Adventures with the DS-1

[zbt]

From Kaba Yaro

We want more second harmonic   

[Rob Strand]

From Kaba Yaro

We want more second harmonic   

The site gives a good account of the differences.

The 220k vs 470k on the second transistor (the gain stage) is a significant difference.

[Rob Strand]

The site gives a good account of the differences.

The 220k vs 470k on the second transistor (the gain stage) is a significant difference.

I'm pretty sure that site is wrong about which transistor the 220k is on.
The 220k shown on the pics is R2 (the base resistor on Q1) not R7 (the CB resistor on Q2).
That's not going to make a big difference.

[zbt]

DS-1 service notes

I double check the schematic between DS-1 and DS-1A, the only differences is

      DS-1    DS-1A
C5    0.47    0.068
C7    250P    100P
C8    1       0.47

Yes the 220k is R2

I don't see any remedy for pull resistor, M5223AL (pull up resistor) and NJM2904L (pull down resistor)
I guess the easiest is to replace the op amp

DS-1 history

looking TA7136 equivalent circuit, it used ccs on output stage, so I think it would be no crossover distortion

and total current Q3 + Q1 + Q9 + Q11 + Q13 = 3.1 -- 4.2 mA (TYP -- MAX)
may be for distortion Q4 and Q5 no need to be linear or same

[Rob Strand]

looking TA7136 equivalent circuit, it used ccs on output stage, so I think it would be no crossover distortion

The output stage of the TA7136 is class A.  It has a current source on the negative side.  So we don't have the crossover distortion from a common class AB and class B output stages used on opamps.  IMHO the TA7136 is still an opamp, it just has a different structure to common opamps.

The discrete opamp used in the WAZA DS1w and the discrete opamp I posted recently both have the current source on the output stage.  It seems to contribute to the shape of the output waveforms.  However it could turn out we don't need it - that's going to need more A/B testing.

https://www.diystompboxes.com/smfforum/index.php?topic=133230.0

One of the DS1 PCB pics I have has an extra ceramic cap between the opamp and the where the wires connect to the PCB.  I can't rememeber what that difference is.  I'd have to go through my old notes.

[PRR]

TA7136 has open-loop gain of ~~90dB  - at 1kHz!! That seems to be a 30MHz gain-bandwidth product, unheard of for a general purpose opamp in the 1970s. But the 3mA A-output does not like less than 10K load at 30V supply, opposed to 2k for about any general purpose opamp of the day. But at 9V the TA7136 should drive 2K or less, with perhaps some small loss of gain.

[Rob Strand]

TA7136 has open-loop gain of ~~90dB  - at 1kHz!! That seems to be a 30MHz gain-bandwidth product, unheard of for a general purpose opamp in the 1970s. But the 3mA A-output does not like less than 10K load at 30V supply, opposed to 2k for about any general purpose opamp of the day. But at 9V the TA7136 should drive 2K or less, with perhaps some small loss of gain.

If you look at the TA7136 test circuit for open loop bandwidth you can see the gain at 1kHz is very much set by the test circuit.    Feedback resistor 820k,  cap to ground 10uF.   XC (@ 1kHz) = 15.9 ohm,   gain (@ 1kHz) ~ 820k / 15.9 = 94dB.   The actual gain is a little less than calculated due to the opamp gain. 

There is an open loop gain plot in the datasheet, you can see the rising gain at low frequencies due to the test circuit cap.   The peak gain, which is more like the LF gain, is higher than the 92dB spec at 1kHz

The wide bandwidth is only for the uncompensate opamp.  When you add the required compensation caps the bandwidth drops.  (In the same vein as the LM301A.)   According to simulations of the uncompensate opamp the roll-off is 2nd order at high frequencies, which lowers the 0dB frequency but it's still over 15MHz.

[Rob Strand]

DS-1 service notes

I double check the schematic between DS-1 and DS-1A, the only differences is

Here's the deal with the two ceramic cap vs three ceramic caps I mentioned earlier.  I don't think I've ever seen it documented.

This only applies to TA7136AP based units (so DS-1 and not DS-1A) and the difference implies MIJ vs MIT units.

The non TA7136AP units have a single ceramic cap in that area.   You have to check the opamp to refine the version further.
The very early units have two ceramic caps but there's a few other differences, and the PCB is different.



There is a potential feedback stability issue when the drive is backed off and the opamp operates in unity gain.   I'm not convinced how successful the MIT revision is but it does something towards the problem.

