An old review of the NAD3020 records that the measured response of the RIAA section is irregular having a degree of bass-cut and treble-boost Tone control adjustment helped trim this but a 'flat' response could not be achieved. I suggest simulating some other values in the feedback network.
Last edited:
Member
Joined 2009
Paid Member
I think people have been thrown-off by this NAD circuit. I haven't seen it used much by the DIY crowd but I think they're missing out. There is no LTP anywhere in it at all. Look at the schematic I posted above, taken from the NAD service manual.
Q402 and Q404 are the input stage. They behave just in the same way as a conventional LTP in acting as an error amplifier. The input signal goes to the base of Q402 and the feedback signal to the base of Q404. The standard LTP is a 'folded' version of this topology and was invented first because the tube guys didn't have p-type tubes. Christopher Rush corrected this omission and invented the npn-pnp version, or so I'm lead to believe. The dissadvantage of this Rush Cascode input stage is that it doesn't provide equal dc-voltages at the two base junctions so it wouldn't be convenient for a dc-coupled amplifier. We don't need dc-coupling here. The advantage of this Rush Cascode is that the series nature of the two devices ensures the current is balanced between both devices, both in terms of dc and ac. In terms of implementation, there's no requirement to match the devices.
The output from the 'unfolded LTP' Rush Cascode is to the base of Q406 which is emitter coupled to the output stage. There are often multiple ways to look at something. I see two ways here. One way is to view Q406 and Q408 as an emitter coupled amplifier with the output at the collector of Q408 loaded on a CCS formed from Q410. Another way to look at this is to view Q408 and Q410 as a pair of current sources and Q406 as an emitter follower that drives the upper current source.
Anyhow, I find it to be an elegant if not somewhat quirky topology. Without an LTP 😉
It's all in post #53, see the link to Andyc 's website for further details. I am thinking that I should focus on matching the RIAA capacitors between channels as my first priority. Once that's achieved and I have a measurement of the capacitances (I have a c-meter) then I calculate the resistor values that I want. So I should plan to buy a few closely spaced valued resistors to pick from.
Do you think I'm on the right track with the schematic in Post #215 and the overall block scheme in Post #219 ?
Last edited:
On the right track.
I am not sure what the ground switching is about for the inputs. Might you have problems if a connected device is floating wrt mains ground and if the signal switch closes slightly before the ground switch?
I wouldn't expect your discrete buffer to out-perform an IC buffer (as I said earlier).
Your complicated shunt regulator is OTT in my opinion. An LM317/337 would be fine. Also, the shunt regulator circuit seems to allow cross-talk between the positive and negative rails.
I don't have an opinion about the RIAA - you've borrowed a working NAD circuit. Ok.
It depends what your goals are. If you want simple and effective you make choices and if you want to try things out because they are novel and interesting to you then you make other choices. It's all good.
Regarding the Mark Levinson...no, the idea was not to expect you to make something just like it! It is a little hard to see details in this article but you may pick up some ideas.
Last edited:
Member
Joined 2009
Paid Member
I'm trying to avoid any possibility of ground loops. If I common together the grounds then I could create ground loops that include two or more of the external sources.
What scenario would this create a problem, are you thinking of pops-and-clicks?
Note that all signal inputs are connected to their respective signal grounds through a 100k resistor. I was wondering if I might have a problem with pops-and-clicks when switching takes place, whether it be input selection or other controls. I thought it might be necessary to have any kind of control switching temporarily mute the output (e.g. for half a second) until transients have settled.
Interesting question and shows you have looked carefully at the design - I assume you are referring to the zener diodes not returning their current to ground. I think you're right, I'll amend the schematic.
Last edited:
I have done that. What i havent is to find out what parameters affecting various perceived sound differences. When the parameters are found, there would be no need to spend too many times in experiment.
But i havent found the need anyway. I just build what i can find on the net and compare by ears and do some non critical tweakings mainly with bypassing and size of capacitance.
One output of my experiments that differs from you is that i have a negative feeling towards opamps, something that you find strange.
nevertheless you ask (post #169) what output impedance an opamp-based voltage regulator is able to achieve? why do you care how well opamp-based circuits perform?
OK I have must have skipped post 53 and see you are on track with that.
Sources credit this circuit to Bjorn Erik Edvardsen - the party you were thinking of I think is Tomlinson Holman. The reference to LTP with floating emitters is due to Stan Curtis a designer who reviewed the NAD3020.
