Member
Joined 2009
Paid Member
I was looking at the Sowter 8985 (50 H, 260 Ohm, 40 mA max. current). But there are others that would work, too (Electraprint, Lundahl, etc.)
Why another gyrator? There is no gyrator in the circuit. Are you referring to the constant-current source (CCS) in the plate load of the driver tube? A CCS (or a gyrator) might be nice in this position because it's smaller, cheaper, more linear and less susceptible to EMI than a real choke. However, a CCS (or a gyrator) would require a substantially higher B+ to accommodate for the voltage swing at the 45 plate. The advantage of a real choke is that it can store energy, which allows it to swing the AC signal higher than the B+ at the 45 plate. A CCS (or a gyrator) can't do that. I prefer to keep the B+ below 500 VDC or so.
yes, the gyrator should provide a firm DC reference for the CCS, which itself is actually now a mu-follower. There are more than one way of creating this gyrator. For DC coupling, going from a simple cascoded CCS (as you draw in post #59) to using a gyrator reference is a very nice improvement. If you measure the DC on the plate for the cascoded CCS circuit vs. gyrator circuit, you will find the gyrator is far more stable. I don't need to tell you how useful a firm DC reference on the grid of your 45 tube will be.
The way Ale Moglia builds a gyrator is to use another CCS (the smaller LND150) on top of a resistor load. Simple yet effective. Fluctuations in B+ don't change the DC reference in this circuit, which is lovely if you are doing DC coupling. The only downside to Ale's gyrator is that you might get a drift if the LND150 gets warm (ie. you place the LND150 too close to the heat sink of the upper DN2540 or STP3NK60ZFP, etc). I have experienced this then had to change my layout...
You can also entertain the idea of using the lower 'FET as a mu-follower.
Ian
Last edited:
I just read Ale Moglias description of his "gyrator" design again. And again. And again. And... there is light at the other end of the tunnel!
Here's my way of looking at the circuit (just in case I'll forget how it works I can come back here). The section made up from M1+M2+R1+R2+P1 is a small cascode CCS. The current out of this CCS flows through R4 to GND, so R4 forms a constant DC voltage reference (ignoring any thermal drift). This constant voltage is applied to the gate of J1 via R6 and sets the DC voltage the source of J1. D1 is a protection diode that doesn't do much during normal operation of the circuit, so I'll just ignore it for now.
The capacitor C1 does not feed back any DC to the gate of J1, so the DC voltage at the output of the whole circuit (i.e., the DC voltage at the source of J1) is controlled only by the voltage across R4. M3 decouples J1 from the raw B+ and drops most of the raw B+ voltage.
If the load (=tube) varies at AC and the impedance of C1 is small compared to R4, the voltage at the source of J1 is fed back to its gate; the J1 gate voltage is not controlled by the DC reference (R4) anymore. This means that J1+C1+R7+M3+R3 now form a cascode CCS.
In other words:
(A) At DC (when the impedance of C1 is larger than R6), the circuit controls the voltage on top of R7 to a fixed value (set by R4/P1).
(B) At AC (when the impedance of C1 is less than R6), the circuit works as a cascode CCS that sets the current through the tube to a fixed value (set by R7).
Please let me know if something is not right... anyways, applying all this to the 45 direct coupled amp, the first feature (A) is nice because it allows setting the voltage on top of R7 and thus provides some control over the voltage at the plate of the driver tube, i.e., the bias voltage of the 45 power tube. The second feature (B) is nice because the fixed current through the driver tube ("infinite load impedance") optimizes the linearity of the driver stage. (I know that you tried to explain this to me before, but it took me a while to understand -- most likely because the "gyrator" term is wired to something else in my brain).
Last edited:
You’re right. The circuit is a hybrid mu-follower. Perhaps the confusion is that we usually call it “gyrator” due to the equivalent inductor-like response of the CCS load. As R7 is “bootstrapped” in AC provides a high impedance load to the triode which gets the most linear response out of the stage.
The low-impedance output is achieved when the output is taken from the mu node. The most important thing for me is that: it sounds really good as is very neutral and doesn’t add any unwanted harmonics by the driver.
Also you can degenerate the cathode bias resistor without compromising much the stage gain and get even better linearity. This is good when driving large volts like in your case. Recently I posted some examples of very nice triodes (6e5p, 6e6p, 6j49p-dr, 6j52p and others) which perform brilliantly at 200Vpp. These are good examples to drive a 45 like in your case.
Ale
The low-impedance output is achieved when the output is taken from the mu node. The most important thing for me is that: it sounds really good as is very neutral and doesn’t add any unwanted harmonics by the driver.
Also you can degenerate the cathode bias resistor without compromising much the stage gain and get even better linearity. This is good when driving large volts like in your case. Recently I posted some examples of very nice triodes (6e5p, 6e6p, 6j49p-dr, 6j52p and others) which perform brilliantly at 200Vpp. These are good examples to drive a 45 like in your case.
Ale
The protection zener (or back to back zeners, depending on the circuit) are there to protect the gate during turn-on and turn-off. Otherwise I think you now understand why this circuit is especially attractive for 'classic' direct coupled designs.
M3 dissipates the most energy, and is capable of dropping a LOT of volts, depending on the current, so long as the heat sink is big enough. I hook up the input/driver stage directly to the B+, with no RC dropping network. Its just not needed imho.
Check out mouser part nr. 532-529902B25G for an Aavid thermalloy part that I find quite decent. Aavid provides thermal resistance values for their parts, which is very useful for calculations.
Due to the exceptionally high effective load, gain is nearly equivalent to the mu of the valve/tube. Internal resistance of the valve/tube pretty much defines the output impedance at the plate if you decide to use it instead of the mu-follower node. Its important to note that the effective load declines with frequency, but since a cascode is used, it is still quite impressive at 20kHz, and higher.