One of the schematics has R21 = 100k.  I'm fairly convinced that's a typo it should be R21=10k.

[m4268588]

Small differences in detail.

       9.88V
         ^
         |
        10k
         |
  +-470k-+
  |      C
--+-----B
  |      E
100k     |
  |      22
  |      |
 _________
 /////////

	2SC732-BL,	2SC2240-GR
hFE	365	362
C(V)	4.28	4.13
B(V)	0.626	0.584
E(mV)	11.8	11.9

[m4268588]

I realized that the diode clip should work with the smallest signal input.

[antonis]

Shouldn't both R14 + C10 be  across D4/D5 to form a Zobel network..??

[Rob Strand]

Shouldn't both R14 + C10 be  across D4/D5 to form a Zobel network..??

The simplest Zobel network is a cap in series with a resistor.  R14 and C10 more or less in series.   The diode can be removed for small signals,  C9 is a large and pretty much short at high frequencies, the tone control is ignored.  When the diodes conduct there is an outright load of R14, which also reduces the loop gain of the amp.  (I'm just using term Zobel network as a generic catchall term here.)

The main point is the unity gain stability is helped by the load.   Maybe not wise to raise the value of R14.

The TA7136AP datasheet offers very little guidance regarding stabilizing the amplifier.  The RIAA and tape-head amps with a gain of 100 are the only examples.  Apparently tone controls with the TA7136AP aren't reliably stable.  With the gain control down the DS-1 is operating at unity gain - that's is the worst case to stabilize.  There's nothing in the datasheet at all about using the amplifier in unity gain never mind stabilizing it.   The only hint is the RIAA network goes to unity gain at high frequencies *BUT* the network loads the amp with 1k ohm at high frequencies.

After making a spice model for the TA7136AP, it's clear the chip is hard to stabilize.   I suspect you need to start adding resistors in series with the caps to help it more.   

Back in the day any designs other than the examples would have to be stabilized by experiment and it's likely not to be successful.

When I see the addition of C24 it's like someone at Boss had another go at stabilizing it.   Maybe they had problems with later devices oscillating in production.

[Rob Strand]

Shouldn't both R14 + C10 be  across D4/D5 to form a Zobel network..??

Your post made me realized I really need to look at the stability of the TA7136AP in detail again.

I did a quick check and the added cap C24 looks like it makes the stability *worse*, but not by much.

One issue I had, and still have, is how much input capacitance the chip has.   The 100k resistor (R11 on the DS-1 schematic) on the TA7136AP inverting input screws up the stability.  If the chip's input capacitance is remotely high it's very hard to stabilize.

[FleshOnGear]

This is all fascinating, if maybe just a bit over my head as far as the theory goes. I just want to make sure I get this straight. I'm guessing that R11 is there to keep the op amp relatively balanced, considering R10. So, is the issue that the op amp's open loop gain has a certain phase shift, then the capacitance of the inverting input along with R11 pushes the phase shift further? Ending up with positive feedback? Could a cap across R11 help? Thanks for any clarification.

[Rob Strand]

I just want to make sure I get this straight. I'm guessing that R11 is there to keep the op amp relatively balanced, considering R10.

I used to think that as well.  R11 does balance the impedances and balance the DC offsets but the offsets are small and not really relevant at all for audio.   You do see a lot of audio designs following the text book but it doesn't achieve anything - R11 could be a left over from that type of thinking.

After that I was thinking the balancing helps reduce scratchy pot noise when the gain pot is varied.   However, it doesn't help with that either.   The DC drop across the pot, which causes the scratchness, just adds to the existing DC offset.

The down side of R11 is it adds quite a bit of noise and it degrades the stability of the feedback loop.

So you start to ask yourself.  Why keep R11 (100K) at all?   More so, why add it to the later designs as well?

The only rational thought was it adds some protection to the input stage.   At 9V even that might not be justified.   For the non TA7136 designs that reasoning doesn't make much either because they use input stages which handle higher voltages.

So, is the issue that the op amp's open loop gain has a certain phase shift, then the capacitance of the inverting input along with R11 pushes the phase shift further? Ending up with positive feedback? Could a cap across R11 help? Thanks for any clarification.

Yes, that's the problem.   Yes a cap across R11 is the solution to the stability issue.   I'm trying to analyze the circuit as is so R11 and its affect must be left in.    From a design perspective it would help to remove it or at least reduce it to 10k or less.

