Date: Mon, 22 Feb 93 00:07:41 EST From: Johannes Chiu Subject: Re:relfection on curcio Thanks for the people who have commented on the article I posted earlier. I guess there are some things I need to clarify, though. >First of all, the ringing is quite real, but it is electronic, not >mechanical in the 6DJ8 tubes. Tube dampers won't help it. In fact, after I have applied the bypass cap at the base of the transistor, I could not hear and ringing coming from the speaker. The high frequency buzz I heard after the mod is in fact due to mechanical vibration and can be heard if I put my ears close to the amp. The ST-70 has a lousy mechanical structure. For instance, the bottom cover of the amp used to vibrate, if I supported the amp on spikes. My solution was to take a piece of plywood about the same size as the bottom cover, and bonded the plywood to the bottom plate using blue-tak. Now, I have removed my tube dampers from the pre-amp and applied it to the 6dJ8s, and the buzz is gone. >However, rather than pursue phase lead compensation, I'm trying to >achieve the stabilization in the forward loop, by adding a rolloff. >The reason for this is that phase lead compenstation works great into >an 8 ohm resistor. Unfortunately, it can be seriously compromized by >the loudspeaker load, which may be inductive, capacitive, or both. >I've found that when you add 1 uF in parallel with the 8 ohm load, the >phase lead compenstation isn't very effective. > ... some stuff deleted... >Using feedback in the forward loop is also recommended in an excellent >article by Norman Crowhurst in _Audiocraft_, sometime in 1956 or 1957 >(reference not handy here at the office). He admits that your amp may >not be flat to 20 KHz anymore, but will sound much better due to >stable operation. > >I am now working on how to use forward loop compenstation on the >Curcio 70. What is needed is an RC network between the plates of the >top 6DJ8's. This is capable of producing a very good margin of >stability. I haven't wired up anything permanently and listened to >it, since I could not do full stability plots. Until last weekend, I >only had an audio oscillator that went up to 100 KHz, I just got one >that goes to 2 MHz. Now I can know how much R and C to use to get >full unconditional stability. (Or as close as I can get with >reasonable frequency response.) > >From what I understand, your objective is to stabilize the circuit. However, I have to remind you that we have two feeback loops, one is from input-cascode_amp-output_tubes-transformer-input, the other is from cascode_amp_plates-transistor-cascode_amp_cathodes. For convenience sake, we can consider the stability of the two loops separately. Since we are dealing with a physical system, if any of the loops is unstable, then the system is unstable even if one loop is not in your input-output transfer function. You are assuming that it is the first loop that is unstable. With my article, I was trying to argue that it is in fact the second loop that causes instability. If the second loop is unstable, you can stabilize it by changing the first loop. This is because the two loops are linked together. The way they are linked probably depends on the CMRR, meaning that changes in your first loop will affect the second loop much more than the other way around. No matter if you change the forward or the feeback loop of the first loop, you are in a way low-pass filtering the feeback signal so that the second loop is less likely to see any high frequency input. This works fine, of course, but it changes your input-output transfer function of the amp. >I don't, however, think that this [CM-]feedback loop has much effect on >high-requency stability. However, I may try it and see. Oscilliscope >pictures are easy to take, even if they do cost $1 each. (Thank you, >Polaroid. I've already spent more on film than on the scope camera!) > By putting a cap at the base of the transistor, I was trying to stabilize the second loop. According to my rough calculations, the second loop has a gain of somewhere between 10 to 20 dBs. The high frequency poles I could identify are: 1) Due to the base-emitter capacitance 2) Due to the collector-emiiter capacitance 3) Double pole due to the two plate-cathode capacitance The double poles are in the 500Mhz region, while the two other poles are in the low Mhz regions. What is probably more important is that the two lower poles are probably stacked not too far from each other. I believe this is why the loop was unstable. The cap now places another pole at the very low frequencies, so that by the time we reach the other poles, the loop gain is way below 1. Of course, I opted for an overkill solution. It is probably safe to place the pole somewhere above the audible frequencies. Then, you still have common-mode feedback over the audible range, and it should change little in the sound. Since this mod stabilized my amp, it is reasonable to assume that the common mode feedback loop is to blame. To prove my point, if the instability was in the first loop, any changes I made in the second loop has little effect on the first loop. Also, remember that I did remove the cap across the global feedback resistor. So, if the cap at the transistor killed the ringing, it has to be instability in the second loop. I think it won't be easy to find out exactly how stable the CM-feedback loop is. To do that, the only way I can think of is to break the loop and inject a signal at the one end, then observe the gain phase relationship at high frequencies. Of course, any instability should show by observing the input output relationship, but it is a less optimal solution. >I wonder if bypassing the current source with a capacitor could be >counter-productive. The reason is that this adds a third time >constant at low frequencies. The bypasses on the cascode grids are >the first, the coupling capacitors are the second. > >Much of the