solid state pentode

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Here is the schematic for the first pass of the SSP preamp.

This is based on the self bias CCS from Gary Pimm. I had been using his CCSs for tube designs for awhile and was thinking about how they worked, its essentially a voltage reference and a follower that keeps that voltage constant across a fixed resistor, hence a CCS. With Gary's design you can vary the current by adjusting the voltage reference, I thought, well what would happen if you couple an audio signal onto that reference, you now have a CCS whose current is proportional to the audio signal, otherwise known as a transconductance amp!

Because Gary's CCS has such a low capacitance at the reference point its perfect for audio use, because of the way the cascoded CCS works the usual high and nonlinear capacitance of MOSFETs is negated.

In the circuit Q1 and Q2 are the main CCS, Q3 and Q4 form a low current CCS used to generate the voltage reference betwen Q1 and Q2. In this design I used 9V battery for the primary reference, but you can do it the way Gary does and use another resistor below R14 instead of the battery, but the input impedance is a little low that way, which is why I chose the battery for a first attempt.

Given the about 4V threshold on Q1 the voltage across R1 is 5V which means there is an idle current of 5mA. The transconductance is 1/R1 so in this case its 1000uS (micro Siemens) which with the anode resistor of 2.2K gives a voltage gain of 2.2. Notice that the voltage gain is R20/R1 no matter what the reference voltage is. One problem with this design is that the voltage gain also multiplies the voltage across R1, so in this case there is an 11V drop across R20, you have to take that into account when designing the power supply. If you go for more gain you also go for more voltage drop across R20.

The design sound better with more current, my present version runs at 30mA instead of 5mA, this sounds better but of course needs a lot more current. This also lets you go with a much lower output impedance. Say R1 = 150 and R20 = 330.

The input attenuator is a bit unusual, I wanted to have a balance control but didn't want a separate attenuator for that. What I came up with is a stepped shunt mode design, with the series resistor adjustable. When adjusting the balance the series resistor is decreased slightly in one channel but not the other, slightly changing the gain for one channel. This lets you have 2 db steps for the main volume control and .5 or 1 db steps for the balance control.

In subsequent tweaks I added a follower stage using almost exactly the same design, but without the battery or R20, the gate of Q1 gets connected directly to the drain of the first stage Q2 and the output is across the second stage R1. By using a low value for the second R1 it draws about 20mA which makes a VERY low distortion output stage that can drive very low impedance loads. It even makes a very good headphone amp!

This preamp sounds amazing, its very clear and open, incredibly smooth, yet highly detailed and lively

I also have a several power amp designs using this concept, but thats another post!

John S
 

Attachments

  • transconductancepreamp2.gif
    transconductancepreamp2.gif
    35.9 KB · Views: 2,225
Gary Pimm has pc boards available, these are the "self bias CCS", these are small little boards (you can see pictures of them on his web site). I built my first 4 or so by hand on perf board, then gave up and bought a bunch of boards from him when I realized I was going to be using a lot of them. You'll have to talk to Gary about board layout files if you want to do your own using his layout. I've done several boards using the circuit but laying it out myself for some of the more complicated circuits where it was going to get a little expensive to have a whole bunch of his little boards.

The really only two things to look out for when laying out the board, the resistors connected to the gates are "gate stoppers" and need to be close to the gate, the other is that there can be significant voltages on these circuits, one of mine is running at over 100V (they are designed for 500V!) so make sure the traces have a little more than normal "minimum spacing" for the board process.

John S.
 
Hi John

Thanks you for sharing the schematic! I have a dumb question. I recently got a power unit with 4 heatsinks, each containing 6 of those Mosfets

75 amp, 55volt, 0.007 ohm, N-Channel ultra fets, TO-220AB

My idea is to breadboard your preamp using this mosfets, which are already nicely mounted and so on. I looked at the basic data, and would say it would work, if I don't go over the 36V from the schematic.

Someday I also want to try the other mosfets, but right now I am short on money and time, but would love to give this solid state pentode a try!

Thank you for the attention

Erik
 
MOSFET choices

ERik,
Q3 and Q4 can not be conventional MOSFET's, The part is a depletion mode MOSFET. The parts you hace that are "75 amp, 55volt, 0.007 ohm, N-Channel ultra fets, TO-220AB" would be better suited to a power amp or power supply. One important characteristic for Q1 and Q2 is low capacitance, the best parts for that application, the IRF 820 has an input capacitance of 360pF while a device just like you have is 3200pF, never mind the ultraFET's are meant to be driven by logic levels, not really intended for linear use. The good news is that the FET's stated in the design are under $1USD each so you don't need to be rich to buy them.
 
Hi Miket

Thank you very much for the reply! Actually I already thought that those mosfets would not be ideal, I even opened the datasheet of the LND and saw a diode within the symbol... But then, I am a complete noob on this.

I will spare some of those heatsinks to build regulated Power Supplies for cathode heaters, that seems to be a nice use!

Erik
 
Mouser has the LND150 in stock, they have 2200 of them right now, that ouht to keep the DIY community going for a little while!

The IRF820s can be had in two versions, the original with the metal tab connected to the drain (I think its the drain) and the insullated tab version (IRFI820G). If you need to use a heatsink the insullated one is nice, you don't need any special hardware to insullate it from the heatsink. Mouser sells the regular version but you have to get the insullated one from DigiKey. It also costs several times as much, but given that you don't need insullators and shoulder washers or nylon screws they are worth it.

For the schematic as posted you don't need a heatsink, if you go for the 30mA version you might want one on Q1.

BTW I'm running a hybrid version of this as a 30 watt SE parafeed power amp with an 813 pentode as a third element in the CCS (see Gary's pages for how to do this).

Gary has a power amp using just SSPs that sounds really good.

Its an incredibly flexible circuit, besides preamps and power amps it makes a very good high voltage regulator, headphone amp, DAC output stage etc. I built a fully differential DAC output stage using these techniques, it sounds wonderful but takes a lot of transistors!
Not to mention the over 100V power supply and 25 watts. A little unusual for a DAC but boy does it sound good.

John S.
 
Sending the package from US to Indonesia will cost me about $ 60 because supertex doesn't have distributors here.

that's why I'm looking for the substitution that already available in our local store. been trying to find out it's equivalent but no luck so far.

Thanks for your response,


regards,
Gede
 
Hi,
A little aside question. I am curious about Gary's choice of FETs for the original CCS. It isn't clear to me how the circuit eliminates the effect of the device parasitics.

I would normally expect bipolars to be used for CCSs as they have much lower parasitic capacitance and generally higher output resistance than similarly powerful FETs.

Were these originally chosen for high voltage or some other reason? I haven't seen the original application in a tube amp.
 
i am curious, also. looks like the lnd150 cascode ccs is used to supply apprx 0.3mA. if you have to use jfets, couldn't you also use a pair of cheaper 2n5457 jfets? i understand the VDS max is a LOT lower, so you might have some negative impacts on noise performance and capacitance of the jfets. hmm, maybe not with only 36v across a pair ...

mlloyd1

traderbam said:
Hi,
A little aside question. I am curious about Gary's choice of FETs for the original CCS. It isn't clear to me how the circuit eliminates the effect of the device parasitics.

I would normally expect bipolars to be used for CCSs as they have much lower parasitic capacitance and generally higher output resistance than similarly powerful FETs.

Were these originally chosen for high voltage or some other reason? I haven't seen the original application in a tube amp.
 
Sorry I've taken so long to reply, things have gotten very busy recently.

The capacitance on the input is low becasue the voltages on the source and drain of Q1 track the input voltage. If the gate goes up a volt and both the source and drain also go up by a volt, there is no potential difference so no charge is transfered hence zero effetive capacitance. Of course in reality its not perfect, due to the non infinite transconductance a small offset voltage does develope so it does still see a small capacitance, but its MUCH lower than the device capacitance.

Q1 is a source follower so the voltage across R1 tracks the gate. Q3 and Q4 form a small CCS which develops a constant voltage aross R14 (referenced to the source of Q1) which couples to the gate of Q2, this causes the drain of Q1 to also follow the input voltage. The voltage gain of the circuit shows up across Q2 so its isolated from Q1.

C5 was originally so large because that forms a "slow turn on" effect whih is useful for tube circuits. In this case its actually a hinderance to start up so slowly, I'm using 2.2uf in my current version.

On the parts selection question, the circuit was originally designed for tube circuits running at hundreds of volts so parts with high voltage capability were chosen. In this much lower voltage application you can certainly use lower voltage parts. I don't have any experience with other parts, but you should be able to use just about any small signal depletion mode mosfet . R17 will probably need to be adjusted so you wind up with 9-10V across R14. (or you could adjust R14, the value is not critical) I would probably go with at least a 60V transistor here.

Besides the voltage rating the LND150s have a fairly tight range on the threshold voltage which means that you usually do not need to adjust R17 or 14 to get the voltage across R14 right. Most other parts have a wider manufacturing range which might lead to tweaking one or the other resistor value on a circuit by circuit basis, but since you would be adjusting the ressitors anyway that probably should not be an issue.

John S.
 
I've spent a few months analyzing the ef86 , much still to be learned there. Think I saw your link on Cvac a while back. Also studying the modern texts ie "Guitar Overdrive' , 'Hi-fi Preamps' etc. Also Looking at pentode patents from 50's 60's too.

Still it seems much is open/unknown due to three grids on example ef86.

ie dynamic grid interactions, compression , supply range and so on. I built a custom test fixture to analyze all.

So now I have even more questions than answers!

The solid state, hybrid and software emulations are revalent from standpoint of work by others can give additional data.

The 'multiplier' ability of valve pentodes seems under utilized.

Thanks
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.