If you use the mu follower node, its wise to make C1 a decent quality metalized polypropylene capacitor, soviet POI, etc. If you direct couple at the plate, C1 can be cheaper (and smaller) film cap.
I find the mu follower node output the best, but direct coupling off the plate is surprisingly good if the input/driver has low enough plate resistance.
Ian
Well, almost. What's the preferred method to set the operating point of the driver tube? In my previous schematic with the simple cascode CCS, the idea was to set the (DC) current by the CCS, and the grid-to-cathode voltage fixed by the battery bias. If I use the "Ale Moglia load" to set the plate voltage, the fixed battery bias will force the DC current according to the tube characteristics (whatever they are). Would you leave this, or is it better to do the typical autobias with a resistor instead of the battery to allow some bias compensation if/when the tube characteristics drift away during aging? If not bypassed, the resistor would provide some local feedback reducing the gain a little bit and improving the linearity (as mentioned above by Ale). A bypass capacitor would remove the bias resistor from the AC, but would be in the audio signal path. I just seem to like battery bias for no good reason (except that it worked really well in another amp I built recently).
Yes, I saw these. Really nice and linear curves. However, some of those tubes would need biasing at a rather low grid voltage to work at around 130 V plate voltage (to keep the B+ of the whole amp down). I usually try to stay away from near-zero or even positive grid voltages during high signal swings in order to avoid grid current. The ECC88 seems to allow a bit more headroom with this. I also read that these tubes tend to be microphonic. Do you have any experience with this?
Hi Matthias
Regarding your chosen 130v plate voltage and grid voltage/bias question:
One of my favorite indirectly heated triodes is the EC8010. Also look at triode strapped pentodes such as E810F, E280F and the DA3. If you want less gain, there are even more triode strapped pentodes you might consider.
To prevent problems with microphonics (esp. for E810F), ferrite beads work a charm. I didn't find the need for them except for the E810F though. Unfortunately, I do not have many Russians ones to try. There is a lot of choice out there though.
Ian
Regarding your chosen 130v plate voltage and grid voltage/bias question:
One of my favorite indirectly heated triodes is the EC8010. Also look at triode strapped pentodes such as E810F, E280F and the DA3. If you want less gain, there are even more triode strapped pentodes you might consider.
To prevent problems with microphonics (esp. for E810F), ferrite beads work a charm. I didn't find the need for them except for the E810F though. Unfortunately, I do not have many Russians ones to try. There is a lot of choice out there though.
Ian
Last edited:
I find that a good non-inductive cathode resistor works perfectly fine. I'm using mills or caddock, but to be honest any decent quality nichrome film is perfectly acceptable.
For a gain similar to your ECC88, look at the triode strapped E55L for superb linearity and TONS of headroom. Also ridiculously low internal resistance, so its an ideal candidate for direct coupling. Unfortunately no longer inexpensive. Maybe there is a Russian equivalent.
Tektronix used them in oscilloscopes...
Tektronix used them in oscilloscopes...
Last edited:
If grid current starts at -1 V(gk) and allowing a grid signal swing of +/-2 V, the grid bias should be at least -3 V(gk). More is better, especially if the audio source provides more signal amplitude. Looking at the curves of these tubes (and some others like the 6e5p and its siblings, or the E55L), I don't see how this could be achieved at 130 V(plate) or so.
I just looked at the 6H30 curves. This could be biased at 130 V(plate) with -5 to -6 Vgk. Or maybe even at 100 V(plate) with -4 Vgk. I have no experience with this tube, but this looks interesting.
In fact, we don't need all the gain provided by the types you mentioned. Which ones would you suggest?
I can see how ferrite beads suppress high-frequency electronic noise. But microphonics (mechanical noise)? How?
Matthias, do you really think you need so much bias? I don't think so. This stage is running in Class A. I don't think you need to bias on top of the most positive potential grid swing. Plus consumer audio is something like 1.5v rms so has a peak at of about 0.9V
I am usually happy with a bias point around 2v or so. I know people who are happy with even less than this. Someone can call me out on this but distortion measurements I did previously were pretty decent...
What is your input signal like? Maybe it would be higher if you are using a pre-amp. I don't use a pre-amp.
Pete Millet did a list a while ago. Its worth checking out: High Gm driver pentodes
There are seriously more. Or if the internal resistance is not a problem, just stick to triodes. Usually I am looking for the ones that can do higher gain. The C3M is one that is highly regarded for low gain (in triode).
Correct. Then I must revise my comment and say that I only experienced RF interference with the E810F. It was solved with ferrite beads.
If you look at my builds, you will see that I always have the PSU on-board. I dislike umbilical cords. Nonetheless, I have not experienced problems with microphonics. Maybe I am just lucky? Or maybe especially careful with how I design and build my PSU's? hard to say...
Last edited:
Member
Joined 2009
Paid Member
Maybe I just said 'peak' in post #78 ? not peak-to-peak...
Anyway, I think cd line level is less than 1.5vrms, but for the sake of correctness, lets look at this closer
Peak-to-peak = rms * root-of-2 ≈ 1.5 × 1.414 = 2.121
The 'peak' of this peak-to-peak value of 2.121 is 2.121/2 = 1.06V which is indeed a little more than 0.9V
Of course this is all for a sine wave, and many of us know that transients can cause greater peaks. But the point is that this stage is running deep in Class A, so the incoming AC signal is not changing DC bias one little bit. A sudden transient is not causing a sudden grid-leak.
Maybe we need a real engineer to weigh in on this though.
Ian
Last edited: