Hi,
Don't overdo it on cancellation. It sounds worse and becomes extremely sensitive to tube variation.
Unbypassed resistors add noise. That is one of the reasons neither CD-77/777 or DP-777 have very low noise.
Gomes always uses signal from lower anode, not upper cathode.
As you know from the CD-77 it uses an ECC81 in the first stage and a 5687 in the second.
So we are clearly doing some more, as Doctor Who likes to call it, "jiggery pokery"...
Ciao T
Don't overdo it on cancellation. It sounds worse and becomes extremely sensitive to tube variation.
Unbypassed resistors add noise. That is one of the reasons neither CD-77/777 or DP-777 have very low noise.
Gomes always uses signal from lower anode, not upper cathode.
As you know from the CD-77 it uses an ECC81 in the first stage and a 5687 in the second.
So we are clearly doing some more, as Doctor Who likes to call it, "jiggery pokery"...
Ciao T
Simulated the CCS with DN2540, both straight and cascoded... don't get to THD specs as with the tube...
You still don´t get it about simulations?
1 The .four is not to trust!
2 THD is uninteresting, harmonic spectra is!
But if you believe Spice is correct, the spectra of the Gomes is not nice as the higher overtones doesn´t diminish the higher they get. How it is IRL I don´t know but probably better.
And as I said to get a better picture you must try many different tubemodels. Why not try what T used in his design, a ECC81.
In a world were the amp with the lowest THD sounded best, it would be easy to choose😉.
1 The .four is not to trust!
2 THD is uninteresting, harmonic spectra is!
But if you believe Spice is correct, the spectra of the Gomes is not nice as the higher overtones doesn´t diminish the higher they get. How it is IRL I don´t know but probably better.
And as I said to get a better picture you must try many different tubemodels. Why not try what T used in his design, a ECC81.
In a world were the amp with the lowest THD sounded best, it would be easy to choose😉.
Hi,
Attached a real Gomes stage, outputting 2V RMS into 10K. Note that this one deliberately did not have any aggressive distortion cancellation applied. Plus the ECC81 is a less linear Valve than the 12AY7.
As we can see, 2nd HD is around 0.3%, 3rd is 0.003%, 4th is MIA, 5th is 0.002% & 6th and higher at 0.0005% or lower. The spike above 10KHz is from the environment, not the test gear, maybe a CRT Monitor leaking in.
We can cancel the 2nd HD a lot more, but 3rd and 5th are then higher and the HD becomes more tube sample dependent. And it does not sound as good.
Using 6072A/12AY7 lowers 2nd HD appreciably, as it is basically a linearised and "audio grade" ECC81. Other ECC81 analogs, such as 5965 also give lower 2nd HD.
Ciao T
Attached a real Gomes stage, outputting 2V RMS into 10K. Note that this one deliberately did not have any aggressive distortion cancellation applied. Plus the ECC81 is a less linear Valve than the 12AY7.
As we can see, 2nd HD is around 0.3%, 3rd is 0.003%, 4th is MIA, 5th is 0.002% & 6th and higher at 0.0005% or lower. The spike above 10KHz is from the environment, not the test gear, maybe a CRT Monitor leaking in.
We can cancel the 2nd HD a lot more, but 3rd and 5th are then higher and the HD becomes more tube sample dependent. And it does not sound as good.
Using 6072A/12AY7 lowers 2nd HD appreciably, as it is basically a linearised and "audio grade" ECC81. Other ECC81 analogs, such as 5965 also give lower 2nd HD.
Ciao T
Attachments
Hi,
I hope you know that much of what John does is theoretcal/simulated only, that he massively overemphasises PSRR over any other considerations and rarely if references what he does to evidence of what is audible?
I had twice the occasion of rescuing projects a friend and DIY Kit Vendor had based on stuff posted at that site which simply did not work properly... I ended up completely redesigning the projects as the original circuits neither worked well electrically in reality nor sounded any good...
So any of his blog posts should be taken "cum grano salis magnitudinem montis".
Ciao T
I hope you know that much of what John does is theoretcal/simulated only, that he massively overemphasises PSRR over any other considerations and rarely if references what he does to evidence of what is audible?
I had twice the occasion of rescuing projects a friend and DIY Kit Vendor had based on stuff posted at that site which simply did not work properly... I ended up completely redesigning the projects as the original circuits neither worked well electrically in reality nor sounded any good...
So any of his blog posts should be taken "cum grano salis magnitudinem montis".
Ciao T
Hi,
I must add, there is a TDA1541 DAC included in this plot, running "sans un filtre numérique", so it is likely to add to some of the higher order stuff.
Here by comparison a TDA1541 with Op-Amp (LM6181 current feedback) analog stage:
/lm6181_harm_1kHz.gif)
This is probably about as as it gets for the TDA1541.
Ciao T
I must add, there is a TDA1541 DAC included in this plot, running "sans un filtre numérique", so it is likely to add to some of the higher order stuff.
Here by comparison a TDA1541 with Op-Amp (LM6181 current feedback) analog stage:
This is probably about as as it gets for the TDA1541.
Ciao T
Makes you wonder why we mess around with tubes 🙂
Thorsten, what's your view on using a Mosfet CCS to replace the top left triode as suggested earlier?
Also i will change the operating point. Not comfortable to work with 400v in a Dac... Should be possible to get good results with lower voltage, also on 6072a.
Thorsten, what's your view on using a Mosfet CCS to replace the top left triode as suggested earlier?
