|An experimental RIAA preamp using no capacitors in the direct signal path
***Revised with 7721 (D3A in Europe) triode connected output tube. Sept 2004
Background and Purpose
The purpose of this report is to describe an experimental RIAA phono preamp that uses several different concepts to derive a high quality RIAA compensated phono preamp. Traditional RIAA compensation is done by way of resistors and capacitors to shape the frequency response. There are alternatives that use L-C-R networks to derive this compensation. (For instance Tango EQ-600 is an LCR network that provides RIAA compensation). In this preamp, we will use L-R only networks. In addition, several different concepts are applied:
The resulting preamp weighs in at 18 pounds for the preamp and an additional 5 pounds for the power supply. It's big!
There are many possible approaches to deriving RIAA compensation. The most common is to use resistors and capacitors to derive the needed poles and zeros. With "common" values, it is practical to get to a very good realization of the RIAA curve. Moreover, as the tube characteristics change over age, it is easy to pick realistic values for this circuit to be relatively immune to the change. (One problem ignored by many people is the fact that as a tube ages, the transconductance decreases and the plate resistance increases. The decrease in transconductance we can deal with... turn up the volume control. The change in plate resistance alters the pole-zeros of the RIAA in degrees that vary with the design. Low impedance RC circuits are more affected than high impedance circuits.)
Another method of deriving RIAA is to use a "bulk" L-C-R RIAA network. The advantage of this approach is that if either the source or the load resistance is fixed, the frequency response of this network does not change with variation in the "other" resistance. So, for instance, if the load is fixed, the driving impedance of the tube can change and nothing much happens to the response. The disadvantage of this approach is that for practical values, relatively low impedances are needed. For instance, the Tango EQ-600 uses 600 ohm source and load impedance, and imbeds inductors of 45mH and 1.8Hy. Note that the inductors have to be made of relatively special material, since the inductance value can not be allowed to change with signal level. Most normal inductors vary A LOT with change in signal level, but we normally don't care, since we're usually concerned that the inductance is "at least" a certain value, not "exactly" a certain value. Also note that if this same circuit were scaled up to more common vacuum tube impedances, the values get much less practical. For example, if the Tango circuit were scaled to 10,000 ohms, the inductors become 750mH and 30Hy. The 750mH is practical, but the 30Hy implementation becomes difficult. Certainly 30Hy is easy to obtain, as long as you're looking for "at least" 30Hy, but as soon as you demand "exactly" 30Hy over a wide signal level range, this is much more difficult.
Another approach to RIAA compensation is to use L-R circuitry. This is even more difficult to implement than the L-C-R approach, as you still have the same inductor problem mentioned above, but you also have the change in response due to changing impedance level to deal with. Nevertheless, this is the approach taken in this design. The inductors used are Lundahl parts available from Benny Glass in Europe. These nice parts are mu metal based and maintain constant inductance with signal level. The part(s) each have two windings that can be used in series or parallel. One part provides 45 or 180 mH and the other provides 450mH or 1.8 Hy. In the design shown here, I am using only the one type: the 450mH/1.8H part. Note that the values of these inductors *could* be used to implement a L-C-R design similar to the Tango.
The 2123Hz pole (and an ultrasonic zero) is derived at the secondary of the interstage coupling transformer (see schematic below). The 50Hz pole/500Hz zero is derived at the secondary of the output transformer.
Tubes used in this design
The 6GU5 is a "TV RF amplifier" tube. It is very interesting in both triode and pentode connection. It is a remote cutoff device, but in triode connection is quite good when transformer coupled. I am running it at about 200 volts and about 10mA. The stage gain is slightly over 100 with the load I have used. (see schematic). The distortion in this mode is actually quite good.
Note that even at highest cartridge outputs, distortion is low. For instance, with a 50mV (!) MM cartridge, the output will be 5 volts from this stage (at the plate) and distortion only 0.2%. Also with only 0.5pF capacitance grid to plate, Miller effect translates this to only about 50pF, which is perfectly acceptable.
The 6GK5 is similar in transconductance, but somewhat lower in MU (75) so its plate resistance is only about 5k. Over the course of listening to this preamp, I was still unhappy with the output stage (see below) so I changed the output tube to a 7721 (D3A in Europe). This tube offers about the same mu as 6GK5 when triode connected, but it has higher transconductance and thus lower plate resistance. It has MUCH lower distortion. With these 2 devices in the experimental circuit, equivalent input noise (measured A weighted) is only about 0.3uV, which is actually low enough so that a MC cartridge can directly be driven with no transformer. (Although I have made provision for the traditional step up transformer in the design). This is a very quiet preamp.
