sonic qualities, just trying to keep things stable. "Where is that weird oscillation pattern coming from?"
The concept of 'ground' is a fleeting subject.
There is no universal ground.
For example: Grounds can lie.
NEVER TRUST A 'GROUND' !!! TREAT ALL WITH SUPREME SUSPICION.
Really, when you think of it, I ain't got any dirt in my amps, so there are NO grounds in any of my amps.
Grounds are common connections to which signals and power are referred or referenced. Nothing more than that. That is textbook.
Grounds do not LOCK DOWN signals or currents. Grounds can move.
As for a 'layout tip' - you are right. I can say that for single chip situations, you can consider it thus. And that was my idea as a 'trick' to improve the 627, which I like (if it has to be an IC). Apply it to the gainclone chip too. I bet that the original Gaincard has the speaker return coming back to the electro cap x 2 junction.
Joe R.
i think i'm sort of getting the point here. looking at a 'ground' purely as a current return path and/or point of reference explains many, at first sight, complicated (not only stability) problems.
Non EE people (like me) tend to forget/overlook this pretty often. Looking at a preamp pcb i am designing i see a number of stupid layout errors already...
/m
Non EE people (like me) tend to forget/overlook this pretty often. Looking at a preamp pcb i am designing i see a number of stupid layout errors already...
/m
matjans said:Looking at a preamp pcb i am designing i see a number of stupid layout errors already...
/m
If you're designing, you're still in time to fix it.
If it's op-amps, don't forget to include a place for a resistor from output to V-!
Joe's trick seams easy to adap without too many changes on an existing circuit, but on an already tight layout I guess those caps will be floating with a leg in the air.
Re: Re: joe's tip
Don't know about the original gaincard but the 47labs 4717 shigaraki seems to have it that way:
follow the green wires from the black speaker post to the space in between the two caps. Pictures borrowed from this dutch review. in the review the pics are clickable which will give you higher res pics.
stupid Q : carlos, wont't you get a negative dc offset by putting a resistor from v- to output?
Joe Rasmussen said:
Don't know about the original gaincard but the 47labs 4717 shigaraki seems to have it that way:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
follow the green wires from the black speaker post to the space in between the two caps. Pictures borrowed from this dutch review. in the review the pics are clickable which will give you higher res pics.
stupid Q : carlos, wont't you get a negative dc offset by putting a resistor from v- to output?
Joe Rasmussen said:
Do not want to be much offensive, but according to Burr-Brown's datasheet the 637 has much worse +Vs PSRR than the 627, even for high frequencies (datasheet, page 5, PSR vs frequency curves). The difference is nearly 20dB.
Please take notice that +Vs PSRR is almost always different than -Vs PSRR for every opamp.
I may have misunderstood what Joe is saying here but I thought the return wire from the speaker always went to the junction of the two 1000 uF caps anyway (as shown below where the thicker blue wire is the one in question).
An externally hosted image should be here but it was not working when we last tested it.
i wasn't joking
Re: signal returns, grounding, , decoupling
IMO a must read on this topic for everyone designing with IC-amplifiers:
Analog-Devices Application Note AN-345:
Grounding for Low- and High-Frequency Circuits: Know Your Ground and Signal Paths for ...
and more detailed:
Analog-Devices Application Note AN-202:
An IC Amplifier Users Guide to Decoupling, Grounding,and Making Things Go Right for a Change
Take a look especially at figures 3a, 3b (and 3c for driving an IGC with a buffer) and the corresponding text. I believe this adresses what Joe is talking about.
IMO a must read on this topic for everyone designing with IC-amplifiers:
Analog-Devices Application Note AN-345:
Grounding for Low- and High-Frequency Circuits: Know Your Ground and Signal Paths for ...
and more detailed:
Analog-Devices Application Note AN-202:
An IC Amplifier Users Guide to Decoupling, Grounding,and Making Things Go Right for a Change
Take a look especially at figures 3a, 3b (and 3c for driving an IGC with a buffer) and the corresponding text. I believe this adresses what Joe is talking about.
Re: Re: Re: joe's tip
Point taken. I was merely supposing. I didn't check the spec sheet, so I got it back to front. But in either case, the principle is till the same, PSR deteriorates at HF and less able to suppress HF return currents. This is one area where feedback does a good job but also needs a bit of help.
Lot to be said for groundplanes and we have used them at VSEI. But they shouldn't, as I'm sure you would agree, be viewed as a perfect cure but as one to put in one's armaments. But in later years we have preferred in straight audio (AF) circuits to analyse loops, then find the best point to reference and 'tap' that point via a wire (or even better a foil or 300x0.1mm tin stranded copper) that has zilch inductance etc., then make sure that neither AC or DC (ideally) actually flows through this connection. If you do it right, everything forms loops that are stable.
But as always, everybody has to decide what or how they feel best able to deal with it. But having a suspicious mind is a good thing, IMHO.
Joe R.
PMA said:
Point taken. I was merely supposing. I didn't check the spec sheet, so I got it back to front. But in either case, the principle is till the same, PSR deteriorates at HF and less able to suppress HF return currents. This is one area where feedback does a good job but also needs a bit of help.
Lot to be said for groundplanes and we have used them at VSEI. But they shouldn't, as I'm sure you would agree, be viewed as a perfect cure but as one to put in one's armaments. But in later years we have preferred in straight audio (AF) circuits to analyse loops, then find the best point to reference and 'tap' that point via a wire (or even better a foil or 300x0.1mm tin stranded copper) that has zilch inductance etc., then make sure that neither AC or DC (ideally) actually flows through this connection. If you do it right, everything forms loops that are stable.
But as always, everybody has to decide what or how they feel best able to deal with it. But having a suspicious mind is a good thing, IMHO.
Joe R.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.