Newnes Dictionary of Electronics by SW Amos and RS Amos describes a long- tailed pair as a phase splitting circuit in which push-pull signals are obtained from the outputs of two similar active devices with a common resistance in their emitter (cathode or source). The authors note the circuit can have applications other than phase splitting - which allows how connected to the discretion of the designer - which I see here.
Holman was responsible for the Advent M300 15 watt receiver. An American reviewer thought a John Curl RIAA design was a little better than the Advent equivalent. Holman also did the APT1 when he left Advent to form his own company and later to develop THX sound systems. I think these two Holman's are worth a look and tracking down reviews - the circuits are available online.
What makes the 'LTP' in this NAD circuit somewhat unusual is the connection of Q406 collector to ground and Q408 collector to the output and main feedback loop.
Since signal transfer is between transistor emitters and this is bi-directional it is interesting to consider how such between Q406 and Q408 emitters impacts on Q404 collector and base signals.
I still see R418 (82R) as a weak point for entry of artifacts due to the power supply or induced by other circuit modules on the line. The current drawn by
R418 is 6 m.A. - if it does not involve too much change to the circuit a constant current diode or FET set up to act as such could be a replacement for R418.
I cannot say why this circuit has not attracted a lot of interest - there are probably simpler circuits that might do as good a job. On noise performance how about BD437 and BD438 for Q402 and Q404?
On your last question my approach would be to use separate modules so these could be swapped out for possible future development. This will allow the sort of tuning, trying better devices and general tinkering as has been suggested by Mark Johnson without having a major overhaul.
Because I believed you know what you are doing, so I was interested if you can make a good regulator from OP07. Since I care for anything best in audio 🙂 (even if it is opamp or whatever)
When you first posted about your pre power supply you only mentioned PSSR but then with Bigun's circuit you seemed to be interested with low output impedance also, so I thought you might have created a good power supply because I thought Zout is the most important parameter.
Soon after that I browsed to check the Zout of several opamp-based super regulators (for reference or comparison) I have built (but not favored) and found out that they are actually very low. So I think I don't know what parameters make a good sounding power supply.
I mentioned LM317 because I like its sound. How I know about making a good sounding circuit is actually by knowing what I perceive and then find out why it is so. I have visually checked the internal circuit of LM317 but it doesn't tell me anything. May be if simulated I will find out something, but like I said I have no urgency to study power supply design.
I just build other people designs. Including yours if you post the design. BTW, my favorite opamp for power supply was OPA627, next was a very cheap noisy one (I forget the model).
Would someone tell my why low output impedance matters? I could imagine 10R with carefully placed decoupling or whatever it is best called would do the same job. Surely the impedance or current use of a circuit is a factor? Is not every ground connection and track throwing it away?
Here is a thought. Surely a CLCLC series of Pi filters do a better job if low current? The use of the CMRR to finish the job. Even using a crap op amp like LM324N usually shows grounding to be 100% the main problem. I often use LM324N at the start of a project. It is as good an indicator of trouble as any other op amp. 90% of the time there is trouble which often need simple fixes. As a preamp becomes more complex these troubles become harder to solve. I don't ever remember an LM317 being the cause except if it oscillated. It seems to be better than many on that.
Here is a thought. Surely a CLCLC series of Pi filters do a better job if low current? The use of the CMRR to finish the job. Even using a crap op amp like LM324N usually shows grounding to be 100% the main problem. I often use LM324N at the start of a project. It is as good an indicator of trouble as any other op amp. 90% of the time there is trouble which often need simple fixes. As a preamp becomes more complex these troubles become harder to solve. I don't ever remember an LM317 being the cause except if it oscillated. It seems to be better than many on that.
@Gareth
I need to think about those ground connections. It's like when you connect an RCA to your amp - can cause clicks/thumps. The issue is connecting the signal before the ground. If you are trying to avoid possible ground loops why not parallel the ground switches with resistors so there is always a connection but through a ground lift?
Shunt regulator. Yes, the common bias resistors between pos and neg is not ideal. I would consider J505 current diodes for your CCS to make it more compact: eliminates R1, Q3,Q4,R205. Have you analyzed the stability of your circuit...it seems to rely on the output cap to slow everything down. Why did you choose 33uF?
I need to think about those ground connections. It's like when you connect an RCA to your amp - can cause clicks/thumps. The issue is connecting the signal before the ground. If you are trying to avoid possible ground loops why not parallel the ground switches with resistors so there is always a connection but through a ground lift?