The overall distortion of the preamp is not too bad:
But, when the output stage is altered to the triode connected 7721, the overall distortion is really low:
Positive Current Feedback
One of the ways that can be used to make the L-R RIAA compensation less sensitive to tube aging (and variation from device to device) is to provide transformer coupling and step the secondary way down. This burns gain but lowers effective impedance. With a potential gain of about 8000 in the 2 stages of this preamp, I can actually afford to step down the output quite a bit, even considering the 20dB loss that RIAA compensation needs as well. In the design, the interstage and output transformers are stepped down.
There is also another way of lowering AND compensating the output impedance, and that is by the use of positive current feedback. Take a quick look at the schematic just below. Note that there is a common path for the two cathodes to ground at D4. By coupling the cathodes together like this and properly phasing the interstage transformer, there is a small degree of positive current feedback introduced. Positive current feedback LOWERS output impedance. Since the system frequency response (RIAA) is highest at low frequencies, the feedback occurs only at these frequencies, and there is correspondingly no loss in resolution due to the feedback.
The feedback is through the common D4 / R11. The dynamic resistance of the diode is only a couple of ohms, so there is little feedback in any event. However, the dynamic resistance of the diode varies with current. As the current DECREASES, the dynamic resistance INCREASES. This is very fortuitous. As the tube(s) age, their transconductance decreases and so does the tube operating current. Normally, this means that the plate resistance increases, which would adversely affect the L-R RIAA compensation network. But, with the increased diode resistance, this increases the amount of positive feedback, and lowers the output resistance, compensating the change. Note that the 7721 now used as output tube has higher gm than the 6GK5, so that I had to change the value of R11.
In SPICE simulation, as I allowed the tubes to degrade to 75% of their initial values... essentially no change. Then allowed them to degrade to 50% of their initial values; less than 1dB change in gain, less than 1dB change in frequency response. Thus, overall, this circuit is immune to aging affects. (gee, I wish I were) ;-p
Note that ther are no capacitors in the signal path. In the early phase of design I had intended the LL1623 (a large output transformer) to be used as the interstage transformer. Unfortunately, there was too much variation in inductance with signal level, and the frequency response varied with level. Placing the 1660 as IT cured that. The reason for the 1623 at all was to take advantage of the low winding resistance. O well. I suspect 1660s can be used for both transformers, but R9 will probably need to be adjusted for proper RIAA compensation (due to the added transformer resistance).
Note that R4 provides an additional zero to either compensate for dropoff in transformer response at ultrasonic frequencies, or provide compensation for the so called 50kHz zero. In reality it is doing a little of both.
That odd A, B, C stuff on the input is actually relatively straightforward. For high level MM cartridges, connect the input to to C. For low level MC cartridges, connect the input to A and jumper B to C. This avoids switches/relays/added wiring in low level signal points.
The output impedance of this thing is less than 700 ohms (primarily R9. Due to the positive feedback, the effective impedance at the secondary of the transformer is about 50 ohms.
Since the secondary of the output effectively isolates the circuit, either unbalanced or "balanced" (read, unreferenced) output is provided. That's what S1 is doing... providing a ground reference if wanted.
Here's the measured frequency response of this critter:
The slight wobble around 10kHz is part of the 1623 characteristics.
With the 7721 in the output stage, the bias had to be altered slightly. An additional diode is in series with the cathode string to the output tube to provide about 2.28V bias. This causes the 7721 triode connected part to consume about the same current as the 6GK5. (actually, perhaps 1mA more). The higher gm and lower plate resistance required a change to the positive feedback resistor to 39.2 ohms (16.9 ohms is shown above with the 6GK5). The overall gain is about the same, and the lower plate resistance slightly improves the overall frequency response. Here's the response with the 7721 output stage:
Here's the power supply I used for this thing:
This thing is useable on either 120 or 240 volt, 50 or 60Hz mains. I used two of those relatively cheap (<20USD) toroid power transformers available from Digi-Key. I am told these parts are available worldwide. Since I intended for this to be part of a fully integrated preamp-lineamp-poweramp system, I wanted to be able to control the power supply without turning the thing on and off in the traditional manner. (Sorta like HP printers... they're on even when off). That's what the relay and diodes are doing at the output of the first reansformer.
The filament voltage is derived via two regulators. The first converts the (about) 23 volts to a regulated 18 volts. The second regulator converts the regulated 18 volts into 12.6 volts at even lower hum level. The purpose of this is indeed to remove even the last traces of hum. I "grounded" the + output of the 18 volt regulator (which is otherwise floating) so that I could also use the negative 18 volts for a bias source in other parts of the design (not phono preamp related).
The secondary of the first transformer backward feeds a second toroidal transformer. This provides a raw voltage of approximately (~) 270 volts for the high voltage needed. (Note that the ~270 volts looks like -270 volt in the picture above). A voltage regulator with two independent outputs at about 215 volt DC is then provided. I provided two jacks so that I could use the outputs in multiple places. In the full system, the second output will go to the main system power supply.
Here's links to PDFs of the preamp and power supply schematics. These are of much higher quality than the reduced size GIFs above which are used for descriptive purposes.