virtue of the Dynaco and Curcio designs over the classic >Williamson circuit is that they have superb low frequency stability. >The Williamson had three time constants (in addition to the output >transformer's contribution, of course), and was notorious for low >frequency stability problems. (The softer power supply didn't help. >It was also notorious for high frequency stability problems, that were >transformer and layout dependent.) > >Adding a third time constant to the Curcio Stereo 70 may have negative >effects on low frequency stability. (Of course, 100 uF is so large >that it may be adequately staggered from the others.) Still, since >the lows of the Curcio 70 are so great, I at least would be loathe to >do anything to compromise them. > >Stability problems at low frequencies certainly could cause shifty >imaging. > I believe a mistake was made here. To talk about time constants is confusing, because it doesn't tell you whether it is a pole or a zero. In order to have flat midband gain, any low frequency pole has to have a even lower frequency zero. From the root locus, one can see that the poles will move towards the zeros, until the zero and pole cancels. It is quite difficult to make the poles move to the right half plane before it cancels with the zero. Furthermore, we are dealing with two loops. The added pole in the second loop does not mean we automatically get a pole in the first loop. The pole could turn to a zero, or the pole will be accompanied by a zero, or the pole might not even show up in the first loop due to pole zero cancellation. >The advantage of this is rather simple, it improves the symmmetry of >the input stage. Unless you are willing to match the driver tubes >with a curve tracer, you're not going to get great balance. This type >of phase splitter, unlike the split-load one, suffers imbalance if the >tubes don't match. (By the way, the screen fedback in the stock >Stereo 70 is there to compensate for the imbalance of the split-load >phase inverter at high frequencies. It's negative feeback, but only >from one side of the push-pull stage.) ....some stuff deleted.... >Thus, by going to a static bias on the current sourse, the balance of >the input stage will be reduced. This hurts two ways: > >1. Increased second harmonic distortion due to the imblance. > >2. Less effective feedback because the two inputs are not of equal >(but opposite) gain. This is probably an order of magnitude less >significant than (1). > >Of course, second harmonic distortion is very nice sounding. (Look at >the Ongaku amplifier.) This can be a matter of taste. Also, the >recent 40W class A amplifier uses a fixed current source for it's diff >amp. I'm not claiming either way is right or wrong. > It's tremendously easy to reduce distortion by applying more feedback. Just recently, I dug up an article where the author took a cathode follower, added a gain stage before that and applied feedback all around it. What he got was a "cathode follower" with output impendance of 5 Ohms, positive peak current drive of 200mA, and THD well below 0.01% even for 100 V RMS output. He used it to drive the grids of a 807 triode up to +40V. Well, fact is that today there are no commercially available all tube pre-amps with such an output impedance. So, my approach to feedback is that whenever I can live without it, I try to get rid of it. >... acheiving stability on the Curcio will help this. However, there >are so many other benefits (lows, detail, imaging specificity) to the >Curcio mod, I'm glad I've done it. Someday it will be stable, and >sound even better. The amp sounded detailed when I first listened to it, but I think that is because it runs quieter than the stock ST-70. I think the original Curcio mod obscures some low level information to achieve this quieteness. Also, regarding image specificity, at first I also thought it was okay, but now I believe with CM-feedback, it is way better. ------------------------------------------------------------------------ Johannes S. Chiu ta-jc155@cunixa.cc.columbia.edu Electrical Engineering Columbia University ------------------------------------------------------------------------ ------------------------------ Subject: Re: Curcio ST-70 Date: Wed, 24 Feb 93 18:10:33 -0500 From: bill@verdix.com There was a discussion of CM feedback regarding the Curcio amp. In my opinion, CM feedback was of minor or no concern in the design of the amp. What Curcio wanted was a good current source for the differential pair without a negative supply voltage or an input cap. This puts only a few volts across the entire current source circuit. It's a bit tricky to build a stable current source in that case, so feedback is used. CM feedback may have been considered a bonus. Regarding low frequency stability, such problems are related to the power supply. The Curcio amp uses regulated supplies. Bill s ------------------------------ Date: Wed, 18 May 1994 14:55:00 -0400 From: "henry (h.) pasternack" Subject: "Daniel Curcio" ST-70 mod. Andre Francois writes: > On the subject of kits, there was a reveiw of several ST70 mods in an >old Audio Amateur issue. They recommended the Daniel Curcio mod above >all else. Mr. Curcio's name is Joe. "Daniel" is the name of a preamplifier he designed. Curcio's design philosophy is to use highly regulated plate and screen supplies. He is a proponent of very detailed, lean tube sound. His designs are sometimes criticized for being too bright or even harsh. I have heard reports of high-frequency instability in Curcio's regulators and amplifier circuits. He tends to run his components at the limit of thermal and electrical capacity. Nevertheless, his designs have also received a fair amount of praise. The Curcio circuits occupy somewhat of a niche in the domain of tube designs. I would say you should think carefully about what you're trying to accomplish before choosing one of his kits or mods. -Henry