I take that back, I think it as organized the best it could have been. After all, you have to get the neat stuff like cable termination and local oscillation in there ithout re-teaching basic electronics. The book is so specific to audio amplifiers, so universal amp theory is not there to segue beteen empirical topics. I had to think about it to realize hy it seemed this ay.
The difficulty I see is in grasping the interconnectedness of every aspect of design of a given amplifier. It's function, it's form, it's interface, it's features, it's economics, reliability and so on. The idea is that you can define every possible requirement the amp ill have, and kno based on that hich topology ill give you the best combination of features. This is very much a multi-level activity as every change in component affects the layout, stability, nominal operating point, and most of the other components. riting a book ell enough to alk a person through all of this for any given set of requirements ould be difficult. In the end the brain becomes the final book that the designer goes by, and sometimes pages stick together.
But I alays think there must be some basic idea that ould provide an intuitive structure upon hich every topic e discuss here could hang like ornaments on a christmas tree, ithout lack of space or unnecessary empty spaces. And from such a structure e could easily identify patterns hereby e ould discover more topics, and topologies no one had put together before.
The difficulty I see is in grasping the interconnectedness of every aspect of design of a given amplifier. It's function, it's form, it's interface, it's features, it's economics, reliability and so on. The idea is that you can define every possible requirement the amp ill have, and kno based on that hich topology ill give you the best combination of features. This is very much a multi-level activity as every change in component affects the layout, stability, nominal operating point, and most of the other components. riting a book ell enough to alk a person through all of this for any given set of requirements ould be difficult. In the end the brain becomes the final book that the designer goes by, and sometimes pages stick together.
But I alays think there must be some basic idea that ould provide an intuitive structure upon hich every topic e discuss here could hang like ornaments on a christmas tree, ithout lack of space or unnecessary empty spaces. And from such a structure e could easily identify patterns hereby e ould discover more topics, and topologies no one had put together before.
I don't know if it will be a problem, because it is not an area where I have much expertise, hence the concern!
My instrumentation is the usual - dual trace CRO, Wien B. oscillator, 16 bit sound card etc. so I am not equipped if any hi impedance sophistication is required.
Surely the philosophy is not "low current gain" but rather "no ill defined quiescent currents". The low current gain is a side effect of the way you have solved that problem. I want to stay with your "no ill defined currents" philosophy but maybe with lower impedance if that is possible.
So what happens if we scale Q9 and Q12 (from post #1) to increase their current to a multiple of Q2 and Q3? Then I think the current in the LTP will still be balanced so the extra flows in the Baxandall super pair. Do you think that will work?
Best wishes
David
Last edited:
"So the connection goes directly to the PSU but isn't connected to the PSU I can't see how what I've drawn is different from what is shown in your e-amp write up, page 13 and page 15:""
The connection is not made on the PCB Ground - it is returned to 0V as shown the picture you posted, and you can see it on the PCB pic you also posted.
The connection is not made on the PCB Ground - it is returned to 0V as shown the picture you posted, and you can see it on the PCB pic you also posted.
Hi David,
Here's a second thought about your principle concern: The performance of the super TIS does not rely on high impedance loads. The low THD is achieved by removing nonlinear impedances: Cob and Early effect. Consequently, as there are no other loads involved, the output impedance of the TIS becomes indeed extreme high, tens of MOhms. But, as it appeared, we don't need this. I've simmed the circuit again, this time with a (linear!) load of 100k respectively 100pF and guess what? THD20k was still below 1ppm. So, don't worry.
Cheers,
E.
Edit: Our posts just crossed
Last edited:
Hi David,
I think it will work, that is, the quiescent current of the VAS is still under control. But..... the AC gain from the 'left' LTP leg, respectively 'right' LTP leg to the output will be quit different, i.e. unbalanced. That means no longer cancellation of the Early effect and much worse PSRR. So forget it.
Best 'ishes' ,
E.
This is improved Layout for the same TT amp and it's single side PCB. The OPS grounding are made on the back side with the wires following + and . trace and then as twisted pair connected to GND. Now it is easy to make twisted wires connection from the middle of the OPS part of the PCB to the PSU.
Please comment.
dado
Please comment.
dado
Attachments
![DADO-TT-TMC-7-2-f-1.LAY].jpg](/community/data/attachments/280/280049-71deda01b3bb454b838352a0f2f9380f.jpg)
Lower part is input stage(by HF I suppose you meant this) in lower part of the PCB. There are C multipliers and C9, C14 are in the bases and C17, C18 after multipliers. As this part is pure A class, transients are low longer zero trace should be no problem, and gnd should be graunded separately.
This is similar to my approach as well. If you take the input ground back to the common ground directly from the input connector, then you will be creating a large loop - and given the sensitivity (its the input), are likely to get noise. The loop is in effect in series with the input signal in this scenario. My signal ground has the input, feedback and front end filter all joined together at the front end of the amp on the PCB, and then this connects to the board ground, and then back to the power supply ground. I have a 3.3 Ohms resistor in between on the amplifer board between the signal ground and the board ground. The 3 way input terminal block allows me to bypass this 3.3 Ohm resistor with a wire link back to where the board ground exits via a terminal block, but I found it made no difference, so the link is not fitted. I guess the correct thing to do would be to replace the 3.3 Ohm with a link but I have not gotten around to that.
BTW, might be an idea to split this discussion off onto a new thread - are we not hijacking Edmonds thread here?
Last edited:
Dadod,
no, not front end decoupling.
Output stage decoupling comes in frequency variable forms.
Electrolytics for medium frequencies where longer traces with attendant inductance are not too much of a problem. Your traces are still far too long for MF decoupling.
The output stage also needs high frequency (HF) decoupling, usually provided by small ceramic capacitors with extremely low trace lengths, of the order of 10mm, for the total loop length from drain/collector to cap to decoupling ground to power ground and then to the opposite polarity via the second cap.
I can't see the HF decoupling.
Long traces defeat HF decoupling.
If you have ceramics across the MF decoupling caps, located at the end of 3cm or 4cm traces, then they are completely useless.
no, not front end decoupling.
Output stage decoupling comes in frequency variable forms.
Electrolytics for medium frequencies where longer traces with attendant inductance are not too much of a problem. Your traces are still far too long for MF decoupling.
The output stage also needs high frequency (HF) decoupling, usually provided by small ceramic capacitors with extremely low trace lengths, of the order of 10mm, for the total loop length from drain/collector to cap to decoupling ground to power ground and then to the opposite polarity via the second cap.
I can't see the HF decoupling.
Long traces defeat HF decoupling.
If you have ceramics across the MF decoupling caps, located at the end of 3cm or 4cm traces, then they are completely useless.
Last edited:
![DADO-TT-TMC-7-2-g.LAY].jpg](/community/data/attachments/280/280077-7a050a802463c525a455b957db8eb3d4.jpg)
When you say "my TT thread", do you mean this one?
That looks a good place for the ceramic decoupling capacitors* but only if you place the emitter traces of the output transistors on top of (or underneath) the collector traces and bring the zobel ground to the ceramics' ground connection.
* at this voltage you are presumably going to use X7R - bear in mind that these have high voltage coefficients. I'm measuring some 0805 X7R 1 uF caps right now and one of them loses 90% of its capacitance at rated voltage (and this is not some no-name brand either)! The best one loses just under 50% capacitance at rated voltage. Make sure you don't resonate the ceramics with the electrolytics - you may need series RC in parallel with the ceramics to damp any such resonance.
Last edited:
At rated voltage - you sure about that? The curves I've seen IIRC show capacitance dropping off very significantly when applied voltage is much less than rated.
Thank you Harry,
I am to sure what type I use as here in Croatia they don't say exactly what type is, but I think is X7R or X5R. According this document X7R should not lose so much: http://www.wolfsonmicro.com/documents/uploads/misc/en/WAN0176.pdf
When I said TT thread I meant this one: http://www.diyaudio.com/forums/solid-state/182554-thermaltrak-tmc-amp.html as this is my working amp with PCB as shown before I tried to improve it. Other TT thread is new amp with high high current C multiplier on the same PCB board and the layout is not ready yet. I hope I will learn wrom this discusion well enough to design better layout for that amp as PCB dimesion is not restricted.
You said to put in parallel with ceramics, series RC. What value for C and what kind? I use 0.1u for ceramics.
Dado
Hi Harry,
I often like to make sure there is an electrolytic somewhere on a line that is bypassed with an electrolytic to prevent resonance, since the ESR of the electrolytic helps damp any resonance, especially among ceramics separated by trace inductance. I've never seen a resonance problem between an electrolytic and a ceramic.
Cheers,
Bob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.