izzy wizzy wrote:
It is tempting to do this input stage for a whole bunch of reasons.
Although a cascode is similar to a pentode, can it be loaded down like a pentode? Fer instance, is there any chance of it driving a low Z EQ network of say about 20k? Maybe I ought to get LTSPICE out to have a fiddle.
Stephen
I would definitely try it - if nothing else it is educational ! - but I think it will be in my system , in the long term .
No, you can't load it down like a pentode, not from what I heard ; because I used it with the 20k network first, just to see. It sounded a bit muddy and toneless .
Ok, I will say that currently there is a little bit too much 2nd-harmonic , so I will try to refine the operating point, and maybe try the Mullard E180F tonight . I suppose there is about 16V across D-S at the moment .
If I find in the longer term it needs a bit more 'polish' , then I'm tempted to try JC's 'folded cascode Gm amp' circuit , which uses ideas from John Curl . James linked me to some writing by Curl today, and he seems like some genuine hero . He reckons one of the problems with transistors tends to be 7th-harmonic, maybe I'm finding an operating point with less 5th and 7th-harmonic, not third .
I could also try some of the dual-FET 2SK389's perhaps .
Bear in mind you need to buy a few, eg. 20 to 50 , and test and match them, as they vary hugely . Oh, or buy from someone offering matched pairs/sets on eg. DIY Audio or likewise .
"Once you find out ... the Circumstances ; then you can go out"
Yes...
I think Nick and James and I will try to find out more about this set-up .
Credit due to Nick who was the instigator of me doing this modification .
This document by John Curl is a great read , just worth reading the first 5-6 pages, if nothing else :
I can understand why JC likes him . I do warm to a guy who says "Class D ! Give me a break ! ..... When I really want to hear quality , I put on a record - sometimes one found in the trash of someone else . Digital essentially sucks ! When DVD or SACD can do 'Live Dead' then I will convert "
Likwise I need to understand the folded cascode circuit a bit better .
Now if I can get this front-end a little bit better ( not much needed ) then I will launch into the next bit , which is a 6N6P ( active loaded ) cathode-follower feeding Dave Slagle's 1.5k RIAA chokes . I have them upstairs ...
MJ
"Once you find out ... the Circumstances ; then you can go out"
OK, now with the Mullard E180F's back in there , they sound better than the Russian ones, with the current D-S setting of 16V . The sound is more sophisticated , a little less 2nd-H too, I think .
I will say that now, any difference in the midrange, compared to where I was with the Step-ups, could be easily put down to the shortcomings of the 100k RIAA network compared to the 20k RIAA .
The low bass continues to be excellent .
"Once you find out ... the Circumstances ; then you can go out"
IslandPink wrote:We are led to believe from the basic application of jFETS ( eg. http://www.tkhifi.com/div/Erno_Borbely_ ... icel_1.pdf ) that biasing the jFET above about 6V in this configuration should be plenty to ensure low distortion . He shows THD results as well, for various signal levels . However... there are often problems with using THD as the measure ; it is un-discriminating .
Today, I decided to add a preset pot to the voltage divider , adding extra R to the 6.8k in the voltage divider . I put in a 10k pot there, so I could try out voltages of 8V to 20V approx, across the FET .
My first effort this evening was with 14.5V across the FET .
Still with the Russian E180F's ....and ...
Now we are cooking with Gaz . I have not felt the need to adjust anything all evening, and have been playing music .
I was going to say, maybe there wasn't enough Volts across the jFet, possiblee
"No matter how fast light travels it finds that the darkness has always got there first, and is waiting for it."
I don't know yet where the best link to 'folded cascode' would be, but if you look at JC's blog item about the 'folded cascode hybrid Gm' circuit, then you'll see he's using the P-channel FETs from a negative supply, under the D3a. It must have something to do with balancing distortion products in the two devices .
The John Curl pdf linked just above is just magic - something well worth reading for anyone interested in audio . Print it off and take a week's holiday in the Lake District !
"Once you find out ... the Circumstances ; then you can go out"
Ok, here's where I am today ( 7-30pm - can't guarantee it much beyond this ) . Before you all cry out "where's the coupling cap" , I admit, I missed it off It's just AFTER the RIAA and before the bias voltage-divider for st. 2 . Of course, it is a Rel-PCU , value 0.1uF .
IslandPink wrote:Now if I can get this front-end a little bit better ( not much needed ) then I will launch into the next bit , which is a 6N6P ( active loaded ) cathode-follower feeding Dave Slagle's 1.5k RIAA chokes . I have them upstairs ...
The LCR route is what I was aiming for but ran out of money. I wanted to pop it in in place of the 20k EQ. The Slagle coils are the way I want to go.
Over the years, I've noted that once people get to an LCR phono, the vast majority of them stop messing with their phono stages. Never heard one so don't know what the fuss is about but it's the kinda fuss I want to mess about with one day
izzy wizzy wrote:
The LCR route is what I was aiming for but ran out of money. I wanted to pop it in in place of the 20k EQ. The Slagle coils are the way I want to go.
Over the years, I've noted that once people get to an LCR phono, the vast majority of them stop messing with their phono stages. Never heard one so don't know what the fuss is about but it's the kinda fuss I want to mess about with one day
cheers,
Stephen
Nick can speak for himself but he has moved away from LCR. Though it may not have been worse than this approach, it hasn't imprisoned him to not move on.
When Nick brought his LCR to an event it sounded great. Peter brought a solid state one he was possibly deciding whether to reverse engineer and rip off, and I thought it was almost the same as Nick's LCR. It may have been all Jfet, who knows? It had no valves.
Nobody else praised it so I may be the odd one out or they just didn't want to praise Peter's offering. Don't know which.
i probably will regret this, but mark emailed me and insistently questioned everything... (i'm exaggerating... my wife hates it) you all have been so overly kind towards me in this discussion, which i have just read, i feel obliged to answer one or two. please understand that i am a very opinionated fellow among very opinionated fellows. i don't have any interest in debating the merits of cap types or the sound of tubes. please use whatever you prefer. but i do love a good puzzle... it is very satisfying to get to some new problems for a change. the old ones are so boring and, well, problematic.
raising the voltage across the fets has several effects, not all of them good. the input impedance is changed and the gate to drain C increases and becomes more voltage variable. but the transfer characteristic is longer (straighter in the middle) which makes the simple distortions go down. 1/f and flicker noise go up, as well as johnson noise (temperature increases with drain voltage and drain current): uncorrelated distortions add, times the root/sum. and while fets are potentially an order of magnitude quieter than these high Gm VHF tubes are, when you are dealing with a MC phono cartridge or a ribbon mic, one does need all the help possible. if you mean to avoid a transformer in front, this is where the rubber meet the road.
clearly, lowering the voltage across the fet means a shorter, more curved transfer characteristic: more 2nd harmonic. but at 3 volts and under, there is almost no flicker noise in a long channel fet (audio fet). the 1/f noise is lowest, and the signal variable C at the input is minimized. at the signal levels one is concerned with, the increase in 2nd isn't huge, but it is real. depending upon the fet it can be 3 or 4 times what it is at 7 volts. but it is not that simple. non-linear distortion and modulation distortion are what make gadgets sound electronic. most interesting to me is that the third drops away. but not the higher order stuff.
the straighter slopes at higher voltages also reduce the higher order stuff, but the noise is measurably higher. some kind of balance is asked for... but where to make your stand?
i have made a stage where i couldn't tell the difference between a short and the output of the shorted amp stage. it was a fet/bipolar hybrid cascode. 8 2SK170s, and 1 2SC1775. 2.8 volts across the fets. 10 volts on the collector and a 60mA current source for a load. i folded the output with a 1K resistor to ground. i can only guess, but the eq. noise res. must have been 10 ohms or less. it measured perfect, open loop. i don't have a UPD or an audio precision... i have no idea what the THD or intermod was. very very low. it sounds very good, but not as good as a D3a Gm amp. and using the pentode with an input transformer gets me to 30 ohms eq. noise resistance. that is under 1 nV/root Hz for a tube front end.
the folded cascode is a pain in the ass. but it can be made to sound better than the standard one. the curvature of the transfer characteristic at 3 volts or less can be "compensated" with a deliberate curvature of the tube's slope. it is a complimentary circuit rather than compound circuit. like the standard cascode, the modulation sensitive aspects of the fets (input level sensitive variable capacitance) are reduced. obviously if you go balanced, the even order stuff falls away... the work you do to lower the odd order and uncorrelated stuff really comes to the fore.
that is my two cents. pfets have more C (less mobility in P material than in N) but i made my case for using them anyway above. it basically allows you to build a pentode with a distortion canceling transfer characteristic. but it is a lot of work. you can always make an ordinary fet/tube cascode, and make an adjustable bias with a range from 1.5 volts to 20... and see where it gets you. most would be happy, i'm guessing.
i will give a plug for the Gm amp. it works for both cascodes and pentodes. cascodes mu is lower (mu squared) but this is not much of an issue open loop. if you are using a feedback design (heavens above!) the "blameless" nature of the circuit means you get the most out of it. i can only recommend it.
and finally, most engineers and audio hackers are so comfortable in the thevenin world view (voltage model). i'd like to give a bit of a plug for the norton vision. yes, ultimately, they become the same outcome. but there is one important difference. in the voltage model, we mainly look "up" into the load. for practical devices, tubes, fets, bipolar transistors... this often means trying to get them to act like short circuits pushing on the work we intend for them to do. a low plate resistance or voltage feedback to reduce output impedance become crucial... none of the devices we must use are particularly effective no impedance devices. coming from the other direction (open circuit pulling on the work) puts less reliance on ideology, and more on practice. an infinite impedance and a 800K ohm impedance aren't as demanding a difference with the devices we must use when we are dealing with 600 ohm load.
it is often easier to flip the value system on its head when looking for a new perspective.
Nick wrote:LCR still does something that RC doesn't, but the one I have is behind the curve now, one day I will have time to bring it up to date.
I knew I was going to be in trouble when I posted here about LCR being the end Don't know the topology. Most LCR phonos use high gm valves for a low Z circuit. Thomas Mayer reckons it's the low Z nature of the thing that is the key more than the LCR bit. Also it might have something to do with how sensitive you are to some component type colourations.
OTOH, maybe it's the sheer investment in it that means it is a final destination for many