Also i will change the operating point. Not comfortable to work with 400v in a Dac... Should be possible to get good results with lower voltage, also on 6072a.
Hey T,
It´s in the opposite direction. Your sim looks better in the higher order area than the sim. Just by looking at the fourier-analysis you can tell there is something funny going on.
Can also agree about the blogs. This also goes for the .four-sims above.
It´s in the opposite direction. Your sim looks better in the higher order area than the sim. Just by looking at the fourier-analysis you can tell there is something funny going on.
Can also agree about the blogs. This also goes for the .four-sims above.
Hi,
As good as it get's for "measured performance"... 😉
If I thought it had merit, I would have used in the AMR designs, wouldn't I?
Ciao T
As good as it get's for "measured performance"... 😉
If I thought it had merit, I would have used in the AMR designs, wouldn't I?
Ciao T
Hi,
Mine was the real thing on an Ap2, not a sim... 😉
The high order stuff looks better because the lower order stuff looks worse.
Ciao T
Mine was the real thing on an Ap2, not a sim... 😉
The high order stuff looks better because the lower order stuff looks worse.
Ciao T
Typo, Off course I know yours is the real thing!
Wrong, all orders considered from first and up. The sim is better in the lowerorder and worse in the high order even without what the TDA1541 adds.
Only wanted to point out that one should take the simmed .four "cum grano salis magnitudinem montis".
Wrong, all orders considered from first and up. The sim is better in the lowerorder and worse in the high order even without what the TDA1541 adds.
Only wanted to point out that one should take the simmed .four "cum grano salis magnitudinem montis".
Hi,
No, I meant if I adjust the distortion cancellation in the circuit, I can bring 2nd HD down to around 0.03% (that becomes very tube dependent though), but higher order stuff will increase...
Essentially while not feedback similar principles apply, with feedback, until get well past 30dB NFB you are increasing higher order harmonics over the original level while reducing lower order ones.
There kind of seems to be an analog to the 1st law of thermodynamics, distortion energy cannot be "destroyed", only shifted from harmonic to another or from one domain into the other... (I appreciate that this merely a first order approximation and does not hold for a large amount of feedback)
Ciao T
No, I meant if I adjust the distortion cancellation in the circuit, I can bring 2nd HD down to around 0.03% (that becomes very tube dependent though), but higher order stuff will increase...
Essentially while not feedback similar principles apply, with feedback, until get well past 30dB NFB you are increasing higher order harmonics over the original level while reducing lower order ones.
There kind of seems to be an analog to the 1st law of thermodynamics, distortion energy cannot be "destroyed", only shifted from harmonic to another or from one domain into the other... (I appreciate that this merely a first order approximation and does not hold for a large amount of feedback)
Ciao T
Hey T,
Sorry, you seem to talk about something else than me😉.
The subject I talked about in that post, was that the sims aren´t reliable and does not have with reality to do.
The discussion about cancellation and NFB you should take with studiostevus. To, as you say, bring down the even harmonics wont help much though.
Sorry, you seem to talk about something else than me😉.
The subject I talked about in that post, was that the sims aren´t reliable and does not have with reality to do.
The discussion about cancellation and NFB you should take with studiostevus. To, as you say, bring down the even harmonics wont help much though.
Hi,
Well, IF the Sim was given sufficiently realistic models (including realistic sample variations for each tube) the correlation between simulation and reality would be quite horrorshow, but alas, in most cases sim's tend to work GO* principle.
Ciao T
* GO is short for Garbage Out - the cause for this being obvious.
Well, IF the Sim was given sufficiently realistic models (including realistic sample variations for each tube) the correlation between simulation and reality would be quite horrorshow, but alas, in most cases sim's tend to work GO* principle.
Ciao T
* GO is short for Garbage Out - the cause for this being obvious.
Alright, fair enough, but that would mean that all desk research is pointless, and it would not even highlight relative differences between topologies...
Case in point: i have done simulations of a gomes stage and a mu follower, both using 6072a (same model).
The former simulates quite nicely (2h < -90db, 3h not visible), while the latter a lot worse (2h -40db, 3h -45db, etc)...
The former simulates quite nicely (2h < -90db, 3h not visible), while the latter a lot worse (2h -40db, 3h -45db, etc)...
Hi,
Simply be aware of limitations and for example base your tube models on several brands and samples, curve trace and spice modeled, with one "average" model and a min and max deviation one.
Then try the different models in the same sim... You may be surprised.
The high amount of HD cancellation apparent, results from having two truly identical tubes, something only possible in simulators.
I actually learned to NOT go for aggressive distortion cancellation from Sim's used that way, but confirmed the sim results in reality...
Ciao T
Simply be aware of limitations and for example base your tube models on several brands and samples, curve trace and spice modeled, with one "average" model and a min and max deviation one.
Then try the different models in the same sim... You may be surprised.
The high amount of HD cancellation apparent, results from having two truly identical tubes, something only possible in simulators.
I actually learned to NOT go for aggressive distortion cancellation from Sim's used that way, but confirmed the sim results in reality...
Ciao T
This is the alternative mu follower I have doctored... THD of 0.22%... no cancellation between tubes, lower B+.... biased with the TDA1541 offset
I am not convinced.... Thorsten, you have surely tried both mu follower and gomes topologies, and have advocated the gomes typology several times on this forum. Which one to pursue?
I am not convinced.... Thorsten, you have surely tried both mu follower and gomes topologies, and have advocated the gomes typology several times on this forum. Which one to pursue?
Last edited:
- Status
- Not open for further replies.