My take on the phonoclone
I'm sure you know the huge phonoclone-Thread http://www.diyaudio.com/forums/showt...398&highlight= concerning the fabulous design of rjm.
A few weeks back, I gave it a try and build it on veroboard. I noticed two things immediatly:
1) Resolution was quiet average.
2) Rhythm (be it transient response or whatever) was very, very good.
So I decided to squeeze some more performance out of this nice little circuit. The result, however, is not that simple (and not that cheap to build) as before.
Every now and then, I made some sims or put a scope at the circuit, but every step off from the original design was confirmed by close listening for a week minimum with various kinds of music, and most steps were taken back one or more times to confirm my listening impressions. Not all were beneficial, of course.
But let me describe the steps I took, at least the major ones.
1) Imagine my first version is rjm's original circuit, but build on veroboard and fired with lead acid batteries (+/- 12V)
My first change was to use the famous jung/didden super regulator with the AOS-boards.
What a difference! The soundstage was much more focussed, instruments have a defined and narrow place at the 'stage', the overall sound is much cleaner with tight and fast bass response. Going back to batteries made the sound muddy and fluffy.
2) Replacing the output cap with a servo loop needed a bit of tweaking. The shown values give the best results, I could only find them through listening, since the sim showed only very minor differences were the sound changed substantly (from unusable to very good) with different values for R9/R6. The bass sounds more tight and deeper, even compared to an 47uF foil cap from mundorf.
The mids and highs 'de-sweetened' with the servo, I found them much more real then. Overall, a big step forward.
3) Then I tried some fancy mod with a 'frozen cartridge', that is injected in the feedback path of the AD811, see
At first, I thought to hear a big increase of resolution, but several switches back and forth assured me that I most probably was fooling myself with wishful thinking. So, it stayed out.
4) Op-amp choice. That was the most surprising task, never would I have thought of the shown combination (AD811 as first stage, AD797 as second) as best possible! Which did I try?
At first, the OP27 of the second stage stayed. The AD811 was indeed the first I tried (due to the above linked thread), but I wasn't so sure, a bit better, but lacking something. The second best was the OPA627, which I found surprising as well (in this configuration). Very sweet highs, but a bit muddy in the lower mids.
I tried OPA604 (baaad...), LT1028 (doesn't work, just fancy noises), AD797 (Ok, but not without radio broadcast, and beneath OPA627 and AD811) and AD8761 (?) (nothing special). At this point, I would probably have stayed with the op27.
Then I checked 2nd stage (all of the mentioned ones in the first).
OPA627: boring! Somewhat extended highs, but fluffy. No punch either. OP27 is better here.
LT1028: way better, but way too soft as well. Choice between OP27 and 1028 would be a matter of taste.
OPA604: simply bad.
AD797: A true winner by a far margin! Endless Resolution, and rhythmycally involving. What a chip!
The first stages op-amp choice was a match between AD811 and OP627 then, because they were both good, but very differently sounding (not surprising, they are very different topologywise).
While the OPA627 has nicer trebles, the AD811 is more complete and a 'groove monster', so it was the winner after two weeks. Only thing I have to try is the AD844, but it hasn’t arrived yet. There are other CF-amps, but nearly all have way too much offset (and cancelling did affect the sound, if possible at all).
5) Class A biasing. The shown 4k are better than higher values, I did not go lower as the chip gets hot enough after a few hours. A more elaborate solution with res+FET sounded worse, so I did not try more fancy solutions with two FETs. The shown decoupling scheme for the AD811 lowers noise significantly when cart is connected and may sound a bit better, too.
I have attached the actual circuit and am looking forward to rebuild it nicely (groundplane, without IC-sockets and the like).
Feel free to comment or point me to a mistake if you see one.
thanks for reading,
If I remember you were building the ono too, if so did you compare with your phonoclone?
No, I did not build the ono, sorry :smash:
I have never thought of a cartridge as a current generator, some MM types might approximate that, but no MC types do. I would reconfigure as a normal non-inverting voltage amplifier. Note that the cartridge is being loaded into a virtual earth and I wonder what that is doing to the frequency response of the cartridge, not to mention its electromechanical damping. I assume this would seriously affect the Q of the mechanical system. I am not sure that is a good or bad thing, but different from what the cartridge designer had in mind.. :)
MC typically have very low dcr so you would get a lot more gain, but the system would almost certainly be highly mechanically overdamped as the coils would be looking into a short circuit. This might not matter much in very low output mc due to their very low source impedances, but again I am not totally sure about this... Rationally speaking it seems like this might be appropriate for very low output mc types, perhaps better than some other possibilities.
