DAC AD1862: Almost THT, I2S input, NOS, R-2R

@Zoran I wonder how 6E5P is going to sound in this application. I tried it as a driver for 6S33S and I did not like the sound at all - 6S33S sounds amazing with other tubes.
In general, high transconductance tubes, with wide open anodes, sounds more as solid state than real boxed classic anode construction (like ECC40 among others). Actually 6C33S is one of them 🙂
BUT
there is a use of 6E5P in very small signal environment. Opposite than in driver arrangement.
There is present small signals, and no capacitive load at the output as present with driver?
Based on this major diff. expected sound outcome can be diff. too?
...
I will post another topology, also simple with grounded grid. For this aplication high transconductance is a must. And at laeast we can compare simulated results...
.
The best way s to try?
.
Advantages of these high transconductance tubes are:
good amplification capabilities - lower Riv value
low internal resistance and ouput impedance as parallel Ri II Rload, under 1K = avoiding additional buffer
(but still cant drive solid state amps with 10K at the input...)
correct phase after phase shift @ Riv
Simplicity (if we lokk at the other side of Power Supply 😎)
 
Last edited:
  • Like
Reactions: andyjevans
Everything for me is tonality - acoustic instruments, voices, Steinway pianos.
Yes...
I suggest to try Germanium IV and small signal circuits 😍
I was in state of shock btw...
few days ago i have conversation with a friend we listening some Ge devices and talk infront of speakers.
Does not listened seriously at all, just chat with some vocal music present.
One of us said - how is voices from speakers somehow naturally matched with our live human voices in room...
I cant explain why...
ECC40 is king in my system, on the basis of tonality.
It has boxed, closed anode. Classic construction. Linear. Robust glass baloon, better than noval package.
Good amplification factor for IV following circuit.
BUT typical, high Ri. Not so high, still high to incloude a buffer after amplifying section. Could be with same type of internal construction with 5687 tube. ECC40 cant drive smaller Rin of the next stage than 220K.
It can drive without buffer higher Rin of next Tube or JFET devices only...

If you design on the basis of theory you would use a valve with a lower Rp, like 3K, and put 6mA or more through it. But that won't necessarily give you the right tonality. Only listening to the valve stage in your system and with your familiar test tracks will tell you what it actually sounds like.
It is not like guessing and telling a stories without techical arguments first? Every tube has strict electrical parameters and should be firs consult with reasoning this party. Then context where it will be put into, then internal topology etc etc. Before listening.
 
Last edited:
  • Like
Reactions: andyjevans
Hi Zoran. As you say, you can't design a tube stage without the maths. But in my case, when I have the right operating point for the setup I'm using, I go to the listening stage. I may then go back and change the operating point and listen again. This usually happens with the anode resistor. I like to have 5x the Rp, so for instance 39K to 47K for 11K Rp. I may need to increase the B+ if I have the right transformer, but if I have 325V B+ which is typical, then I may slowly decrease the anode resistor to around 34K to increase the current. I find that the value of the anode resistor is just as important as the current. The anode resistor value was absolutely critical with the 26 tube - it had to be 5x the Rp. Other tubes can be more tolerant.

Higher transconductance tubes can be easier to work with and you have a lower output impedance. I haven't really found any that I like as much as the ECC40, which as you say needs to drive a following tube stage. The E80CC is good, but ECC40 is better to my ears. I bought a 6463 to play with since Audio Note used it and people seem to like it, but I haven't listened to it yet. Gain (mu) is only 22 so borderline. Plenty of fun to be had with tubes!
 
In general, high transconductance tubes, with wide open anodes, sounds more as solid state than real boxed classic anode construction (like ECC40 among others). Actually 6C33S is one of them 🙂
6S33S sounds like as solid state if you go beyond 200V B+. Try it at 200V!

I am tempted to try 6E5P in IV. But also 6AC7 which sound I really like. Need more time!
 
  • Like
Reactions: andyjevans
I listened several different amps with 6C33S and they was OK in general? No one was bad. They has the charm.
I think one of them was some itteration of the one from romy site?
Didnt made any amp with 6C33S.
But personally, I am more to the side of 10Y, 2A3 etc amps.
.
I am tempted to try 6E5P in IV. But also 6AC7 which sound I really like. Need more time!
as I wrote 6E5P in iV is in total different position then as the driver. BUT maybe is better to use 6E6P.
.
I will try to make some IV ampli circuit for 6AC7 it looks OK as pentode.
 
  • Like
Reactions: andyjevans
Hi,
I paste here a post of Thorsten Loesch from end of August as I asked for tubes for TDA1541A as well. Migth interrest some of you too for tryingg to translate to the AD1862 nd else :

"
If people interrested, I could propose as well for instance on a standalone pcb an hybrid-Mufollower with the ECC88/E188CC/6N23P-EV tube if readers are interrested.

I would suggest 6N3/5670 or even better 6J1/6AK5 instead. Hybrid Mu... Hmmm. Make it a real one please.

6N6/6N30/5687WB (pins are a LITTLE different!) with E-822 (8.2mA) CRD or whatever is desired as current in the cathode as upper part, extra MOSFET CCS to pull more current out of the cathode.

Then our signal tube below.

Maybe 4 sockets for the signal tube(s), 2 x 7pin for 6AK5, 1 x 9 pin for 5670 and 1 x 9 pin for ECC88 style tubes. Connect grids dnd snodes together, separate cathode resistors to get bias "just so".

And one more 9 pin socket for 6N30/5687 with a switch to select.

....

....

Resistor I/U conversion with BJT low noise 2mA offset CCS and resistor between AoL/R and AGND. "
 
@diyiggy It's good to know what other people liked. From my experience just knowing what tube might sound good is not enough, same tube may sound very different in different designs and some of them are very sensitive to the operating point. Experimenting with a single tube may take weeks. Life is too short!

I tried LL1660 (inspired by Lampizator) with 6DJ7 and 5687. I think I like Michelag/Grunf design better even with cheap caps. Experimenting with these transformers now https://www.aliexpress.us/item/2251...st_main.5.562d18024ffe8E&gatewayAdapt=glo2usa. I will post here if I find anything interesting.
 
+100%

I tried mu folllower Moglia style thanks to @gaszto (with some very sligthy layout mods).

I have an expensive E188CC/7803.

But tubes are very too much expensive for I and I don't think ultimatly it is better, and we learned now they could need some sand to. OT or good DC coupling migth be expensive, it adds a traffo more. Even the Nichicon Muse ES bipolar here are colored despite the very low noise floor, they add someting when used with solid states parts.

I like also the fact than some have tried both tube and sand, I have the feeling there is a progression line in this thread.

All designs have drawbacks. But I am against transformers after a dac chip. Most are not good enough and anyway it needs some active to sound good at the end.
It is interesting some prefered CFA here despite having some problems for that I/V task. The fact is current conveyor and buff at input and not sensible to input capacitance is interresting.

I surmise most will judge an I/V only on the tonal balance and of course according their whole hifi layout. Less people talks aboout others issues like bass dynamic, harmonics fatigue, soundstage. And for sure it i almost impossible when you are not in the room with the guy !

I regret there iss no more "cousu main haute couture" made discrete i/v stage. But I am not able to do this. Playing with op amps is the most I can accheive, lol. I focus on the layout quality of DACs though. I spend a lot of time about this the last 10 months.

Problem is certainly most can not design sota post DAC stages, me the first, some think they can, but a simple opa amps flows their 20 years experience.

Reality is hard. Those PCM chips can dig a lot of the newest SD chips if more great care will be shared w/o finishing to bad copies of DACS as we see here 99% of the time (Distinction, Gabster,, Audio Notes, etc). And when some good guys tries (I mean the few that are knowing like T Loesch, it is trolled by egotic beotians and soon mods are flooding the baby in the bath trigger the shower for the peace, putting all the good minds in the same bag.

I may finish with a mortage and buy a Mola Mola or multibit channel PCM DAC (that should have been done here from several years, crazy no one focused on that nowadays and one has to look at bad MINIDSP Flex 8 with so-so trade offs to stay cheap - while the goal is brillant).
 
Last edited:
I made some model for 6AC7 in triode mode. Based on Tom Schlangen measurements.
It could be used as amplification segment after Riv. I will try to simulate in spice elements in the circuit.
The tube in triode mode is linear, just a little less than 6E6P, higher internal resistance stil enough for driving high Rin of the next stage. For universal use nedds a buffer...
Can be set on lower voltages or a bit higher with most common present power suplies under 350V.
.
6AC7 tools settings 01.png


6AC7 tools settings 02.png


Code:
****  6AC7_triode ******************************************
* Created on 12/22/2024 09:28 using paint_kit.jar 3.1
* www.dmitrynizh.com/tubeparams_image.htm
* Plate Curves image file:
* Data source link: 6AC7_triode measured by Tom Schlangen
*----------------------------------------------------------------------------------
.SUBCKT 6AC7_triode 1 2 3 ; Plate Grid Cathode
+ PARAMS: CCG=11.7P  CGP=3.4P CCP=5.7P RGI=2000
+ MU=52.6 KG1=285 KP=248 KVB=102 VCT=0.388 EX=1.358
*----------------------------------------------------------------------------------
* Vp_MAX=400 Ip_MAX=30 Vg_step=0.5 Vg_start=0.5 Vg_count=17
* Rp=10000 Vg_ac=0.085 P_max=3 Vg_qui=-2.4 Vp_qui=180
* X_MIN=61 Y_MIN=56 X_SIZE=844 Y_SIZE=485 FSZ_X=1711 FSZ_Y=674 XYGrid=true
* showLoadLine=y showIp=y isDHT=n isPP=n isAsymPP=n showDissipLimit=y
* showIg1=n gridLevel2=n isInputSnapped=n 
* XYProjections=y harmonicPlot=y dissipPlot=y
*----------------------------------------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LOG(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G  ; TO AVOID FLOATING NODES
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RCP 1 3 1G   ; TO AVOID FLOATING NODES
C1 2 3 {CCG} ; CATHODE-GRID
C2 2 1 {CGP} ; GRID=PLATE
C3 1 3 {CCP} ; CATHODE-PLATE
D3 5 3 DX ; POSITIVE GRID CURRENT
R1 2 5 {RGI} ; POSITIVE GRID CURRENT
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS 6AC7_triode
*$
 
6Ж4
Russian tube equivalent to 6AC7. Very good tube, excellent build, afordable, I suggest this tube as 6AC7.
On this link we can see the internal structure.
I made sice model based on tube datasheets. Two original measurements in triode mode are virtualy is tha same.
Again, tube is very linear, low distortion in this Riv aplication, moderate ower supply, no expensive elements. Etc...
this tube has separate pins for each electrode and can be configured on 2 ways in triode mode.
I presume that in factory measurements G3-K, G2-A.

.
6G4_ss.jpg

This is shape of the anode surface i captured from the video. This is open 2 plates anode.
6Ж4 anode.png


For now, static graphs and spice model only. Latter I will try to put model in spice...
.
6Ž4 orig tools 01.png


6Ž4 orig tools 02.png


Code:
**** 6Ž4_triode  ******************************************
* Created on 12/22/2024 15:15 using paint_kit.jar 3.1
* www.dmitrynizh.com/tubeparams_image.htm
* Plate Curves image file:
* Data source link: origial datas triode mode anode chrs
*----------------------------------------------------------------------------------
.SUBCKT 6Ž4_triode 1 2 3 ; Plate Grid Cathode
+ PARAMS: CCG=11.7P  CGP=3.4P CCP=5.7P RGI=2000
+ MU=41 KG1=270 KP=278 KVB=1260 VCT=-0.2 EX=1.4
*----------------------------------------------------------------------------------
* Vp_MAX=360 Ip_MAX=40 Vg_step=1 Vg_start=1 Vg_count=8
* Rp=10000 Vg_ac=0.09 P_max=3 Vg_qui=-1.2 Vp_qui=125
* X_MIN=82 Y_MIN=32 X_SIZE=1161 Y_SIZE=513 FSZ_X=1883 FSZ_Y=661 XYGrid=true
* showLoadLine=y showIp=y isDHT=n isPP=n isAsymPP=n showDissipLimit=y
* showIg1=n gridLevel2=n isInputSnapped=n
* XYProjections=y harmonicPlot=n dissipPlot=n
*----------------------------------------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LOG(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G  ; TO AVOID FLOATING NODES
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RCP 1 3 1G   ; TO AVOID FLOATING NODES
C1 2 3 {CCG} ; CATHODE-GRID
C2 2 1 {CGP} ; GRID=PLATE
C3 1 3 {CCP} ; CATHODE-PLATE
D3 5 3 DX ; POSITIVE GRID CURRENT
R1 2 5 {RGI} ; POSITIVE GRID CURRENT
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS 6Ž4_troide
*$
 
Last edited:
As I mentioned I did some measurement:
Raspberry Pi 4 (i2s out) - IanCanada FiFoPi v2 - my AD 1865 DAC HAT (link) - 39R I/V resistor - Sowter 1495 (1:9) - DIY Forn End 2022
Measurement has been done by Cosmos E1DA + Cosmos Scaler

44_1kHz.png

96_kHz.png

192_kHz.png

AD1865_Sowter_FR2022.png


Disregard the crosstalk, sometimes it gives bad values. Cosmos E1DA switching to mono if no signal at channel 0. That's the reason why L-R and R-L are different. In the reallity it should better than -78dB
1734904111982.png

AD1865_Sowter_FR2022_THD.jpg

Basically THD is according to AD 1865 datasheet.
 
Last edited:
As I promised, one grounded grid IV circuit with 6E5P tube. But can be 6E6P used too. Key word is transconductance.
These JFETS on the schematics are best performeres at least in spice... Circuit is iesy to set to 0mv to the DAC Ioutput.
Distortion is very very low, and for almpst any value of this time, external Riv. Minimum of elements required. No complementary, no push-pull no globala FB. All SE...
.
BUT the Output is out of phase. -180deg.
That is for OK for use with tube preamps, one stage, which are also inverting phase, so the right phase will be on amplifier...
.
Output impedance at 1 K to about 100ohm is equal to RIV value. And it can be used without buffer.
For RIV = 1200 ohms Zout = 1200 Ohms, Vout p-p=2.4Vp-p with 2mAp-p DAC Iout current.
.
Best version of this topology was with classic ECC/PCC88 parallel sections, one tube per channel.
.
6E5P_GroundedGrid_IV.jpg
 
  • Like
Reactions: skrstic
One standard topology with russian 6Ж4 pentode in triode mode.
.
Very good behaviour, low Zout, Linear, small disttortion, in-phase, very good amplification Riv can be lower,
Moderate Vb (power suply voltage), very low phase drift at the end of BW but with given values only,
with other values of C the phase will be hardly degraded and sound too.
So 6Ж4 is very useful tube.
;
For all AD, and PCM old DACs with 2mAp-p RIV is 68-to 91 ohm
For PCM63 with double Io 4mAp-p, RIV value is 1/2 for same Output Vp-p, for instance 33-47 ohms.

.

6Z4_after_Riv_SCH.png
 
  • Like
Reactions: skrstic
One standard topology with russian 6Ж4 pentode in triode mode.
.
Very good behaviour, low Zout, Linear, small disttortion, in-phase, very good amplification Riv can be lower,
Moderate Vb (power suply voltage), very low phase drift at the end of BW but with given values only,
with other values of C the phase will be hardly degraded and sound too.
So 6Ж4 is very useful tube.
;
For all AD, and PCM old DACs with 2mAp-p RIV is 68-to 91 ohm
For PCM63 with double Io 4mAp-p, RIV value is 1/2 for same Output Vp-p, for instance 33-47 ohms.

.

View attachment 1397747
Very good for ccs, this one.

The other one with grounded grid, those jfets serve as ccs (sink) ?
 
@Zoran Rp for 6AC7 is 1M (as per datasheet), I cannot find a datasheet with Rp for 6Ж4. From your model I see it's 10k. It's interesting, since 6AC7 and 6Ж4 should be very similar. @andyjevans mentioned 5x rule - I wonder, how does the model look like for RL at 50k?
Just for clarify:
Rp is derived from R plate that actually mean Internal resistance of the tube and real name is Ri
R on P (as plate), is R load at Anode. And that is actual Anode load resistance. Mark is RL because it is more clear.
Ri = internal resitance of tube
RL = external load resistance at anode
...
Ri of pentode is very high, but the very same pentode connected in triode mode, has significantly lower Ri,
Also significantly lower amplification factor...
But as I wrote this is not Load resitance.
...
So
It is not true that rule 1 : 5 = Ri : RL exists. Rather some mathematical deprivation for maximum efficiency in literature is 1 : 2 = Ri : RL
But
that is not for minimum THD, Optimum static load line, considering Input signal value, considering output signal, etc.
...
In short words we have an context where we need to put and set the tube, and we have some pre-designed parameters for tube to fit with own characteristics...
And it is not guessing at all if we want success in sound outcome.
.
Take a look at the staic chrs of tools software graphs where RL is present and You can see very narrow region of real opperation of the tube regarding of input signal...
 
Very good for ccs, this one.
If You think about Gyrator (sort of CCS in Anode, as the Load) then yes.
But it is an active circuit with dozen active elements of JFETs MOSFETS and BJTs and dozen of also additional pasive elements. On the othe side we have one tube... Considering this fact that is more solid state circuit than tube?
And all that for repalcing, with a good reason, complicated and non faceable L as the load...
With specific outcome of sound signature of very high level load almost parallel with voltage axes.
.
As for more tube design we have SRPP or Mu Follower circuits, that actively load the anode, and circuit has in the same time low output resistance.
.
For instace we can choose higher RL than 10K to match higher already present power supply of say 350V
Without changin a lot sound of the circuit.
That will be 350V - 317V = 33V and with 18,17mA resulting in 1816 ohms additional to 10K
about 11.8 to 12K R load for 350V PS
Bare in mind additional dissipation too 🙂

The other one with grounded grid, those jfets serve as ccs (sink) ?
Yes, these 4391 are Constant current sources in negative supply branch for stability and better PSRR. Cascoded and paralleled for match the total current flow and minimize dissipation. With changing the bias trough we can set the off-set to 0mV @ Iout pin of the dac. Protection diodes are added at the DAC Ioutput.
(In sims they are giving even slight better results than witout... No reasoun not to believe that in real world will be the same)
Other JFET in cathode is for biasing the tube.
Transconductance of the tube is keyword with that circuit and 6Z4 has satisfied.
Note that total Vb is moderately low and achievable.
.
This circuit is interesting because for very low values of Rload as RIV we can obtain external RIV that is not directly loading Ioutput of DAC. In this type DAC has going to very low impedance input of the circuit.
 
Last edited:
  • Like
Reactions: Michelag
Just for clarify:
Rp is derived from R plate that actually mean Internal resistance of the tube and real name is Ri
R on P (as plate), is R load at Anode. And that is actual Anode load resistance. Mark is RL because it is more clear.
Ri = internal resitance of tube
RL = external load resistance at anode
...
Ri of pentode is very high, but the very same pentode connected in triode mode, has significantly lower Ri,
Also significantly lower amplification factor...
But as I wrote this is not Load resitance.
...
So
It is not true that rule 1 : 5 = Ri : RL exists. Rather some mathematical deprivation for maximum efficiency in literature is 1 : 2 = Ri : RL
But
that is not for minimum THD, Optimum static load line, considering Input signal value, considering output signal, etc.
...
In short words we have an context where we need to put and set the tube, and we have some pre-designed parameters for tube to fit with own characteristics...
And it is not guessing at all if we want success in sound outcome.
.
Take a look at the staic chrs of tools software graphs where RL is present and You can see very narrow region of real opperation of the tube regarding of input signal...
Thank you for clarifying, I missed the part that plate resistance is specified in the pentode mode http://www.r-type.org/pdfs/6ac7.pdf. I think the 1:5 rule is based on listening tests - I am definitely going to try it with different tubes.
 
  • Like
Reactions: andyjevans