The frequency region concerning is way above audio.    When you have a low pass filter in the feedback loop there is phase-shift at 1/10 the cut-off frequency, f3= 1/(2piRC).  Those two aspects are at odds with each other.  So with R=R11=100k it doesn't take much capacitance to have an effect - it's just asking for trouble.

One of the problems coming up with spice models for opamps is you have to assign capacitances and sizes to the transistors inside the chip.   It's a bit of a shot in the dark.    I'll admit my TA7136 spice model isn't quite there and it doesn't help that I don't have any TA7136 devices to measure.

[Rob Strand]

FWIW, I replaced some of the transistor models with models from similar discrete IC simulations.   The new models have lower input capacitance.   As expected the 100k (R11) has far less impact on the stability of the feedback loop.    Do the new models represent reality for the chip?  I don't know but given the benefit of the doubt it does seem possible the 100k can be present without it going unstable or nearly unstable.

[ElectricDruid]

I don't know but given the benefit of the doubt it does seem possible the 100k can be present without it going unstable or nearly unstable.

To summarise, "May not blow up in all circumstances" still doesn't sound like a ringing endorsement! 

[Rob Strand]

To summarise, "May not blow up in all circumstances" still doesn't sound like a ringing endorsement!

At least it's only a simulated blow up.  We know the Boss actual devices seem to work.  However, the audio forum guys have complaints even some commercial devices using TA7136's are marginal.

[zbt]

C1 R2, no diff < 20Hz
C7 > 20kHz for bat ear

C5 R8 not quite sure
470nF 10k  33.86Hz  DS-1
68nF 10k  234.05Hz  DS-1A

R9 22ohm value so low
IE = IC = (9.88V - 4.13V) / 10k = 0.575mA
re = 25/0.575 = 43.478
rb = hfe(re+RE) = 362(43+22) = 23530
47nF rb//100k = 47nF 19k = 178.23
(not include 470k)

assuming 2SC732-BL much lower hfe (250) (DS-1)
rb = hfe(re+RE) = 250(43+22) = 16250
47nF 14k = 241.9Hz

DS-1  in 241Hz out 34Hz 
DS-1A in 178Hz out 234Hz

DS-1 cut at input, DS-1A cut at output?

For C8 I'm a little confused if it's combined with R11,
will the filter become like the T model?

this from boss-ds1-analysis

C8 R13 VR1
  1uF 4k7 104k7  33.86Hz  1.52Hz  DS-1
470nF 4k7 104k7  72.05Hz  3.234Hz DS-1A
not like DOD 250 or Tube screamer?

C1 and C8 can equate its value with DS-1
C5 due to hfe adjustment?
C7 stability adjustment

Can we be sure that the only factor left is the IC?

YouTube
may be I am just wrong, cause is all the same for me 

[Rob Strand]

C1 R2, no diff < 20Hz
C7 > 20kHz for bat ear

C5 R8 not quite sure
470nF 10k  33.86Hz  DS-1
68nF 10k  234.05Hz  DS-1A


assuming 2SC732-BL much lower hfe (250) (DS-1)
rb = hfe(re+RE) = 250(43+22) = 16250
47nF 14k = 241.9Hz

DS-1  in 241Hz out 34Hz 
DS-1A in 178Hz out 234Hz


Can we be sure that the only factor left is the IC?

YouTube
may be I am just wrong, cause is all the same for me 

You have a few bugs in your analysis.  Check the part values against the original schematics.
Also the 2SC732-BL device is high gain.  m4268588's measurement looks in range
 https://www.diystompboxes.com/smfforum/index.php?topic=133112.msg1299055#msg1299055

Not being a DS-1 connoisseur I'm a little skeptical about the post filtering 470n+100k vs 68n+100k.  Sometimes you can hear such changes on some setups.

FYI, I put up a detailed analysis of the DS-1 transistor gain stage in this post.   It's not as straight forward as you might expect.  Also, things change when the stage clips.

https://www.diystompboxes.com/smfforum/index.php?topic=133175.msg1298133#msg1298133


The general complaint is the non TA7136 sound more fizzy.   Some worst than others and the complaints follow the opamp type more than anything.  We know different opamps can have subtle differences in the high end.  So that is likely.  I've seen people complain about the MIT vs MIJ TA7136 units - IMHO that's debatable.

Also opamps with lower slew-rates can sound smoother.   In the DS-1 circuit the TA7136 has a low slew rate.  So if I had to guess I'd say that was a factor.