It is generally not a bad idea to provide some voltage gain in the first stage as this helps to establish a good noise figure before you apply equalization using potentially noisy passive components.
I'm not too comfortable with the quality of this design, but I admit I design with op-amps all day long for ATE applications, no longer do audio design professionally, and use tubes for my own designs. This design strikes me as far from optimum, but YMMV.
One possible thought is you might want to try a two pole integrator in your dc servo. I have not had good results in the past with the single pole type. Stagger the values by about an order of magnitude, both poles can be relatively to extremely close together, try to taste.
Also I would use smaller value resistors in the servo divider back to the AD797, don't change the ratio..
The 2.21K resistor(s) contribute(s) almost as much noise as the input voltage noise of the AD797. The signal level at the output of the AD811 is going to be on the order of millivolts, certainly less than 100mV even at 20kHz with a typical MM cartridge with a dcr of say 600 ohms, so even a 221 ohm input resistor at the AD797 is not going to result in excessive currents and linearity problems.
I would scale the resistors and capacitors in the feedback network with an eye to reducing resistor johnson noise within reason. A 750K resistor will generate almost 100nVrtHz, and even though shunted at high frequencies it will generate low frequency noise. I would scale by at least an order of magnitude for low noise, but be cognizant of quality issues in the equalizer network capacitors.
edit: additional thoughts added
thanks for sharing your thougts. Let me respond to some of them.
1) lower load resistor for U1: I thought about it previously, and I will possibly try that later on. The feedback network of U2 would as well be lowered then to achieve the same gain. On the other hand, the light feedback loading may be one reason for the good sound. We'll see.
2) two pole integrator for servo: do you have an example? Are you simply talking of dividing up R9 and insert an additional cap? What benefit would you want?
3) Of course, I use a MC cart here. Though one with a low Z would probably be optimum, I use a Denon DL103 (~40Ohm) at the moment. When it comes to inverted topo, people say that carts that perform well with input transformers might perform well with this design. See my links for experiences of other people with this topology.
Just add a 100K resistor and 1.0uF cap ahead of the existing bias servo integrator and that should do the trick. Tweaking of values (aka tc's) might be necessary to retain the tight bass response you are looking for, overall subsonic behavior will be a lot better due to the steeper slope the 2nd order network gives you.
I have since looked at the other thread where the virtues of this approach are discussed so I have a better feel for the intent at least.
The DL-103 I suspect is a good match for this design, and probably any comparable transformer matched cartridges would work pretty well.
I am still curious about the effect of the virtual earth input (transimpedance amplifier) on the damping of the mechanical system. I wonder whether it tames the typical ultrasonic peak in the transfer function of mc type cartridges, I would assume it does..
I would also recommend taking a look at a combination of passive and active equalization in your next go around. The basic transimpedance concept is an interesting one, and suitable enough for low output mc cartridges..
Lots of interesting audio ideas have come out of MIT over the years, this phono stage concept, lightbulb based speaker protection (read simple compressor) and of course Dr. Bose :xeye:
1) The question of mechanical damping and its relevance to freq. response or other domains is behind my capabilities (I'm no EE or the like). What effects would you expect are present, at least potentially?
2) I sim'ed an elaborate servo loop, yes, the slope is steeper, but there is also a resonance area, don't know if it would affect the sound.
I think I have to try...
(don't mind different opamps here, they are just present to get the sim running)
...and the results:
Your servo circuit is not quite what I had in mind when I suggested the change. Basically what you now have is an op-amp with an inverting gain of 2 and some frequency response shaping inside the loop.
I can well see why there is a huge resonance in the pass band and I expect the servo won't really work at all, overall gain including the attenuation at the op-amp input is far less than unity, for effective servo operation you want fairly high voltage gain.
Refering to your current schematic I would remove R10, this will restore the integrator loop gain.. Change R4 to 100K, C2 to 1.0uF, remove C5, and replace R9 with a short. This will give you the classical 2nd order integrator I was talking about, and should have very good subsonic behavior.
You don't need to use a laplace source to model the cartridge, just a simple voltage generator with a 40 resistor in series, unless you want to do eq which appears to be the case here.
I still recommend scaling the components in the feedback network to reduce johnson noise, reduce resistors to 0.1X current values, increase caps 10X for a worthwhile reduction in noise.
Most probably, this fiddling with the servo is a bit overkill, but:
according to the sim, R10 dampens the 'spike' of the pole. The output network (R9/C5) should filter any potential audability of the servo further, but the integration is indeed affected above 1k, so I leave it out.
I will try other values for the riaa, but I have to order some parts first.
The plot is taken without R9/C5 but with R10, green is freq. response at output, blue is output of the servoamp
|All times are GMT. The time now is 10:44 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio