This design is intended to be the more economical little brother of the Muscovite http://www.diyaudio.com/forums/analogue-source/213769-muscovite-6s3p-tube-phonostage.html. Like the Muscovite it offers high gain in the region of 54dB providing output levels comparable to most modern digital sources. (4mV - 5mVrms in @ 1kHz)
The prototype achieves RIAA accuracy of +/- 0.5dB over 20 - 20kHz with random parts from the scrap parts bin, with careful measurement and trimming accuracy of +/-0.1dB should be achievable.
This design is intended for the direct use of MM/HOMC type cartridges, and LOMC through step up transformers.
Only two tubes are used per channel excluding the PSU which could be tubed or a Salas SSHV2. No audio chokes are required.
Tube complement is a single 6ж9п (6Z9P) and 6c3п-EB (6S3P-EV) per channel. Passive equalization is used.
Edit 2014-04-13: If you are considering building this design have a look at the Muscovite Mini II thread as well.
The prototype achieves RIAA accuracy of +/- 0.5dB over 20 - 20kHz with random parts from the scrap parts bin, with careful measurement and trimming accuracy of +/-0.1dB should be achievable.
This design is intended for the direct use of MM/HOMC type cartridges, and LOMC through step up transformers.
Only two tubes are used per channel excluding the PSU which could be tubed or a Salas SSHV2. No audio chokes are required.
Tube complement is a single 6ж9п (6Z9P) and 6c3п-EB (6S3P-EV) per channel. Passive equalization is used.
Edit 2014-04-13: If you are considering building this design have a look at the Muscovite Mini II thread as well.
Schematic and Comments
The design is relatively straight forward although a fair amount of time was spent characterizing the 6Z9P pentode for this application, after the 6ж52п (6Z52P) was initially selected and discarded as being too inconsistent and requiring far too much current to achieve its high transconductance. (It works better as a triode IMO)
The first stage is pentode connected and provides ~44dB of gain. Bias of 1.4V is provided by a pair of 1N4148 diodes in series, the screen supply is resistively dropped to a target value of 150V, but vary from 140 - 160V depending on the tube. (More later) The tube is resistively loaded, and a 12.1K resistor able to dissipate at least 4W is required - recommend combining 2W metal films in series (preferably) or parallel to get the required wattage and value.
The second stage is a 6S3P triode with fixed bias at ~ -1.2V provided by a single lithium iron AA cell (V3 in the schematic) per channel. (Life of battery measured in years.) Gain is approximately 30dB. A 4W 15K resistive provides loading for the output stage. (See first stage for resistor recommendations.)
EQ is passively implemented and is based on the Lipshitz single stage topology.
I have constructed and listened to a single channel version that both sounded good and matches LTSpice predicted performance both AC and DC quite well.
Remaining questions to be resolved revolve around whether or not some sort of channel to channel gain trim or selected tubes will be necessary.
Incidentally the tubes used are still quite inexpensive. I recommend the 6Z9P-E, and 6S3P-EV grades.
http://www.mif.pg.gda.pl/homepages/frank/sheets/112/6/6Z9PE.pdf
http://www.mif.pg.gda.pl/homepages/frank/sheets/112/6/6S3PEV.pdf
The design is relatively straight forward although a fair amount of time was spent characterizing the 6Z9P pentode for this application, after the 6ж52п (6Z52P) was initially selected and discarded as being too inconsistent and requiring far too much current to achieve its high transconductance. (It works better as a triode IMO)
The first stage is pentode connected and provides ~44dB of gain. Bias of 1.4V is provided by a pair of 1N4148 diodes in series, the screen supply is resistively dropped to a target value of 150V, but vary from 140 - 160V depending on the tube. (More later) The tube is resistively loaded, and a 12.1K resistor able to dissipate at least 4W is required - recommend combining 2W metal films in series (preferably) or parallel to get the required wattage and value.
The second stage is a 6S3P triode with fixed bias at ~ -1.2V provided by a single lithium iron AA cell (V3 in the schematic) per channel. (Life of battery measured in years.) Gain is approximately 30dB. A 4W 15K resistive provides loading for the output stage. (See first stage for resistor recommendations.)
EQ is passively implemented and is based on the Lipshitz single stage topology.
I have constructed and listened to a single channel version that both sounded good and matches LTSpice predicted performance both AC and DC quite well.
Remaining questions to be resolved revolve around whether or not some sort of channel to channel gain trim or selected tubes will be necessary.
Incidentally the tubes used are still quite inexpensive. I recommend the 6Z9P-E, and 6S3P-EV grades.
http://www.mif.pg.gda.pl/homepages/frank/sheets/112/6/6Z9PE.pdf
http://www.mif.pg.gda.pl/homepages/frank/sheets/112/6/6S3PEV.pdf
Attachments
My Prototype
Having spent days characterizing the 6Z9P I decided that before I invested too much time and money in the effort I had better build a prototype to determine whether or not the performance predictions were accurate, and more importantly with a pentode front end that the end result was worth listening to. (It was)
The Muscovite Mini is intended to be much less costly to build than the Muscovite while not being a huge compromise in performance. It is a few dB noisier, but offers generally similar performance.
I had a friend over today who thought it was more open and dynamic than the reference D3A/5842 gyrator pre-amp used to provide the other channel.
I now just need to order an appropriate chassis and build a stereo version where remaining questions about 6Z9P matching issues will be answered. (Are matched tubes or a provision for gain trimming required or is it easy to find pairs that just work?)
The monophonic prototype was build on an old piece of one of our kitchen cabinets. Two critical areas have some shielding - underneath the RIAA network and underneath the RCA input jack and environs.
Having spent days characterizing the 6Z9P I decided that before I invested too much time and money in the effort I had better build a prototype to determine whether or not the performance predictions were accurate, and more importantly with a pentode front end that the end result was worth listening to. (It was)
The Muscovite Mini is intended to be much less costly to build than the Muscovite while not being a huge compromise in performance. It is a few dB noisier, but offers generally similar performance.
I had a friend over today who thought it was more open and dynamic than the reference D3A/5842 gyrator pre-amp used to provide the other channel.
I now just need to order an appropriate chassis and build a stereo version where remaining questions about 6Z9P matching issues will be answered. (Are matched tubes or a provision for gain trimming required or is it easy to find pairs that just work?)
The monophonic prototype was build on an old piece of one of our kitchen cabinets. Two critical areas have some shielding - underneath the RIAA network and underneath the RCA input jack and environs.
Attachments
Didn't do the Arthur Loesch trick for the coupling cap between the first and second stage? Also, why is there R3 and R4? At first glance, it looks like you'd only need one of them.
Not sure which Arthur Loesch trick you are referring to, he had several. This topopology maximizes available gain, and yeah I would admit I can throw away several dB without much of an issue other than on SNR. Unfortunately it also maximizes the required value of that coupling capacitor. Re tricks: I'm not sold on the benefits of applying large dc voltages to film caps, no measurable or audible benefit demonstrated to me by its proponents. One of his not so good ideas was minimizing the value of both capacitors and maximizing resistor values in the RIAA circuit - demonstrably (audibly and measurably) noisier. You do need to minimize interactions between the RIAA and the driving amplifier stage when using cascode, pentode or wimpy high rp triodes as the varying load impedance modulates the gain - and yes you can fix most of this in the EQ. (Not my favorite approach)
R3 and R4 form a voltage divider for the bias voltage. Lithium iron batteries when fresh have a cell voltage of 1.9V and I wanted a bias voltage of -1.2V..
He puts the RC coupling after the EQ network.
Thanks for the resistor explanation; that had me scratching my head a bit.
Yes, you could actually bias the second stage with an IR led in the cathode. The battery gives you a little flexibility in that it is fairly easy to change the operating point either by installing a regular alkaline battery or tweaking the resistor values. The reduction in gain/increase in rp would be insignificant, and the cathode current is >10mA so an LED should work fine. (Note I have not accessed the potential effect of large variations in cathode signal current on bias LED performance so there could be a linearity hit.)
Yes I remember that, the argument there was to apply DC to the EQ caps polarizing them, and to allow the use of a much smaller coupling cap. Unfortunately in most instances this implementation throws away 15% or more of the available gain (assuming 200K and 1meg resistors) depending on the ratio of the first RIAA resistor in network and the grid resistor. (Implicit voltage divider, and some of the resistor values he used in the RIAA network are much larger than anything I would consider)
While Arthur and I were friends we did not always see eye to eye and some design choices he made led to significantly poorer noise performance - this is one of them. (He seemed not to mind significant levels of hum and hiss, and my speaker system is far more efficient than what he was using so noise is a big issue to me.) He felt that capacitors were inherently the most flawed passive components and strived to use the smallest values he could get away with based on the premise that lower value capacitors have superior electrical performance.
Tube Regulated Supply (Single)
Here is a single tube regulated supply intended to power two channels. It is quite similar to a single channel supply for the Muscovite. The plate supply transformer is a readily available and inexpensive EI type (XPWR228-120) from Edcor.
The supply is capable of supplying 60mA at 300V. Broadband noise and ripple on the output should be
<1mVpp.
Probably not a bad idea to select a 75V zener at around 3mA goal being to select one within +/- 1% of nominal. Alternately if you want to reduce the contribution of zener noise to the supply output you can stack five 15V zeners in series or if you feel extreme thirteen 5.6V zeners and three 1N4148 - you may need to increase the current through the zeners for optimum performance - easiest way to do this would be just to use add a resistor between the zeners and the 300V output. Additional current should be limited to <=10mA to avoid frying the power transformer. (Or use a different transformer) Again make sure to verify that the output voltage is correct regardless of the reference you use.
A separate DC filament supply will be posted later.
Note: The Salas SSHV2 is an excellent supply option, though not one that I personally plan to pursue. I am currently using one to power the analog section of my diy DAC where it works great. You'd set up the CCS for ~70mA and the shunt for 300V, be sure to follow the recommendations for heat sinking. For set up a 25W 5.6K (6K) resistor is probably a good idea. Here: http://www.diyaudio.com/forums/group-buys/206033-gb-salas-sshv2-regulator.html (Note the current group buy is closed.)
http://www.diyaudio.com/forums/group-buys/206033-gb-salas-sshv2-regulator.html
Here is a single tube regulated supply intended to power two channels. It is quite similar to a single channel supply for the Muscovite. The plate supply transformer is a readily available and inexpensive EI type (XPWR228-120) from Edcor.
The supply is capable of supplying 60mA at 300V. Broadband noise and ripple on the output should be
<1mVpp.
Probably not a bad idea to select a 75V zener at around 3mA goal being to select one within +/- 1% of nominal. Alternately if you want to reduce the contribution of zener noise to the supply output you can stack five 15V zeners in series or if you feel extreme thirteen 5.6V zeners and three 1N4148 - you may need to increase the current through the zeners for optimum performance - easiest way to do this would be just to use add a resistor between the zeners and the 300V output. Additional current should be limited to <=10mA to avoid frying the power transformer. (Or use a different transformer) Again make sure to verify that the output voltage is correct regardless of the reference you use.
A separate DC filament supply will be posted later.
Note: The Salas SSHV2 is an excellent supply option, though not one that I personally plan to pursue. I am currently using one to power the analog section of my diy DAC where it works great. You'd set up the CCS for ~70mA and the shunt for 300V, be sure to follow the recommendations for heat sinking. For set up a 25W 5.6K (6K) resistor is probably a good idea. Here: http://www.diyaudio.com/forums/group-buys/206033-gb-salas-sshv2-regulator.html (Note the current group buy is closed.)
http://www.diyaudio.com/forums/group-buys/206033-gb-salas-sshv2-regulator.html
Attachments
A suitable chassis has been purchased on eBay for the phono pre-amplifier, and I will do the same for the power supply when I decide exactly what I am going to use to power it long term.
Simplicity is attractive so I may in fact go with the single EL34 regulator presented a couple of posts back.
The filament supply should be clean and well regulated DC, the supply will be referenced to ground through a pair of 330 resistors forming a voltage divider across the filament supply - since the cathodes are essentially grounded there is no significant issue with noise pick up, and the filament of the 6S3P does not like being very positive relative to the cathode at least according to the datasheet so we won't go there...
Total filament current will be on the order of 1.2A so an LT1085 should handle this with no problem. An LM317 should be OK although it would be operating continuously at 80% of its rated current which makes me a little uncomfortable. A toroid with an 8 - 9V secondary at a couple of amps should be adequate. (20 - 25VA)
Simplicity is attractive so I may in fact go with the single EL34 regulator presented a couple of posts back.
The filament supply should be clean and well regulated DC, the supply will be referenced to ground through a pair of 330 resistors forming a voltage divider across the filament supply - since the cathodes are essentially grounded there is no significant issue with noise pick up, and the filament of the 6S3P does not like being very positive relative to the cathode at least according to the datasheet so we won't go there...
Total filament current will be on the order of 1.2A so an LT1085 should handle this with no problem. An LM317 should be OK although it would be operating continuously at 80% of its rated current which makes me a little uncomfortable. A toroid with an 8 - 9V secondary at a couple of amps should be adequate. (20 - 25VA)
Kevin, the 6BC4 still sels for a dollar at ESRC....what do you think? can they be used here as well?
http://frank.pocnet.net/sheets/049/6/6BC4.pdf
http://frank.pocnet.net/sheets/049/6/6BC4.pdf
No idea, not familiar with this type. You'd be blazing a new trail. It looks promising enough, transconductance is about half so it may be a bit noisier, rp and mu are quite similar.
I'm not really satisfied at this point with the apparent low frequency noise performance of the 6ж9п / 6Z9P.. (FFT measurement) I need to build the pre-amplifier in a proper chassis and with a proper low noise power supply to determine whether I have a problem with 1/F noise or not.
Fall back plan is either the 6688 or D3A either of which would slot in with relatively trivial changes.
I'm not really satisfied at this point with the apparent low frequency noise performance of the 6ж9п / 6Z9P.. (FFT measurement) I need to build the pre-amplifier in a proper chassis and with a proper low noise power supply to determine whether I have a problem with 1/F noise or not.
Fall back plan is either the 6688 or D3A either of which would slot in with relatively trivial changes.
Remembering, of course, that your PSR is something between zero and pathetic. That's one reason I abandoned pentodes for MM stages, despite their attractively low input capacitance- the design problem just got transferred to the power supply.
I've got the low noise supply design thing down pat, I'm much less convinced that I yet understand all of the pitfalls of using a pentode in this application. My reasons for this were the high realizable gain - here about 44dB and the non-issue with miller capacitance. Concerns actually were partition noise (seemingly not a problem) and inconsistent sample to sample transconductance, and hence gain..
Transconductance is indeed likely to vary, both at g1 and g2. Likewise, the current ratio between plate and screen. Since the screen is an input, any noise there is amplified. And since CMR at the plate is essentially nonexistent, that's added in as well. It's doable (else cascodes wouldn't work, either!), but it's not trivial (yes, I know you know all these things😀).
Interestingly enough I decided after a quick simulation that shifting the op point of the 6ж9п / 6Z9P by replacing the two diode bias string with a red led having a vf of 2.0V was a good idea. This reduced the plate current by roughly 30% and resulted in a measured reduction in LF noise of 10dB.. Overall gain was not significantly affected.
I've also changed the value of C7 back to 10uF as the smaller value resulted in excessive LF roll-off causing a significant response error between 20 - 50Hz.
Note that the prototyping technique (building on kitchen cabinet scraps) and the not quite optimal PSU do exert a significant penalty.
In an ideal world the LF noise floor at 20Hz would be ~20dB worse than at 1kHz by virtue of the RIAA equalization, this design is about 5dB worse than that due to 1/F noise (and other noise mechanisms) in the pentode and a suspected contribution from the power supply. Even in the unimproved version of the design listening I was not aware of any unusual issue with the LF performance - yes it was significantly noisier than the Muscovite, but not exceptionally or obtrusively so.
I've also changed the value of C7 back to 10uF as the smaller value resulted in excessive LF roll-off causing a significant response error between 20 - 50Hz.
Note that the prototyping technique (building on kitchen cabinet scraps) and the not quite optimal PSU do exert a significant penalty.
In an ideal world the LF noise floor at 20Hz would be ~20dB worse than at 1kHz by virtue of the RIAA equalization, this design is about 5dB worse than that due to 1/F noise (and other noise mechanisms) in the pentode and a suspected contribution from the power supply. Even in the unimproved version of the design listening I was not aware of any unusual issue with the LF performance - yes it was significantly noisier than the Muscovite, but not exceptionally or obtrusively so.
Attachments
Hi Kevin,
The first mic amp I built used 6j9p-e in triode mode. I had exactly the same experience noise-wise as you when I increased the bias from a little over a volt to over 2 volts. I found it difficult to find really matched ones( I burned in about 20) and even then there was significant change over a period of many hours. If you tap the chassis you will certainly get a silvery shimmer but I didn't hear any microphony in actual use. Nevertheless, I use it in the 2nd stage of my (differential) phono amp, finished 5 years ago and I have no urge to change. First stage uses...........6s3p-ev, my favourite small signal valve!
One of my reasons for using differential was to cut down Miller capacitance.
The first mic amp I built used 6j9p-e in triode mode. I had exactly the same experience noise-wise as you when I increased the bias from a little over a volt to over 2 volts. I found it difficult to find really matched ones( I burned in about 20) and even then there was significant change over a period of many hours. If you tap the chassis you will certainly get a silvery shimmer but I didn't hear any microphony in actual use. Nevertheless, I use it in the 2nd stage of my (differential) phono amp, finished 5 years ago and I have no urge to change. First stage uses...........6s3p-ev, my favourite small signal valve!
One of my reasons for using differential was to cut down Miller capacitance.
Last edited:
Sounds like we have similar tastes. 😀 I used the 6S3P-EV in my Muscovite design and believe that is my best sounding/performing design to date.
I have 10 of the 6ж9п / 6Z9P in my stash currently and found of the 8 I have tested that they fell into two groups of transconductance range.
I'm finding that I still have some issues with the FFT hardware - in this instance generated by the computer that controls it all. I was having trouble setting the output level of the generator to low values, and could see the amplitude changing when observing on a scope, none of my DMM however would report values below 10mV - turns out to be RF from the computer that is confusing them. (The 2002 is capable of accurate measurements below 1mVrms.) I think the phono stage is rectifying some of this noise as well and that is what is generating the line frequency noise spectra amongst other things.
I have 10 of the 6ж9п / 6Z9P in my stash currently and found of the 8 I have tested that they fell into two groups of transconductance range.
I'm finding that I still have some issues with the FFT hardware - in this instance generated by the computer that controls it all. I was having trouble setting the output level of the generator to low values, and could see the amplitude changing when observing on a scope, none of my DMM however would report values below 10mV - turns out to be RF from the computer that is confusing them. (The 2002 is capable of accurate measurements below 1mVrms.) I think the phono stage is rectifying some of this noise as well and that is what is generating the line frequency noise spectra amongst other things.
The change in biasing that I implemented was due to raising the output voltage of the PSU overnight from one day of a recording session to the next. My job on this (classical piano CD) was in a producer capacity but the chief engineer wanted to set some of my gear up to run alongside the usual Neumann etc. stuff for his own interest. He did not know that I had reconfigured the power supply but came out of the control room to ask me what I had done as the noise reduction had been quite marked. Since then, I have always biased 6j9p-e at over 2 volts with about 150V anode to cathode.
It's a very nice looking valve and seems terribly well made.
It's a very nice looking valve and seems terribly well made.
- Status
- Not open for further replies.