Shunt regulator. Yes, the common bias resistors between pos and neg is not ideal. I would consider J505 current diodes for your CCS to make it more compact: eliminates R1, Q3,Q4,R205. Have you analyzed the stability of your circuit...it seems to rely on the output cap to slow everything down. Why did you choose 33uF?
If think so. If the input signal causes Q402 emitter to draw more current it will source this from Q404 emitter which do likewise by drawing current from Q406 base - causing Q406 emitter to draw more of the current provided by R418.
In consequence there will be less current available for Q408 emitter which will reduce the base-emitter voltage - and base current. The change of state will allow more current from D402 and D404 to travel via R424 to Q410 base and this transistor will conduct more as Q408 conducts less.
If the input is a regular ac these changes will reverse as the signal gets to the crossing point - which is a dc in this instance with the ac superimposed - then there will be feedback from Q408 emitter to Q406 emitter.
If the input signal causes less current to be drawn from Q404 the process reverses.
Q408 collector has a lot of connections to it including the next stage in the chain for this reason and the fact the RC network R428/C422 needs a high impedance buffer to make sure this zeros correctly.
As far as matching emitter and collector currents of complementary transistors for Q402 and Q404 is concerned whether it matters or not if the devices are from the same broad family- I think this would be less of an issue with medium power complements like BD437 and BD438.
That's the right question. 🙂 Pursuit of very low Z through wide-band active circuits can lead to serious side-effects that will counteract any benefit the low Z brings.
On the NAD style RIAA. Sort any EQ errors as best you can. Often people use COG/MPO or polystyrene caps. If all active try a 2 uS passive filter at the end ( scratches then sound nicer, more tick than tock ). If it were 10nF 200R it could be a useful PSU hum blocker also. As my brother pointed out if everything about a PSU is low Z it can cause a PSU hum loop. His feeling was the output of a stage is where to lift things a bit.
I had never liked RC. I mean in an amplifier with RC, the sound would always be preferred when the RC removed. So I thought that the problem was high power supply output impedance...
But then I found out that LC, RC really needs exact capacitance (or inductance)! But what is the secret formula here??
Many said that the problem is the feedback? May be power supply is not so different with amplifier?
What is a power amplifier if not a high current voltage regulator with 20kHz bandwidth voltage adjustment? Voltage regulators usually have single ended outputs. All the same design challenges apply.
Member
Joined 2009
Paid Member
I figure we are looking at this board as a phono module, you can take away the relays but you need everything else. The phono needs a regulated supply due to low PSRR, which means the supply will be heard if it's poor. Shunt regulators appear to have a good reputation in this regard, the Salas Simplistic is a good example. My shunt design is almost a bjt version of that one. The output buffer helps the phono drive the next stage since the output from the phono is relatively high impedance.
I'll have to run a sim on the regulator stability though.
I'll have to run a sim on the regulator stability though.
Last edited:
Have a look at the Linear Audio article by John Walton link. They measured the performance of 13 different voltage regulators, then hooked up those regulators to a reference preamp and listened to music (single blind). Listening preferences were then correlated to regulator measurements. One of their findings was that output impedance was positively correlated with listening preference, namely: better sound tended to go hand in hand with lower output impedance. The correlation was not perfect; there was a bad sounding regulator with below-average output impedance and there was a good sound regulator with above-average output impedance.
Member
Joined 2009
Paid Member
I also suspect that output impedance is important, afterall, if it's low enough it implies there will be no voltage drop across the power supply as a result of signal current flowing through the rails. It maybe that for some regulators some small amount of voltage drop across the supply that correlates with the signal is somewhat pleasurable from a sonic perspective - and there are power amplifiers that exploit this by allowing some of the output stage current ripple on the power rails to interfere with the front end circuits. No doubt it's the same with capacitor choices, their impact on the sound depends on signal current induced voltage drops interacting with less than perfect PSRR. So the shunt regulator for this phono-module is necessarily 'busy' in order to achieve decent output impedance - limited of course by the simple circuitry employed which limits the OLG of the thing as frequency rises. I could design a regulator that is technically 'better' but there are diminishing returns so what I have here I hope is going to provide for pleasing results.
(that article is not free and given that the internet is full of free stuff I don't pay)
(that article is not free and given that the internet is full of free stuff I don't pay)
Last edited:
- Status
- Not open for further replies.